Zsh has command line editing, builtin spelling correction, programmable command completion, shell functions (with autoloading), a history mechanism, and a host of other features.
ftp://ftp.zsh.org/pub/ https://www.zsh.org/pub/)
The up-to-date source code is available via Git from Sourceforge. See https://sourceforge.net/projects/zsh/ for details. A summary of instructions for the archive can be found at http://zsh.sourceforge.net/.
To subscribe or unsubscribe, send mail to the associated administrative address for the mailing list.
YOU ONLY NEED TO JOIN ONE OF THE MAILING LISTS AS THEY ARE NESTED. All submissions to zsh-announce are automatically forwarded to zsh-users. All submissions to zsh-users are automatically forwarded to zsh-workers.
If you have problems subscribing/unsubscribing to any of the mailing lists, send mail to <listmaster@zsh.org>. The mailing lists are maintained by Karsten Thygesen <karthy@kom.auc.dk>.
The mailing lists are archived; the archives can be accessed via the administrative addresses listed above. There is also a hypertext archive, maintained by Geoff Wing <gcw@zsh.org>, available at https://www.zsh.org/mla/.
If there are any remaining arguments after option processing, and neither of the options -c or -s was supplied, the first argument is taken as the file name of a script containing shell commands to be executed. If the option PATH_SCRIPT is set, and the file name does not contain a directory path (i.e. there is no `/' in the name), first the current directory and then the command path given by the variable PATH are searched for the script. If the option is not set or the file name contains a `/' it is used directly.
After the first one or two arguments have been appropriated as described above, the remaining arguments are assigned to the positional parameters.
For further options, which are common to invocation and the set builtin, see zshoptions(1).
The long option `--emulate' followed (in a separate word) by an emulation mode may be passed to the shell. The emulation modes are those described for the emulate builtin, see zshbuiltins(1). The `--emulate' option must precede any other options (which might otherwise be overridden), but following options are honoured, so may be used to modify the requested emulation mode. Note that certain extra steps are taken to ensure a smooth emulation when this option is used compared with the emulate command within the shell: for example, variables that conflict with POSIX usage such as path are not defined within the shell.
Options may be specified by name using the -o option. -o acts like a single-letter option, but takes a following string as the option name. For example,
zsh -x -o shwordsplit scr
runs the script scr, setting the XTRACE option by the corresponding letter `-x' and the SH_WORD_SPLIT option by name. Options may be turned off by name by using +o instead of -o. -o can be stacked up with preceding single-letter options, so for example `-xo shwordsplit' or `-xoshwordsplit' is equivalent to `-x -o shwordsplit'.
Options may also be specified by name in GNU long option style, `--option-name'. When this is done, `-' characters in the option name are permitted: they are translated into `_', and thus ignored. So, for example, `zsh --sh-word-split' invokes zsh with the SH_WORD_SPLIT option turned on. Like other option syntaxes, options can be turned off by replacing the initial `-' with a `+'; thus `+-sh-word-split' is equivalent to `--no-sh-word-split'. Unlike other option syntaxes, GNU-style long options cannot be stacked with any other options, so for example `-x-shwordsplit' is an error, rather than being treated like `-x --shwordsplit'.
The special GNU-style option `--version' is handled; it sends to standard output the shell's version information, then exits successfully. `--help' is also handled; it sends to standard output a list of options that can be used when invoking the shell, then exits successfully.
Option processing may be finished, allowing following arguments that start with `-' or `+' to be treated as normal arguments, in two ways. Firstly, a lone `-' (or `+') as an argument by itself ends option processing. Secondly, a special option `--' (or `+-'), which may be specified on its own (which is the standard POSIX usage) or may be stacked with preceding options (so `-x-' is equivalent to `-x --'). Options are not permitted to be stacked after `--' (so `-x-f' is an error), but note the GNU-style option form discussed above, where `--shwordsplit' is permitted and does not end option processing.
Except when the sh/ksh emulation single-letter options are in effect, the option `-b' (or `+b') ends option processing. `-b' is like `--', except that further single-letter options can be stacked after the `-b' and will take effect as normal.
In sh and ksh compatibility modes the following parameters are not special and not initialized by the shell: ARGC, argv, cdpath, fignore, fpath, HISTCHARS, mailpath, MANPATH, manpath, path, prompt, PROMPT, PROMPT2, PROMPT3, PROMPT4, psvar, status, watch.
The usual zsh startup/shutdown scripts are not executed. Login shells source /etc/profile followed by $HOME/.profile. If the ENV environment variable is set on invocation, $ENV is sourced after the profile scripts. The value of ENV is subjected to parameter expansion, command substitution, and arithmetic expansion before being interpreted as a pathname. Note that the PRIVILEGED option also affects the execution of startup files.
The following options are set if the shell is invoked as sh or ksh: NO_BAD_PATTERN, NO_BANG_HIST, NO_BG_NICE, NO_EQUALS, NO_FUNCTION_ARGZERO, GLOB_SUBST, NO_GLOBAL_EXPORT, NO_HUP, INTERACTIVE_COMMENTS, KSH_ARRAYS, NO_MULTIOS, NO_NOMATCH, NO_NOTIFY, POSIX_BUILTINS, NO_PROMPT_PERCENT, RM_STAR_SILENT, SH_FILE_EXPANSION, SH_GLOB, SH_OPTION_LETTERS, SH_WORD_SPLIT. Additionally the BSD_ECHO and IGNORE_BRACES options are set if zsh is invoked as sh. Also, the KSH_OPTION_PRINT, LOCAL_OPTIONS, PROMPT_BANG, PROMPT_SUBST and SINGLE_LINE_ZLE options are set if zsh is invoked as ksh.
These restrictions are enforced after processing the startup files. The startup files should set up PATH to point to a directory of commands which can be safely invoked in the restricted environment. They may also add further restrictions by disabling selected builtins.
Restricted mode can also be activated any time by setting the RESTRICTED option. This immediately enables all the restrictions described above even if the shell still has not processed all startup files.
A shell Restricted Mode is an outdated way to restrict what users may do: modern systems have better, safer and more reliable ways to confine user actions, such as chroot jails, containers and zones.
A restricted shell is very difficult to implement safely. The feature may be removed in a future version of zsh.
It is important to realise that the restrictions only apply to the shell, not to the commands it runs (except for some shell builtins). While a restricted shell can only run the restricted list of commands accessible via the predefined `PATH' variable, it does not prevent those commands from running any other command.
As an example, if `env' is among the list of allowed commands, then it allows the user to run any command as `env' is not a shell builtin command and can run arbitrary executables.
So when implementing a restricted shell framework it is important to be fully aware of what actions each of the allowed commands or features (which may be regarded as modules) can perform.
Many commands can have their behaviour affected by environment variables. Except for the few listed above, zsh does not restrict the setting of environment variables.
If a `perl', `python', `bash', or other general purpose interpreted script it treated as a restricted command, the user can work around the restriction by setting specially crafted `PERL5LIB', `PYTHONPATH', `BASHENV' (etc.) environment variables. On GNU systems, any command can be made to run arbitrary code when performing character set conversion (including zsh itself) by setting a `GCONV_PATH' environment variable. Those are only a few examples.
Bear in mind that, contrary to some other shells, `readonly' is not a security feature in zsh as it can be undone and so cannot be used to mitigate the above.
A restricted shell only works if the allowed commands are few and carefully written so as not to grant more access to users than intended. It is also important to restrict what zsh module the user may load as some of them, such as `zsh/system', `zsh/mapfile' and `zsh/files', allow bypassing most of the restrictions.
Commands are then read from $ZDOTDIR/.zshenv. If the shell is a login shell, commands are read from /etc/zprofile and then $ZDOTDIR/.zprofile. Then, if the shell is interactive, commands are read from /etc/zshrc and then $ZDOTDIR/.zshrc. Finally, if the shell is a login shell, /etc/zlogin and $ZDOTDIR/.zlogin are read.
When a login shell exits, the files $ZDOTDIR/.zlogout and then /etc/zlogout are read. This happens with either an explicit exit via the exit or logout commands, or an implicit exit by reading end-of-file from the terminal. However, if the shell terminates due to exec'ing another process, the logout files are not read. These are also affected by the RCS and GLOBAL_RCS options. Note also that the RCS option affects the saving of history files, i.e. if RCS is unset when the shell exits, no history file will be saved.
If ZDOTDIR is unset, HOME is used instead. Files listed above as being in /etc may be in another directory, depending on the installation.
As /etc/zshenv is run for all instances of zsh, it is important that it be kept as small as possible. In particular, it is a good idea to put code that does not need to be run for every single shell behind a test of the form `if [[ -o rcs ]]; then ...' so that it will not be executed when zsh is invoked with the `-f' option.
Any of these files may be pre-compiled with the zcompile builtin command (see zshbuiltins(1)). If a compiled file exists (named for the original file plus the .zwc extension) and it is newer than the original file, the compiled file will be used instead.
When it starts, the shell reads commands from various files. These can be created or edited to customize the shell. See the section Startup/Shutdown Files in zsh(1).
If no personal initialization files exist for the current user, a function is run to help you change some of the most common settings. It won't appear if your administrator has disabled the zsh/newuser module. The function is designed to be self-explanatory. You can run it by hand with `autoload -Uz zsh-newuser-install; zsh-newuser-install -f'. See also the section User Configuration Functions in zshcontrib(1).
Interaction with the shell uses the builtin Zsh Line Editor, ZLE. This is described in detail in zshzle(1).
The first decision a user must make is whether to use the Emacs or Vi editing mode as the keys for editing are substantially different. Emacs editing mode is probably more natural for beginners and can be selected explicitly with the command bindkey -e.
A history mechanism for retrieving previously typed lines (most simply with the Up or Down arrow keys) is available; note that, unlike other shells, zsh will not save these lines when the shell exits unless you set appropriate variables, and the number of history lines retained by default is quite small (30 lines). See the description of the shell variables (referred to in the documentation as parameters) HISTFILE, HISTSIZE and SAVEHIST in zshparam(1). Note that it's currently only possible to read and write files saving history when the shell is interactive, i.e. it does not work from scripts.
The shell now supports the UTF-8 character set (and also others if supported by the operating system). This is (mostly) handled transparently by the shell, but the degree of support in terminal emulators is variable. There is some discussion of this in the shell FAQ, http://www.zsh.org/FAQ/. Note in particular that for combining characters to be handled the option COMBINING_CHARS needs to be set. Because the shell is now more sensitive to the definition of the character set, note that if you are upgrading from an older version of the shell you should ensure that the appropriate variable, either LANG (to affect all aspects of the shell's operation) or LC_CTYPE (to affect only the handling of character sets) is set to an appropriate value. This is true even if you are using a single-byte character set including extensions of ASCII such as ISO-8859-1 or ISO-8859-15. See the description of LC_CTYPE in zshparam(1).
Completion is a feature present in many shells. It allows the user to type only a part (usually the prefix) of a word and have the shell fill in the rest. The completion system in zsh is programmable. For example, the shell can be set to complete email addresses in arguments to the mail command from your ~/.abook/addressbook; usernames, hostnames, and even remote paths in arguments to scp, and so on. Anything that can be written in or glued together with zsh can be the source of what the line editor offers as possible completions.
Zsh has two completion systems, an old, so called compctl completion (named after the builtin command that serves as its complete and only user interface), and a new one, referred to as compsys, organized as library of builtin and user-defined functions. The two systems differ in their interface for specifying the completion behavior. The new system is more customizable and is supplied with completions for many commonly used commands; it is therefore to be preferred.
The completion system must be enabled explicitly when the shell starts. For more information see zshcompsys(1).
Apart from completion, the line editor is highly extensible by means of shell functions. Some useful functions are provided with the shell; they provide facilities such as:
See the section `ZLE Functions' in zshcontrib(1) for descriptions of these.
The shell has a large number of options for changing its behaviour. These cover all aspects of the shell; browsing the full documentation is the only good way to become acquainted with the many possibilities. See zshoptions(1).
The shell has a rich set of patterns which are available for file matching (described in the documentation as `filename generation' and also known for historical reasons as `globbing') and for use when programming. These are described in the section `Filename Generation' in zshexpn(1).
Of particular interest are the following patterns that are not commonly supported by other systems of pattern matching:
Although the syntax of zsh is in ways similar to the Korn shell, and therefore more remotely to the original UNIX shell, the Bourne shell, its default behaviour does not entirely correspond to those shells. General shell syntax is introduced in the section `Shell Grammar' in zshmisc(1).
One commonly encountered difference is that variables substituted onto the command line are not split into words. See the description of the shell option SH_WORD_SPLIT in the section `Parameter Expansion' in zshexpn(1). In zsh, you can either explicitly request the splitting (e.g. ${=foo}) or use an array when you want a variable to expand to more than one word. See the section `Array Parameters' in zshparam(1).
The most convenient way of adding enhancements to the shell is typically by writing a shell function and arranging for it to be autoloaded. Functions are described in the section `Functions' in zshmisc(1). Users changing from the C shell and its relatives should notice that aliases are less used in zsh as they don't perform argument substitution, only simple text replacement.
A few general functions, other than those for the line editor described above, are provided with the shell and are described in zshcontrib(1). Features include:
The first word is the command to be executed, and the remaining words, if any, are arguments to the command. If a command name is given, the parameter assignments modify the environment of the command when it is executed. The value of a simple command is its exit status, or 128 plus the signal number if terminated by a signal. For example,
echo foo
is a simple command with arguments.
A pipeline is either a simple command, or a sequence of two or more simple commands where each command is separated from the next by `|' or `|&'. Where commands are separated by `|', the standard output of the first command is connected to the standard input of the next. `|&' is shorthand for `2>&1 |', which connects both the standard output and the standard error of the command to the standard input of the next. The value of a pipeline is the value of the last command, unless the pipeline is preceded by `!' in which case the value is the logical inverse of the value of the last command. For example,
echo foo | sed 's/foo/bar/'
is a pipeline, where the output (`foo' plus a newline) of the first command will be passed to the input of the second.
If a pipeline is preceded by `coproc', it is executed as a coprocess; a two-way pipe is established between it and the parent shell. The shell can read from or write to the coprocess by means of the `>&p' and `<&p' redirection operators or with `print -p' and `read -p'. A pipeline cannot be preceded by both `coproc' and `!'. If job control is active, the coprocess can be treated in other than input and output as an ordinary background job.
A sublist is either a single pipeline, or a sequence of two or more pipelines separated by `&&' or `||'. If two pipelines are separated by `&&', the second pipeline is executed only if the first succeeds (returns a zero status). If two pipelines are separated by `||', the second is executed only if the first fails (returns a nonzero status). Both operators have equal precedence and are left associative. The value of the sublist is the value of the last pipeline executed. For example,
dmesg | grep panic && print yes
is a sublist consisting of two pipelines, the second just a simple command which will be executed if and only if the grep command returns a zero status. If it does not, the value of the sublist is that return status, else it is the status returned by the print (almost certainly zero).
A list is a sequence of zero or more sublists, in which each sublist is terminated by `;', `&', `&|', `&!', or a newline. This terminator may optionally be omitted from the last sublist in the list when the list appears as a complex command inside `(...)' or `{...}'. When a sublist is terminated by `;' or newline, the shell waits for it to finish before executing the next sublist. If a sublist is terminated by a `&', `&|', or `&!', the shell executes the last pipeline in it in the background, and does not wait for it to finish (note the difference from other shells which execute the whole sublist in the background). A backgrounded pipeline returns a status of zero.
More generally, a list can be seen as a set of any shell commands whatsoever, including the complex commands below; this is implied wherever the word `list' appears in later descriptions. For example, the commands in a shell function form a special sort of list.
The -c option clears the environment.
The -l option is equivalent to the - precommand modifier, to treat the replacement command as a login shell; the command is executed with a - prepended to its argv[0] string. This flag has no effect if used together with the -a option.
The -a option is used to specify explicitly the argv[0] string (the name of the command as seen by the process itself) to be used by the replacement command and is directly equivalent to setting a value for the ARGV0 environment variable.
The term consists of one or more newline or ; which terminate the words, and are optional when the `in word' is omitted.
More than one parameter name can appear before the list of words. If N names are given, then on each execution of the loop the next N words are assigned to the corresponding parameters. If there are more names than remaining words, the remaining parameters are each set to the empty string. Execution of the loop ends when there is no remaining word to assign to the first name. It is only possible for in to appear as the first name in the list, else it will be treated as marking the end of the list.
The repeat syntax is disabled by default when the shell starts in a mode emulating another shell. It can be enabled with the command `enable -r repeat'
Note further that, unless the SH_GLOB option is set, the whole pattern with alternatives is treated by the shell as equivalent to a group of patterns within parentheses, although white space may appear about the parentheses and the vertical bar and will be stripped from the pattern at those points. White space may appear elsewhere in the pattern; this is not stripped. If the SH_GLOB option is set, so that an opening parenthesis can be unambiguously treated as part of the case syntax, the expression is parsed into separate words and these are treated as strict alternatives (as in other shells).
If the list that is executed is terminated with ;& rather than ;;, the following list is also executed. The rule for the terminator of the following list ;;, ;& or ;| is applied unless the esac is reached.
If the list that is executed is terminated with ;| the shell continues to scan the patterns looking for the next match, executing the corresponding list, and applying the rule for the corresponding terminator ;;, ;& or ;|. Note that word is not re-expanded; all applicable patterns are tested with the same word.
Optional newlines or semicolons may appear after the always; note, however, that they may not appear between the preceding closing brace and the always.
An `error' in this context is a condition such as a syntax error which causes the shell to abort execution of the current function, script, or list. Syntax errors encountered while the shell is parsing the code do not cause the always-list to be executed. For example, an erroneously constructed if block in try-list would cause the shell to abort during parsing, so that always-list would not be executed, while an erroneous substitution such as ${*foo*} would cause a run-time error, after which always-list would be executed.
An error condition can be tested and reset with the special integer variable TRY_BLOCK_ERROR. Outside an always-list the value is irrelevant, but it is initialised to -1. Inside always-list, the value is 1 if an error occurred in the try-list, else 0. If TRY_BLOCK_ERROR is set to 0 during the always-list, the error condition caused by the try-list is reset, and shell execution continues normally after the end of always-list. Altering the value during the try-list is not useful (unless this forms part of an enclosing always block).
Regardless of TRY_BLOCK_ERROR, after the end of always-list the normal shell status $? is the value returned from try-list. This will be non-zero if there was an error, even if TRY_BLOCK_ERROR was set to zero.
The following executes the given code, ignoring any errors it causes. This is an alternative to the usual convention of protecting code by executing it in a subshell.
{
# code which may cause an error
} always {
# This code is executed regardless of the error.
(( TRY_BLOCK_ERROR = 0 ))
}
# The error condition has been reset.
When a try block occurs outside of any function, a return or a exit encountered in try-list does not cause the execution of always-list. Instead, the shell exits immediately after any EXIT trap has been executed. Otherwise, a return command encountered in try-list will cause the execution of always-list, just like break and continue.
If the option SH_GLOB is set for compatibility with other shells, then whitespace may appear between the left and right parentheses when there is a single word; otherwise, the parentheses will be treated as forming a globbing pattern in that case.
In any of the forms above, a redirection may appear outside the function body, for example
func() { ... } 2>&1
The redirection is stored with the function and applied whenever the function is executed. Any variables in the redirection are expanded at the point the function is executed, but outside the function scope.
The short versions below only work if sublist is of the form `{ list }' or if the SHORT_LOOPS option is set. For the if, while and until commands, in both these cases the test part of the loop must also be suitably delimited, such as by `[[ ... ]]' or `(( ... ))', else the end of the test will not be recognized. For the for, repeat, case and select commands no such special form for the arguments is necessary, but the other condition (the special form of sublist or use of the SHORT_LOOPS option) still applies.
if [[ -o ignorebraces ]] {
print yes
}
works, but
if true { # Does not work!
print yes
}
does not, since the test is not suitably delimited.
do done esac then elif else fi for case if while function repeat time until select coproc nocorrect foreach end ! [[ { } declare export float integer local readonly typeset
Additionally, `}' is recognized in any position if neither the IGNORE_BRACES option nor the IGNORE_CLOSE_BRACES option is set.
Fatal errors found in non-interactive shells include:
If the POSIX_BUILTINS option is set, more errors associated with shell builtin commands are treated as fatal, as specified by the POSIX standard.
A word is defined as:
Alias expansion is done on the shell input before any other expansion except history expansion. Therefore, if an alias is defined for the word foo, alias expansion may be avoided by quoting part of the word, e.g. \foo. Any form of quoting works, although there is nothing to prevent an alias being defined for the quoted form such as \foo as well.
When POSIX_ALIASES is set, only plain unquoted strings are eligible for aliasing. The alias builtin does not reject ineligible aliases, but they are not expanded.
For use with completion, which would remove an initial backslash followed by a character that isn't special, it may be more convenient to quote the word by starting with a single quote, i.e. 'foo; completion will automatically add the trailing single quote.
Although aliases can be used in ways that bend normal shell syntax, not every string of non-white-space characters can be used as an alias.
Any set of characters not listed as a word above is not a word, hence no attempt is made to expand it as an alias, no matter how it is defined (i.e. via the builtin or the special parameter aliases described in the section THE ZSH/PARAMETER MODULE in zshmodules(1)). However, as noted in the case of POSIX_ALIASES above, the shell does not attempt to deduce whether the string corresponds to a word at the time the alias is created.
For example, an expression containing an = at the start of a command line is an assignment and cannot be expanded as an alias; a lone = is not an assignment but can only be set as an alias using the parameter, as otherwise the = is taken part of the syntax of the builtin command.
It is not presently possible to alias the `((' token that introduces arithmetic expressions, because until a full statement has been parsed, it cannot be distinguished from two consecutive `(' tokens introducing nested subshells. Also, if a separator such as && is aliased, \&& turns into the two tokens \& and &, each of which may have been aliased separately. Similarly for \<<, \>|, etc.
There is a commonly encountered problem with aliases illustrated by the following code:
alias echobar='echo bar'; echobar
This prints a message that the command echobar could not be found. This happens because aliases are expanded when the code is read in; the entire line is read in one go, so that when echobar is executed it is too late to expand the newly defined alias. This is often a problem in shell scripts, functions, and code executed with `source' or `.'. Consequently, use of functions rather than aliases is recommended in non-interactive code.
Note also the unhelpful interaction of aliases and function definitions:
alias func='noglob func'
func() {
echo Do something with $*
}
Because aliases are expanded in function definitions, this causes the following command to be executed:
noglob func() {
echo Do something with $*
}
which defines noglob as well as func as functions with the body given. To avoid this, either quote the name func or use the alternative function definition form `function func'. Ensuring the alias is defined after the function works but is problematic if the code fragment might be re-executed.
A string enclosed between `$'' and `'' is processed the same way as the string arguments of the print builtin, and the resulting string is considered to be entirely quoted. A literal `'' character can be included in the string by using the `\'' escape.
All characters enclosed between a pair of single quotes ('') that is not preceded by a `$' are quoted. A single quote cannot appear within single quotes unless the option RC_QUOTES is set, in which case a pair of single quotes are turned into a single quote. For example,
print ''''
outputs nothing apart from a newline if RC_QUOTES is not set, but one single quote if it is set.
Inside double quotes (""), parameter and command substitution occur, and `\' quotes the characters `\', ``', `"', `$', and the first character of $histchars (default `!').
The following may appear anywhere in a simple command or may precede or follow a complex command. Expansion occurs before word or digit is used except as noted below. If the result of substitution on word produces more than one filename, redirection occurs for each separate filename in turn.
If any character of word is quoted with single or double quotes or a `\', no interpretation is placed upon the characters of the document. Otherwise, parameter and command substitution occurs, `\' followed by a newline is removed, and `\' must be used to quote the characters `\', `$', ``' and the first character of word.
Note that word itself does not undergo shell expansion. Backquotes in word do not have their usual effect; instead they behave similarly to double quotes, except that the backquotes themselves are passed through unchanged. (This information is given for completeness and it is not recommended that backquotes be used.) Quotes in the form $'...' have their standard effect of expanding backslashed references to special characters.
If <<- is used, then all leading tabs are stripped from word and from the document.
If one of the above is preceded by a digit, then the file descriptor referred to is that specified by the digit instead of the default 0 or 1. The order in which redirections are specified is significant. The shell evaluates each redirection in terms of the (file descriptor, file) association at the time of evaluation. For example:
... 1>fname 2>&1
first associates file descriptor 1 with file fname. It then associates file descriptor 2 with the file associated with file descriptor 1 (that is, fname). If the order of redirections were reversed, file descriptor 2 would be associated with the terminal (assuming file descriptor 1 had been) and then file descriptor 1 would be associated with file fname.
The `|&' command separator described in Simple Commands & Pipelines in zshmisc(1) is a shorthand for `2>&1 |'.
The various forms of process substitution, `<(list)', and `=(list)' for input and `>(list)' for output, are often used together with redirection. For example, if word in an output redirection is of the form `>(list)' then the output is piped to the command represented by list. See Process Substitution in zshexpn(1).
When the shell is parsing arguments to a command, and the shell option IGNORE_BRACES is not set, a different form of redirection is allowed: instead of a digit before the operator there is a valid shell identifier enclosed in braces. The shell will open a new file descriptor that is guaranteed to be at least 10 and set the parameter named by the identifier to the file descriptor opened. No whitespace is allowed between the closing brace and the redirection character. For example:
... {myfd}>&1
This opens a new file descriptor that is a duplicate of file descriptor 1 and sets the parameter myfd to the number of the file descriptor, which will be at least 10. The new file descriptor can be written to using the syntax >&$myfd. The file descriptor remains open in subshells and forked external executables.
The syntax {varid}>&-, for example {myfd}>&-, may be used to close a file descriptor opened in this fashion. Note that the parameter given by varid must previously be set to a file descriptor in this case.
It is an error to open or close a file descriptor in this fashion when the parameter is readonly. However, it is not an error to read or write a file descriptor using <&$param or >&$param if param is readonly.
If the option CLOBBER is unset, it is an error to open a file descriptor using a parameter that is already set to an open file descriptor previously allocated by this mechanism. Unsetting the parameter before using it for allocating a file descriptor avoids the error.
Note that this mechanism merely allocates or closes a file descriptor; it does not perform any redirections from or to it. It is usually convenient to allocate a file descriptor prior to use as an argument to exec. The syntax does not in any case work when used around complex commands such as parenthesised subshells or loops, where the opening brace is interpreted as part of a command list to be executed in the current shell.
The following shows a typical sequence of allocation, use, and closing of a file descriptor:
integer myfd
exec {myfd}>~/logs/mylogfile.txt
print This is a log message. >&$myfd
exec {myfd}>&-
Note that the expansion of the variable in the expression >&$myfd occurs at the point the redirection is opened. This is after the expansion of command arguments and after any redirections to the left on the command line have been processed.
date >foo >bar
writes the date to two files, named `foo' and `bar'. Note that a pipe is an implicit redirection; thus
date >foo | cat
writes the date to the file `foo', and also pipes it to cat.
Note that the shell opens all the files to be used in the multio process immediately, not at the point they are about to be written.
Note also that redirections are always expanded in order. This happens regardless of the setting of the MULTIOS option, but with the option in effect there are additional consequences. For example, the meaning of the expression >&1 will change after a previous redirection:
date >&1 >output
In the case above, the >&1 refers to the standard output at the start of the line; the result is similar to the tee command. However, consider:
date >output >&1
As redirections are evaluated in order, when the >&1 is encountered the standard output is set to the file output and another copy of the output is therefore sent to that file. This is unlikely to be what is intended.
If the MULTIOS option is set, the word after a redirection operator is also subjected to filename generation (globbing). Thus
: > *
will truncate all files in the current directory, assuming there's at least one. (Without the MULTIOS option, it would create an empty file called `*'.) Similarly, you can do
echo exit 0 >> *.sh
If the user tries to open a file descriptor for reading more than once, the shell opens the file descriptor as a pipe to a process that copies all the specified inputs to its output in the order specified, provided the MULTIOS option is set. It should be noted that each file is opened immediately, not at the point where it is about to be read: this behaviour differs from cat, so if strictly standard behaviour is needed, cat should be used instead.
Thus
sort <foo <fubar
or even
sort <f{oo,ubar}
is equivalent to `cat foo fubar | sort'.
Expansion of the redirection argument occurs at the point the redirection is opened, at the point described above for the expansion of the variable in >&$myfd.
Note that a pipe is an implicit redirection; thus
cat bar | sort <foo
is equivalent to `cat bar foo | sort' (note the order of the inputs).
If the MULTIOS option is unset, each redirection replaces the previous redirection for that file descriptor. However, all files redirected to are actually opened, so
echo Hello > bar > baz
when MULTIOS is unset will truncate `bar', and write `Hello' into `baz'.
There is a problem when an output multio is attached to an external program. A simple example shows this:
cat file >file1 >file2 cat file1 file2
Here, it is possible that the second `cat' will not display the full contents of file1 and file2 (i.e. the original contents of file repeated twice).
The reason for this is that the multios are spawned after the cat process is forked from the parent shell, so the parent shell does not wait for the multios to finish writing data. This means the command as shown can exit before file1 and file2 are completely written. As a workaround, it is possible to run the cat process as part of a job in the current shell:
{ cat file } >file >file2
Here, the {...} job will pause to wait for both files to be written.
If the parameter NULLCMD is not set or the option CSH_NULLCMD is set, an error is caused. This is the csh behavior and CSH_NULLCMD is set by default when emulating csh.
If the option SH_NULLCMD is set, the builtin `:' is inserted as a command with the given redirections. This is the default when emulating sh or ksh.
Otherwise, if the parameter NULLCMD is set, its value will be used as a command with the given redirections. If both NULLCMD and READNULLCMD are set, then the value of the latter will be used instead of that of the former when the redirection is an input. The default for NULLCMD is `cat' and for READNULLCMD is `more'. Thus
< file
shows the contents of file on standard output, with paging if that is a terminal. NULLCMD and READNULLCMD may refer to shell functions.
Otherwise, the shell searches each element of $path for a directory containing an executable file by that name. If the search is unsuccessful, the shell prints an error message and returns a nonzero exit status.
If execution fails because the file is not in executable format, and the file is not a directory, it is assumed to be a shell script. /bin/sh is spawned to execute it. If the program is a file beginning with `#!', the remainder of the first line specifies an interpreter for the program. The shell will execute the specified interpreter on operating systems that do not handle this executable format in the kernel.
If no external command is found but a function command_not_found_handler exists the shell executes this function with all command line arguments. The return status of the function becomes the status of the command. If the function wishes to mimic the behaviour of the shell when the command is not found, it should print the message `command not found: cmd' to standard error and return status 127. Note that the handler is executed in a subshell forked to execute an external command, hence changes to directories, shell parameters, etc. have no effect on the main shell.
Functions execute in the same process as the caller and share all files and present working directory with the caller. A trap on EXIT set inside a function is executed after the function completes in the environment of the caller.
The return builtin is used to return from function calls.
Function identifiers can be listed with the functions builtin. Functions can be undefined with the unfunction builtin.
A function can be marked as undefined using the autoload builtin (or `functions -u' or `typeset -fu'). Such a function has no body. When the function is first executed, the shell searches for its definition using the elements of the fpath variable. Thus to define functions for autoloading, a typical sequence is:
fpath=(~/myfuncs $fpath) autoload myfunc1 myfunc2 ...
The usual alias expansion during reading will be suppressed if the autoload builtin or its equivalent is given the option -U. This is recommended for the use of functions supplied with the zsh distribution. Note that for functions precompiled with the zcompile builtin command the flag -U must be provided when the .zwc file is created, as the corresponding information is compiled into the latter.
For each element in fpath, the shell looks for three possible files, the newest of which is used to load the definition for the function:
If element already includes a .zwc extension (i.e. the extension was explicitly given by the user), element is searched for the definition of the function without comparing its age to that of other files; in fact, there does not need to be any directory named element without the suffix. Thus including an element such as `/usr/local/funcs.zwc' in fpath will speed up the search for functions, with the disadvantage that functions included must be explicitly recompiled by hand before the shell notices any changes.
In summary, the order of searching is, first, in the parents of directories in fpath for the newer of either a compiled directory or a directory in fpath; second, if more than one of these contains a definition for the function that is sought, the leftmost in the fpath is chosen; and third, within a directory, the newer of either a compiled function or an ordinary function definition is used.
If the KSH_AUTOLOAD option is set, or the file contains only a simple definition of the function, the file's contents will be executed. This will normally define the function in question, but may also perform initialization, which is executed in the context of the function execution, and may therefore define local parameters. It is an error if the function is not defined by loading the file.
Otherwise, the function body (with no surrounding `funcname() {...}') is taken to be the complete contents of the file. This form allows the file to be used directly as an executable shell script. If processing of the file results in the function being re-defined, the function itself is not re-executed. To force the shell to perform initialization and then call the function defined, the file should contain initialization code (which will be executed then discarded) in addition to a complete function definition (which will be retained for subsequent calls to the function), and a call to the shell function, including any arguments, at the end.
For example, suppose the autoload file func contains
func() { print This is func; }
print func is initialized
then `func; func' with KSH_AUTOLOAD set will produce both messages on the first call, but only the message `This is func' on the second and subsequent calls. Without KSH_AUTOLOAD set, it will produce the initialization message on the first call, and the other message on the second and subsequent calls.
It is also possible to create a function that is not marked as autoloaded, but which loads its own definition by searching fpath, by using `autoload -X' within a shell function. For example, the following are equivalent:
myfunc() {
autoload -X
}
myfunc args...
and
unfunction myfunc # if myfunc was defined autoload myfunc myfunc args...
In fact, the functions command outputs `builtin autoload -X' as the body of an autoloaded function. This is done so that
eval "$(functions)"
produces a reasonable result. A true autoloaded function can be identified by the presence of the comment `# undefined' in the body, because all comments are discarded from defined functions.
To load the definition of an autoloaded function myfunc without executing myfunc, use:
autoload +X myfunc
If no name is given for a function, it is `anonymous' and is handled specially. Either form of function definition may be used: a `()' with no preceding name, or a `function' with an immediately following open brace. The function is executed immediately at the point of definition and is not stored for future use. The function name is set to `(anon)'.
Arguments to the function may be specified as words following the closing brace defining the function, hence if there are none no arguments (other than $0) are set. This is a difference from the way other functions are parsed: normal function definitions may be followed by certain keywords such as `else' or `fi', which will be treated as arguments to anonymous functions, so that a newline or semicolon is needed to force keyword interpretation.
Note also that the argument list of any enclosing script or function is hidden (as would be the case for any other function called at this point).
Redirections may be applied to the anonymous function in the same manner as to a current-shell structure enclosed in braces. The main use of anonymous functions is to provide a scope for local variables. This is particularly convenient in start-up files as these do not provide their own local variable scope.
For example,
variable=outside
function {
local variable=inside
print "I am $variable with arguments $*"
} this and that
print "I am $variable"
outputs the following:
I am inside with arguments this and that I am outside
Note that function definitions with arguments that expand to nothing, for example `name=; function $name { ... }', are not treated as anonymous functions. Instead, they are treated as normal function definitions where the definition is silently discarded.
For the functions below, it is possible to define an array that has the same name as the function with `_functions' appended. Any element in such an array is taken as the name of a function to execute; it is executed in the same context and with the same arguments as the basic function. For example, if $chpwd_functions is an array containing the values `mychpwd', `chpwd_save_dirstack', then the shell attempts to execute the functions `chpwd', `mychpwd' and `chpwd_save_dirstack', in that order. Any function that does not exist is silently ignored. A function found by this mechanism is referred to elsewhere as a `hook function'. An error in any function causes subsequent functions not to be run. Note further that an error in a precmd hook causes an immediately following periodic function not to run (though it may run at the next opportunity).
If any of the hook functions returns status 1 (or any non-zero value other than 2, though this is not guaranteed for future versions of the shell) the history line will not be saved, although it lingers in the history until the next line is executed, allowing you to reuse or edit it immediately.
If any of the hook functions returns status 2 the history line will be saved on the internal history list, but not written to the history file. In case of a conflict, the first non-zero status value is taken.
A hook function may call `fc -p ...' to switch the history context so that the history is saved in a different file from the that in the global HISTFILE parameter. This is handled specially: the history context is automatically restored after the processing of the history line is finished.
The following example function works with one of the options INC_APPEND_HISTORY or SHARE_HISTORY set, in order that the line is written out immediately after the history entry is added. It first adds the history line to the normal history with the newline stripped, which is usually the correct behaviour. Then it switches the history context so that the line will be written to a history file in the current directory.
zshaddhistory() {
print -sr -- ${1%%$'\n'}
fc -p .zsh_local_history
}
The functions below are treated specially but do not have corresponding hook arrays.
If a function of this form is defined and null, the shell and processes spawned by it will ignore SIGNAL.
The return status from the function is handled specially. If it is zero, the signal is assumed to have been handled, and execution continues normally. Otherwise, the shell will behave as interrupted except that the return status of the trap is retained.
Programs terminated by uncaught signals typically return the status 128 plus the signal number. Hence the following causes the handler for SIGINT to print a message, then mimic the usual effect of the signal.
TRAPINT() {
print "Caught SIGINT, aborting."
return $(( 128 + $1 ))
}
The functions TRAPZERR, TRAPDEBUG and TRAPEXIT are never executed inside other traps.
The functions beginning `TRAP' may alternatively be defined with the trap builtin: this may be preferable for some uses. Setting a trap with one form removes any trap of the other form for the same signal; removing a trap in either form removes all traps for the same signal. The forms
TRAPNAL() {
# code
}
('function traps') and
trap ' # code ' NAL
('list traps') are equivalent in most ways, the exceptions being the following:
[1] 1234
indicating that the job which was started asynchronously was job number 1 and had one (top-level) process, whose process ID was 1234.
If a job is started with `&|' or `&!', then that job is immediately disowned. After startup, it does not have a place in the job table, and is not subject to the job control features described here.
If you are running a job and wish to do something else you may hit the key ^Z (control-Z) which sends a TSTP signal to the current job: this key may be redefined by the susp option of the external stty command. The shell will then normally indicate that the job has been `suspended', and print another prompt. You can then manipulate the state of this job, putting it in the background with the bg command, or run some other commands and then eventually bring the job back into the foreground with the foreground command fg. A ^Z takes effect immediately and is like an interrupt in that pending output and unread input are discarded when it is typed.
A job being run in the background will suspend if it tries to read from the terminal.
Note that if the job running in the foreground is a shell function, then suspending it will have the effect of causing the shell to fork. This is necessary to separate the function's state from that of the parent shell performing the job control, so that the latter can return to the command line prompt. As a result, even if fg is used to continue the job the function will no longer be part of the parent shell, and any variables set by the function will not be visible in the parent shell. Thus the behaviour is different from the case where the function was never suspended. Zsh is different from many other shells in this regard.
One additional side effect is that use of disown with a job created by suspending shell code in this fashion is delayed: the job can only be disowned once any process started from the parent shell has terminated. At that point, the disowned job disappears silently from the job list.
The same behaviour is found when the shell is executing code as the right hand side of a pipeline or any complex shell construct such as if, for, etc., in order that the entire block of code can be managed as a single job. Background jobs are normally allowed to produce output, but this can be disabled by giving the command `stty tostop'. If you set this tty option, then background jobs will suspend when they try to produce output like they do when they try to read input.
When a command is suspended and continued later with the fg or wait builtins, zsh restores tty modes that were in effect when it was suspended. This (intentionally) does not apply if the command is continued via `kill -CONT', nor when it is continued with bg.
There are several ways to refer to jobs in the shell. A job can be referred to by the process ID of any process of the job or by one of the following:
The shell learns immediately whenever a process changes state. It normally informs you whenever a job becomes blocked so that no further progress is possible. If the NOTIFY option is not set, it waits until just before it prints a prompt before it informs you. All such notifications are sent directly to the terminal, not to the standard output or standard error.
When the monitor mode is on, each background job that completes triggers any trap set for CHLD.
When you try to leave the shell while jobs are running or suspended, you will be warned that `You have suspended (running) jobs'. You may use the jobs command to see what they are. If you do this or immediately try to exit again, the shell will not warn you a second time; the suspended jobs will be terminated, and the running jobs will be sent a SIGHUP signal, if the HUP option is set.
To avoid having the shell terminate the running jobs, either use the nohup command (see nohup(1)) or the disown builtin.
Certain jobs are run asynchronously by the shell other than those explicitly put into the background; even in cases where the shell would usually wait for such jobs, an explicit exit command or exit due to the option ERR_EXIT will cause the shell to exit without waiting. Examples of such asynchronous jobs are process substitution, see the section PROCESS SUBSTITUTION in the zshexpn(1) manual page, and the handler processes for multios, see the section MULTIOS in the zshmisc(1) manual page.
The let builtin command takes arithmetic expressions as arguments; each is evaluated separately. Since many of the arithmetic operators, as well as spaces, require quoting, an alternative form is provided: for any command which begins with a `((', all the characters until a matching `))' are treated as a quoted expression and arithmetic expansion performed as for an argument of let. More precisely, `((...))' is equivalent to `let "..."'. The return status is 0 if the arithmetic value of the expression is non-zero, 1 if it is zero, and 2 if an error occurred.
For example, the following statement
(( val = 2 + 1 ))
is equivalent to
let "val = 2 + 1"
both assigning the value 3 to the shell variable val and returning a zero status.
Integers can be in bases other than 10. A leading `0x' or `0X' denotes hexadecimal and a leading `0b' or `0B' binary. Integers may also be of the form `base#n', where base is a decimal number between two and thirty-six representing the arithmetic base and n is a number in that base (for example, `16#ff' is 255 in hexadecimal). The base# may also be omitted, in which case base 10 is used. For backwards compatibility the form `[base]n' is also accepted.
An integer expression or a base given in the form `base#n' may contain underscores (`_') after the leading digit for visual guidance; these are ignored in computation. Examples are 1_000_000 or 0xffff_ffff which are equivalent to 1000000 and 0xffffffff respectively.
It is also possible to specify a base to be used for output in the form `[#base]', for example `[#16]'. This is used when outputting arithmetical substitutions or when assigning to scalar parameters, but an explicitly defined integer or floating point parameter will not be affected. If an integer variable is implicitly defined by an arithmetic expression, any base specified in this way will be set as the variable's output arithmetic base as if the option `-i base' to the typeset builtin had been used. The expression has no precedence and if it occurs more than once in a mathematical expression, the last encountered is used. For clarity it is recommended that it appear at the beginning of an expression. As an example:
typeset -i 16 y print $(( [#8] x = 32, y = 32 )) print $x $y
outputs first `8#40', the rightmost value in the given output base, and then `8#40 16#20', because y has been explicitly declared to have output base 16, while x (assuming it does not already exist) is implicitly typed by the arithmetic evaluation, where it acquires the output base 8.
The base may be replaced or followed by an underscore, which may itself be followed by a positive integer (if it is missing the value 3 is used). This indicates that underscores should be inserted into the output string, grouping the number for visual clarity. The following integer specifies the number of digits to group together. For example:
setopt cbases print $(( [#16_4] 65536 ** 2 ))
outputs `0x1_0000_0000'.
The feature can be used with floating point numbers, in which case the base must be omitted; grouping is away from the decimal point. For example,
zmodload zsh/mathfunc print $(( [#_] sqrt(1e7) ))
outputs `3_162.277_660_168_379_5' (the number of decimal places shown may vary).
If the C_BASES option is set, hexadecimal numbers are output in the standard C format, for example `0xFF' instead of the usual `16#FF'. If the option OCTAL_ZEROES is also set (it is not by default), octal numbers will be treated similarly and hence appear as `077' instead of `8#77'. This option has no effect on the output of bases other than hexadecimal and octal, and these formats are always understood on input.
When an output base is specified using the `[#base]' syntax, an appropriate base prefix will be output if necessary, so that the value output is valid syntax for input. If the # is doubled, for example `[##16]', then no base prefix is output.
Floating point constants are recognized by the presence of a decimal point or an exponent. The decimal point may be the first character of the constant, but the exponent character e or E may not, as it will be taken for a parameter name. All numeric parts (before and after the decimal point and in the exponent) may contain underscores after the leading digit for visual guidance; these are ignored in computation.
An arithmetic expression uses nearly the same syntax and associativity of expressions as in C.
In the native mode of operation, the following operators are supported (listed in decreasing order of precedence):
The operators `&&', `||', `&&=', and `||=' are short-circuiting, and only one of the latter two expressions in a ternary operator is evaluated. Note the precedence of the bitwise AND, OR, and XOR operators.
With the option C_PRECEDENCES the precedences (but no other properties) of the operators are altered to be the same as those in most other languages that support the relevant operators:
Note the precedence of exponentiation in both cases is below that of unary operators, hence `-3**2' evaluates as `9', not `-9'. Use parentheses where necessary: `-(3**2)'. This is for compatibility with other shells.
Mathematical functions can be called with the syntax `func(args)', where the function decides if the args is used as a string or a comma-separated list of arithmetic expressions. The shell currently defines no mathematical functions by default, but the module zsh/mathfunc may be loaded with the zmodload builtin to provide standard floating point mathematical functions.
An expression of the form `##x' where x is any character sequence such as `a', `^A', or `\M-\C-x' gives the value of this character and an expression of the form `#name' gives the value of the first character of the contents of the parameter name. Character values are according to the character set used in the current locale; for multibyte character handling the option MULTIBYTE must be set. Note that this form is different from `$#name', a standard parameter substitution which gives the length of the parameter name. `#\' is accepted instead of `##', but its use is deprecated.
Named parameters and subscripted arrays can be referenced by name within an arithmetic expression without using the parameter expansion syntax. For example,
((val2 = val1 * 2))
assigns twice the value of $val1 to the parameter named val2.
An internal integer representation of a named parameter can be specified with the integer builtin. Arithmetic evaluation is performed on the value of each assignment to a named parameter declared integer in this manner. Assigning a floating point number to an integer results in rounding towards zero.
Likewise, floating point numbers can be declared with the float builtin; there are two types, differing only in their output format, as described for the typeset builtin. The output format can be bypassed by using arithmetic substitution instead of the parameter substitution, i.e. `${float}' uses the defined format, but `$((float))' uses a generic floating point format.
Promotion of integer to floating point values is performed where necessary. In addition, if any operator which requires an integer (`&', `|', `^', `<<', `>>' and their equivalents with assignment) is given a floating point argument, it will be silently rounded towards zero except for `~' which rounds down.
Users should beware that, in common with many other programming languages but not software designed for calculation, the evaluation of an expression in zsh is taken a term at a time and promotion of integers to floating point does not occur in terms only containing integers. A typical result of this is that a division such as 6/8 is truncated, in this being rounded towards 0. The FORCE_FLOAT shell option can be used in scripts or functions where floating point evaluation is required throughout.
Scalar variables can hold integer or floating point values at different times; there is no memory of the numeric type in this case.
If a variable is first assigned in a numeric context without previously being declared, it will be implicitly typed as integer or float and retain that type either until the type is explicitly changed or until the end of the scope. This can have unforeseen consequences. For example, in the loop
for (( f = 0; f < 1; f += 0.1 )); do # use $f done
if f has not already been declared, the first assignment will cause it to be created as an integer, and consequently the operation `f += 0.1' will always cause the result to be truncated to zero, so that the loop will fail. A simple fix would be to turn the initialization into `f = 0.0'. It is therefore best to declare numeric variables with explicit types.
When no option named option exists, and the POSIX_BUILTINS option hasn't been set, return 3 with a warning. If that option is set, return 1 with no warning.
If the option BASH_REMATCH is not set the scalar parameter MATCH is set to the substring that matched the pattern and the integer parameters MBEGIN and MEND to the index of the start and end, respectively, of the match in string, such that if string is contained in variable var the expression `${var[$MBEGIN,$MEND]}' is identical to `$MATCH'. The setting of the option KSH_ARRAYS is respected. Likewise, the array match is set to the substrings that matched parenthesised subexpressions and the arrays mbegin and mend to the indices of the start and end positions, respectively, of the substrings within string. The arrays are not set if there were no parenthesised subexpressions. For example, if the string `a short string' is matched against the regular expression `s(...)t', then (assuming the option KSH_ARRAYS is not set) MATCH, MBEGIN and MEND are `short', 3 and 7, respectively, while match, mbegin and mend are single entry arrays containing the strings `hor', `4' and `6', respectively.
If the option BASH_REMATCH is set the array BASH_REMATCH is set to the substring that matched the pattern followed by the substrings that matched parenthesised subexpressions within the pattern.
For compatibility, if there is a single argument that is not syntactically significant, typically a variable, the condition is treated as a test for whether the expression expands as a string of non-zero length. In other words, [[ $var ]] is the same as [[ -n $var ]]. It is recommended that the second, explicit, form be used where possible.
Normal shell expansion is performed on the file, string and pattern arguments, but the result of each expansion is constrained to be a single word, similar to the effect of double quotes.
Filename generation is not performed on any form of argument to conditions. However, it can be forced in any case where normal shell expansion is valid and when the option EXTENDED_GLOB is in effect by using an explicit glob qualifier of the form (#q) at the end of the string. A normal glob qualifier expression may appear between the `q' and the closing parenthesis; if none appears the expression has no effect beyond causing filename generation. The results of filename generation are joined together to form a single word, as with the results of other forms of expansion.
This special use of filename generation is only available with the [[ syntax. If the condition occurs within the [ or test builtin commands then globbing occurs instead as part of normal command line expansion before the condition is evaluated. In this case it may generate multiple words which are likely to confuse the syntax of the test command.
For example,
[[ -n file*(#qN) ]]
produces status zero if and only if there is at least one file in the current directory beginning with the string `file'. The globbing qualifier N ensures that the expression is empty if there is no matching file.
Pattern metacharacters are active for the pattern arguments; the patterns are the same as those used for filename generation, see zshexpn(1), but there is no special behaviour of `/' nor initial dots, and no glob qualifiers are allowed.
In each of the above expressions, if file is of the form `/dev/fd/n', where n is an integer, then the test applied to the open file whose descriptor number is n, even if the underlying system does not support the /dev/fd directory.
In the forms which do numeric comparison, the expressions exp undergo arithmetic expansion as if they were enclosed in $((...)).
For example, the following:
[[ ( -f foo || -f bar ) && $report = y* ]] && print File exists.
tests if either file foo or file bar exists, and if so, if the value of the parameter report begins with `y'; if the complete condition is true, the message `File exists.' is printed.
If the PROMPT_SUBST option is set, the prompt string is first subjected to parameter expansion, command substitution and arithmetic expansion. See zshexpn(1).
Certain escape sequences may be recognised in the prompt string.
If the PROMPT_BANG option is set, a `!' in the prompt is replaced by the current history event number. A literal `!' may then be represented as `!!'.
If the PROMPT_PERCENT option is set, certain escape sequences that start with `%' are expanded. Many escapes are followed by a single character, although some of these take an optional integer argument that should appear between the `%' and the next character of the sequence. More complicated escape sequences are available to provide conditional expansion.
In addition, if the system supports the POSIX gettimeofday system call, %. provides decimal fractions of a second since the epoch with leading zeroes. By default three decimal places are provided, but a number of digits up to 9 may be given following the %; hence %6. outputs microseconds, and %9. outputs nanoseconds. (The latter requires a nanosecond-precision clock_gettime; systems lacking this will return a value multiplied by the appropriate power of 10.) A typical example of this is the format `%D{%H:%M:%S.%.}'.
The GNU extension %N is handled as a synonym for %9..
Additionally, the GNU extension that a `-' between the % and the format character causes a leading zero or space to be stripped is handled directly by the shell for the format characters d, f, H, k, l, m, M, S and y; any other format characters are provided to the system's strftime(3) with any leading `-' present, so the handling is system dependent. Further GNU (or other) extensions are also passed to strftime(3) and may work if the system supports them.
A positive numeric argument between the % and the { is treated as described for %G below.
Multiple uses of %G accumulate in the obvious fashion; the position of the %G is unimportant. Negative integers are not handled.
Note that when prompt truncation is in use it is advisable to divide up output into single characters within each %{...%} group so that the correct truncation point can be found.
The left parenthesis may be preceded or followed by a positive integer n, which defaults to zero. A negative integer will be multiplied by -1, except as noted below for `l'. The test character x may be any of the following:
The numeric argument, which in the third form may appear immediately after the `[', specifies the maximum permitted length of the various strings that can be displayed in the prompt. In the first two forms, this numeric argument may be negative, in which case the truncation length is determined by subtracting the absolute value of the numeric argument from the number of character positions remaining on the current prompt line. If this results in a zero or negative length, a length of 1 is used. In other words, a negative argument arranges that after truncation at least n characters remain before the right margin (left margin for RPROMPT).
The forms with `<' truncate at the left of the string, and the forms with `>' truncate at the right of the string. For example, if the current directory is `/home/pike', the prompt `%8<..<%/' will expand to `..e/pike'. In this string, the terminating character (`<', `>' or `]'), or in fact any character, may be quoted by a preceding `\'; note when using print -P, however, that this must be doubled as the string is also subject to standard print processing, in addition to any backslashes removed by a double quoted string: the worst case is therefore `print -P "%<\\\\<<..."'.
If the string is longer than the specified truncation length, it will appear in full, completely replacing the truncated string.
The part of the prompt string to be truncated runs to the end of the string, or to the end of the next enclosing group of the `%(' construct, or to the next truncation encountered at the same grouping level (i.e. truncations inside a `%(' are separate), which ever comes first. In particular, a truncation with argument zero (e.g., `%<<') marks the end of the range of the string to be truncated while turning off truncation from there on. For example, the prompt `%10<...<%~%<<%# ' will print a truncated representation of the current directory, followed by a `%' or `#', followed by a space. Without the `%<<', those two characters would be included in the string to be truncated. Note that `%-0<<' is not equivalent to `%<<' but specifies that the prompt is truncated at the right margin.
Truncation applies only within each individual line of the prompt, as delimited by embedded newlines (if any). If the total length of any line of the prompt after truncation is greater than the terminal width, or if the part to be truncated contains embedded newlines, truncation behavior is undefined and may change in a future version of the shell. Use `%-n(l.true-text.false-text)' to remove parts of the prompt when the available space is less than n.
The following sections explain the types of expansion in detail.
Immediately before execution, each command is saved in the history list, the size of which is controlled by the HISTSIZE parameter. The one most recent command is always retained in any case. Each saved command in the history list is called a history event and is assigned a number, beginning with 1 (one) when the shell starts up. The history number that you may see in your prompt (see EXPANSION OF PROMPT SEQUENCES in zshmisc(1)) is the number that is to be assigned to the next command.
The first character is followed by an optional event designator (see the section `Event Designators') and then an optional word designator (the section `Word Designators'); if neither of these designators is present, no history expansion occurs.
Input lines containing history expansions are echoed after being expanded, but before any other expansions take place and before the command is executed. It is this expanded form that is recorded as the history event for later references.
History expansions do not nest.
By default, a history reference with no event designator refers to the same event as any preceding history reference on that command line; if it is the only history reference in a command, it refers to the previous command. However, if the option CSH_JUNKIE_HISTORY is set, then every history reference with no event specification always refers to the previous command.
For example, `!' is the event designator for the previous command, so `!!:1' always refers to the first word of the previous command, and `!!$' always refers to the last word of the previous command. With CSH_JUNKIE_HISTORY set, then `!:1' and `!$' function in the same manner as `!!:1' and `!!$', respectively. Conversely, if CSH_JUNKIE_HISTORY is unset, then `!:1' and `!$' refer to the first and last words, respectively, of the same event referenced by the nearest other history reference preceding them on the current command line, or to the previous command if there is no preceding reference.
The character sequence `^foo^bar' (where `^' is actually the second character of the histchars parameter) repeats the last command, replacing the string foo with bar. More precisely, the sequence `^foo^bar^' is synonymous with `!!:s^foo^bar^', hence other modifiers (see the section `Modifiers') may follow the final `^'. In particular, `^foo^bar^:G' performs a global substitution.
If the shell encounters the character sequence `!"' in the input, the history mechanism is temporarily disabled until the current list (see zshmisc(1)) is fully parsed. The `!"' is removed from the input, and any subsequent `!' characters have no special significance.
A less convenient but more comprehensible form of command history support is provided by the fc builtin.
Note that a `%' word designator works only when used in one of `!%', `!:%' or `!?str?:%', and only when used after a !? expansion (possibly in an earlier command). Anything else results in an error, although the error may not be the most obvious one.
This transformation is agnostic about what is in the filesystem, i.e. is on the logical, not the physical directory. It takes place in the same manner as when changing directories when neither of the options CHASE_DOTS or CHASE_LINKS is set. For example, `/before/here/../after' is always transformed to `/before/after', regardless of whether `/before/here' exists or what kind of object (dir, file, symlink, etc.) it is.
Note: on systems that do not have a realpath(3) library function, symbolic links are not resolved, so on those systems `a' and `A' are equivalent.
Note: foo:A and realpath(foo) are different on some inputs. For realpath(foo) semantics, see the `P` modifier.
Unlike realpath(3), non-existent trailing components are permitted and preserved.
The forms `gs/l/r' and `s/l/r/:G' perform global substitution, i.e. substitute every occurrence of r for l. Note that the g or :G must appear in exactly the position shown.
See further notes on this form of substitution below.
The s/l/r/ substitution works as follows. By default the left-hand side of substitutions are not patterns, but character strings. Any character can be used as the delimiter in place of `/'. A backslash quotes the delimiter character. The character `&', in the right-hand-side r, is replaced by the text from the left-hand-side l. The `&' can be quoted with a backslash. A null l uses the previous string either from the previous l or from the contextual scan string s from `!?s'. You can omit the rightmost delimiter if a newline immediately follows r; the rightmost `?' in a context scan can similarly be omitted. Note the same record of the last l and r is maintained across all forms of expansion.
Note that if a `&' is used within glob qualifiers an extra backslash is needed as a & is a special character in this case.
Also note that the order of expansions affects the interpretation of l and r. When used in a history expansion, which occurs before any other expansions, l and r are treated as literal strings (except as explained for HIST_SUBST_PATTERN below). When used in parameter expansion, the replacement of r into the parameter's value is done first, and then any additional process, parameter, command, arithmetic, or brace references are applied, which may evaluate those substitutions and expansions more than once if l appears more than once in the starting value. When used in a glob qualifier, any substitutions or expansions are performed once at the time the qualifier is parsed, even before the `:s' expression itself is divided into l and r sides.
If the option HIST_SUBST_PATTERN is set, l is treated as a pattern of the usual form described in the section FILENAME GENERATION below. This can be used in all the places where modifiers are available; note, however, that in globbing qualifiers parameter substitution has already taken place, so parameters in the replacement string should be quoted to ensure they are replaced at the correct time. Note also that complicated patterns used in globbing qualifiers may need the extended glob qualifier notation (#q:s/.../.../) in order for the shell to recognize the expression as a glob qualifier. Further, note that bad patterns in the substitution are not subject to the NO_BAD_PATTERN option so will cause an error.
When HIST_SUBST_PATTERN is set, l may start with a # to indicate that the pattern must match at the start of the string to be substituted, and a % may appear at the start or after an # to indicate that the pattern must match at the end of the string to be substituted. The % or # may be quoted with two backslashes.
For example, the following piece of filename generation code with the EXTENDED_GLOB option:
print -r -- *.c(#q:s/#%(#b)s(*).c/'S${match[1]}.C'/)
takes the expansion of *.c and applies the glob qualifiers in the (#q...) expression, which consists of a substitution modifier anchored to the start and end of each word (#%). This turns on backreferences ((#b)), so that the parenthesised subexpression is available in the replacement string as ${match[1]}. The replacement string is quoted so that the parameter is not substituted before the start of filename generation.
The following f, F, w and W modifiers work only with parameter expansion and filename generation. They are listed here to provide a single point of reference for all modifiers.
Note that `<<(list)' is not a special syntax; it is equivalent to `< <(list)', redirecting standard input from the result of process substitution. Hence all the following documentation applies. The second form (with the space) is recommended for clarity.
In the case of the < or > forms, the shell runs the commands in list as a subprocess of the job executing the shell command line. If the system supports the /dev/fd mechanism, the command argument is the name of the device file corresponding to a file descriptor; otherwise, if the system supports named pipes (FIFOs), the command argument will be a named pipe. If the form with > is selected then writing on this special file will provide input for list. If < is used, then the file passed as an argument will be connected to the output of the list process. For example,
paste <(cut -f1 file1) <(cut -f3 file2) | tee >(process1) >(process2) >/dev/null
cuts fields 1 and 3 from the files file1 and file2 respectively, pastes the results together, and sends it to the processes process1 and process2.
If =(...) is used instead of <(...), then the file passed as an argument will be the name of a temporary file containing the output of the list process. This may be used instead of the < form for a program that expects to lseek (see lseek(2)) on the input file.
There is an optimisation for substitutions of the form =(<<<arg), where arg is a single-word argument to the here-string redirection <<<. This form produces a file name containing the value of arg after any substitutions have been performed. This is handled entirely within the current shell. This is effectively the reverse of the special form $(<arg) which treats arg as a file name and replaces it with the file's contents.
The = form is useful as both the /dev/fd and the named pipe implementation of <(...) have drawbacks. In the former case, some programmes may automatically close the file descriptor in question before examining the file on the command line, particularly if this is necessary for security reasons such as when the programme is running setuid. In the second case, if the programme does not actually open the file, the subshell attempting to read from or write to the pipe will (in a typical implementation, different operating systems may have different behaviour) block for ever and have to be killed explicitly. In both cases, the shell actually supplies the information using a pipe, so that programmes that expect to lseek (see lseek(2)) on the file will not work.
Also note that the previous example can be more compactly and efficiently written (provided the MULTIOS option is set) as:
paste <(cut -f1 file1) <(cut -f3 file2) \ > >(process1) > >(process2)
The shell uses pipes instead of FIFOs to implement the latter two process substitutions in the above example.
There is an additional problem with >(process); when this is attached to an external command, the parent shell does not wait for process to finish and hence an immediately following command cannot rely on the results being complete. The problem and solution are the same as described in the section MULTIOS in zshmisc(1). Hence in a simplified version of the example above:
paste <(cut -f1 file1) <(cut -f3 file2) > >(process)
(note that no MULTIOS are involved), process will be run asynchronously as far as the parent shell is concerned. The workaround is:
{ paste <(cut -f1 file1) <(cut -f3 file2) } > >(process)
The extra processes here are spawned from the parent shell which will wait for their completion.
Another problem arises any time a job with a substitution that requires a temporary file is disowned by the shell, including the case where `&!' or `&|' appears at the end of a command containing a substitution. In that case the temporary file will not be cleaned up as the shell no longer has any memory of the job. A workaround is to use a subshell, for example,
(mycmd =(myoutput)) &!
as the forked subshell will wait for the command to finish then remove the temporary file.
A general workaround to ensure a process substitution endures for an appropriate length of time is to pass it as a parameter to an anonymous shell function (a piece of shell code that is run immediately with function scope). For example, this code:
() {
print File $1:
cat $1
} =(print This be the verse)
outputs something resembling the following
File /tmp/zsh6nU0kS: This be the verse
The temporary file created by the process substitution will be deleted when the function exits.
Note in particular the fact that words of unquoted parameters are not automatically split on whitespace unless the option SH_WORD_SPLIT is set; see references to this option below for more details. This is an important difference from other shells. However, as in other shells, null words are elided from unquoted parameters' expansions.
With default options, after the assignments:
array=("first word" "" "third word")
scalar="only word"
then $array substitutes two words, `first word' and `third word', and $scalar substitutes a single word `only word'. Note that second element of array was elided. Scalar parameters can be elided too if their value is null (empty). To avoid elision, use quoting as follows: "$scalar" for scalars and "${array[@]}" or "${(@)array}" for arrays. (The last two forms are equivalent.)
Parameter expansions can involve flags, as in `${(@kv)aliases}', and other operators, such as `${PREFIX:-"/usr/local"}'. Parameter expansions can also be nested. These topics will be introduced below. The full rules are complicated and are noted at the end.
In the expansions discussed below that require a pattern, the form of the pattern is the same as that used for filename generation; see the section `Filename Generation'. Note that these patterns, along with the replacement text of any substitutions, are themselves subject to parameter expansion, command substitution, and arithmetic expansion. In addition to the following operations, the colon modifiers described in the section `Modifiers' in the section `History Expansion' can be applied: for example, ${i:s/foo/bar/} performs string substitution on the expansion of parameter $i.
In the following descriptions, `word' refers to a single word substituted on the command line, not necessarily a space delimited word.
If name is an array parameter, and the KSH_ARRAYS option is not set, then the value of each element of name is substituted, one element per word. Otherwise, the expansion results in one word only; with KSH_ARRAYS, this is the first element of an array. No field splitting is done on the result unless the SH_WORD_SPLIT option is set. See also the flags = and s:string:.
In any of the above expressions that test a variable and substitute an alternate word, note that you can use standard shell quoting in the word value to selectively override the splitting done by the SH_WORD_SPLIT option and the = flag, but not splitting by the s:string: flag.
In the following expressions, when name is an array and the substitution is not quoted, or if the `(@)' flag or the name[@] syntax is used, matching and replacement is performed on each array element separately.
a=(1 2 3 4); b=(a b); print ${a:^b}
will output `1 a 2 b'. For `:^^', then the input is repeated until all of the longer array has been used up and the above will output `1 a 2 b 3 a 4 b'.
Either or both inputs may be a scalar, they will be treated as an array of length 1 with the scalar as the only element. If either array is empty, the other array is output with no extra elements inserted.
Currently the following code will output `a b' and `1' as two separate elements, which can be unexpected. The second print provides a workaround which should continue to work if this is changed.
a=(a b); b=(1 2); print -l "${a:^b}"; print -l "${${a:^b}}"
If offset is non-negative, then if the variable name is a scalar substitute the contents starting offset characters from the first character of the string, and if name is an array substitute elements starting offset elements from the first element. If length is given, substitute that many characters or elements, otherwise the entire rest of the scalar or array.
A positive offset is always treated as the offset of a character or element in name from the first character or element of the array (this is different from native zsh subscript notation). Hence 0 refers to the first character or element regardless of the setting of the option KSH_ARRAYS.
A negative offset counts backwards from the end of the scalar or array, so that -1 corresponds to the last character or element, and so on.
When positive, length counts from the offset position toward the end of the scalar or array. When negative, length counts back from the end. If this results in a position smaller than offset, a diagnostic is printed and nothing is substituted.
The option MULTIBYTE is obeyed, i.e. the offset and length count multibyte characters where appropriate.
offset and length undergo the same set of shell substitutions as for scalar assignment; in addition, they are then subject to arithmetic evaluation. Hence, for example
print ${foo:3}
print ${foo: 1 + 2}
print ${foo:$(( 1 + 2))}
print ${foo:$(echo 1 + 2)}
all have the same effect, extracting the string starting at the fourth character of $foo if the substitution would otherwise return a scalar, or the array starting at the fourth element if $foo would return an array. Note that with the option KSH_ARRAYS $foo always returns a scalar (regardless of the use of the offset syntax) and a form such as ${foo[*]:3} is required to extract elements of an array named foo.
If offset is negative, the - may not appear immediately after the : as this indicates the ${name:-word} form of substitution. Instead, a space may be inserted before the -. Furthermore, neither offset nor length may begin with an alphabetic character or & as these are used to indicate history-style modifiers. To substitute a value from a variable, the recommended approach is to precede it with a $ as this signifies the intention (parameter substitution can easily be rendered unreadable); however, as arithmetic substitution is performed, the expression ${var: offs} does work, retrieving the offset from $offs.
For further compatibility with other shells there is a special case for array offset 0. This usually accesses the first element of the array. However, if the substitution refers to the positional parameter array, e.g. $@ or $*, then offset 0 instead refers to $0, offset 1 refers to $1, and so on. In other words, the positional parameter array is effectively extended by prepending $0. Hence ${*:0:1} substitutes $0 and ${*:1:1} substitutes $1.
The pattern may begin with a `#', in which case the pattern must match at the start of the string, or `%', in which case it must match at the end of the string, or `#%' in which case the pattern must match the entire string. The repl may be an empty string, in which case the final `/' may also be omitted. To quote the final `/' in other cases it should be preceded by a single backslash; this is not necessary if the `/' occurs inside a substituted parameter. Note also that the `#', `%' and `#% are not active if they occur inside a substituted parameter, even at the start.
If, after quoting rules apply, ${name} expands to an array, the replacements act on each element individually. Note also the effect of the I and S parameter expansion flags below; however, the flags M, R, B, E and N are not useful.
For example,
foo="twinkle twinkle little star" sub="t*e" rep="spy"
print ${foo//${~sub}/$rep}
print ${(S)foo//${~sub}/$rep}
Here, the `~' ensures that the text of $sub is treated as a pattern rather than a plain string. In the first case, the longest match for t*e is substituted and the result is `spy star', while in the second case, the shortest matches are taken and the result is `spy spy lispy star'.
If the option POSIX_IDENTIFIERS is not set, and spec is a simple name, then the braces are optional; this is true even for special parameters so e.g. $#- and $#* take the length of the string $- and the array $* respectively. If POSIX_IDENTIFIERS is set, then braces are required for the # to be treated in this fashion.
Internally, each such expansion is converted into the equivalent list for brace expansion. E.g., ${^var} becomes {$var[1],$var[2],...}, and is processed as described in the section `Brace Expansion' below: note, however, the expansion happens immediately, with any explicit brace expansion happening later. If word splitting is also in effect the $var[N] may themselves be split into different list elements.
Note that splitting is applied to word in the assignment forms of spec before the assignment to name is performed. This affects the result of array assignments with the A flag.
In nested substitutions, note that the effect of the ~ applies to the result of the current level of substitution. A surrounding pattern operation on the result may cancel it. Hence, for example, if the parameter foo is set to *, ${~foo//\*/*.c} is substituted by the pattern *.c, which may be expanded by filename generation, but ${${~foo}//\*/*.c} substitutes to the string *.c, which will not be further expanded.
If a ${...} type parameter expression or a $(...) type command substitution is used in place of name above, it is expanded first and the result is used as if it were the value of name. Thus it is possible to perform nested operations: ${${foo#head}%tail} substitutes the value of $foo with both `head' and `tail' deleted. The form with $(...) is often useful in combination with the flags described next; see the examples below. Each name or nested ${...} in a parameter expansion may also be followed by a subscript expression as described in Array Parameters in zshparam(1).
Note that double quotes may appear around nested expressions, in which case only the part inside is treated as quoted; for example, ${(f)"$(foo)"} quotes the result of $(foo), but the flag `(f)' (see below) is applied using the rules for unquoted expansions. Note further that quotes are themselves nested in this context; for example, in "${(@f)"$(foo)"}", there are two sets of quotes, one surrounding the whole expression, the other (redundant) surrounding the $(foo) as before.
If the MULTIBYTE option is set and the number is greater than 127 (i.e. not an ASCII character) it is treated as a Unicode character.
This assigns an array parameter with `${...=...}', `${...:=...}' or `${...::=...}'. If this flag is repeated (as in `AA'), assigns an associative array parameter. Assignment is made before sorting or padding; if field splitting is active, the word part is split before assignment. The name part may be a subscripted range for ordinary arrays; when assigning an associative array, the word part must be converted to an array, for example by using `${(AA)=name=...}' to activate field splitting.
Surrounding context such as additional nesting or use of the value in a scalar assignment may cause the array to be joined back into a single string again.
Quoting using one of the q family of flags does not work for this purpose since quotes are not stripped from non-pattern characters by GLOB_SUBST. In other words,
pattern=${(q)str}
[[ $str = ${~pattern} ]]
works if $str is `a*b' but not if it is `a b', whereas
pattern=${(b)str}
[[ $str = ${~pattern} ]]
is always true for any possible value of $str.
If used with a nested parameter or command substitution, the result of that will be taken as a parameter name in the same way. For example, if you have `foo=bar' and `bar=baz', the strings ${(P)foo}, ${(P)${foo}}, and ${(P)$(echo bar)} will be expanded to `baz'.
Likewise, if the reference is itself nested, the expression with the flag is treated as if it were directly replaced by the parameter name. It is an error if this nested substitution produces an array with more than one word. For example, if `name=assoc' where the parameter assoc is an associative array, then `${${(P)name}[elt]}' refers to the element of the associative subscripted `elt'.
If this flag is given twice, the resulting words are quoted in single quotes and if it is given three times, the words are quoted in double quotes; in these forms no special handling of unprintable or invalid characters is attempted. If the flag is given four times, the words are quoted in single quotes preceded by a $. Note that in all three of these forms quoting is done unconditionally, even if this does not change the way the resulting string would be interpreted by the shell.
If a q- is given (only a single q may appear), a minimal form of single quoting is used that only quotes the string if needed to protect special characters. Typically this form gives the most readable output.
If a q+ is given, an extended form of minimal quoting is used that causes unprintable characters to be rendered using $'...'. This quoting is similar to that used by the output of values by the typeset family of commands.
Note that this is done very late, even later than the `(s)' flag. So to access single words in the result use nested expansions as in `${${(z)foo}[2]}'. Likewise, to remove the quotes in the resulting words use `${(Q)${(z)foo}}'.
The following flags (except p) are followed by one or more arguments as shown. Any character, or the matching pairs `(...)', `{...}', `[...]', or `<...>', may be used in place of a colon as delimiters, but note that when a flag takes more than one argument, a matched pair of delimiters must surround each argument.
Alternatively, with this option string arguments may be in the form $var in which case the value of the variable is substituted. Note this form is strict; the string argument does not undergo general parameter expansion.
For example,
sep=:
val=a:b:c
print ${(ps.$sep.)val}
splits the variable on a :.
[[ "?" = ${(~j.|.)array} ]]
treats `|' as a pattern and succeeds if and only if $array contains the string `?' as an element. The ~ may be repeated to toggle the behaviour; its effect only lasts to the end of the parenthesised group.
The arguments :string1: and :string2: are optional; neither, the first, or both may be given. Note that the same pairs of delimiters must be used for each of the three arguments. The space to the left will be filled with string1 (concatenated as often as needed) or spaces if string1 is not given. If both string1 and string2 are given, string2 is inserted once directly to the left of each word, truncated if necessary, before string1 is used to produce any remaining padding.
If either of string1 or string2 is present but empty, i.e. there are two delimiters together at that point, the first character of $IFS is used instead.
If the MULTIBYTE option is in effect, the flag m may also be given, in which case widths will be used for the calculation of padding; otherwise individual multibyte characters are treated as occupying one unit of width.
If the MULTIBYTE option is not in effect, each byte in the string is treated as occupying one unit of width.
Control characters are always assumed to be one unit wide; this allows the mechanism to be used for generating repetitions of control characters.
If the m is repeated, the character either counts zero (if it has zero width), else one. For printable character strings this has the effect of counting the number of glyphs (visibly separate characters), except for the case where combining characters themselves have non-zero width (true in certain alphabets).
Left and right padding may be used together. In this case the strategy is to apply left padding to the first half width of each of the resulting words, and right padding to the second half. If the string to be padded has odd width the extra padding is applied on the left.
For historical reasons, the usual behaviour that empty array elements are retained inside double quotes is disabled for arrays generated by splitting; hence the following:
line="one::three"
print -l "${(s.:.)line}"
produces two lines of output for one and three and elides the empty field. To override this behaviour, supply the `(@)' flag as well, i.e. "${(@s.:.)line}".
The following flags are meaningful with the ${...#...} or ${...%...} forms. The S and I flags may also be used with the ${.../...} forms.
% str="aXbXc"
% echo ${(S)str#X*}
abXc
% echo ${(S)str##X*}
a
%
With % or %%, search for the match that starts closest to the end of the string:
% str="aXbXc"
% echo ${(S)str%X*}
aXbc
% echo ${(S)str%%X*}
aXb
%
(Note that % and %% don't search for the match that ends closest to the end of the string, as one might expect.)
With substitution via ${.../...} or ${...//...}, specifies non-greedy matching, i.e. that the shortest instead of the longest match should be replaced:
% str="abab"
% echo ${str/*b/_}
_
% echo ${(S)str/*b/_}
_ab
%
The exprth match is counted such that there is either one or zero matches from each starting position in the string, although for global substitution matches overlapping previous replacements are ignored. With the ${...%...} and ${...%%...} forms, the starting position for the match moves backwards from the end as the index increases, while with the other forms it moves forward from the start.
Hence with the string
which switch is the right switch for Ipswich?
Here is a summary of the rules for substitution; this assumes that braces are present around the substitution, i.e. ${...}. Some particular examples are given below. Note that the Zsh Development Group accepts no responsibility for any brain damage which may occur during the reading of the following rules.
Note that, unless the `(P)' flag is present, the flags and any subscripts apply directly to the value of the nested substitution; for example, the expansion ${${foo}} behaves exactly the same as ${foo}. When the `(P)' flag is present in a nested substitution, the other substitution rules are applied to the value before it is interpreted as a name, so ${${(P)foo}} may differ from ${(P)foo}.
At each nested level of substitution, the substituted words undergo all forms of single-word substitution (i.e. not filename generation), including command substitution, arithmetic expansion and filename expansion (i.e. leading ~ and =). Thus, for example, ${${:-=cat}:h} expands to the directory where the cat program resides. (Explanation: the internal substitution has no parameter but a default value =cat, which is expanded by filename expansion to a full path; the outer substitution then applies the modifier :h and takes the directory part of the path.)
At the outermost level of substitution, the `(P)' flag (rule 4.) ignores these transformations and uses the unmodified value of the parameter as the name to be replaced. This is usually the desired behavior because padding may make the value syntactically illegal as a parameter name, but if capitalization changes are desired, use the ${${(P)foo}} form (rule 25.).
If the value so far names a parameter that has internal flags (rule 2.), those internal flags are applied to the new value after replacement.
If no `(s)', `(f)' or `=' was given, but the word is not quoted and the option SH_WORD_SPLIT is set, the word is split on occurrences of any of the characters in $IFS. Note this step, too, takes place at all levels of a nested substitution.
If a single word is not required, this rule is skipped.
Strictly speaking, the removal happens later as the same happens with other forms of substitution; the point to note here is simply that it occurs after any of the above parameter operations.
The following illustrates the rules for nested parameter expansions. Suppose that $foo contains the array (bar baz):
As an example of the rules for word splitting and joining, suppose $foo contains the array `(ax1 bx1)'. Then
The substitution `$(cat foo)' may be replaced by the faster `$(<foo)'. In this case foo undergoes single word shell expansions (parameter expansion, command substitution and arithmetic expansion), but not filename generation.
If the option GLOB_SUBST is set, the result of any unquoted command substitution, including the special form just mentioned, is eligible for filename generation.
An expression of the form `{n1..n2}', where n1 and n2 are integers, is expanded to every number between n1 and n2 inclusive. If either number begins with a zero, all the resulting numbers will be padded with leading zeroes to that minimum width, but for negative numbers the - character is also included in the width. If the numbers are in decreasing order the resulting sequence will also be in decreasing order.
An expression of the form `{n1..n2..n3}', where n1, n2, and n3 are integers, is expanded as above, but only every n3th number starting from n1 is output. If n3 is negative the numbers are output in reverse order, this is slightly different from simply swapping n1 and n2 in the case that the step n3 doesn't evenly divide the range. Zero padding can be specified in any of the three numbers, specifying it in the third can be useful to pad for example `{-99..100..01}' which is not possible to specify by putting a 0 on either of the first two numbers (i.e. pad to two characters).
An expression of the form `{c1..c2}', where c1 and c2 are single characters (which may be multibyte characters), is expanded to every character in the range from c1 to c2 in whatever character sequence is used internally. For characters with code points below 128 this is US ASCII (this is the only case most users will need). If any intervening character is not printable, appropriate quotation is used to render it printable. If the character sequence is reversed, the output is in reverse order, e.g. `{d..a}' is substituted as `d c b a'.
If a brace expression matches none of the above forms, it is left unchanged, unless the option BRACE_CCL (an abbreviation for `brace character class') is set. In that case, it is expanded to a list of the individual characters between the braces sorted into the order of the characters in the ASCII character set (multibyte characters are not currently handled). The syntax is similar to a [...] expression in filename generation: `-' is treated specially to denote a range of characters, but `^' or `!' as the first character is treated normally. For example, `{abcdef0-9}' expands to 16 words 0 1 2 3 4 5 6 7 8 9 a b c d e f.
Note that brace expansion is not part of filename generation (globbing); an expression such as */{foo,bar} is split into two separate words */foo and */bar before filename generation takes place. In particular, note that this is liable to produce a `no match' error if either of the two expressions does not match; this is to be contrasted with */(foo|bar), which is treated as a single pattern but otherwise has similar effects.
To combine brace expansion with array expansion, see the ${^spec} form described in the section Parameter Expansion above.
A `~' by itself is replaced by the value of $HOME. A `~' followed by a `+' or a `-' is replaced by current or previous working directory, respectively.
A `~' followed by a number is replaced by the directory at that position in the directory stack. `~0' is equivalent to `~+', and `~1' is the top of the stack. `~+' followed by a number is replaced by the directory at that position in the directory stack. `~+0' is equivalent to `~+', and `~+1' is the top of the stack. `~-' followed by a number is replaced by the directory that many positions from the bottom of the stack. `~-0' is the bottom of the stack. The PUSHD_MINUS option exchanges the effects of `~+' and `~-' where they are followed by a number.
If the function zsh_directory_name exists, or the shell variable zsh_directory_name_functions exists and contains an array of function names, then the functions are used to implement dynamic directory naming. The functions are tried in order until one returns status zero, so it is important that functions test whether they can handle the case in question and return an appropriate status.
A `~' followed by a string namstr in unquoted square brackets is treated specially as a dynamic directory name. Note that the first unquoted closing square bracket always terminates namstr. The shell function is passed two arguments: the string n (for name) and namstr. It should either set the array reply to a single element which is the directory corresponding to the name and return status zero (executing an assignment as the last statement is usually sufficient), or it should return status non-zero. In the former case the element of reply is used as the directory; in the latter case the substitution is deemed to have failed. If all functions fail and the option NOMATCH is set, an error results.
The functions defined as above are also used to see if a directory can be turned into a name, for example when printing the directory stack or when expanding %~ in prompts. In this case each function is passed two arguments: the string d (for directory) and the candidate for dynamic naming. The function should either return non-zero status, if the directory cannot be named by the function, or it should set the array reply to consist of two elements: the first is the dynamic name for the directory (as would appear within `~[...]'), and the second is the prefix length of the directory to be replaced. For example, if the trial directory is /home/myname/src/zsh and the dynamic name for /home/myname/src (which has 16 characters) is s, then the function sets
reply=(s 16)
The directory name so returned is compared with possible static names for parts of the directory path, as described below; it is used if the prefix length matched (16 in the example) is longer than that matched by any static name.
It is not a requirement that a function implements both n and d calls; for example, it might be appropriate for certain dynamic forms of expansion not to be contracted to names. In that case any call with the first argument d should cause a non-zero status to be returned.
The completion system calls `zsh_directory_name c' followed by equivalent calls to elements of the array zsh_directory_name_functions, if it exists, in order to complete dynamic names for directories. The code for this should be as for any other completion function as described in zshcompsys(1).
As a working example, here is a function that expands any dynamic names beginning with the string p: to directories below /home/pws/perforce. In this simple case a static name for the directory would be just as effective.
zsh_directory_name() {
emulate -L zsh
setopt extendedglob
local -a match mbegin mend
if [[ $1 = d ]]; then
# turn the directory into a name
if [[ $2 = (#b)(/home/pws/perforce/)([^/]##)* ]]; then
typeset -ga reply
reply=(p:$match[2] $(( ${#match[1]} + ${#match[2]} )) )
else
return 1
fi
elif [[ $1 = n ]]; then
# turn the name into a directory
[[ $2 != (#b)p:(?*) ]] && return 1
typeset -ga reply
reply=(/home/pws/perforce/$match[1])
elif [[ $1 = c ]]; then
# complete names
local expl
local -a dirs
dirs=(/home/pws/perforce/*(/:t))
dirs=(p:${^dirs})
_wanted dynamic-dirs expl 'dynamic directory' compadd -S\] -a dirs
return
else
return 1
fi
return 0
}
It is also possible to define directory names using the -d option to the hash builtin.
When the shell prints a path (e.g. when expanding %~ in prompts or when printing the directory stack), the path is checked to see if it has a named directory as its prefix. If so, then the prefix portion is replaced with a `~' followed by the name of the directory. The shorter of the two ways of referring to the directory is used, i.e. either the directory name or the full path; the name is used if they are the same length. The parameters $PWD and $OLDPWD are never abbreviated in this fashion.
If a word begins with an unquoted `=' and the EQUALS option is set, the remainder of the word is taken as the name of a command. If a command exists by that name, the word is replaced by the full pathname of the command.
Filename expansion is performed on the right hand side of a parameter assignment, including those appearing after commands of the typeset family. In this case, the right hand side will be treated as a colon-separated list in the manner of the PATH parameter, so that a `~' or an `=' following a `:' is eligible for expansion. All such behaviour can be disabled by quoting the `~', the `=', or the whole expression (but not simply the colon); the EQUALS option is also respected.
If the option MAGIC_EQUAL_SUBST is set, any unquoted shell argument in the form `identifier=expression' becomes eligible for file expansion as described in the previous paragraph. Quoting the first `=' also inhibits this.
The word is replaced with a list of sorted filenames that match the pattern. If no matching pattern is found, the shell gives an error message, unless the NULL_GLOB option is set, in which case the word is deleted; or unless the NOMATCH option is unset, in which case the word is left unchanged.
In filename generation, the character `/' must be matched explicitly; also, a `.' must be matched explicitly at the beginning of a pattern or after a `/', unless the GLOB_DOTS option is set. No filename generation pattern matches the files `.' or `..'. In other instances of pattern matching, the `/' and `.' are not treated specially.
Another set of named classes is handled internally by the shell and is not sensitive to the locale:
Note that the square brackets are additional to those enclosing the whole set of characters, so to test for a single alphanumeric character you need `[[:alnum:]]'. Named character sets can be used alongside other types, e.g. `[[:alpha:]0-9]'.
Be careful when using other wildcards adjacent to patterns of this form; for example, <0-9>* will actually match any number whatsoever at the start of the string, since the `<0-9>' will match the first digit, and the `*' will match any others. This is a trap for the unwary, but is in fact an inevitable consequence of the rule that the longest possible match always succeeds. Expressions such as `<0-9>[^[:digit:]]*' can be used instead.
Note that grouping cannot extend over multiple directories: it is an error to have a `/' within a group (this only applies for patterns used in filename generation). There is one exception: a group of the form (pat/)# appearing as a complete path segment can match a sequence of directories. For example, foo/(a*/)#bar matches foo/bar, foo/any/bar, foo/any/anyother/bar, and so on.
For example,
foo="a_string_with_a_message"
if [[ $foo = (a|an)_(#b)(*) ]]; then
print ${foo[$mbegin[1],$mend[1]]}
fi
prints `string_with_a_message'. Note that the first set of parentheses is before the (#b) and does not create a backreference.
Backreferences work with all forms of pattern matching other than filename generation, but note that when performing matches on an entire array, such as ${array#pattern}, or a global substitution, such as ${param//pat/repl}, only the data for the last match remains available. In the case of global replacements this may still be useful. See the example for the m flag below.
The numbering of backreferences strictly follows the order of the opening parentheses from left to right in the pattern string, although sets of parentheses may be nested. There are special rules for parentheses followed by `#' or `##'. Only the last match of the parenthesis is remembered: for example, in `[[ abab = (#b)([ab])# ]]', only the final `b' is stored in match[1]. Thus extra parentheses may be necessary to match the complete segment: for example, use `X((ab|cd)#)Y' to match a whole string of either `ab' or `cd' between `X' and `Y', using the value of $match[1] rather than $match[2].
If the match fails none of the parameters is altered, so in some cases it may be necessary to initialise them beforehand. If some of the backreferences fail to match -- which happens if they are in an alternate branch which fails to match, or if they are followed by # and matched zero times -- then the matched string is set to the empty string, and the start and end indices are set to -1.
Pattern matching with backreferences is slightly slower than without.
For example,
arr=(veldt jynx grimps waqf zho buck)
print ${arr//(#m)[aeiou]/${(U)MATCH}}
forces all the matches (i.e. all vowels) into uppercase, printing `vEldt jynx grImps wAqf zhO bUck'.
Unlike backreferences, there is no speed penalty for using match references, other than the extra substitutions required for the replacement strings in cases such as the example shown.
Another use is in parameter substitution; for example `${array/(#s)A*Z(#e)}' will remove only elements of an array which match the complete pattern `A*Z'. There are other ways of performing many operations of this type, however the combination of the substitution operations `/' and `//' with the `(#s)' and `(#e)' flags provides a single simple and memorable method.
Note that assertions of the form `(^(#s))' also work, i.e. match anywhere except at the start of the string, although this actually means `anything except a zero-length portion at the start of the string'; you need to use `(""~(#s))' to match a zero-length portion of the string not at the start.
For example, the test string fooxx can be matched by the pattern (#i)FOOXX, but not by (#l)FOOXX, (#i)FOO(#I)XX or ((#i)FOOX)X. The string (#ia2)readme specifies case-insensitive matching of readme with up to two errors.
When using the ksh syntax for grouping both KSH_GLOB and EXTENDED_GLOB must be set and the left parenthesis should be preceded by @. Note also that the flags do not affect letters inside [...] groups, in other words (#i)[a-z] still matches only lowercase letters. Finally, note that when examining whole paths case-insensitively every directory must be searched for all files which match, so that a pattern of the form (#i)/foo/bar/... is potentially slow.
Thus, the pattern (#a3)abcd matches dcba, with the errors occurring by using the first rule twice and the second once, grouping the string as [d][cb][a] and [a][bc][d].
Non-literal parts of the pattern must match exactly, including characters in character ranges: hence (#a1)??? matches strings of length four, by applying rule 4 to an empty part of the pattern, but not strings of length two, since all the ? must match. Other characters which must match exactly are initial dots in filenames (unless the GLOB_DOTS option is set), and all slashes in filenames, so that a/bc is two errors from ab/c (the slash cannot be transposed with another character). Similarly, errors are counted separately for non-contiguous strings in the pattern, so that (ab|cd)ef is two errors from aebf.
When using exclusion via the ~ operator, approximate matching is treated entirely separately for the excluded part and must be activated separately. Thus, (#a1)README~READ_ME matches READ.ME but not READ_ME, as the trailing READ_ME is matched without approximation. However, (#a1)README~(#a1)READ_ME does not match any pattern of the form READ?ME as all such forms are now excluded.
Apart from exclusions, there is only one overall error count; however, the maximum errors allowed may be altered locally, and this can be delimited by grouping. For example, (#a1)cat((#a0)dog)fox allows one error in total, which may not occur in the dog section, and the pattern (#a1)cat(#a0)dog(#a1)fox is equivalent. Note that the point at which an error is first found is the crucial one for establishing whether to use approximation; for example, (#a1)abc(#a0)xyz will not match abcdxyz, because the error occurs at the `x', where approximation is turned off.
Entire path segments may be matched approximately, so that `(#a1)/foo/d/is/available/at/the/bar' allows one error in any path segment. This is much less efficient than without the (#a1), however, since every directory in the path must be scanned for a possible approximate match. It is best to place the (#a1) after any path segments which are known to be correct.
As a shorthand, `**/' is equivalent to `(*/)#'; note that this therefore matches files in the current directory as well as subdirectories. Thus:
ls -ld -- (*/)#bar
or
ls -ld -- **/bar
does a recursive directory search for files named `bar' (potentially including the file `bar' in the current directory). This form does not follow symbolic links; the alternative form `***/' does, but is otherwise identical. Neither of these can be combined with other forms of globbing within the same path segment; in that case, the `*' operators revert to their usual effect.
Even shorter forms are available when the option GLOB_STAR_SHORT is set. In that case if no / immediately follows a ** or *** they are treated as if both a / plus a further * are present. Hence:
setopt GLOBSTARSHORT ls -ld -- **.c
is equivalent to
ls -ld -- **/*.c
If the option BARE_GLOB_QUAL is set, then a trailing set of parentheses containing no `|' or `(' characters (or `~' if it is special) is taken as a set of glob qualifiers. A glob subexpression that would normally be taken as glob qualifiers, for example `(^x)', can be forced to be treated as part of the glob pattern by doubling the parentheses, in this case producing `((^x))'.
If the option EXTENDED_GLOB is set, a different syntax for glob qualifiers is available, namely `(#qx)' where x is any of the same glob qualifiers used in the other format. The qualifiers must still appear at the end of the pattern. However, with this syntax multiple glob qualifiers may be chained together. They are treated as a logical AND of the individual sets of flags. Also, as the syntax is unambiguous, the expression will be treated as glob qualifiers just as long any parentheses contained within it are balanced; appearance of `|', `(' or `~' does not negate the effect. Note that qualifiers will be recognised in this form even if a bare glob qualifier exists at the end of the pattern, for example `*(#q*)(.)' will recognise executable regular files if both options are set; however, mixed syntax should probably be avoided for the sake of clarity. Note that within conditions using the `[[' form the presence of a parenthesised expression (#q...) at the end of a string indicates that globbing should be performed; the expression may include glob qualifiers, but it is also valid if it is simply (#q). This does not apply to the right hand side of pattern match operators as the syntax already has special significance.
A qualifier may be any one of the following:
If the qualifier `f' is followed by any other character anything up to the next matching character (`[', `{', and `<' match `]', `}', and `>' respectively, any other character matches itself) is taken as a list of comma-separated sub-specs. Each sub-spec may be either an octal number as described above or a list of any of the characters `u', `g', `o', and `a', followed by a `=', a `+', or a `-', followed by a list of any of the characters `r', `w', `x', `s', and `t', or an octal digit. The first list of characters specify which access rights are to be checked. If a `u' is given, those for the owner of the file are used, if a `g' is given, those of the group are checked, a `o' means to test those of other users, and the `a' says to test all three groups. The `=', `+', and `-' again says how the modes are to be checked and have the same meaning as described for the first form above. The second list of characters finally says which access rights are to be expected: `r' for read access, `w' for write access, `x' for the right to execute the file (or to search a directory), `s' for the setuid and setgid bits, and `t' for the sticky bit.
Thus, `*(f70?)' gives the files for which the owner has read, write, and execute permission, and for which other group members have no rights, independent of the permissions for other users. The pattern `*(f-100)' gives all files for which the owner does not have execute permission, and `*(f:gu+w,o-rx:)' gives the files for which the owner and the other members of the group have at least write permission, and for which other users don't have read or execute permission.
In the first form, the first character after the `e' will be used as a separator and anything up to the next matching separator will be taken as the string; `[', `{', and `<' match `]', `}', and `>', respectively, while any other character matches itself. Note that expansions must be quoted in the string to prevent them from being expanded before globbing is done. string is then executed as shell code. The string globqual is appended to the array zsh_eval_context the duration of execution.
During the execution of string the filename currently being tested is available in the parameter REPLY; the parameter may be altered to a string to be inserted into the list instead of the original filename. In addition, the parameter reply may be set to an array or a string, which overrides the value of REPLY. If set to an array, the latter is inserted into the command line word by word.
For example, suppose a directory contains a single file `lonely'. Then the expression `*(e:'reply=(${REPLY}{1,2})':)' will cause the words `lonely1' and `lonely2' to be inserted into the command line. Note the quoting of string.
The form +cmd has the same effect, but no delimiters appear around cmd. Instead, cmd is taken as the longest sequence of characters following the + that are alphanumeric or underscore. Typically cmd will be the name of a shell function that contains the appropriate test. For example,
nt() { [[ $REPLY -nt $NTREF ]] }
NTREF=reffile
ls -ld -- *(+nt)
lists all files in the directory that have been modified more recently than reffile.
Any fractional part of the difference between the access time and the current part in the appropriate units is ignored in the comparison. For instance, `echo *(ah-5)' would echo files accessed within the last five hours, while `echo *(ah+5)' would echo files accessed at least six hours ago, as times strictly between five and six hours are treated as five hours.
If this flag is directly followed by a size specifier `k' (`K'), `m' (`M'), or `p' (`P') (e.g. `Lk-50') the check is performed with kilobytes, megabytes, or blocks (of 512 bytes) instead. (On some systems additional specifiers are available for gigabytes, `g' or `G', and terabytes, `t' or `T'.) If a size specifier is used a file is regarded as "exactly" the size if the file size rounded up to the next unit is equal to the test size. Hence `*(Lm1)' matches files from 1 byte up to 1 Megabyte inclusive. Note also that the set of files "less than" the test size only includes files that would not match the equality test; hence `*(Lm-1)' only matches files of zero size.
Implies oN when no oc qualifier is used.
The default sorting is n (by name) unless the Y glob qualifier is used, in which case it is N (unsorted).
oe and o+ are special cases; they are each followed by shell code, delimited as for the e glob qualifier and the + glob qualifier respectively (see above). The code is executed for each matched file with the parameter REPLY set to the name of the file on entry and globsort appended to zsh_eval_context. The code should modify the parameter REPLY in some fashion. On return, the value of the parameter is used instead of the file name as the string on which to sort. Unlike other sort operators, oe and o+ may be repeated, but note that the maximum number of sort operators of any kind that may appear in any glob expression is 12.
A typical use for this is to prepend an option before all occurrences of a file name; for example, the pattern `*(P:-f:)' produces the command line arguments `-f file1 -f file2 ...'
If the modifier ^ is active, then string will be appended instead of prepended. Prepending and appending is done independently so both can be used on the same glob expression; for example by writing `*(P:foo:^P:bar:^P:baz:)' which produces the command line arguments `foo baz file1 bar ...'
More than one of these lists can be combined, separated by commas. The whole list matches if at least one of the sublists matches (they are `or'ed, the qualifiers in the sublists are `and'ed). Some qualifiers, however, affect all matches generated, independent of the sublist in which they are given. These are the qualifiers `M', `T', `N', `D', `n', `o', `O' and the subscripts given in brackets (`[...]').
If a `:' appears in a qualifier list, the remainder of the expression in parenthesis is interpreted as a modifier (see the section `Modifiers' in the section `History Expansion'). Each modifier must be introduced by a separate `:'. Note also that the result after modification does not have to be an existing file. The name of any existing file can be followed by a modifier of the form `(:...)' even if no actual filename generation is performed, although note that the presence of the parentheses causes the entire expression to be subjected to any global pattern matching options such as NULL_GLOB. Thus:
ls -ld -- *(-/)
lists all directories and symbolic links that point to directories, and
ls -ld -- *(-@)
lists all broken symbolic links, and
ls -ld -- *(%W)
lists all world-writable device files in the current directory, and
ls -ld -- *(W,X)
lists all files in the current directory that are world-writable or world-executable, and
print -rC1 /tmp/foo*(u0^@:t)
outputs the basename of all root-owned files beginning with the string `foo' in /tmp, ignoring symlinks, and
ls -ld -- *.*~(lex|parse).[ch](^D^l1)
lists all files having a link count of one whose names contain a dot (but not those starting with a dot, since GLOB_DOTS is explicitly switched off) except for lex.c, lex.h, parse.c and parse.h.
print -rC1 b*.pro(#q:s/pro/shmo/)(#q.:s/builtin/shmiltin/)
demonstrates how colon modifiers and other qualifiers may be chained together. The ordinary qualifier `.' is applied first, then the colon modifiers in order from left to right. So if EXTENDED_GLOB is set and the base pattern matches the regular file builtin.pro, the shell will print `shmiltin.shmo'.
The attributes of a parameter determine the type of its value, often referred to as the parameter type or variable type, and also control other processing that may be applied to the value when it is referenced. The value type may be a scalar (a string, an integer, or a floating point number), an array (indexed numerically), or an associative array (an unordered set of name-value pairs, indexed by name, also referred to as a hash).
Named scalar parameters may have the exported, -x, attribute, to copy them into the process environment, which is then passed from the shell to any new processes that it starts. Exported parameters are called environment variables. The shell also imports environment variables at startup time and automatically marks the corresponding parameters as exported. Some environment variables are not imported for reasons of security or because they would interfere with the correct operation of other shell features.
Parameters may also be special, that is, they have a predetermined meaning to the shell. Special parameters cannot have their type changed or their readonly attribute turned off, and if a special parameter is unset, then later recreated, the special properties will be retained.
To declare the type of a parameter, or to assign a string or numeric value to a scalar parameter, use the typeset builtin.
The value of a scalar parameter may also be assigned by writing:
name=value
In scalar assignment, value is expanded as a single string, in which the elements of arrays are joined together; filename expansion is not performed unless the option GLOB_ASSIGN is set.
When the integer attribute, -i, or a floating point attribute, -E or -F, is set for name, the value is subject to arithmetic evaluation. Furthermore, by replacing `=' with `+=', a parameter can be incremented or appended to. See the section `Array Parameters' and Arithmetic Evaluation (in zshmisc(1)) for additional forms of assignment.
Note that assignment may implicitly change the attributes of a parameter. For example, assigning a number to a variable in arithmetic evaluation may change its type to integer or float, and with GLOB_ASSIGN assigning a pattern to a variable may change its type to an array.
To reference the value of a parameter, write `$name' or `${name}'. See Parameter Expansion in zshexpn(1) for complete details. That section also explains the effect of the difference between scalar and array assignment on parameter expansion.
set -A name value ...
name=(value ...)
name=([key]=value ...)
If no parameter name exists, an ordinary array parameter is created. If the parameter name exists and is a scalar, it is replaced by a new array.
In the third form, key is an expression that will be evaluated in arithmetic context (in its simplest form, an integer) that gives the index of the element to be assigned with value. In this form any elements not explicitly mentioned that come before the largest index to which a value is assigned are assigned an empty string. The indices may be in any order. Note that this syntax is strict: [ and ]= must not be quoted, and key may not consist of the unquoted string ]=, but is otherwise treated as a simple string. The enhanced forms of subscript expression that may be used when directly subscripting a variable name, described in the section Array Subscripts below, are not available.
The syntaxes with and without the explicit key may be mixed. An implicit key is deduced by incrementing the index from the previously assigned element. Note that it is not treated as an error if latter assignments in this form overwrite earlier assignments.
For example, assuming the option KSH_ARRAYS is not set, the following:
array=(one [3]=three four)
causes the array variable array to contain four elements one, an empty string, three and four, in that order.
In the forms where only value is specified, full command line expansion is performed.
In the [key]=value form, both key and value undergo all forms of expansion allowed for single word shell expansions (this does not include filename generation); these are as performed by the parameter expansion flag (e) as described in zshexpn(1). Nested parentheses may surround value and are included as part of the value, which is joined into a plain string; this differs from ksh which allows the values themselves to be arrays. A future version of zsh may support that. To cause the brackets to be interpreted as a character class for filename generation, and therefore to treat the resulting list of files as a set of values, quote the equal sign using any form of quoting. Example:
name=([a-z]'='*)
To append to an array without changing the existing values, use one of the following:
name+=(value ...)
name+=([key]=value ...)
In the second form key may specify an existing index as well as an index off the end of the old array; any existing value is overwritten by value. Also, it is possible to use [key]+=value to append to the existing value at that index.
Within the parentheses on the right hand side of either form of the assignment, newlines and semicolons are treated the same as white space, separating individual values. Any consecutive sequence of such characters has the same effect.
Ordinary array parameters may also be explicitly declared with:
typeset -a name
Associative arrays must be declared before assignment, by using:
typeset -A name
When name refers to an associative array, the list in an assignment is interpreted as alternating keys and values:
set -A name key value ...
name=(key value ...)
name=([key]=value ...)
Note that only one of the two syntaxes above may be used in any given assignment; the forms may not be mixed. This is unlike the case of numerically indexed arrays.
Every key must have a value in this case. Note that this assigns to the entire array, deleting any elements that do not appear in the list. The append syntax may also be used with an associative array:
name+=(key value ...)
name+=([key]=value ...)
This adds a new key/value pair if the key is not already present, and replaces the value for the existing key if it is. In the second form it is also possible to use [key]+=value to append to the existing value at that key. Expansion is performed identically to the corresponding forms for normal arrays, as described above.
To create an empty array (including associative arrays), use one of:
set -A name
name=()
Subscripts may be used inside braces used to delimit a parameter name, thus `${foo[2]}' is equivalent to `$foo[2]'. If the KSH_ARRAYS option is set, the braced form is the only one that works, as bracketed expressions otherwise are not treated as subscripts.
If the KSH_ARRAYS option is not set, then by default accesses to an array element with a subscript that evaluates to zero return an empty string, while an attempt to write such an element is treated as an error. For backward compatibility the KSH_ZERO_SUBSCRIPT option can be set to cause subscript values 0 and 1 to be equivalent; see the description of the option in zshoptions(1).
The same subscripting syntax is used for associative arrays, except that no arithmetic expansion is applied to exp. However, the parsing rules for arithmetic expressions still apply, which affects the way that certain special characters must be protected from interpretation. See Subscript Parsing below for details.
A subscript of the form `[*]' or `[@]' evaluates to all elements of an array; there is no difference between the two except when they appear within double quotes. `"$foo[*]"' evaluates to `"$foo[1] $foo[2] ..."', whereas `"$foo[@]"' evaluates to `"$foo[1]" "$foo[2]" ...'. For associative arrays, `[*]' or `[@]' evaluate to all the values, in no particular order. Note that this does not substitute the keys; see the documentation for the `k' flag under Parameter Expansion Flags in zshexpn(1) for complete details. When an array parameter is referenced as `$name' (with no subscript) it evaluates to `$name[*]', unless the KSH_ARRAYS option is set in which case it evaluates to `${name[0]}' (for an associative array, this means the value of the key `0', which may not exist even if there are values for other keys).
A subscript of the form `[exp1,exp2]' selects all elements in the range exp1 to exp2, inclusive. (Associative arrays are unordered, and so do not support ranges.) If one of the subscripts evaluates to a negative number, say -n, then the nth element from the end of the array is used. Thus `$foo[-3]' is the third element from the end of the array foo, and `$foo[1,-1]' is the same as `$foo[*]'.
Subscripting may also be performed on non-array values, in which case the subscripts specify a substring to be extracted. For example, if FOO is set to `foobar', then `echo $FOO[2,5]' prints `ooba'. Note that some forms of subscripting described below perform pattern matching, and in that case the substring extends from the start of the match of the first subscript to the end of the match of the second subscript. For example,
string="abcdefghijklm"
print ${string[(r)d?,(r)h?]}
prints `defghi'. This is an obvious generalisation of the rule for single-character matches. For a single subscript, only a single character is referenced (not the range of characters covered by the match).
Note that in substring operations the second subscript is handled differently by the r and R subscript flags: the former takes the shortest match as the length and the latter the longest match. Hence in the former case a * at the end is redundant while in the latter case it matches the whole remainder of the string. This does not affect the result of the single subscript case as here the length of the match is irrelevant.
A subscript may be used on the left side of an assignment like so:
name[exp]=value
In this form of assignment the element or range specified by exp is replaced by the expression on the right side. An array (but not an associative array) may be created by assignment to a range or element. Arrays do not nest, so assigning a parenthesized list of values to an element or range changes the number of elements in the array, shifting the other elements to accommodate the new values. (This is not supported for associative arrays.)
This syntax also works as an argument to the typeset command:
typeset "name[exp]"=value
The value may not be a parenthesized list in this case; only single-element assignments may be made with typeset. Note that quotes are necessary in this case to prevent the brackets from being interpreted as filename generation operators. The noglob precommand modifier could be used instead.
To delete an element of an ordinary array, assign `()' to that element. To delete an element of an associative array, use the unset command:
unset "name[exp]"
The flags s, n and b take an argument; the delimiter is shown below as `:', but any character, or the matching pairs `(...)', `{...}', `[...]', or `<...>', may be used, but note that `<...>' can only be used if the subscript is inside a double quoted expression or a parameter substitution enclosed in braces as otherwise the expression is interpreted as a redirection.
The flags currently understood are:
If a search through an ordinary array failed, the search sets the subscript to one past the end of the array, and hence ${array[(r)pattern]} will substitute the empty string. Thus the success of a search can be tested by using the (i) flag, for example (assuming the option KSH_ARRAYS is not in effect):
[[ ${array[(i)pattern]} -le ${#array} ]]
If KSH_ARRAYS is in effect, the -le should be replaced by -lt.
Note that in subscripts with both `r' and `R' pattern characters are active even if they were substituted for a parameter (regardless of the setting of GLOB_SUBST which controls this feature in normal pattern matching). The flag `e' can be added to inhibit pattern matching. As this flag does not inhibit other forms of substitution, care is still required; using a parameter to hold the key has the desired effect:
key2='original key'
print ${array[(Re)$key2]}
This flag can also be used to force * or @ to be interpreted as a single key rather than as a reference to all values. It may be used for either purpose on the left side of an assignment.
See Parameter Expansion Flags (zshexpn(1)) for additional ways to manipulate the results of array subscripting.
This discussion applies mainly to associative array key strings and to patterns used for reverse subscripting (the `r', `R', `i', etc. flags), but it may also affect parameter substitutions that appear as part of an arithmetic expression in an ordinary subscript.
To avoid subscript parsing limitations in assignments to associative array elements, use the append syntax:
aa+=('key with "*strange*" characters' 'value string')
The basic rule to remember when writing a subscript expression is that all text between the opening `[' and the closing `]' is interpreted as if it were in double quotes (see zshmisc(1)). However, unlike double quotes which normally cannot nest, subscript expressions may appear inside double-quoted strings or inside other subscript expressions (or both!), so the rules have two important differences.
The first difference is that brackets (`[' and `]') must appear as balanced pairs in a subscript expression unless they are preceded by a backslash (`\'). Therefore, within a subscript expression (and unlike true double-quoting) the sequence `\[' becomes `[', and similarly `\]' becomes `]'. This applies even in cases where a backslash is not normally required; for example, the pattern `[^[]' (to match any character other than an open bracket) should be written `[^\[]' in a reverse-subscript pattern. However, note that `\[^\[\]' and even `\[^[]' mean the same thing, because backslashes are always stripped when they appear before brackets!
The same rule applies to parentheses (`(' and `)') and braces (`{' and `}'): they must appear either in balanced pairs or preceded by a backslash, and backslashes that protect parentheses or braces are removed during parsing. This is because parameter expansions may be surrounded by balanced braces, and subscript flags are introduced by balanced parentheses.
The second difference is that a double-quote (`"') may appear as part of a subscript expression without being preceded by a backslash, and therefore that the two characters `\"' remain as two characters in the subscript (in true double-quoting, `\"' becomes `"'). However, because of the standard shell quoting rules, any double-quotes that appear must occur in balanced pairs unless preceded by a backslash. This makes it more difficult to write a subscript expression that contains an odd number of double-quote characters, but the reason for this difference is so that when a subscript expression appears inside true double-quotes, one can still write `\"' (rather than `\\\"') for `"'.
To use an odd number of double quotes as a key in an assignment, use the typeset builtin and an enclosing pair of double quotes; to refer to the value of that key, again use double quotes:
typeset -A aa typeset "aa[one\"two\"three\"quotes]"=QQQ print "$aa[one\"two\"three\"quotes]"
It is important to note that the quoting rules do not change when a parameter expansion with a subscript is nested inside another subscript expression. That is, it is not necessary to use additional backslashes within the inner subscript expression; they are removed only once, from the innermost subscript outwards. Parameters are also expanded from the innermost subscript first, as each expansion is encountered left to right in the outer expression.
A further complication arises from a way in which subscript parsing is not different from double quote parsing. As in true double-quoting, the sequences `\*', and `\@' remain as two characters when they appear in a subscript expression. To use a literal `*' or `@' as an associative array key, the `e' flag must be used:
typeset -A aa aa[(e)*]=star print $aa[(e)*]
A last detail must be considered when reverse subscripting is performed. Parameters appearing in the subscript expression are first expanded and then the complete expression is interpreted as a pattern. This has two effects: first, parameters behave as if GLOB_SUBST were on (and it cannot be turned off); second, backslashes are interpreted twice, once when parsing the array subscript and again when parsing the pattern. In a reverse subscript, it's necessary to use four backslashes to cause a single backslash to match literally in the pattern. For complex patterns, it is often easiest to assign the desired pattern to a parameter and then refer to that parameter in the subscript, because then the backslashes, brackets, parentheses, etc., are seen only when the complete expression is converted to a pattern. To match the value of a parameter literally in a reverse subscript, rather than as a pattern, use `${(q)name}' (see zshexpn(1)) to quote the expanded value.
Note that the `k' and `K' flags are reverse subscripting for an ordinary array, but are not reverse subscripting for an associative array! (For an associative array, the keys in the array itself are interpreted as patterns by those flags; the subscript is a plain string in that case.)
One final note, not directly related to subscripting: the numeric names of positional parameters (described below) are parsed specially, so for example `$2foo' is equivalent to `${2}foo'. Therefore, to use subscript syntax to extract a substring from a positional parameter, the expansion must be surrounded by braces; for example, `${2[3,5]}' evaluates to the third through fifth characters of the second positional parameter, but `$2[3,5]' is the entire second parameter concatenated with the filename generation pattern `[3,5]'.
The parameters *, @ and argv are arrays containing all the positional parameters; thus `$argv[n]', etc., is equivalent to simply `$n'. Note that the options KSH_ARRAYS or KSH_ZERO_SUBSCRIPT apply to these arrays as well, so with either of those options set, `${argv[0]}' is equivalent to `$1' and so on.
Positional parameters may be changed after the shell or function starts by using the set builtin, by assigning to the argv array, or by direct assignment of the form `n=value' where n is the number of the positional parameter to be changed. This also creates (with empty values) any of the positions from 1 to n that do not already have values. Note that, because the positional parameters form an array, an array assignment of the form `n=(value ...)' is allowed, and has the effect of shifting all the values at positions greater than n by as many positions as necessary to accommodate the new values.
When a parameter is read or assigned to, the innermost existing parameter of that name is used. (That is, the local parameter hides any less-local parameter.) However, assigning to a non-existent parameter, or declaring a new parameter with export, causes it to be created in the outermost scope.
Local parameters disappear when their scope ends. unset can be used to delete a parameter while it is still in scope; any outer parameter of the same name remains hidden.
Special parameters may also be made local; they retain their special attributes unless either the existing or the newly-created parameter has the -h (hide) attribute. This may have unexpected effects: there is no default value, so if there is no assignment at the point the variable is made local, it will be set to an empty value (or zero in the case of integers). The following:
typeset PATH=/new/directory:$PATH
is valid for temporarily allowing the shell or programmes called from it to find the programs in /new/directory inside a function.
Note that the restriction in older versions of zsh that local parameters were never exported has been removed.
The following parameters are automatically set by the shell:
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
The values of RANDOM form an intentionally-repeatable pseudo-random sequence; subshells that reference RANDOM will result in identical pseudo-random values unless the value of RANDOM is referenced or seeded in the parent shell in between subshell invocations.
Unlike other special parameters, the type of the SECONDS parameter can be changed using the typeset command. Only integer and one of the floating point types are allowed. For example, `typeset -F SECONDS' causes the value to be reported as a floating point number. The value is available to microsecond accuracy, although the shell may show more or fewer digits depending on the use of typeset. See the documentation for the builtin typeset in zshbuiltins(1) for more details.
Note that it is possible that an interrupt arrives during the execution of the always block; this interrupt is also propagated.
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
Note that the variable does not give information on syntactic context such as pipelines or subshells. Use $ZSH_SUBSHELL to detect subshells.
The context is one of the following:
In cases where there are two parameters with an upper- and lowercase form of the same name, such as path and PATH, the lowercase form is an array and the uppercase form is a scalar with the elements of the array joined together by colons. These are similar to tied parameters created via `typeset -T'. The normal use for the colon-separated form is for exporting to the environment, while the array form is easier to manipulate within the shell. Note that unsetting either of the pair will unset the other; they retain their special properties when recreated, and recreating one of the pair will recreate the other.
This parameter may be profitably set in some circumstances, e.g. for slow modems dialing into a communications server, or on a slow wide area network. It should be set to the baud rate of the slowest part of the link for best performance.
The characters must be in the ASCII character set; any attempt to set histchars to characters with a locale-dependent meaning will be rejected with an error message.
Note that HISTORY_IGNORE defines a single pattern: to specify alternatives use the `(first|second|...)' syntax.
Compare the HIST_NO_STORE option or the zshaddhistory hook, either of which would prevent such commands from being added to the interactive history at all. If you wish to use HISTORY_IGNORE to stop history being added in the first place, you can define the following hook:
zshaddhistory() {
emulate -L zsh
## uncomment if HISTORY_IGNORE
## should use EXTENDED_GLOB syntax
# setopt extendedglob
[[ $1 != ${~HISTORY_IGNORE} ]]
}
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
If the parameter is unset, the default is used. Note this has a different effect from setting the parameter to an empty string.
For backward compatibility, if the SUNKEYBOARDHACK option is explicitly set, the value of KEYBOARD_HACK reverts to backquote. If the option is explicitly unset, this variable is set to empty.
These parameters only exist if the installation supports dynamic module loading.
If this is made local, it is not implicitly set to 0, but may be explicitly set locally.
The actions available at the prompt are [nyae]:
A star may be inserted between the percent sign and flags printing time (e.g., `%*E'); this causes the time to be printed in `hh:mm:ss.ttt' format (hours and minutes are only printed if they are not zero). Alternatively, `m' or `u' may be used (e.g., `%mE') to produce time output in milliseconds or microseconds, respectively.
If it contains the single word `all', then all login/logout events are reported. If it contains the single word `notme', then all events are reported as with `all' except $USERNAME.
An entry in this list may consist of a username, an `@' followed by a remote hostname, and a `%' followed by a line (tty). Any of these may be a pattern (be sure to quote this during the assignment to watch so that it does not immediately perform file generation); the setting of the EXTENDED_GLOB option is respected. Any or all of these components may be present in an entry; if a login/logout event matches all of them, it is reported.
For example, with the EXTENDED_GLOB option set, the following:
watch=('^(pws|barts)')
causes reports for activity associated with any user other than pws or barts.
NOTE: The `%m' and `%M' escapes will work only if there is a host name field in the utmp on your machine. Otherwise they are treated as ordinary strings.
The test character x may be any one of `l', `n', `m' or `M', which indicate a `true' result if the corresponding escape sequence would return a non-empty value; or it may be `a', which indicates a `true' result if the watched user has logged in, or `false' if he has logged out. Other characters evaluate to neither true nor false; the entire expression is omitted in this case.
If the result is `true', then the true-text is formatted according to the rules above and printed, and the false-text is skipped. If `false', the true-text is skipped and the false-text is formatted and printed. Either or both of the branches may be empty, but both separators must be present in any case.
This two-element array contains the terminal escape sequences for enabling and disabling the feature. These escape sequences are used to enable bracketed paste when ZLE is active and disable it at other times. Unsetting the parameter has the effect of ensuring that bracketed paste remains disabled.
These variables can contain the sets of characters that will cause the suffix to be removed. If ZLE_REMOVE_SUFFIX_CHARS is set, those characters will cause the suffix to be removed; if ZLE_SPACE_SUFFIX_CHARS is set, those characters will cause the suffix to be removed and replaced by a space.
If ZLE_REMOVE_SUFFIX_CHARS is not set, the default behaviour is equivalent to:
ZLE_REMOVE_SUFFIX_CHARS=$' \t\n;&|'
If ZLE_REMOVE_SUFFIX_CHARS is set but is empty, no characters have this behaviour. ZLE_SPACE_SUFFIX_CHARS takes precedence, so that the following:
ZLE_SPACE_SUFFIX_CHARS=$'&|'
causes the characters `&' and `|' to remove the suffix but to replace it with a space.
To illustrate the difference, suppose that the option AUTO_REMOVE_SLASH is in effect and the directory DIR has just been completed, with an appended /, following which the user types `&'. The default result is `DIR&'. With ZLE_REMOVE_SUFFIX_CHARS set but without including `&' the result is `DIR/&'. With ZLE_SPACE_SUFFIX_CHARS set to include `&' the result is `DIR &'.
Note that certain completions may provide their own suffix removal or replacement behaviour which overrides the values described here. See the completion system documentation in zshcompsys(1).
Typically this will be used to set the value to 0 so that the prompt appears flush with the right hand side of the screen. This is not the default as many terminals do not handle this correctly, in particular when the prompt appears at the extreme bottom right of the screen. Recent virtual terminals are more likely to handle this case correctly. Some experimentation is necessary.
The sense of an option name may be inverted by preceding it with `no', so `setopt No_Beep' is equivalent to `unsetopt beep'. This inversion can only be done once, so `nonobeep' is not a synonym for `beep'. Similarly, `tify' is not a synonym for `nonotify' (the inversion of `notify').
Some options also have one or more single letter names. There are two sets of single letter options: one used by default, and another used to emulate sh/ksh (used when the SH_OPTION_LETTERS option is set). The single letter options can be used on the shell command line, or with the set, setopt and unsetopt builtins, as normal Unix options preceded by `-'.
The sense of the single letter options may be inverted by using `+' instead of `-'. Some of the single letter option names refer to an option being off, in which case the inversion of that name refers to the option being on. For example, `+n' is the short name of `exec', and `-n' is the short name of its inversion, `noexec'.
In strings of single letter options supplied to the shell at startup, trailing whitespace will be ignored; for example the string `-f ' will be treated just as `-f', but the string `-f i' is an error. This is because many systems which implement the `#!' mechanism for calling scripts do not strip trailing whitespace.
For example, suppose /foo/bar is a link to the directory /alt/rod. Without this option set, `cd /foo/bar/..' changes to /foo; with it set, it changes to /alt. The same applies if the current directory is /foo/bar and `cd ..' is used. Note that all other symbolic links in the path will also be resolved.
Also, if the option is set, the conditions under which the shell prints the new directory after changing to it are modified. It is no longer restricted to interactive shells (although printing of the directory stack with pushd is still limited to interactive shells); and any use of a component of CDPATH, including a `.' but excluding an empty component that is otherwise treated as `.', causes the directory to be printed.
Note that when the pattern matcher is used, matching control (for example, case-insensitive or anchored matching) cannot be used. This limitation only applies when the current word contains a pattern; simply turning on the GLOB_COMPLETE option does not have this effect.
For example, with both options unset a function may be defined in the following fashion:
args() { echo $# }
while if either option is set, this does not work and something equivalent to the following is required:
args() { echo $#; }
This option respects the setting of the KSH_TYPESET option. In other words, if both options are in effect, arguments looking like assignments will not undergo word splitting.
The option is on by default if the shell was compiled with MULTIBYTE_SUPPORT; otherwise it is off by default and has no effect if turned on.
If the option is off a single byte is always treated as a single character. This setting is designed purely for examining strings known to contain raw bytes or other values that may not be characters in the current locale. It is not necessary to unset the option merely because the character set for the current locale does not contain multibyte characters.
The option does not affect the shell's editor, which always uses the locale to determine multibyte characters. This is because the character set displayed by the terminal emulator is independent of shell settings.
For example, the following code produces a warning for the assignment inside the function nested as that overrides the value within toplevel
toplevel() {
local foo="in fn"
nested
}
nested() {
foo="in nested"
}
setopt warn_nested_var
toplevel
`: <beginning time>:<elapsed seconds>;<command>'.
When writing out a copy of the history file, zsh preserves the old file's permissions and group information, but will refuse to write out a new file if it would change the history file's owner.
This option is only useful if INC_APPEND_HISTORY and SHARE_HISTORY are turned off. The three options should be considered mutually exclusive.
This option both imports new commands from the history file, and also causes your typed commands to be appended to the history file (the latter is like specifying INC_APPEND_HISTORY, which should be turned off if this option is in effect). The history lines are also output with timestamps ala EXTENDED_HISTORY (which makes it easier to find the spot where we left off reading the file after it gets re-written).
By default, history movement commands visit the imported lines as well as the local lines, but you can toggle this on and off with the set-local-history zle binding. It is also possible to create a zle widget that will make some commands ignore imported commands, and some include them.
If you find that you want more control over when commands get imported, you may wish to turn SHARE_HISTORY off, INC_APPEND_HISTORY or INC_APPEND_HISTORY_TIME (see above) on, and then manually import commands whenever you need them using `fc -RI'.
This option is set by default for backward compatibility; it is not recommended that its behaviour be relied upon. Note that the builtin export always sets both the -x and -g flags, and hence its effect extends beyond the scope of the enclosing function; this is the most portable way to achieve this behaviour.
If the option is not set, and the option APPEND_CREATE is also not set, `>>!' or `>>|' must be used to create a file. If either option is set, `>>' may be used.
The shell variable CORRECT_IGNORE may be set to a pattern to match words that will never be offered as corrections.
The shell variable CORRECT_IGNORE_FILE may be set to a pattern to match file names that will never be offered as corrections.
Also, if this option is set and the Zsh Line Editor is used, widgets implemented by shell functions can be bound to EOF (normally Control-D) without printing the normal warning message. This works only for normal widgets, not for completion widgets.
Note that subdirectories of the current directory are always searched for executables specified in this form. This takes place before any search indicated by this option, and regardless of whether `.' or the current directory appear in the command search path.
The check is omitted if the commands run from the previous command line included a `jobs' command, since it is assumed the user is aware that there are background or suspended jobs. A `jobs' command run from one of the hook functions defined in the section SPECIAL FUNCTIONS in zshmisc(1) is not counted for this purpose.
This option has no effect unless CHECK_JOBS is set.
When the option is not set, the MONITOR option is unset on entry to subshells, so that job control is no longer active. When the option is set, the MONITOR option and job control remain active in the subshell, but note that the subshell has no access to jobs in the parent shell.
When the option is not set, jobs put in the background or foreground with bg or fg are displayed with the same information that would be reported by jobs. When the option is set, only the text is printed. The output from jobs itself is not affected by the option.
When the option is not set, job information from the parent shell is saved for output within a subshell (for example, within a pipeline). When the option is set, the output of jobs is empty until a job is started within the subshell.
In previous versions of the shell, it was necessary to enable POSIX_JOBS in order for the builtin command wait to return the status of background jobs that had already exited. This is no longer the case.
When a partial line is preserved, by default you will see an inverse+bold character at the end of the partial line: a `%' for a normal user or a `#' for root. If set, the shell parameter PROMPT_EOL_MARK can be used to customize how the end of partial lines are shown.
NOTE: if the PROMPT_CR option is not set, enabling this option will have no effect. This option is on by default.
When this option is set, aliases can be used for defining functions.
For example, consider the following definitions as they might occur in a startup file.
alias foo=bar
foo() {
print This probably does not do what you expect.
}
Here, foo is expanded as an alias to bar before the () is encountered, so the function defined would be named bar. By default this is instead an error in native mode. Note that quoting any part of the function name, or using the keyword function, avoids the problem, so is recommended when the function name can also be an alias.
The behaviour is also disabled inside DEBUG traps. In this case the option is handled specially: it is unset on entry to the trap. If the option DEBUG_BEFORE_CMD is set, as it is by default, and the option ERR_EXIT is found to have been set on exit, then the command for which the DEBUG trap is being executed is skipped. The option is restored after the trap exits.
Non-zero status in a command list containing && or || is ignored for commands not at the end of the list. Hence
false && true
does not trigger exit.
Exiting due to ERR_EXIT has certain interactions with asynchronous jobs noted in the section JOBS in zshmisc(1).
Normally this option inherits the behaviour of ERR_EXIT that code followed by `&&' `||' does not trigger a return. Hence in the following:
summit || true
no return is forced as the combined effect always has a zero return status.
Note. however, that if summit in the above example is itself a function, code inside it is considered separately: it may force a return from summit (assuming the option remains set within summit), but not from the enclosing context. This behaviour is different from ERR_EXIT which is unaffected by function scope.
unsetopt localtraps
trap - INT
fn() { setopt localtraps; trap '' INT; sleep 3; }
will restore normal handling of SIGINT after the function exits.
Sequences of digits indicating a numeric base such as the `08' component in `08#77' are always interpreted as decimal, regardless of leading zeroes.
If this option is not set, the shell will report an error when a append redirection (>>) is used on a file that does not already exists (the traditional zsh behaviour of NO_CLOBBER). If the option is set, no error is reported (POSIX behaviour).
Alters the way arguments to the typeset family of commands, including declare, export, float, integer, local and readonly, are processed. Without this option, zsh will perform normal word splitting after command and parameter expansion in arguments of an assignment; with it, word splitting does not take place in those cases.
If neither this option nor KSH_ARRAYS is set, accesses to an element of an array or string with subscript zero return an empty element or string, while attempts to set element zero of an array or string are treated as an error. However, attempts to set an otherwise valid subscript range that includes zero will succeed. For example, if KSH_ZERO_SUBSCRIPT is not set,
array[0]=(element)
is an error, while
array[0,1]=(element)
is not and will replace the first element of the array.
This option is for compatibility with older versions of the shell and is not recommended in new code.
Alias expansion takes place while text is being read; hence when this option is set it does not take effect until the end of any function or other piece of shell code parsed as one unit. Note this may cause differences from other shells even when the option is in effect. For example, when running a command with `zsh -c', or even `zsh -o posixaliases -c', the entire command argument is parsed as one unit, so aliases defined within the argument are not available even in later lines. If in doubt, avoid use of aliases in non-interactive code.
emulate sh -o POSIX_ARGZERO
Note that NO_POSIX_ARGZERO has no effect unless FUNCTION_ARGZERO was already enabled upon entry to the function or script.
In addition, various error conditions associated with the above builtins or exec cause a non-interactive shell to exit and an interactive shell to return to its top-level processing.
Furthermore, functions and shell builtins are not executed after an exec prefix; the command to be executed must be an external command found in the path.
Furthermore, the getopts builtin behaves in a POSIX-compatible fashion in that the associated variable OPTIND is not made local to functions.
Moreover, the warning and special exit code from [[ -o non_existent_option ]] are suppressed.
In addition, setting this option limits the effect of parameter substitution with no braces, so that the expression $# is treated as the parameter $# even if followed by a valid parameter name. When it is unset, zsh allows expressions of the form $#name to refer to the length of $name, even for special variables, for example in expressions such as $#- and $#*.
Another difference is that with the option set assignment to an unset variable in arithmetic context causes the variable to be created as a scalar rather than a numeric type. So after `unset t; (( t = 3 ))'. without POSIX_IDENTIFIERS set t has integer type, while with it set it has scalar type.
When the option is unset and multibyte character support is enabled (i.e. it is compiled in and the option MULTIBYTE is set), then additionally any alphanumeric characters in the local character set may be used in identifiers. Note that scripts and functions written with this feature are not portable, and also that both options must be set before the script or function is parsed; setting them during execution is not sufficient as the syntax variable=value has already been parsed as a command rather than an assignment.
If multibyte character support is not compiled into the shell this option is ignored; all octets with the top bit set may be used in identifiers. This is non-standard but is the traditional zsh behaviour.
When this option is not set, null characters embedded within strings of the form $'...' are treated as ordinary characters. The entire string is maintained within the shell and output to files where necessary, although owing to restrictions of the library interface the string is truncated at the null character in file names, environment variables, or in arguments to external programs.
When this option is set, the $'...' expression is truncated at the null character. Note that remaining parts of the same string beyond the termination of the quotes are not truncated.
For example, the command line argument a$'b\0c'd is treated with the option off as the characters a, b, null, c, d, and with the option on as the characters a, b, d.
#!/bin/zsh -p
The option is enabled automatically on startup if the effective user (group) ID is not equal to the real user (group) ID. In this case, turning the option off causes the effective user and group IDs to be set to the real user and group IDs. Be aware that if that fails the shell may be running with different IDs than was intended so a script should check for failure and act accordingly, for example:
unsetopt privileged || exit
The PRIVILEGED option disables sourcing user startup files. If zsh is invoked as `sh' or `ksh' with this option set, /etc/suid_profile is sourced (after /etc/profile on interactive shells). Sourcing ~/.profile is disabled and the contents of the ENV variable is ignored. This option cannot be changed using the -m option of setopt and unsetopt, and changing it inside a function always changes it globally regardless of the LOCAL_OPTIONS option.
If this option is set, the pattern test [[:WORD:]] matches a zero-width punctuation character on the assumption that it will be used as part of a word in combination with a word character. Otherwise the base shell does not handle combining characters specially.
Note that although this is on by default in ksh emulation it only provides superficial compatibility with the ksh line editor and reduces the effectiveness of the zsh line editor. As it has no effect on shell syntax, many users may wish to disable this option when using ksh emulation interactively.
Some shell builtin commands take options as described in individual entries; these are often referred to in the list below as `flags' to avoid confusion with shell options, which may also have an effect on the behaviour of builtin commands. In this introductory section, `option' always has the meaning of an option to a command that should be familiar to most command line users.
Typically, options are single letters preceded by a hyphen (-). Options that take an argument accept it either immediately following the option letter or after white space, for example `print -C3 {1..9}' or `print -C 3 {1..9}' are equivalent. Arguments to options are not the same as arguments to the command; the documentation indicates which is which. Options that do not take an argument may be combined in a single word, for example `print -rca -- *' and `print -r -c -a -- *' are equivalent.
Some shell builtin commands also take options that begin with `+' instead of `-'. The list below makes clear which commands these are.
Options (together with their individual arguments, if any) must appear in a group before any non-option arguments; once the first non-option argument has been found, option processing is terminated.
All builtin commands other than `echo' and precommand modifiers, even those that have no options, can be given the argument `--' to terminate option processing. This indicates that the following words are non-option arguments, but is otherwise ignored. This is useful in cases where arguments to the command may begin with `-'. For historical reasons, most builtin commands (including `echo') also recognize a single `-' in a separate word for this purpose; note that this is less standard and use of `--' is recommended.
If file does not contain a slash, or if PATH_DIRS is set, the shell looks in the components of $path to find the directory containing file. Files in the current directory are not read unless `.' appears somewhere in $path. If a file named `file.zwc' is found, is newer than file, and is the compiled form (created with the zcompile builtin) of file, then commands are read from that file instead of file.
If any arguments arg are given, they become the positional parameters; the old positional parameters are restored when the file is done executing. However, if no arguments are given, the positional parameters remain those of the calling context, and no restoring is done.
If file was not found the return status is 127; if file was found but contained a syntax error the return status is 126; else the return status is the exit status of the last command executed.
If the -s flag is present, define a suffix alias: if the command word on a command line is in the form `text.name', where text is any non-empty string, it is replaced by the text `value text.name'. Note that name is treated as a literal string, not a pattern. A trailing space in value is not special in this case. For example,
alias -s ps='gv --'
will cause the command `*.ps' to be expanded to `gv -- *.ps'. As alias expansion is carried out earlier than globbing, the `*.ps' will then be expanded. Suffix aliases constitute a different name space from other aliases (so in the above example it is still possible to create an alias for the command ps) and the two sets are never listed together.
For each name with no value, print the value of name, if any. With no arguments, print all currently defined aliases other than suffix aliases. If the -m flag is given the arguments are taken as patterns (they should be quoted to preserve them from being interpreted as glob patterns), and the aliases matching these patterns are printed. When printing aliases and one of the -g, -r or -s flags is present, restrict the printing to global, regular or suffix aliases, respectively; a regular alias is one which is neither a global nor a suffix alias. Using `+' instead of `-', or ending the option list with a single `+', prevents the values of the aliases from being printed.
If the -L flag is present, then print each alias in a manner suitable for putting in a startup script. The exit status is nonzero if a name (with no value) is given for which no alias has been defined.
For more on aliases, include common problems, see the section ALIASING in zshmisc(1).
If name consists of an absolute path, the function is defined to load from the file given (searching as usual for dump files in the given location). The name of the function is the basename (non-directory part) of the file. It is normally an error if the function is not found in the given location; however, if the option -d is given, searching for the function defaults to $fpath. If a function is loaded by absolute path, any functions loaded from it that are marked for autoload without an absolute path have the load path of the parent function temporarily prepended to $fpath.
If the option -r or -R is given, the function is searched for immediately and the location is recorded internally for use when the function is executed; a relative path is expanded using the value of $PWD. This protects against a change to $fpath after the call to autoload. With -r, if the function is not found, it is silently left unresolved until execution; with -R, an error message is printed and command processing aborted immediately the search fails, i.e. at the autoload command rather than at function execution..
The flag -X may be used only inside a shell function. It causes the calling function to be marked for autoloading and then immediately loaded and executed, with the current array of positional parameters as arguments. This replaces the previous definition of the function. If no function definition is found, an error is printed and the function remains undefined and marked for autoloading. If an argument is given, it is used as a directory (i.e. it does not include the name of the function) in which the function is to be found; this may be combined with the -d option to allow the function search to default to $fpath if it is not in the given location.
The flag +X attempts to load each name as an autoloaded function, but does not execute it. The exit status is zero (success) if the function was not previously defined and a definition for it was found. This does not replace any existing definition of the function. The exit status is nonzero (failure) if the function was already defined or when no definition was found. In the latter case the function remains undefined and marked for autoloading. If ksh-style autoloading is enabled, the function created will contain the contents of the file plus a call to the function itself appended to it, thus giving normal ksh autoloading behaviour on the first call to the function. If the -m flag is also given each name is treated as a pattern and all functions already marked for autoload that match the pattern are loaded.
With the -t flag, turn on execution tracing; with -T, turn on execution tracing only for the current function, turning it off on entry to any called functions that do not also have tracing enabled.
With the -U flag, alias expansion is suppressed when the function is loaded.
With the -w flag, the names are taken as names of files compiled with the zcompile builtin, and all functions defined in them are marked for autoloading.
The flags -z and -k mark the function to be autoloaded using the zsh or ksh style, as if the option KSH_AUTOLOAD were unset or were set, respectively. The flags override the setting of the option at the time the function is loaded.
Note that the autoload command makes no attempt to ensure the shell options set during the loading or execution of the file have any particular value. For this, the emulate command can be used:
emulate zsh -c 'autoload -Uz func'
arranges that when func is loaded the shell is in native zsh emulation, and this emulation is also applied when func is run.
Some of the functions of autoload are also provided by functions -u or functions -U, but autoload is a more comprehensive interface.
Otherwise, if arg begins with a slash, attempt to change to the directory given by arg.
If arg does not begin with a slash, the behaviour depends on whether the current directory `.' occurs in the list of directories contained in the shell parameter cdpath. If it does not, first attempt to change to the directory arg under the current directory, and if that fails but cdpath is set and contains at least one element attempt to change to the directory arg under each component of cdpath in turn until successful. If `.' occurs in cdpath, then cdpath is searched strictly in order so that `.' is only tried at the appropriate point.
The order of testing cdpath is modified if the option POSIX_CD is set, as described in the documentation for the option.
If no directory is found, the option CDABLE_VARS is set, and a parameter named arg exists whose value begins with a slash, treat its value as the directory. In that case, the parameter is added to the named directory hash table.
The second form of cd substitutes the string new for the string old in the name of the current directory, and tries to change to this new directory.
The third form of cd extracts an entry from the directory stack, and changes to that directory. An argument of the form `+n' identifies a stack entry by counting from the left of the list shown by the dirs command, starting with zero. An argument of the form `-n' counts from the right. If the PUSHD_MINUS option is set, the meanings of `+' and `-' in this context are swapped. If the POSIX_CD option is set, this form of cd is not recognised and will be interpreted as the first form.
If the -q (quiet) option is specified, the hook function chpwd and the functions in the array chpwd_functions are not called. This is useful for calls to cd that do not change the environment seen by an interactive user.
If the -s option is specified, cd refuses to change the current directory if the given pathname contains symlinks. If the -P option is given or the CHASE_LINKS option is set, symbolic links are resolved to their true values. If the -L option is given symbolic links are retained in the directory (and not resolved) regardless of the state of the CHASE_LINKS option.
See also the section `Precommand Modifiers' in zshmisc(1).
With the option -p, name ... refer to elements of the shell's pattern syntax as described in the section `Filename Generation'. Certain elements can be disabled separately, as given below.
Note that patterns not allowed by the current settings for the options EXTENDED_GLOB, KSH_GLOB and SH_GLOB are never enabled, regardless of the setting here. For example, if EXTENDED_GLOB is not active, the pattern ^ is ineffective even if `disable -p "^"' has not been issued. The list below indicates any option settings that restrict the use of the pattern. It should be noted that setting SH_GLOB has a wider effect than merely disabling patterns as certain expressions, in particular those involving parentheses, are parsed differently.
The following patterns may be disabled; all the strings need quoting on the command line to prevent them from being interpreted immediately as patterns and the patterns are shown below in single quotes as a reminder.
If the jobs are currently stopped and the AUTO_CONTINUE option is not set, a warning is printed containing information about how to make them running after they have been disowned. If one of the latter two forms is used, the jobs will automatically be made running, independent of the setting of the AUTO_CONTINUE option.
The -E flag, or the BSD_ECHO option, can be used to disable these escape sequences. In the latter case, -e flag can be used to enable them.
Note that for standards compliance a double dash does not terminate option processing; instead, it is printed directly. However, a single dash does terminate option processing, so the first dash, possibly following options, is not printed, but everything following it is printed as an argument. The single dash behaviour is different from other shells. For a more portable way of printing text, see printf, and for a more controllable way of printing text within zsh, see print.
With single argument set up zsh options to emulate the specified shell as much as possible. csh will never be fully emulated. If the argument is not one of the shells listed above, zsh will be used as a default; more precisely, the tests performed on the argument are the same as those used to determine the emulation at startup based on the shell name, see the section COMPATIBILITY in zsh(1) . In addition to setting shell options, the command also restores the pristine state of pattern enables, as if all patterns had been enabled using enable -p.
If the emulate command occurs inside a function that has been marked for execution tracing with functions -t then the xtrace option will be turned on regardless of emulation mode or other options. Note that code executed inside the function by the ., source, or eval commands is not considered to be running directly from the function, hence does not provoke this behaviour.
If the -R switch is given, all settable options are reset to their default value corresponding to the specified emulation mode, except for certain options describing the interactive environment; otherwise, only those options likely to cause portability problems in scripts and functions are altered. If the -L switch is given, the options LOCAL_OPTIONS, LOCAL_PATTERNS and LOCAL_TRAPS will be set as well, causing the effects of the emulate command and any setopt, disable -p or enable -p, and trap commands to be local to the immediately surrounding shell function, if any; normally these options are turned off in all emulation modes except ksh. The -L switch is mutually exclusive with the use of -c in flags.
If there is a single argument and the -l switch is given, the options that would be set or unset (the latter indicated with the prefix `no') are listed. -l can be combined with -L or -R and the list will be modified in the appropriate way. Note the list does not depend on the current setting of options, i.e. it includes all options that may in principle change, not just those that would actually change.
The flags may be any of the invocation-time flags described in the section INVOCATION in zsh(1), except that `-o EMACS' and `-o VI' may not be used. Flags such as `+r'/`+o RESTRICTED' may be prohibited in some circumstances.
If -c arg appears in flags, arg is evaluated while the requested emulation is temporarily in effect. In this case the emulation mode and all options are restored to their previous values before emulate returns. The -R switch may precede the name of the shell to emulate; note this has a meaning distinct from including -R in flags.
Use of -c enables `sticky' emulation mode for functions defined within the evaluated expression: the emulation mode is associated thereafter with the function so that whenever the function is executed the emulation (respecting the -R switch, if present) and all options are set (and pattern disables cleared) before entry to the function, and the state is restored after exit. If the function is called when the sticky emulation is already in effect, either within an `emulate shell -c' expression or within another function with the same sticky emulation, entry and exit from the function do not cause options to be altered (except due to standard processing such as the LOCAL_OPTIONS option). This also applies to functions marked for autoload within the sticky emulation; the appropriate set of options will be applied at the point the function is loaded as well as when it is run.
For example:
emulate sh -c 'fni() { setopt cshnullglob; }
fno() { fni; }'
fno
The two functions fni and fno are defined with sticky sh emulation. fno is then executed, causing options associated with emulations to be set to their values in sh. fno then calls fni; because fni is also marked for sticky sh emulation, no option changes take place on entry to or exit from it. Hence the option cshnullglob, turned off by sh emulation, will be turned on within fni and remain on return to fno. On exit from fno, the emulation mode and all options will be restored to the state they were in before entry to the temporary emulation.
The documentation above is typically sufficient for the intended purpose of executing code designed for other shells in a suitable environment. More detailed rules follow.
enable -p reenables patterns disabled with disable -p. Note that it does not override globbing options; for example, `enable -p "~"' does not cause the pattern character ~ to be active unless the EXTENDED_GLOB option is also set. To enable all possible patterns (so that they may be individually disabled with disable -p), use `setopt EXTENDED_GLOB KSH_GLOB NO_SH_GLOB'.
With -c clear the environment; with -l prepend - to the argv[0] string of the command executed (to simulate a login shell); with -a argv0 set the argv[0] string of the command executed. See the section `Precommand Modifiers' in zshmisc(1).
If the option POSIX_BUILTINS is set, command is never interpreted as a shell builtin command or shell function. This means further precommand modifiers such as builtin and noglob are also not interpreted within the shell. Hence command is always found by searching the command path.
If command is omitted but any redirections are specified, then the redirections will take effect in the current shell.
See notes at the end of the section JOBS in zshmisc(1) for some possibly unexpected interactions of the exit command with jobs.
The first two forms of this command select a range of events from first to last from the history list. The arguments first and last may be specified as a number or as a string. A negative number is used as an offset to the current history event number. A string specifies the most recent event beginning with the given string. All substitutions old=new, if any, are then performed on the text of the events.
In addition to the number range,
If first is not specified, it will be set to -1 (the most recent event), or to -16 if the -l flag is given. If last is not specified, it will be set to first, or to -1 if the -l flag is given. However, if the current event has added entries to the history with `print -s' or `fc -R', then the default last for -l includes all new history entries since the current event began.
When the -l flag is given, the resulting events are listed on standard output. Otherwise the editor program specified by -e ename is invoked on a file containing these history events. If -e is not given, the value of the parameter FCEDIT is used; if that is not set the value of the parameter EDITOR is used; if that is not set a builtin default, usually `vi' is used. If ename is `-', no editor is invoked. When editing is complete, the edited command is executed.
The flag -r reverses the order of the events and the flag -n suppresses event numbers when listing.
Also when listing,
`fc -p' pushes the current history list onto a stack and switches to a new history list. If the -a option is also specified, this history list will be automatically popped when the current function scope is exited, which is a much better solution than creating a trap function to call `fc -P' manually. If no arguments are specified, the history list is left empty, $HISTFILE is unset, and $HISTSIZE & $SAVEHIST are set to their default values. If one argument is given, $HISTFILE is set to that filename, $HISTSIZE & $SAVEHIST are left unchanged, and the history file is read in (if it exists) to initialize the new list. If a second argument is specified, $HISTSIZE & $SAVEHIST are instead set to the single specified numeric value. Finally, if a third argument is specified, $SAVEHIST is set to a separate value from $HISTSIZE. You are free to change these environment values for the new history list however you desire in order to manipulate the new history list.
`fc -P' pops the history list back to an older list saved by `fc -p'. The current list is saved to its $HISTFILE before it is destroyed (assuming that $HISTFILE and $SAVEHIST are set appropriately, of course). The values of $HISTFILE, $HISTSIZE, and $SAVEHIST are restored to the values they had when `fc -p' was called. Note that this restoration can conflict with making these variables "local", so your best bet is to avoid local declarations for these variables in functions that use `fc -p'. The one other guaranteed-safe combination is declaring these variables to be local at the top of your function and using the automatic option (-a) with `fc -p'. Finally, note that it is legal to manually pop a push marked for automatic popping if you need to do so before the function exits.
`fc -R' reads the history from the given file, `fc -W' writes the history out to the given file, and `fc -A' appends the history out to the given file. If no filename is specified, the $HISTFILE is assumed. If the -I option is added to -R, only those events that are not already contained within the internal history list are added. If the -I option is added to -A or -W, only those events that are new since last incremental append/write to the history file are appended/written. In any case, the created file will have no more than $SAVEHIST entries.
The -x option indicates that any functions output will have each leading tab for indentation, added by the shell to show syntactic structure, expanded to the given number num of spaces. num can also be 0 to suppress all indentation.
The -W option turns on the option WARN_NESTED_VAR for the named function or functions only. The option is turned off at the start of nested functions (apart from anonoymous functions) unless the called function also has the -W attribute.
The -c option causes oldfn to be copied to newfn. The copy is efficiently handled internally by reference counting. If oldfn was marked for autoload it is first loaded and if this fails the copy fails. Either function may subsequently be redefined without affecting the other. A typical idiom is that oldfn is the name of a library shell function which is then redefined to call newfn, thereby installing a modified version of the function.
Use of the -M option may not be combined with any of the options handled by typeset -f.
functions -M mathfn defines mathfn as the name of a mathematical function recognised in all forms of arithmetical expressions; see the section `Arithmetic Evaluation' in zshmisc(1). By default mathfn may take any number of comma-separated arguments. If min is given, it must have exactly min args; if min and max are both given, it must have at least min and at most max args. max may be -1 to indicate that there is no upper limit.
By default the function is implemented by a shell function of the same name; if shellfn is specified it gives the name of the corresponding shell function while mathfn remains the name used in arithmetical expressions. The name of the function in $0 is mathfn (not shellfn as would usually be the case), provided the option FUNCTION_ARGZERO is in effect. The positional parameters in the shell function correspond to the arguments of the mathematical function call. The result of the last arithmetical expression evaluated inside the shell function (even if it is a form that normally only returns a status) gives the result of the mathematical function.
If the additional option -s is given to functions -M, the argument to the function is a single string: anything between the opening and matching closing parenthesis is passed to the function as a single argument, even if it includes commas or white space. The minimum and maximum argument specifiers must therefore be 1 if given. An empty argument list is passed as a zero-length string.
functions -M with no arguments lists all such user-defined functions in the same form as a definition. With the additional option -m and a list of arguments, all functions whose mathfn matches one of the pattern arguments are listed.
function +M removes the list of mathematical functions; with the additional option -m the arguments are treated as patterns and all functions whose mathfn matches the pattern are removed. Note that the shell function implementing the behaviour is not removed (regardless of whether its name coincides with mathfn).
For example, the following prints the cube of 3:
zmath_cube() { (( $1 * $1 * $1 )) }
functions -M cube 1 1 zmath_cube
print $(( cube(3) ))
The following string function takes a single argument, including the commas, so prints 11:
stringfn() { (( $#1 )) }
functions -Ms stringfn
print $(( stringfn(foo,bar,rod) ))
Each time it is invoked, getopts places the option letter it finds in the shell parameter name, prepended with a `+' when arg begins with a `+'. The index of the next arg is stored in OPTIND. The option argument, if any, is stored in OPTARG.
The first option to be examined may be changed by explicitly assigning to OPTIND. OPTIND has an initial value of 1, and is normally set to 1 upon entry to a shell function and restored upon exit (this is disabled by the POSIX_BUILTINS option). OPTARG is not reset and retains its value from the most recent call to getopts. If either of OPTIND or OPTARG is explicitly unset, it remains unset, and the index or option argument is not stored. The option itself is still stored in name in this case.
A leading `:' in optstring causes getopts to store the letter of any invalid option in OPTARG, and to set name to `?' for an unknown option and to `:' when a required argument is missing. Otherwise, getopts sets name to `?' and prints an error message when an option is invalid. The exit status is nonzero when there are no more options.
A command name starting with a / is never hashed, whether by explicit use of the hash command or otherwise. Such a command is always found by direct look up in the file system.
Given no arguments, and neither the -r or -f options, the selected hash table will be listed in full.
The -r option causes the selected hash table to be emptied. It will be subsequently rebuilt in the normal fashion. The -f option causes the selected hash table to be fully rebuilt immediately. For the command hash table this hashes all the absolute directories in the PATH, and for the named directory hash table this adds all users' home directories. These two options cannot be used with any arguments.
The -m option causes the arguments to be taken as patterns (which should be quoted) and the elements of the hash table matching those patterns are printed. This is the only way to display a limited selection of hash table elements.
For each name with a corresponding value, put `name' in the selected hash table, associating it with the pathname `value'. In the command hash table, this means that whenever `name' is used as a command argument, the shell will try to execute the file given by `value'. In the named directory hash table, this means that `value' may be referred to as `~name'.
For each name with no corresponding value, attempt to add name to the hash table, checking what the appropriate value is in the normal manner for that hash table. If an appropriate value can't be found, then the hash table will be unchanged.
The -v option causes hash table entries to be listed as they are added by explicit specification. If has no effect if used with -f.
If the -L flag is present, then each hash table entry is printed in the form of a call to hash.
The -Z option replaces the shell's argument and environment space with the given string, truncated if necessary to fit. This will normally be visible in ps (ps(1)) listings. This feature is typically used by daemons, to indicate their state.
On some systems, alternative signal names are allowed for a few signals. Typical examples are SIGCHLD and SIGCLD or SIGPOLL and SIGIO, assuming they correspond to the same signal number. kill -l will only list the preferred form, however kill -l alt will show if the alternative form corresponds to a signal number. For example, under Linux kill -l IO and kill -l POLL both output 29, hence kill -IO and kill -POLL have the same effect.
Many systems will allow process IDs to be negative to kill a process group or zero to kill the current process group.
If limit is not specified, print the current limit placed on resource, otherwise set the limit to the specified value. If the -h flag is given, use hard limits instead of soft limits. If no resource is given, print all limits.
When looping over multiple resources, the shell will abort immediately if it detects a badly formed argument. However, if it fails to set a limit for some other reason it will continue trying to set the remaining limits.
resource can be one of:
Which of these resource limits are available depends on the system. resource can be abbreviated to any unambiguous prefix. It can also be an integer, which corresponds to the integer defined for the resource by the operating system.
If argument corresponds to a number which is out of the range of the resources configured into the shell, the shell will try to read or write the limit anyway, and will report an error if this fails. As the shell does not store such resources internally, an attempt to set the limit will fail unless the -s option is present.
limit is a number, with an optional scaling factor, as follows:
The limit command is not made available by default when the shell starts in a mode emulating another shell. It can be made available with the command `zmodload -F zsh/rlimits b:limit'.
If the -q (quiet) option is specified, the hook function chpwd and the functions in the array $chpwd_functions are not called, and the new directory stack is not printed. This is useful for calls to popd that do not change the environment seen by an interactive user.
The start of the output of each print command is assumed to be aligned with a tab stop. Widths of multibyte characters are handled if the option MULTIBYTE is in effect. This option is ignored if other formatting options are in effect, namely column alignment or printf style, or if output is to a special location such as shell history or the command line editor.
If any of `-m', `-o' or `-O' are used in combination with `-f' and there are no arguments (after the removal process in the case of `-m') then nothing is printed.
Normally, conversion specifications are applied to each argument in order but they can explicitly specify the nth argument is to be used by replacing `%' by `%n$' and `*' by `*n$'. It is recommended that you do not mix references of this explicit style with the normal style and the handling of such mixed styles may be subject to future change.
If arguments remain unused after formatting, the format string is reused until all arguments have been consumed. With the print builtin, this can be suppressed by using the -r option. If more arguments are required by the format than have been specified, the behaviour is as if zero or an empty string had been specified as the argument.
The -v option causes the output to be stored as the value of the parameter name, instead of printed. If name is an array and the format string is reused when consuming arguments then one array element will be used for each use of the format string.
The third form of pushd changes directory by rotating the directory list. An argument of the form `+n' identifies a stack entry by counting from the left of the list shown by the dirs command, starting with zero. An argument of the form `-n' counts from the right. If the PUSHD_MINUS option is set, the meanings of `+' and `-' in this context are swapped.
If the -q (quiet) option is specified, the hook function chpwd and the functions in the array $chpwd_functions are not called, and the new directory stack is not printed. This is useful for calls to pushd that do not change the environment seen by an interactive user.
If the option -q is not specified and the shell option PUSHD_SILENT is not set, the directory stack will be printed after a pushd is performed.
The options -s, -L and -P have the same meanings as for the cd builtin.
Note that despite the mnemonic `key' this option does read full characters, which may consist of multiple bytes if the option MULTIBYTE is set.
If the first argument contains a `?', the remainder of this word is used as a prompt on standard error when the shell is interactive.
The value (exit status) of read is 1 when an end-of-file is encountered, or when -c or -l is present and the command is not called from a compctl function, or as described for -q. Otherwise the value is 0.
The behavior of some combinations of the -k, -p, -q, -u and -z flags is undefined. Presently -q cancels all the others, -p cancels -u, -k cancels -z, and otherwise -z cancels both -p and -u.
The -c or -l flags cancel any and all of -kpquz.
If return was executed from a trap in a TRAPNAL function, the effect is different for zero and non-zero return status. With zero status (or after an implicit return at the end of the trap), the shell will return to whatever it was previously processing; with a non-zero status, the shell will behave as interrupted except that the return status of the trap is retained. Note that the numeric value of the signal which caused the trap is passed as the first argument, so the statement `return $((128+$1))' will return the same status as if the signal had not been trapped.
If the -A flag is specified, name is set to an array containing the given args; if no name is specified, all arrays are printed together with their values.
If +A is used and name is an array, the given arguments will replace the initial elements of that array; if no name is specified, all arrays are printed without their values.
The behaviour of arguments after -A name or +A name depends on whether the option KSH_ARRAYS is set. If it is not set, all arguments following name are treated as values for the array, regardless of their form. If the option is set, normal option processing continues at that point; only regular arguments are treated as values for the array. This means that
set -A array -x -- foo
sets array to `-x -- foo' if KSH_ARRAYS is not set, but sets the array to foo and turns on the option `-x' if it is set.
If the -A flag is not present, but there are arguments beyond the options, the positional parameters are set. If the option list (if any) is terminated by `--', and there are no further arguments, the positional parameters will be unset.
If no arguments and no `--' are given, then the names and values of all parameters are printed on the standard output. If the only argument is `+', the names of all parameters are printed.
For historical reasons, `set -' is treated as `set +xv' and `set - args' as `set +xv -- args' when in any other emulation mode than zsh's native mode.
If no arguments are supplied, the names of all options currently set are printed. The form is chosen so as to minimize the differences from the default options for the current emulation (the default emulation being native zsh, shown as <Z> in zshoptions(1)). Options that are on by default for the emulation are shown with the prefix no only if they are off, while other options are shown without the prefix no and only if they are on. In addition to options changed from the default state by the user, any options activated automatically by the shell (for example, SHIN_STDIN or INTERACTIVE) will be shown in the list. The format is further modified by the option KSH_OPTION_PRINT, however the rationale for choosing options with or without the no prefix remains the same in this case.
If the -m flag is given the arguments are taken as patterns (which should be quoted to protect them from filename expansion), and all options with names matching these patterns are set.
Note that a bad option name does not cause execution of subsequent shell code to be aborted; this is behaviour is different from that of `set -o'. This is because set is regarded as a special builtin by the POSIX standard, but setopt is not.
If the option -p is given arguments are instead removed (popped) from the end rather than the start of the array.
The command attempts to implement POSIX and its extensions where these are specified. Unfortunately there are intrinsic ambiguities in the syntax; in particular there is no distinction between test operators and strings that resemble them. The standard attempts to resolve these for small numbers of arguments (up to four); for five or more arguments compatibility cannot be relied on. Users are urged wherever possible to use the `[[' test syntax which does not have these ambiguities.
If arg is `-', then the specified signals are reset to their defaults, or, if no sig args are present, all traps are reset.
If arg is an empty string, then the specified signals are ignored by the shell (and by the commands it invokes).
If arg is omitted but one or more sig args are provided (i.e. the first argument is a valid signal number or name), the effect is the same as if arg had been specified as `-'.
The trap command with no arguments prints a list of commands associated with each signal.
If sig is ZERR then arg will be executed after each command with a nonzero exit status. ERR is an alias for ZERR on systems that have no SIGERR signal (this is the usual case).
If sig is DEBUG then arg will be executed before each command if the option DEBUG_BEFORE_CMD is set (as it is by default), else after each command. Here, a `command' is what is described as a `sublist' in the shell grammar, see the section SIMPLE COMMANDS & PIPELINES in zshmisc(1). If DEBUG_BEFORE_CMD is set various additional features are available. First, it is possible to skip the next command by setting the option ERR_EXIT; see the description of the ERR_EXIT option in zshoptions(1). Also, the shell parameter ZSH_DEBUG_CMD is set to the string corresponding to the command to be executed following the trap. Note that this string is reconstructed from the internal format and may not be formatted the same way as the original text. The parameter is unset after the trap is executed.
If sig is 0 or EXIT and the trap statement is executed inside the body of a function, then the command arg is executed after the function completes. The value of $? at the start of execution is the exit status of the shell or the return status of the function exiting. If sig is 0 or EXIT and the trap statement is not executed inside the body of a function, then the command arg is executed when the shell terminates; the trap runs before any zshexit hook functions.
ZERR, DEBUG, and EXIT traps are not executed inside other traps. ZERR and DEBUG traps are kept within subshells, while other traps are reset.
Note that traps defined with the trap builtin are slightly different from those defined as `TRAPNAL () { ... }', as the latter have their own function environment (line numbers, local variables, etc.) while the former use the environment of the command in which they were called. For example,
trap 'print $LINENO' DEBUG
will print the line number of a command executed after it has run, while
TRAPDEBUG() { print $LINENO; }
will always print the number zero.
Alternative signal names are allowed as described under kill above. Defining a trap under either name causes any trap under an alternative name to be removed. However, it is recommended that for consistency users stick exclusively to one name or another.
Without options it reports whether the terminal is frozen or not.
Note that, regardless of whether the tty is frozen or not, the shell needs to change the settings when the line editor starts, so unfreezing the tty does not guarantee settings made on the command line are preserved. Strings of commands run between editing the command line will see a consistent tty state. See also the shell variable STTY for a means of initialising the tty before running external commands.
Except as noted below for control flags that change the behavior, a parameter is created for each name that does not already refer to one. When inside a function, a new parameter is created for every name (even those that already exist), and is unset again when the function completes. See `Local Parameters' in zshparam(1). The same rules apply to special shell parameters, which retain their special attributes when made local.
For each name=value assignment, the parameter name is set to value.
If the shell option TYPESET_SILENT is not set, for each remaining name that refers to a parameter that is already set, the name and value of the parameter are printed in the form of an assignment. Nothing is printed for newly-created parameters, or when any attribute flags listed below are given along with the name. Using `+' instead of minus to introduce an attribute turns it off.
If no name is present, the names and values of all parameters are printed. In this case the attribute flags restrict the display to only those parameters that have the specified attributes, and using `+' rather than `-' to introduce the flag suppresses printing of the values of parameters when there is no parameter name.
All forms of the command handle scalar assignment. Array assignment is possible if any of the reserved words declare, export, float, integer, local, readonly or typeset is matched when the line is parsed (N.B. not when it is executed). In this case the arguments are parsed as assignments, except that the `+=' syntax and the GLOB_ASSIGN option are not supported, and scalar values after = are not split further into words, even if expanded (regardless of the setting of the KSH_TYPESET option; this option is obsolete).
Examples of the differences between command and reserved word parsing:
# Reserved word parsing typeset svar=$(echo one word) avar=(several words)
The above creates a scalar parameter svar and an array parameter avar as if the assignments had been
svar="one word" avar=(several words)
On the other hand:
# Normal builtin interface builtin typeset svar=$(echo two words)
The builtin keyword causes the above to use the standard builtin interface to typeset in which argument parsing is performed in the same way as for other commands. This example creates a scalar svar containing the value two and another scalar parameter words with no value. An array value in this case would either cause an error or be treated as an obscure set of glob qualifiers.
Arbitrary arguments are allowed if they take the form of assignments after command line expansion; however, these only perform scalar assignment:
var='svar=val' typeset $var
The above sets the scalar parameter svar to the value val. Parentheses around the value within var would not cause array assignment as they will be treated as ordinary characters when $var is substituted. Any non-trivial expansion in the name part of the assignment causes the argument to be treated in this fashion:
typeset {var1,var2,var3}=name
The above syntax is valid, and has the expected effect of setting the three parameters to the same value, but the command line is parsed as a set of three normal command line arguments to typeset after expansion. Hence it is not possible to assign to multiple arrays by this means.
Note that each interface to any of the commands my be disabled separately. For example, `disable -r typeset' disables the reserved word interface to typeset, exposing the builtin interface, while `disable typeset' disables the builtin. Note that disabling the reserved word interface for typeset may cause problems with the output of `typeset -p', which assumes the reserved word interface is available in order to restore array and associative array values.
Unlike parameter assignment statements, typeset's exit status on an assignment that involves a command substitution does not reflect the exit status of the command substitution. Therefore, to test for an error in a command substitution, separate the declaration of the parameter from its initialization:
# WRONG typeset var1=$(exit 1) || echo "Trouble with var1" # RIGHT typeset var1 && var1=$(exit 1) || echo "Trouble with var1"
To initialize a parameter param to a command output and mark it readonly, use typeset -r param or readonly param after the parameter assignment statement.
If no attribute flags are given, and either no name arguments are present or the flag +m is used, then each parameter name printed is preceded by a list of the attributes of that parameter (array, association, exported, float, integer, readonly, or undefined for autoloaded parameters not yet loaded). If +m is used with attribute flags, and all those flags are introduced with +, the matching parameter names are printed but their values are not.
The following control flags change the behavior of typeset:
If the +g flag is combined with -m, a new local parameter is created for every matching parameter that is not already local. Otherwise -m applies all other flags or assignments to the existing parameters.
Except when assignments are made with name=value, using +m forces the matching parameters and their attributes to be printed, even inside a function. Note that -m is ignored if no patterns are given, so `typeset -m' displays attributes but `typeset -a +m' does not.
-p may be followed by an optional integer argument. Currently only the value 1 is supported. In this case arrays and associative arrays are printed with newlines between indented elements for readability.
Only one of the scalar and array parameters may be assigned an initial value (the restrictions on assignment forms described above also apply).
Both the scalar and the array may be manipulated as normal. If one is unset, the other will automatically be unset too. There is no way of untying the variables without unsetting them, nor of converting the type of one of them with another typeset command; +T does not work, assigning an array to scalar is an error, and assigning a scalar to array sets it to be a single-element array.
Note that both `typeset -xT ...' and `export -T ...' work, but only the scalar will be marked for export. Setting the value using the scalar version causes a split on all separators (which cannot be quoted). It is possible to apply -T to two previously tied variables but with a different separator character, in which case the variables remain joined as before but the separator is changed.
When an existing scalar is tied to a new array, the value of the scalar is preserved but no attribute other than export will be preserved.
Attribute flags that transform the final value (-L, -R, -Z, -l, -u) are only applied to the expanded value at the point of a parameter expansion expression using `$'. They are not applied when a parameter is retrieved internally by the shell for any purpose.
The following attribute flags may be specified:
The width is the count of characters, which may be multibyte characters if the MULTIBYTE option is in effect. Note that the screen width of the character is not taken into account; if this is required, use padding with parameter expansion flags ${(ml...)...} as described in `Parameter Expansion Flags' in zshexpn(1).
When the parameter is expanded, it is filled on the right with blanks or truncated if necessary to fit the field. Note truncation can lead to unexpected results with numeric parameters. Leading zeros are removed if the -Z flag is also set.
This flag has a different meaning when used with -f; see below.
Note that the builtin functions provides the same basic capabilities as typeset -f but gives access to a few extra options; autoload gives further additional options for the case typeset -fu and typeset -fU.
If the POSIX_BUILTINS option is set, the readonly attribute is more restrictive: unset variables can be marked readonly and cannot then be set; furthermore, the readonly attribute cannot be removed from any variable.
It is still possible to change other attributes of the variable though, some of which like -U or -Z would affect the value. More generally, the readonly attribute should not be relied on as a security mechanism.
Note that in zsh (like in pdksh but unlike most other shells) it is still possible to create a local variable of the same name as this is considered a different variable (though this variable, too, can be marked readonly). Special variables that have been made readonly retain their value and readonly attribute when made local.
By default, only soft limits are manipulated. If the -H flag is given use hard limits instead of soft limits. If the -S flag is given together with the -H flag set both hard and soft limits.
If no options are used, the file size limit (-f) is assumed.
If limit is omitted the current value of the specified resources are printed. When more than one resource value is printed, the limit name and unit is printed before each value.
When looping over multiple resources, the shell will abort immediately if it detects a badly formed argument. However, if it fails to set a limit for some other reason it will continue trying to set the remaining limits.
Not all the following resources are supported on all systems. Running ulimit -a will show which are supported.
A resource may also be specified by integer in the form `-N resource', where resource corresponds to the integer defined for the resource by the operating system. This may be used to set the limits for resources known to the shell which do not correspond to option letters. Such limits will be shown by number in the output of `ulimit -a'.
The number may alternatively be out of the range of limits compiled into the shell. The shell will try to read or write the limit anyway, and will report an error if this fails.
The unlimit command is not made available by default when the shell starts in a mode emulating another shell. It can be made available with the command `zmodload -F zsh/rlimits b:unlimit'.
Individual elements of associative array parameters may be unset by using subscript syntax on name, which should be quoted (or the entire command prefixed with noglob) to protect the subscript from filename generation.
If the -m flag is specified the arguments are taken as patterns (should be quoted) and all parameters with matching names are unset. Note that this cannot be used when unsetting associative array elements, as the subscript will be treated as part of the pattern.
The -v flag specifies that name refers to parameters. This is the default behaviour.
unset -f is equivalent to unfunction.
It is possible to wait for recent processes (specified by process ID, not by job) that were running in the background even if the process has exited. Typically the process ID will be recorded by capturing the value of the variable $! immediately after the process has been started. There is a limit on the number of process IDs remembered by the shell; this is given by the value of the system configuration parameter CHILD_MAX. When this limit is reached, older process IDs are discarded, least recently started processes first.
Note there is no protection against the process ID wrapping, i.e. if the wait is not executed soon enough there is a chance the process waited for is the wrong one. A conflict implies both process IDs have been generated by the shell, as other processes are not recorded, and that the user is potentially interested in both, so this problem is intrinsic to process IDs.
If name is not an alias, built-in command, external command, shell function, hashed command, or a reserved word, the exit status shall be non-zero, and -- if -v, -c, or -w was passed -- a message will be written to standard output. (This is different from other shells that write that message to standard error.)
whence is most useful when name is only the last path component of a command, i.e. does not include a `/'; in particular, pattern matching only succeeds if just the non-directory component of the command is passed.
The first form (without the -c, -a or -t options) creates a compiled file. If only the file argument is given, the output file has the name `file.zwc' and will be placed in the same directory as the file. The shell will load the compiled file instead of the normal function file when the function is autoloaded; see the section `Autoloading Functions' in zshmisc(1) for a description of how autoloaded functions are searched. The extension .zwc stands for `zsh word code'.
If there is at least one name argument, all the named files are compiled into the output file given as the first argument. If file does not end in .zwc, this extension is automatically appended. Files containing multiple compiled functions are called `digest' files, and are intended to be used as elements of the FPATH/fpath special array.
The second form, with the -c or -a options, writes the compiled definitions for all the named functions into file. For -c, the names must be functions currently defined in the shell, not those marked for autoloading. Undefined functions that are marked for autoloading may be written by using the -a option, in which case the fpath is searched and the contents of the definition files for those functions, if found, are compiled into file. If both -c and -a are given, names of both defined functions and functions marked for autoloading may be given. In either case, the functions in files written with the -c or -a option will be autoloaded as if the KSH_AUTOLOAD option were unset.
The reason for handling loaded and not-yet-loaded functions with different options is that some definition files for autoloading define multiple functions, including the function with the same name as the file, and, at the end, call that function. In such cases the output of `zcompile -c' does not include the additional functions defined in the file, and any other initialization code in the file is lost. Using `zcompile -a' captures all this extra information.
If the -m option is combined with -c or -a, the names are used as patterns and all functions whose names match one of these patterns will be written. If no name is given, the definitions of all functions currently defined or marked as autoloaded will be written.
Note the second form cannot be used for compiling functions that include redirections as part of the definition rather than within the body of the function; for example
fn1() { { ... } >~/logfile }
can be compiled but
fn1() { ... } >~/logfile
cannot. It is possible to use the first form of zcompile to compile autoloadable functions that include the full function definition instead of just the body of the function.
The third form, with the -t option, examines an existing compiled file. Without further arguments, the names of the original files compiled into it are listed. The first line of output shows the version of the shell which compiled the file and how the file will be used (i.e. by reading it directly or by mapping it into memory). With arguments, nothing is output and the return status is set to zero if definitions for all names were found in the compiled file, and non-zero if the definition for at least one name was not found.
Other options:
When compiling scripts instead of autoloadable functions, it is often desirable to use this option; otherwise the whole file, including the code to define functions which have already been defined, will remain mapped, consequently wasting memory.
The created file always contains two versions of the compiled format, one for big-endian machines and one for small-endian machines. The upshot of this is that the compiled file is machine independent and if it is read or mapped, only one half of the file is actually used (and mapped).
Without arguments the names of all currently loaded binary modules are printed. The -L option causes this list to be in the form of a series of zmodload commands. Forms with arguments are:
The named module is searched for in the same way a command is, using $module_path instead of $path. However, the path search is performed even when the module name contains a `/', which it usually does. There is no way to prevent the path search.
If the module supports features (see below), zmodload tries to enable all features when loading a module. If the module was successfully loaded but not all features could be enabled, zmodload returns status 2.
If the option -s is given, no error is printed if the module was not available (though other errors indicating a problem with the module are printed). The return status indicates if the module was loaded. This is appropriate if the caller considers the module optional.
With -u, zmodload unloads modules. The same name must be given that was given when the module was loaded, but it is not necessary for the module to exist in the file system. The -i option suppresses the error if the module is already unloaded (or was never loaded).
Each module has a boot and a cleanup function. The module will not be loaded if its boot function fails. Similarly a module can only be unloaded if its cleanup function runs successfully.
The standard features are builtins, conditions, parameters and math functions; these are indicated by the prefix `b:', `c:' (`C:' for an infix condition), `p:' and `f:', respectively, followed by the name that the corresponding feature would have in the shell. For example, `b:strftime' indicates a builtin named strftime and p:EPOCHSECONDS indicates a parameter named EPOCHSECONDS. The module may provide other (`abstract') features of its own as indicated by its documentation; these have no prefix.
With -l or -L, features provided by the module are listed. With -l alone, a list of features together with their states is shown, one feature per line. With -L alone, a zmodload -F command that would cause enabled features of the module to be turned on is shown. With -lL, a zmodload -F command that would cause all the features to be set to their current state is shown. If one of these combinations is given with the option -P param then the parameter param is set to an array of features, either features together with their state or (if -L alone is given) enabled features.
With the option -L the module name may be omitted; then a list of all enabled features for all modules providing features is printed in the form of zmodload -F commands. If -l is also given, the state of both enabled and disabled features is output in that form.
A set of features may be provided together with -l or -L and a module name; in that case only the state of those features is considered. Each feature may be preceded by + or - but the character has no effect. If no set of features is provided, all features are considered.
With -e, the command first tests that the module is loaded; if it is not, status 1 is returned. If the module is loaded, the list of features given as an argument is examined. Any feature given with no prefix is simply tested to see if the module provides it; any feature given with a prefix + or - is tested to see if is provided and in the given state. If the tests on all features in the list succeed, status 0 is returned, else status 1.
With -m, each entry in the given list of features is taken as a pattern to be matched against the list of features provided by the module. An initial + or - must be given explicitly. This may not be combined with the -a option as autoloads must be specified explicitly.
With -a, the given list of features is marked for autoload from the specified module, which may not yet be loaded. An optional + may appear before the feature name. If the feature is prefixed with -, any existing autoload is removed. The options -l and -L may be used to list autoloads. Autoloading is specific to individual features; when the module is loaded only the requested feature is enabled. Autoload requests are preserved if the module is subsequently unloaded until an explicit `zmodload -Fa module -feature' is issued. It is not an error to request an autoload for a feature of a module that is already loaded.
When the module is loaded each autoload is checked against the features actually provided by the module; if the feature is not provided the autoload request is deleted. A warning message is output; if the module is being loaded to provide a different feature, and that autoload is successful, there is no effect on the status of the current command. If the module is already loaded at the time when zmodload -Fa is run, an error message is printed and status 1 returned.
zmodload -Fa can be used with the -l, -L, -e and -P options for listing and testing the existence of autoloadable features. In this case -l is ignored if -L is specified. zmodload -FaL with no module name lists autoloads for all modules.
Note that only standard features as described above can be autoloaded; other features require the module to be loaded before enabling.
With -d and one argument, all dependencies for that module are listed. With -d and no arguments, all module dependencies are listed. This listing is by default in a Makefile-like format. The -L option changes this format to a list of zmodload -d commands.
If -d and -u are both used, dependencies are removed. If only one argument is given, all dependencies for that module are removed.
With -ab and no arguments, all autoloaded builtins are listed, with the module name (if different) shown in parentheses after the builtin name. The -L option changes this format to a list of zmodload -a commands.
If -b is used together with the -u option, it removes builtins previously defined with -ab. This is only possible if the builtin is not yet loaded. -i suppresses the error if the builtin is already removed (or never existed).
Autoload requests are retained if the module is subsequently unloaded until an explicit `zmodload -ub builtin' is issued.
If given no condition names, all defined names are listed (as a series of zmodload commands if the -L option is given).
The -uc option removes definitions for autoloaded conditions.
The existence of aliases for modules is completely independent of whether the name resolved is actually loaded as a module: while the alias exists, loading and unloading the module under any alias has exactly the same effect as using the resolved name, and does not affect the connection between the alias and the resolved name which can be removed either by zmodload -R or by redefining the alias. Chains of aliases (i.e. where the first resolved name is itself an alias) are valid so long as these are not circular. As the aliases take the same format as module names, they may include path separators: in this case, there is no requirement for any part of the path named to exist as the alias will be resolved first. For example, `any/old/alias' is always a valid alias.
Dependencies added to aliased modules are actually added to the resolved module; these remain if the alias is removed. It is valid to create an alias whose name is one of the standard shell modules and which resolves to a different module. However, if a module has dependencies, it will not be possible to use the module name as an alias as the module will already be marked as a loadable module in its own right.
Apart from the above, aliases can be used in the zmodload command anywhere module names are required. However, aliases will not be shown in lists of loaded modules with a bare `zmodload'.
Note that zsh makes no distinction between modules that were linked into the shell and modules that are loaded dynamically. In both cases this builtin command has to be used to make available the builtins and other things defined by modules (unless the module is autoloaded on these definitions). This is true even for systems that don't support dynamic loading of modules.
There are two display modes. The first, multiline mode, is the default. It only works if the TERM parameter is set to a valid terminal type that can move the cursor up. The second, single line mode, is used if TERM is invalid or incapable of moving the cursor up, or if the SINGLE_LINE_ZLE option is set. This mode is similar to ksh, and uses no termcap sequences. If TERM is "emacs", the ZLE option will be unset by default.
The parameters BAUD, COLUMNS, and LINES are also used by the line editor. See Parameters Used By The Shell in zshparam(1).
The parameter zle_highlight is also used by the line editor; see Character Highlighting below. Highlighting of special characters and the region between the cursor and the mark (as set with set-mark-command in Emacs mode, or by visual-mode in Vi mode) is enabled by default; consult this reference for more information. Irascible conservatives will wish to know that all highlighting may be disabled by the following setting:
zle_highlight=(none)
In many places, references are made to the numeric argument. This can by default be entered in emacs mode by holding the alt key and typing a number, or pressing escape before each digit, and in vi command mode by typing the number before entering a command. Generally the numeric argument causes the next command entered to be repeated the specified number of times, unless otherwise noted below; this is implemented by the digit-argument widget. See also the Arguments subsection of the Widgets section for some other ways the numeric argument can be modified.
There can be any number of keymaps at any time, and each keymap has one or more names. If all of a keymap's names are deleted, it disappears. bindkey can be used to manipulate keymap names.
Initially, there are eight keymaps:
The `.safe' keymap is special. It can never be altered, and the name can never be removed. However, it can be linked to other names, which can be removed. In the future other special keymaps may be added; users should avoid using names beginning with `.' for their own keymaps.
In addition to these names, either `emacs' or `viins' is also linked to the name `main'. If one of the VISUAL or EDITOR environment variables contain the string `vi' when the shell starts up then it will be `viins', otherwise it will be `emacs'. bindkey's -e and -v options provide a convenient way to override this default choice.
When the editor starts up, it will select the `main' keymap. If that keymap doesn't exist, it will use `.safe' instead.
In the `.safe' keymap, each single key is bound to self-insert, except for ^J (line feed) and ^M (return) which are bound to accept-line. This is deliberately not pleasant to use; if you are using it, it means you deleted the main keymap, and you should put it back.
The key timeout is also applied when ZLE is reading the bytes from a multibyte character string when it is in the appropriate mode. (This requires that the shell was compiled with multibyte mode enabled; typically also the locale has characters with the UTF-8 encoding, although any multibyte encoding known to the operating system is supported.) If the second or a subsequent byte is not read within the timeout period, the shell acts as if ? were typed and resets the input state.
As well as ZLE commands, key sequences can be bound to other strings, by using `bindkey -s'. When such a sequence is read, the replacement string is pushed back as input, and the command reading process starts again using these fake keystrokes. This input can itself invoke further replacement strings, but in order to detect loops the process will be stopped if there are twenty such replacements without a real command being read.
A key sequence typed by the user can be turned into a command name for use in user-defined widgets with the read-command widget, described in the subsection `Miscellaneous' of the section `Standard Widgets' below.
If a key sequence is defined in a local keymap, it will hide a key sequence in the global keymap that is a prefix of that sequence. An example of this occurs with the binding of iw in viopp as this hides the binding of i in vicmd. However, a longer sequence in the global keymap that shares the same prefix can still apply so for example the binding of ^Xa in the global keymap will be unaffected by the binding of ^Xb in the local keymap.
If a keymap selection is required and none of the options above are used, the `main' keymap is used. Some operations do not permit a keymap to be selected, namely:
If the -L option is also used, list in the form of bindkey commands to create or link the keymaps. `bindkey -lL main' shows which keymap is linked to `main', if any, and hence if the standard emacs or vi emulation is in effect. This option does not show the .safe keymap because it cannot be created in that fashion; however, neither is `bindkey -lL .safe' reported as an error, it simply outputs nothing.
To use a newly created keymap, it should be linked to main. Hence the sequence of commands to create and use a new keymap `mymap' initialized from the emacs keymap (which remains unchanged) is:
bindkey -N mymap emacs bindkey -A mymap main
Note that while `bindkey -A newmap main' will work when newmap is emacs or viins, it will not work for vicmd, as switching from vi insert to command mode becomes impossible.
The following operations act on the `main' keymap if no keymap selection option was given:
When -R is also used, interpret the in-strings as ranges.
When -p is also used, the in-strings specify prefixes. Any binding that has the given in-string as a prefix, not including the binding for the in-string itself, if any, will be removed. For example,
bindkey -rpM viins '^['
will remove all bindings in the vi-insert keymap beginning with an escape character (probably cursor keys), but leave the binding for the escape character itself (probably vi-cmd-mode). This is incompatible with the option -R.
Note that both in-string and out-string are subject to the same form of interpretation, as described below.
When the option -p is used, the in-string must be present. The listing shows all bindings which have the given key sequence as a prefix, not including any bindings for the key sequence itself.
When the -L option is used, the list is in the form of bindkey commands to create the key bindings.
When the -R option is used as noted above, a valid range consists of two characters, with an optional `-' between them. All characters between the two specified, inclusive, are bound as specified.
For either in-string or out-string, the following escape sequences are recognised:
In all other cases, `\' escapes the following character. Delete is written as `^?'. Note that `\M^?' and `^\M?' are not the same, and that (unlike emacs), the bindings `\M-X' and `\eX' are entirely distinct, although they are initialized to the same bindings by `bindkey -m'.
If an array or array slice is being edited, separator characters as defined in $IFS will be shown quoted with a backslash, as will backslashes themselves. Conversely, when the edited text is split into an array, a backslash quotes an immediately following separator character or backslash; no other special handling of backslashes, or any handling of quotes, is performed.
Individual elements of existing array or associative array parameters may be edited by using subscript syntax on name. New elements are created automatically, even without -c.
If the -p flag is given, the following string will be taken as the prompt to display at the left. If the -r flag is given, the following string gives the prompt to display at the right. If the -h flag is specified, the history can be accessed from ZLE. If the -e flag is given, typing ^D (Control-D) on an empty line causes vared to exit immediately with a non-zero return value.
The -M option gives a keymap to link to the main keymap during editing, and the -m option gives a keymap to link to the vicmd keymap during editing. For vi-style editing, this allows a pair of keymaps to override viins and vicmd. For emacs-style editing, only -M is normally needed but the -m option may still be used. On exit, the previous keymaps will be restored.
Vared calls the usual `zle-line-init' and `zle-line-finish' hooks before and after it takes control. Using the -i and -f options, it is possible to replace these with other custom widgets.
If `-t tty' is given, tty is the name of a terminal device to be used instead of the default /dev/tty. If tty does not refer to a terminal an error is reported.
With no options and no arguments, only the return status will be set. It is zero if ZLE is currently active and widgets could be invoked using this builtin command and non-zero otherwise. Note that even if non-zero status is returned, zle may still be active as part of the completion system; this does not allow direct calls to ZLE widgets.
Otherwise, which operation it performs depends on its options:
When combined with the -a option, all widget names are listed, including the builtin ones. In this case the -L option is ignored.
If at least one string is given, and -a is present or -L is not used, nothing will be printed. The return status will be zero if all strings are names of existing widgets and non-zero if at least one string is not a name of a defined widget. If -a is also present, all widget names are used for the comparison including builtin widgets, else only user-defined widgets are used.
If at least one string is present and the -L option is used, user-defined widgets matching any string are listed in the form of zle commands to create the widgets.
yank for indicating that the widget has yanked text into the buffer. If the widget is wrapping an existing internal widget, no further action is necessary, but if it has inserted the text manually, then it should also take care to set YANK_START and YANK_END correctly. yankbefore does the same but is used when the yanked text appears after the cursor.
kill for indicating that text has been killed into the cutbuffer. When repeatedly invoking a kill widget, text is appended to the cutbuffer instead of replacing it, but when wrapping such widgets, it is necessary to call `zle -f kill' to retain this effect.
vichange for indicating that the widget represents a vi change that can be repeated as a whole with `vi-repeat-change'. The flag should be set early in the function before inspecting the value of NUMERIC or invoking other widgets. This has no effect for a widget invoked from insert mode. If insert mode is active when the widget finishes, the change extends until next returning to command mode.
If the optional strings are given they are listed below the prompt in the same way as completion lists are printed. If no strings are given but the -c option is used such a list is cleared.
Note that this option is only useful for widgets that do not exit immediately after using it because the strings displayed will be erased immediately after return from the widget.
This command can safely be called outside user defined widgets; if zle is active, the display will be refreshed, while if zle is not active, the command has no effect. In this case there will usually be no other arguments.
The status is zero if zle was active, else one.
As ZLE uses a stack, if this option is used repeatedly the last string pushed onto the stack will be processed first. However, the characters in each string will be processed in the order in which they appear in the string.
This keymap selection affects the interpretation of following keystrokes within this invocation of ZLE. Any following invocation (e.g., the next command line) will start as usual with the `main' keymap selected.
Installs handler (the name of a shell function) to handle input from file descriptor fd. Installing a handler for an fd which is already handled causes the existing handler to be replaced. Any number of handlers for any number of readable file descriptors may be installed. Note that zle makes no attempt to check whether this fd is actually readable when installing the handler. The user must make their own arrangements for handling the file descriptor when zle is not active.
When zle is attempting to read data, it will examine both the terminal and the list of handled fd's. If data becomes available on a handled fd, zle calls handler with the fd which is ready for reading as the first argument. Under normal circumstances this is the only argument, but if an error was detected, a second argument provides details: `hup' for a disconnect, `nval' for a closed or otherwise invalid descriptor, or `err' for any other condition. Systems that support only the `select' system call always use `err'.
If the option -w is also given, the handler is instead a line editor widget, typically a shell function made into a widget using `zle -N'. In that case handler can use all the facilities of zle to update the current editing line. Note, however, that as handling fd takes place at a low level changes to the display will not automatically appear; the widget should call `zle -R' to force redisplay. As of this writing, widget handlers only support a single argument and thus are never passed a string for error state, so widgets must be prepared to test the descriptor themselves.
If either type of handler produces output to the terminal, it should call `zle -I' before doing so (see below). Handlers should not attempt to read from the terminal.
If no handler is given, but an fd is present, any handler for that fd is removed. If there is none, an error message is printed and status 1 is returned.
If no arguments are given, or the -L option is supplied, a list of handlers is printed in a form which can be stored for later execution.
An fd (but not a handler) may optionally be given with the -L option; in this case, the function will list the handler if any, else silently return status 1.
Note that this feature should be used with care. Activity on one of the fd's which is not properly handled can cause the terminal to become unusable. Removing an fd handler from within a signal trap may cause unpredictable behavior.
Here is a simple example of using this feature. A connection to a remote TCP port is created using the ztcp command; see the description of the zsh/net/tcp module in zshmodules(1). Then a handler is installed which simply prints out any data which arrives on this connection. Note that `select' will indicate that the file descriptor needs handling if the remote side has closed the connection; we handle that by testing for a failed read.
if ztcp pwspc 2811; then
tcpfd=$REPLY
handler() {
zle -I
local line
if ! read -r line <&$1; then
# select marks this fd if we reach EOF,
# so handle this specially.
print "[Read on fd $1 failed, removing.]" >&2
zle -F $1
return 1
fi
print -r - $line
}
zle -F $tcpfd handler
fi
TRAPUSR1() {
# Invalidate zle display
[[ -o zle ]] && zle -I
# Show output
print Hello
}
In general, the trap function may need to test whether zle is active before using this method (as shown in the example), since the zsh/zle module may not even be loaded; if it is not, the command can be skipped.
It is possible to call `zle -I' several times before control is returned to the editor; the display will only be invalidated the first time to minimise disruption.
Note that there are normally better ways of manipulating the display from within zle widgets; see, for example, `zle -R' above.
The returned status is zero if zle was invalidated, even though this may have been by a previous call to `zle -I' or by a system notification. To test if a zle widget may be called at this point, execute zle with no arguments and examine the return status.
`zle -T transformation func' specifies that the given transformation (see below) is effected by shell function func.
`zle -Tr transformation' removes the given transformation if it was present (it is not an error if none was).
`zle -TL' can be used to list all transformations currently in operation.
Currently the only transformation is tc. This is used instead of outputting termcap codes to the terminal. When the transformation is in operation the shell function is passed the termcap code that would be output as its first argument; if the operation required a numeric argument, that is passed as a second argument. The function should set the shell variable REPLY to the transformed termcap code. Typically this is used to produce some simply formatted version of the code and optional argument for debugging or testing. Note that this transformation is not applied to other non-printing characters such as carriage returns and newlines.
With the options -n and -N, the current numeric argument will be saved and then restored after the call to widget; `-n num' sets the numeric argument temporarily to num, while `-N' sets it to the default, i.e. as if there were none.
With the option -K, keymap will be used as the current keymap during the execution of the widget. The previous keymap will be restored when the widget exits.
Normally, calling a widget in this way does not set the special parameter WIDGET and related parameters, so that the environment appears as if the top-level widget called by the user were still active. With the option -w, WIDGET and related parameters are set to reflect the widget being executed by the zle call.
Any further arguments will be passed to the widget; note that as standard argument handling is performed, any general argument list should be preceded by --. If it is a shell function, these are passed down as positional parameters; for builtin widgets it is up to the widget in question what it does with them. Currently arguments are only handled by the incremental-search commands, the history-search-forward and -backward and the corresponding functions prefixed by vi-, and by universal-argument. No error is flagged if the command does not use the arguments, or only uses some of them.
The return status reflects the success or failure of the operation carried out by the widget, or if it is a user-defined widget the return status of the shell function.
A non-zero return status causes the shell to beep when the widget exits, unless the BEEP options was unset or the widget was called via the zle command. Thus if a user defined widget requires an immediate beep, it should call the beep widget directly.
The standard widgets built into ZLE are listed in Standard Widgets below. Other built-in widgets can be defined by other modules (see zshmodules(1)). Each built-in widget has two names: its normal canonical name, and the same name preceded by a `.'. The `.' name is special: it can't be rebound to a different widget. This makes the widget available even when its usual name has been redefined.
User-defined widgets are defined using `zle -N', and implemented as shell functions. When the widget is executed, the corresponding shell function is executed, and can perform editing (or other) actions. It is recommended that user-defined widgets should not have names starting with `.'.
These special parameters are always available in widget functions, but are not in any way special outside ZLE. If they have some normal value outside ZLE, that value is temporarily inaccessible, but will return when the widget function exits. These special parameters in fact have local scope, like parameters created in a function using local.
Inside completion widgets and traps called while ZLE is active, these parameters are available read-only.
Note that the parameters appear as local to any ZLE widget in which they appear. Hence if it is desired to override them this needs to be done within a nested function:
widget-function() {
# $WIDGET here refers to the special variable
# that is local inside widget-function
() {
# This anonymous nested function allows WIDGET
# to be used as a local variable. The -h
# removes the special status of the variable.
local -h WIDGET
}
}
All parameters are read-only.
The default size for the kill ring is eight, however the length may be changed by normal array operations. Any empty string in the kill ring is ignored by the yank-pop command, hence the size of the array effectively sets the maximum length of the kill ring, while the number of non-zero strings gives the current length, both as seen by the user at the command line.
Each string consists of the following parts:
For example,
region_highlight=("P0 20 bold")
specifies that the first twenty characters of the text including any predisplay string should be highlighted in bold.
Note that the effect of region_highlight is not saved and disappears as soon as the line is accepted.
The final highlighting on the command line depends on both region_highlight and zle_highlight; see the section CHARACTER HIGHLIGHTING below for details.
All parameters are read-only.
A typical use of this variable in a widget function is as follows (note the additional function scope is required):
() {
local UNDO_LIMIT_NO=$UNDO_CHANGE_NO
# Perform some form of recursive edit.
}
YANK_ACTIVE is read-only.
ZLE_RECURSIVE is read-only.
Currently, the states shown are the insert mode as set by the overwrite-mode or vi-replace widgets and whether history commands will visit imported entries as controlled by the set-local-history widget. The string contains `insert' if characters to be inserted on the command line move existing characters to the right or `overwrite' if characters to be inserted overwrite existing characters. It contains `localhistory' if only local history commands will be visited or `globalhistory' if imported history commands will also be visited.
The substrings are sorted in alphabetical order so that if you want to test for two specific substrings in a future-proof way, you can do match by doing:
if [[ $ZLE_STATE == *globalhistory*insert* ]]; then ...; fi
There are a few user-defined widgets which are special to the shell. If they do not exist, no special action is taken. The environment provided is identical to that for any other editing widget.
zle-isearch-update() { zle -M "Line $HISTNO"; }
zle -N zle-isearch-update
Note the line output by `zle -M' is not deleted on exit from incremental search. This can be done from a zle-isearch-exit widget:
zle-isearch-exit() { zle -M ""; }
zle -N zle-isearch-exit
zle-line-init() { zle -K vicmd; }
zle -N zle-line-init
(The command inside the function sets the keymap directly; it is equivalent to zle vi-cmd-mode.)
The value $KEYMAP within the function reflects the new keymap. The old keymap is passed as the sole argument.
This can be used for detecting switches between the vi command (vicmd) and insert (usually main) keymaps.
Note that cursor keys are bound to movement keys in all three keymaps; the shell assumes that the cursor keys send the key sequences reported by the terminal-handling library (termcap or terminfo). The key sequences shown in the list are those based on the VT100, common on many modern terminals, but in fact these are not necessarily bound. In the case of the viins keymap, the initial escape character of the sequences serves also to return to the vicmd keymap: whether this happens is determined by the KEYTIMEOUT parameter, see zshparam(1).
If called from a function by the zle command with arguments, the first argument is taken as the string for which to search, rather than the first word in the buffer.
A restricted set of editing functions is available in the mini-buffer. Keys are looked up in the special isearch keymap, and if not found there in the main keymap (note that by default the isearch keymap is empty). An interrupt signal, as defined by the stty setting, will stop the search and go back to the original line. An undefined key will have the same effect. Note that the following always perform the same task within incremental searches and cannot be replaced by user defined widgets, nor can the set of functions be extended. The supported functions are:
In addition, the modifications that were made while in vi insert mode are merged to form a single undo event.
Any character that is not bound to one of the above functions, or self-insert or self-insert-unmeta, will cause the mode to be exited. The character is then looked up and executed in the keymap in effect at that point.
When called from a widget function by the zle command, the incremental search commands can take a string argument. This will be treated as a string of keys, as for arguments to the bindkey command, and used as initial input for the command. Any characters in the string which are unused by the incremental search will be silently ignored. For example,
zle history-incremental-search-backward forceps
will search backwards for forceps, leaving the minibuffer containing the string `forceps'.
The prompt changes to indicate an invalid pattern; this may simply indicate the pattern is not yet complete.
Note that only non-overlapping matches are reported, so an expression with wildcards may return fewer matches on a line than are visible by inspection.
If called from a function by the zle command with arguments, the first argument is taken as the string for which to search, rather than the first word in the buffer.
A restricted set of editing functions is available in the mini-buffer. An interrupt signal, as defined by the stty setting, will stop the search. The functions available in the mini-buffer are: accept-line, backward-delete-char, vi-backward-delete-char, backward-kill-word, vi-backward-kill-word, clear-screen, redisplay, quoted-insert and vi-quoted-insert.
vi-cmd-mode is treated the same as accept-line, and magic-space is treated as a space. Any other character that is not bound to self-insert or self-insert-unmeta will beep and be ignored. If the function is called from vi command mode, the bindings of the current insert mode will be used.
If called from a function by the zle command with arguments, the first argument is taken as the string for which to search, rather than the first word in the buffer.
If called from a function by the zle command with arguments, the first argument is taken as the string for which to search, rather than the first word in the buffer.
When called from a shell function invoked from a user-defined widget, the command can take one to three arguments. The first argument specifies a history offset which applies to successive calls to this widget: if it is -1, the default behaviour is used, while if it is 1, successive calls will move forwards through the history. The value 0 can be used to indicate that the history line examined by the previous execution of the command will be reexamined. Note that negative numbers should be preceded by a `--' argument to avoid confusing them with options.
If two arguments are given, the second specifies the word on the command line in normal array index notation (as a more natural alternative to the numeric argument). Hence 1 is the first word, and -1 (the default) is the last word.
If a third argument is given, its value is ignored, but it is used to signify that the history offset is relative to the current history line, rather than the one remembered after the previous invocations of insert-last-word.
For example, the default behaviour of the command corresponds to
zle insert-last-word -- -1 -1
while the command
zle insert-last-word -- -1 1 -
always copies the first word of the line in the history immediately before the line being edited. This has the side effect that later invocations of the widget will be relative to that line.
If called from a function by the zle command with arguments, the first argument is taken as the string for which to search, rather than the first word in the buffer.
For compatibility with vi, if the command is vi-forward-word or vi-forward-blank-word, the whitespace after the word is not included. If you prefer the more consistent behaviour with the whitespace included use the following key binding:
bindkey -a -s cw dwi
If called from a ZLE widget function in the form `zle copy-region-as-kill string' then string will be taken as the text to copy to the kill buffer. The cursor, the mark and the text on the command line are not used in this case.
With a positive numeric argument N, the word around the cursor, or following it if the cursor is between words, is transposed with the preceding N words. The cursor is put at the end of the resulting group of words.
With a negative numeric argument -N, the effect is the same as using a positive argument N except that the original cursor position is retained, regardless of how the words are rearranged.
Inside a widget function, a call to this function treats the last key of the key sequence which called the widget as the digit.
Inside a widget function, if passed an argument, i.e. `zle universal-argument num', the numeric argument will be set to num; this is equivalent to `NUMERIC=num'.
The function can be used with a command argument inside a user-defined widget. The following code sets the base to 16 and lets the user input a hexadecimal argument until a key out of the digit range is typed:
zle argument-base 16 zle universal-argument
This widget is intended to be called from user-defined widgets to enforce a desired suffix-removal behavior.
This widget is intended to be called from user-defined widgets to enforce a desired suffix-preservation behavior.
When invoked interactively, the pasted text is inserted to the buffer and placed in the cutbuffer. If a numeric argument is given, shell quoting will be applied to the pasted text before it is inserted.
When a named buffer is specified with vi-set-buffer ("x), the pasted text is stored in that named buffer but not inserted.
When called from a widget function as `bracketed-paste name`, the pasted text is assigned to the variable name and no other processing is done.
See also the zle_bracketed_paste parameter.
A restricted set of editing functions is available in the mini-buffer. Keys are looked up in the special command keymap, and if not found there in the main keymap. An interrupt signal, as defined by the stty setting, will abort the function. Note that the following always perform the same task within the executed-named-cmd environment and cannot be replaced by user defined widgets, nor can the set of functions be extended. The allowed functions are: backward-delete-char, vi-backward-delete-char, clear-screen, redisplay, quoted-insert, vi-quoted-insert, backward-kill-word, vi-backward-kill-word, kill-whole-line, vi-kill-line, backward-kill-line, list-choices, delete-char-or-list, complete-word, accept-line, expand-or-complete and expand-or-complete-prefix.
kill-region kills the last word, and vi-cmd-mode is treated the same as accept-line. The space and tab characters, if not bound to one of these functions, will complete the name and then list the possibilities if the AUTO_LIST option is set. Any other character that is not bound to self-insert or self-insert-unmeta will beep and be ignored. The bindings of the current insert mode will be used.
Currently this command may not be redefined or called by name.
Like execute-named-cmd, this command may not be redefined, but it may be called by name.
The following widget, caps-lock, serves as an example.
self-insert-ucase() {
LBUFFER+=${(U)KEYS[-1]}
}
integer stat
zle -N self-insert self-insert-ucase
zle -A caps-lock save-caps-lock
zle -A accept-line caps-lock
zle recursive-edit
stat=$?
zle -A .self-insert self-insert
zle -A save-caps-lock caps-lock
zle -D save-caps-lock
(( stat )) && zle send-break
return $stat
This causes typed letters to be inserted capitalised until either accept-line (i.e. typically the return key) is typed or the caps-lock widget is invoked again; the later is handled by saving the old definition of caps-lock as save-caps-lock and then rebinding it to invoke accept-line. Note that an error from the recursive edit is detected as a non-zero return status and propagated by using the send-break widget.
Otherwise, the prompt is only expanded each time zle starts, and when the display has been interrupted by output from another part of the shell (such as a job notification) which causes the command line to be reprinted.
reset-prompt doesn't alter the special parameter LASTWIDGET.
When a buffer is specified for a cut, change or yank command, the text concerned replaces the previous contents of the specified buffer. If a named buffer is specified using a capital, the newly cut text is appended to the buffer instead of overwriting it. When using the "_ buffer, nothing happens. This can be useful for deleting text without affecting any buffers.
If no buffer is specified for a cut or change command, "1 is used, and the contents of "1 to "8 are each shifted along one buffer; the contents of "9 is lost. If no buffer is specified for a yank command, "0 is used. Finally, a paste command without a specified buffer will paste the text from the most recent command regardless of any buffer that might have been used with that command.
When called from a widget function by the zle command, the buffer can optionally be specified with an argument. For example,
zle vi-set-buffer A
Note that when invoked from vi command mode, the full prior change made in insert mode is reverted, the changes having been merged when command mode was selected.
The line editor has the ability to highlight characters or regions of the line that have a particular significance. This is controlled by the array parameter zle_highlight, if it has been set by the user.
If the parameter contains the single entry none all highlighting is turned off. Note the parameter is still expected to be an array.
Otherwise each entry of the array should consist of a word indicating a context for highlighting, then a colon, then a comma-separated list of the types of highlighting to apply in that context.
The contexts available for highlighting are the following:
When region_highlight is set, the contexts that describe a region -- isearch, region, suffix, and paste -- are applied first, then region_highlight is applied, then the remaining zle_highlight contexts are applied. If a particular character is affected by multiple specifications, the last specification wins.
zle_highlight may contain additional fields for controlling how terminal sequences to change colours are output. Each of the following is followed by a colon and a string in the same form as for key bindings. This will not be necessary for the vast majority of terminals as the defaults shown in parentheses are widely used.
The available types of highlighting are the following. Note that not all types of highlighting are available on all terminals:
Not all terminals support this and, of those that do, not all provide facilities to test the support, hence the user should decide based on the terminal type. Most terminals support the colours black, red, green, yellow, blue, magenta, cyan and white, which can be set by name. In addition. default may be used to set the terminal's default foreground colour. Abbreviations are allowed; b or bl selects black. Some terminals may generate additional colours if the bold attribute is also present.
On recent terminals and on systems with an up-to-date terminal database the number of colours supported may be tested by the command `echotc Co'; if this succeeds, it indicates a limit on the number of colours which will be enforced by the line editor. The number of colours is in any case limited to 256 (i.e. the range 0 to 255).
Some modern terminal emulators have support for 24-bit true colour (16 million colours). In this case, the hex triplet format can be used. This consists of a `#' followed by either a three or six digit hexadecimal number describing the red, green and blue components of the colour. Hex triplets can also be used with 88 and 256 colour terminals via the zsh/nearcolor module (see zshmodules(1)).
Colour is also known as color.
The characters described above as `special' are as follows. The formatting described here is used irrespective of whether the characters are highlighted:
Unprintable multibyte characters are shown as a hexadecimal number between angle brackets. The number is the code point of the character in the wide character set; this may or may not be Unicode, depending on the operating system.
Not all systems support this: for it to work, the system's representation of wide characters must be code values from the Universal Character Set, as defined by IS0 10646 (also known as Unicode).
If zle_highlight is not set or no value applies to a particular context, the defaults applied are equivalent to
zle_highlight=(region:standout special:standout suffix:bold isearch:underline paste:standout)
i.e. both the region and special characters are shown in standout mode.
Within widgets, arbitrary regions may be highlighted by setting the special array parameter region_highlight; see above.
Completion widgets are defined by the -C option to the zle builtin command provided by the zsh/zle module (see zshzle(1)). For example,
zle -C complete expand-or-complete completer
defines a widget named `complete'. The second argument is the name of any of the builtin widgets that handle completions: complete-word, expand-or-complete, expand-or-complete-prefix, menu-complete, menu-expand-or-complete, reverse-menu-complete, list-choices, or delete-char-or-list. Note that this will still work even if the widget in question has been re-bound.
When this newly defined widget is bound to a key using the bindkey builtin command defined in the zsh/zle module (see zshzle(1)), typing that key will call the shell function `completer'. This function is responsible for generating the possible matches using the builtins described below. As with other ZLE widgets, the function is called with its standard input closed.
Once the function returns, the completion code takes over control again and treats the matches in the same manner as the specified builtin widget, in this case expand-or-complete.
The parameters ZLE_REMOVE_SUFFIX_CHARS and ZLE_SPACE_SUFFIX_CHARS are used by the completion mechanism, but are not special. See Parameters Used By The Shell in zshparam(1).
Inside completion widgets, and any functions called from them, some parameters have special meaning; outside these functions they are not special to the shell in any way. These parameters are used to pass information between the completion code and the completion widget. Some of the builtin commands and the condition codes use or change the current values of these parameters. Any existing values will be hidden during execution of completion widgets; except for compstate, the parameters are reset on each function exit (including nested function calls from within the completion widget) to the values they had when the function was entered.
IPREFIX=${PREFIX%%\=*}=
PREFIX=${PREFIX#*=}
causes the part of the prefix up to and including the first equal sign not to be treated as part of a matched string. This can be done automatically by the compset builtin, see below.
If it was set when at least one match equal to the string on the line was generated, the match is accepted.
On exit it may be set to any of the values above (where setting it to the empty string is the same as unsetting it), or to a number, in which case the match whose number is given will be inserted into the command line. Negative numbers count backward from the last match (with `-1' selecting the last match) and out-of-range values are wrapped around, so that a value of zero selects the last match and a value one more than the maximum selects the first. Unless the value of this key ends in a space, the match is inserted as in a menu completion, i.e. without automatically appending a space.
Both menu and automenu may also specify the number of the match to insert, given after a colon. For example, `menu:2' says to start menu completion, beginning with the second match.
Note that a value containing the substring `tab' makes the matches generated be ignored and only the TAB be inserted.
Finally, it may also be set to all, which makes all matches generated be inserted into the line.
If the substring force appears in the value, this makes the list be shown even if there is only one match. Normally, the list would be shown only if there are at least two matches.
The value contains the substring packed if the LIST_PACKED option is set. If this substring is given for all matches added to a group, this group will show the LIST_PACKED behavior. The same is done for the LIST_ROWS_FIRST option with the substring rows.
Finally, if the value contains the string explanations, only the explanation strings, if any, will be listed and if it contains messages, only the messages (added with the -x option of compadd) will be listed. If it contains both explanations and messages both kinds of explanation strings will be listed. It will be set appropriately on entry to a completion widget and may be changed there.
As with old_list, the value of this key will only be used if it is the string keep. If it was set to this value by the widget and there was an old match inserted into the command line, this match will be kept and if the value of the insert key specifies that another match should be inserted, this will be inserted after the old one.
After the widget has exited the value of this key is only used if it was set to keep. In this case the completion code will continue to use this old list. If the widget generated new matches, they will not be used.
Note that the matcher specifications given to the compadd builtin command are not used if this is set to a non-empty string.
On exit, it may be set to single as above. It may also be set to always, or to the empty string or unset; in those cases the cursor will be moved to the end of the string always or never respectively. Any other string is treated as match.
This builtin command can be used to add matches directly and control all the information the completion code stores with each possible match. The return status is zero if at least one match was added and non-zero if no matches were added.
The completion code breaks the string to complete into seven fields in the order:
<ipre><apre><hpre><word><hsuf><asuf><isuf>
The first field is an ignored prefix taken from the command line, the contents of the IPREFIX parameter plus the string given with the -i option. With the -U option, only the string from the -i option is used. The field <apre> is an optional prefix string given with the -P option. The <hpre> field is a string that is considered part of the match but that should not be shown when listing completions, given with the -p option; for example, functions that do filename generation might specify a common path prefix this way. <word> is the part of the match that should appear in the list of completions, i.e. one of the words given at the end of the compadd command line. The suffixes <hsuf>, <asuf> and <isuf> correspond to the prefixes <hpre>, <apre> and <ipre> and are given by the options -s, -S and -I, respectively.
The supported flags are:
If there are fewer display strings than words, the leftover words will be displayed unchanged and if there are more display strings than words, the leftover display strings will be silently ignored.
Within the explanation, the following sequences may be used to specify output attributes as described in the section EXPANSION OF PROMPT SEQUENCES in zshmisc(1): `%B', `%S', `%U', `%F', `%K' and their lower case counterparts, as well as `%{...%}'. `%F', `%K' and `%{...%}' take arguments in the same form as prompt expansion. (Note that the sequence `%G' is not available; an argument to `%{' should be used instead.) The sequence `%%' produces a literal `%'.
These sequences are most often employed by users when customising the format style (see zshcompsys(1)), but they must also be taken into account when writing completion functions, as passing descriptions with unescaped `%' characters to utility functions such as _arguments and _message may produce unexpected results. If arbitrary text is to be passed in a description, it can be escaped using e.g. ${my_str//\%/%%}.
This option may also be used without the -S option; then any automatically added space will be removed when one of the characters in the list is typed.
The array may be the name of an array parameter or a list of literal patterns enclosed in parentheses and quoted, as in `-F "(*?.o *?.h)"'. If the name of an array is given, the elements of the array are taken as the patterns.
Except for the -M flag, if any of these flags is given more than once, the first one (and its argument) will be used.
The options are:
Without the optional number, the longest match is taken, but if number is given, anything up to the numberth match is moved. If the number is negative, the numberth longest match is moved. For example, if PREFIX contains the string `a=b=c', then compset -P '*\=' will move the string `a=b=' into the IPREFIX parameter, but compset -P 1 '*\=' will move only the string `a='.
If the optional end is given, the modification is done only if the current word position is also less than or equal to end. In this case, the words from position end onwards are also removed from the words array.
Both begin and end may be negative to count backwards from the last element of the words array.
If the optional pattern end-pat is also given, and there is an element in the words array matching this pattern, the parameters are modified only if the index of this word is higher than the one given by the CURRENT parameter (so that the matching word has to be after the cursor). In this case, the words starting with the one matching end-pat are also removed from the words array. If words contains no word matching end-pat, the testing and modification is performed as if it were not given.
In all the above cases the return status is zero if the test succeeded and the parameters were modified and non-zero otherwise. This allows one to use this builtin in tests such as:
if compset -P '*\='; then ...
This forces anything up to and including the last equal sign to be ignored by the completion code.
The return status can be used to test if a matching compctl definition was found. It is non-zero if a compctl was found and zero otherwise.
Note that this builtin is defined by the zsh/compctl module.
The following additional condition codes for use within the [[ ... ]] construct are available in completion widgets. These work on the special parameters. All of these tests can also be performed by the compset builtin, but in the case of the condition codes the contents of the special parameters are not modified.
It is possible by use of the -M option of the compadd builtin command to specify how the characters in the string to be completed (referred to here as the command line) map onto the characters in the list of matches produced by the completion code (referred to here as the trial completions). Note that this is not used if the command line contains a glob pattern and the GLOB_COMPLETE option is set or the pattern_match of the compstate special association is set to a non-empty string.
The match-spec given as the argument to the -M option (see `Completion Builtin Commands' above) consists of one or more matching descriptions separated by whitespace. Each description consists of a letter followed by a colon and then the patterns describing which character sequences on the line match which character sequences in the trial completion. Any sequence of characters not handled in this fashion must match exactly, as usual.
The forms of match-spec understood are as follows. In each case, the form with an upper case initial character retains the string already typed on the command line as the final result of completion, while with a lower case initial character the string on the command line is changed into the corresponding part of the trial completion.
If no lpat is given but a ranchor is, this matches the gap between substrings matched by lanchor and ranchor. Unlike lanchor, the ranchor only needs to match the trial completion string.
The b and B forms are similar to l and L with an empty anchor, but need to match only the beginning of the word on the command line or trial completion, respectively.
Each lpat, tpat or anchor is either an empty string or consists of a sequence of literal characters (which may be quoted with a backslash), question marks, character classes, and correspondence classes; ordinary shell patterns are not used. Literal characters match only themselves, question marks match any character, and character classes are formed as for globbing and match any character in the given set.
Correspondence classes are defined like character classes, but with two differences: they are delimited by a pair of braces, and negated classes are not allowed, so the characters ! and ^ have no special meaning directly after the opening brace. They indicate that a range of characters on the line match a range of characters in the trial completion, but (unlike ordinary character classes) paired according to the corresponding position in the sequence. For example, to make any ASCII lower case letter on the line match the corresponding upper case letter in the trial completion, you can use `m:{a-z}={A-Z}' (however, see below for the recommended form for this). More than one pair of classes can occur, in which case the first class before the = corresponds to the first after it, and so on. If one side has more such classes than the other side, the superfluous classes behave like normal character classes. In anchor patterns correspondence classes also behave like normal character classes.
The standard `[:name:]' forms described for standard shell patterns (see the section FILENAME GENERATION in zshexpn(1)) may appear in correspondence classes as well as normal character classes. The only special behaviour in correspondence classes is if the form on the left and the form on the right are each one of [:upper:], [:lower:]. In these cases the character in the word and the character on the line must be the same up to a difference in case. Hence to make any lower case character on the line match the corresponding upper case character in the trial completion you can use `m:{[:lower:]}={[:upper:]}'. Although the matching system does not yet handle multibyte characters, this is likely to be a future extension, at which point this syntax will handle arbitrary alphabets; hence this form, rather than the use of explicit ranges, is the recommended form. In other cases `[:name:]' forms are allowed. If the two forms on the left and right are the same, the characters must match exactly. In remaining cases, the corresponding tests are applied to both characters, but they are not otherwise constrained; any matching character in one set goes with any matching character in the other set: this is equivalent to the behaviour of ordinary character classes.
The pattern tpat may also be one or two stars, `*' or `**'. This means that the pattern on the command line can match any number of characters in the trial completion. In this case the pattern must be anchored (on either side); in the case of a single star, the anchor then determines how much of the trial completion is to be included -- only the characters up to the next appearance of the anchor will be matched. With two stars, substrings matched by the anchor can be matched, too.
Examples:
The keys of the options association defined by the parameter module are the option names in all-lower-case form, without underscores, and without the optional no at the beginning even though the builtins setopt and unsetopt understand option names with upper case letters, underscores, and the optional no. The following alters the matching rules so that the prefix no and any underscore are ignored when trying to match the trial completions generated and upper case letters on the line match the corresponding lower case letters in the words:
compadd -M 'L:|[nN][oO]= M:_= M:{[:upper:]}={[:lower:]}' - \
${(k)options}
The first part says that the pattern `[nN][oO]' at the beginning (the empty anchor before the pipe symbol) of the string on the line matches the empty string in the list of words generated by completion, so it will be ignored if present. The second part does the same for an underscore anywhere in the command line string, and the third part uses correspondence classes so that any upper case letter on the line matches the corresponding lower case letter in the word. The use of the upper case forms of the specification characters (L and M) guarantees that what has already been typed on the command line (in particular the prefix no) will not be deleted.
Note that the use of L in the first part means that it matches only when at the beginning of both the command line string and the trial completion. I.e., the string `_NO_f' would not be completed to `_NO_foo', nor would `NONO_f' be completed to `NONO_foo' because of the leading underscore or the second `NO' on the line which makes the pattern fail even though they are otherwise ignored. To fix this, one would use `B:[nN][oO]=' instead of the first part. As described above, this matches at the beginning of the trial completion, independent of other characters or substrings at the beginning of the command line word which are ignored by the same or other match-specs.
The second example makes completion case insensitive. This is just the same as in the option example, except here we wish to retain the characters in the list of completions:
compadd -M 'm:{[:lower:]}={[:upper:]}' ...
This makes lower case letters match their upper case counterparts. To make upper case letters match the lower case forms as well:
compadd -M 'm:{[:lower:][:upper:]}={[:upper:][:lower:]}' ...
A nice example for the use of * patterns is partial word completion. Sometimes you would like to make strings like `c.s.u' complete to strings like `comp.source.unix', i.e. the word on the command line consists of multiple parts, separated by a dot in this example, where each part should be completed separately -- note, however, that the case where each part of the word, i.e. `comp', `source' and `unix' in this example, is to be completed from separate sets of matches is a different problem to be solved by the implementation of the completion widget. The example can be handled by:
compadd -M 'r:|.=* r:|=*' \ - comp.sources.unix comp.sources.misc ...
The first specification says that lpat is the empty string, while anchor is a dot; tpat is *, so this can match anything except for the `.' from the anchor in the trial completion word. So in `c.s.u', the matcher sees `c', followed by the empty string, followed by the anchor `.', and likewise for the second dot, and replaces the empty strings before the anchors, giving `c[omp].s[ources].u[nix]', where the last part of the completion is just as normal.
With the pattern shown above, the string `c.u' could not be completed to `comp.sources.unix' because the single star means that no dot (matched by the anchor) can be skipped. By using two stars as in `r:|.=**', however, `c.u' could be completed to `comp.sources.unix'. This also shows that in some cases, especially if the anchor is a real pattern, like a character class, the form with two stars may result in more matches than one would like.
The second specification is needed to make this work when the cursor is in the middle of the string on the command line and the option COMPLETE_IN_WORD is set. In this case the completion code would normally try to match trial completions that end with the string as typed so far, i.e. it will only insert new characters at the cursor position rather than at the end. However in our example we would like the code to recognise matches which contain extra characters after the string on the line (the `nix' in the example). Hence we say that the empty string at the end of the string on the line matches any characters at the end of the trial completion.
More generally, the specification
compadd -M 'r:|[.,_-]=* r:|=*' ...
allows one to complete words with abbreviations before any of the characters in the square brackets. For example, to complete veryverylongfile.c rather than veryverylongheader.h with the above in effect, you can just type very.c before attempting completion.
The specifications with both a left and a right anchor are useful to complete partial words whose parts are not separated by some special character. For example, in some places strings have to be completed that are formed `LikeThis' (i.e. the separate parts are determined by a leading upper case letter) or maybe one has to complete strings with trailing numbers. Here one could use the simple form with only one anchor as in:
compadd -M 'r:|[[:upper:]0-9]=* r:|=*' LikeTHIS FooHoo 5foo123 5bar234
But with this, the string `H' would neither complete to `FooHoo' nor to `LikeTHIS' because in each case there is an upper case letter before the `H' and that is matched by the anchor. Likewise, a `2' would not be completed. In both cases this could be changed by using `r:|[[:upper:]0-9]=**', but then `H' completes to both `LikeTHIS' and `FooHoo' and a `2' matches the other strings because characters can be inserted before every upper case letter and digit. To avoid this one would use:
compadd -M 'r:[^[:upper:]0-9]||[[:upper:]0-9]=** r:|=*' \
LikeTHIS FooHoo foo123 bar234
By using these two anchors, a `H' matches only upper case `H's that are immediately preceded by something matching the left anchor `[^[:upper:]0-9]'. The effect is, of course, that `H' matches only the string `FooHoo', a `2' matches only `bar234' and so on.
When using the completion system (see zshcompsys(1)), users can define match specifications that are to be used for specific contexts by using the matcher and matcher-list styles. The values for the latter will be used everywhere.
The first step is to define the widget:
zle -C complete complete-word complete-files
Then the widget can be bound to a key using the bindkey builtin command:
bindkey '^X\t' complete
After that the shell function complete-files will be invoked after typing control-X and TAB. The function should then generate the matches, e.g.:
complete-files () { compadd - * }
This function will complete files in the current directory matching the current word.
This describes the shell code for the `new' completion system, referred to as compsys. It is written in shell functions based on the features described in zshcompwid(1).
The features are contextual, sensitive to the point at which completion is started. Many completions are already provided. For this reason, a user can perform a great many tasks without knowing any details beyond how to initialize the system, which is described below in INITIALIZATION.
The context that decides what completion is to be performed may be
A full context specification contains other elements, as we shall describe.
Besides commands names and contexts, the system employs two more concepts, styles and tags. These provide ways for the user to configure the system's behaviour.
Tags play a dual role. They serve as a classification system for the matches, typically indicating a class of object that the user may need to distinguish. For example, when completing arguments of the ls command the user may prefer to try files before directories, so both of these are tags. They also appear as the rightmost element in a context specification.
Styles modify various operations of the completion system, such as output formatting, but also what kinds of completers are used (and in what order), or which tags are examined. Styles may accept arguments and are manipulated using the zstyle command described in see zshmodules(1).
In summary, tags describe what the completion objects are, and style how they are to be completed. At various points of execution, the completion system checks what styles and/or tags are defined for the current context, and uses that to modify its behavior. The full description of context handling, which determines how tags and other elements of the context influence the behaviour of styles, is described below in COMPLETION SYSTEM CONFIGURATION.
When a completion is requested, a dispatcher function is called; see the description of _main_complete in the list of control functions below. This dispatcher decides which function should be called to produce the completions, and calls it. The result is passed to one or more completers, functions that implement individual completion strategies: simple completion, error correction, completion with error correction, menu selection, etc.
More generally, the shell functions contained in the completion system are of two types:
If the system was installed completely, it should be enough to call the shell function compinit from your initialization file; see the next section. However, the function compinstall can be run by a user to configure various aspects of the completion system.
Usually, compinstall will insert code into .zshrc, although if that is not writable it will save it in another file and tell you that file's location. Note that it is up to you to make sure that the lines added to .zshrc are actually run; you may, for example, need to move them to an earlier place in the file if .zshrc usually returns early. So long as you keep them all together (including the comment lines at the start and finish), you can rerun compinstall and it will correctly locate and modify these lines. Note, however, that any code you add to this section by hand is likely to be lost if you rerun compinstall, although lines using the command `zstyle' should be gracefully handled.
The new code will take effect next time you start the shell, or run .zshrc by hand; there is also an option to make them take effect immediately. However, if compinstall has removed definitions, you will need to restart the shell to see the changes.
To run compinstall you will need to make sure it is in a directory mentioned in your fpath parameter, which should already be the case if zsh was properly configured as long as your startup files do not remove the appropriate directories from fpath. Then it must be autoloaded (`autoload -U compinstall' is recommended). You can abort the installation any time you are being prompted for information, and your .zshrc will not be altered at all; changes only take place right at the end, where you are specifically asked for confirmation.
This section describes the use of compinit to initialize completion for the current session when called directly; if you have run compinstall it will be called automatically from your .zshrc.
To initialize the system, the function compinit should be in a directory mentioned in the fpath parameter, and should be autoloaded (`autoload -U compinit' is recommended), and then run simply as `compinit'. This will define a few utility functions, arrange for all the necessary shell functions to be autoloaded, and will then re-define all widgets that do completion to use the new system. If you use the menu-select widget, which is part of the zsh/complist module, you should make sure that that module is loaded before the call to compinit so that that widget is also re-defined. If completion styles (see below) are set up to perform expansion as well as completion by default, and the TAB key is bound to expand-or-complete, compinit will rebind it to complete-word; this is necessary to use the correct form of expansion.
Should you need to use the original completion commands, you can still bind keys to the old widgets by putting a `.' in front of the widget name, e.g. `.expand-or-complete'.
To speed up the running of compinit, it can be made to produce a dumped configuration that will be read in on future invocations; this is the default, but can be turned off by calling compinit with the option -D. The dumped file is .zcompdump in the same directory as the startup files (i.e. $ZDOTDIR or $HOME); alternatively, an explicit file name can be given by `compinit -d dumpfile'. The next invocation of compinit will read the dumped file instead of performing a full initialization.
If the number of completion files changes, compinit will recognise this and produce a new dump file. However, if the name of a function or the arguments in the first line of a #compdef function (as described below) change, it is easiest to delete the dump file by hand so that compinit will re-create it the next time it is run. The check performed to see if there are new functions can be omitted by giving the option -C. In this case the dump file will only be created if there isn't one already.
The dumping is actually done by another function, compdump, but you will only need to run this yourself if you change the configuration (e.g. using compdef) and then want to dump the new one. The name of the old dumped file will be remembered for this purpose.
If the parameter _compdir is set, compinit uses it as a directory where completion functions can be found; this is only necessary if they are not already in the function search path.
For security reasons compinit also checks if the completion system would use files not owned by root or by the current user, or files in directories that are world- or group-writable or that are not owned by root or by the current user. If such files or directories are found, compinit will ask if the completion system should really be used. To avoid these tests and make all files found be used without asking, use the option -u, and to make compinit silently ignore all insecure files and directories use the option -i. This security check is skipped entirely when the -C option is given.
The security check can be retried at any time by running the function compaudit. This is the same check used by compinit, but when it is executed directly any changes to fpath are made local to the function so they do not persist. The directories to be checked may be passed as arguments; if none are given, compaudit uses fpath and _compdir to find completion system directories, adding missing ones to fpath as necessary. To force a check of exactly the directories currently named in fpath, set _compdir to an empty string before calling compaudit or compinit.
The function bashcompinit provides compatibility with bash's programmable completion system. When run it will define the functions, compgen and complete which correspond to the bash builtins with the same names. It will then be possible to use completion specifications and functions written for bash.
The convention for autoloaded functions used in completion is that they start with an underscore; as already mentioned, the fpath/FPATH parameter must contain the directory in which they are stored. If zsh was properly installed on your system, then fpath/FPATH automatically contains the required directories for the standard functions.
For incomplete installations, if compinit does not find enough files beginning with an underscore (fewer than twenty) in the search path, it will try to find more by adding the directory _compdir to the search path. If that directory has a subdirectory named Base, all subdirectories will be added to the path. Furthermore, if the subdirectory Base has a subdirectory named Core, compinit will add all subdirectories of the subdirectories to the path: this allows the functions to be in the same format as in the zsh source distribution.
When compinit is run, it searches all such files accessible via fpath/FPATH and reads the first line of each of them. This line should contain one of the tags described below. Files whose first line does not start with one of these tags are not considered to be part of the completion system and will not be treated specially.
The tags are:
Each name may also be of the form `cmd=service'. When completing the command cmd, the function typically behaves as if the command (or special context) service was being completed instead. This provides a way of altering the behaviour of functions that can perform many different completions. It is implemented by setting the parameter $service when calling the function; the function may choose to interpret this how it wishes, and simpler functions will probably ignore it.
If the #compdef line contains one of the options -p or -P, the words following are taken to be patterns. The function will be called when completion is attempted for a command or context that matches one of the patterns. The options -p and -P are used to specify patterns to be tried before or after other completions respectively. Hence -P may be used to specify default actions.
The option -N is used after a list following -p or -P; it specifies that remaining words no longer define patterns. It is possible to toggle between the three options as many times as necessary.
When one of the key-sequences is typed, the function in the file will be invoked to generate the matches. Note that a key will not be re-bound if it already was (that is, was bound to something other than undefined-key). The widget created has the same name as the file and can be bound to any other keys using bindkey as usual.
#compdef -K _foo_complete complete-word "^X^C" \ _foo_list list-choices "^X^D"
(all on one line) defines a widget _foo_complete for completion, bound to `^X^C', and a widget _foo_list for listing, bound to `^X^D'.
The # is part of the tag name and no white space is allowed after it. The #compdef tags use the compdef function described below; the main difference is that the name of the function is supplied implicitly.
The special contexts for which completion functions can be defined are:
Default implementations are supplied for each of these contexts. In most cases the context -context- is implemented by a corresponding function _context, for example the context `-tilde-' and the function `_tilde').
The contexts -redirect- and -value- allow extra context-specific information. (Internally, this is handled by the functions for each context calling the function _dispatch.) The extra information is added separated by commas.
For the -redirect- context, the extra information is in the form `-redirect-,op,command', where op is the redirection operator and command is the name of the command on the line. If there is no command on the line yet, the command field will be empty.
For the -value- context, the form is `-value-,name,command', where name is the name of the parameter on the left hand side of the assignment. In the case of elements of an associative array, for example `assoc=(key <TAB>', name is expanded to `name-key'. In certain special contexts, such as completing after `make CFLAGS=', the command part gives the name of the command, here make; otherwise it is empty.
It is not necessary to define fully specific completions as the functions provided will try to generate completions by progressively replacing the elements with `-default-'. For example, when completing after `foo=<TAB>', _value will try the names `-value-,foo,' (note the empty command part), `-value-,foo,-default-' and`-value-,-default-,-default-', in that order, until it finds a function to handle the context.
As an example:
compdef '_files -g "*.log"' '-redirect-,2>,-default-'
completes files matching `*.log' after `2> <TAB>' for any command with no more specific handler defined.
Also:
compdef _foo -value-,-default-,-default-
specifies that _foo provides completions for the values of parameters for which no special function has been defined. This is usually handled by the function _value itself.
The same lookup rules are used when looking up styles (as described below); for example
zstyle ':completion:*:*:-redirect-,2>,*:*' file-patterns '*.log'
is another way to make completion after `2> <TAB>' complete files matching `*.log'.
The following function is defined by compinit and may be called directly.
Alternatively, all the arguments may have the form `cmd=service'. Here service should already have been defined by `cmd1=service' lines in #compdef files, as described above. The argument for cmd will be completed in the same way as service.
The function argument may alternatively be a string containing almost any shell code. If the string contains an equal sign, the above will take precedence. The option -e may be used to specify the first argument is to be evaluated as shell code even if it contains an equal sign. The string will be executed using the eval builtin command to generate completions. This provides a way of avoiding having to define a new completion function. For example, to complete files ending in `.h' as arguments to the command foo:
compdef '_files -g "*.h"' foo
The option -n prevents any completions already defined for the command or context from being overwritten.
The option -d deletes any completion defined for the command or contexts listed.
The names may also contain -p, -P and -N options as described for the #compdef tag. The effect on the argument list is identical, switching between definitions of patterns tried initially, patterns tried finally, and normal commands and contexts.
The parameter $_compskip may be set by any function defined for a pattern context. If it is set to a value containing the substring `patterns' none of the pattern-functions will be called; if it is set to a value containing the substring `all', no other function will be called. Setting $_compskip in this manner is of particular utility when using the -p option, as otherwise the dispatcher will move on to additional functions (likely the default one) after calling the pattern-context one, which can mangle the display of completion possibilities if not handled properly.
The form with -k defines a widget with the same name as the function that will be called for each of the key-sequences; this is like the #compdef -k tag. The function should generate the completions needed and will otherwise behave like the builtin widget whose name is given as the style argument. The widgets usable for this are: complete-word, delete-char-or-list, expand-or-complete, expand-or-complete-prefix, list-choices, menu-complete, menu-expand-or-complete, and reverse-menu-complete, as well as menu-select if the zsh/complist module is loaded. The option -n prevents the key being bound if it is already to bound to something other than undefined-key.
The form with -K is similar and defines multiple widgets based on the same function, each of which requires the set of three arguments name, style and key-sequence, where the latter two are as for -k and the first must be a unique widget name beginning with an underscore.
Wherever applicable, the -a option makes the function autoloadable, equivalent to autoload -U function.
The function compdef can be used to associate existing completion functions with new commands. For example,
compdef _pids foo
uses the function _pids to complete process IDs for the command foo.
Note also the _gnu_generic function described below, which can be used to complete options for commands that understand the `--help' option.
This section gives a short overview of how the completion system works, and then more detail on how users can configure how and when matches are generated.
When completion is attempted somewhere on the command line the completion system begins building the context. The context represents everything that the shell knows about the meaning of the command line and the significance of the cursor position. This takes account of a number of things including the command word (such as `grep' or `zsh') and options to which the current word may be an argument (such as the `-o' option to zsh which takes a shell option as an argument).
The context starts out very generic ("we are beginning a completion") and becomes more specific as more is learned ("the current word is in a position that is usually a command name" or "the current word might be a variable name" and so on). Therefore the context will vary during the same call to the completion system.
This context information is condensed into a string consisting of multiple fields separated by colons, referred to simply as `the context' in the remainder of the documentation. Note that a user of the completion system rarely needs to compose a context string, unless for example a new function is being written to perform completion for a new command. What a user may need to do is compose a style pattern, which is matched against a context when needed to look up context-sensitive options that configure the completion system.
The next few paragraphs explain how a context is composed within the completion function suite. Following that is discussion of how styles are defined. Styles determine such things as how the matches are generated, similarly to shell options but with much more control. They are defined with the zstyle builtin command (see zshmodules(1)).
The context string always consists of a fixed set of fields, separated by colons and with a leading colon before the first. Fields which are not yet known are left empty, but the surrounding colons appear anyway. The fields are always in the order :completion:function:completer:command:argument:tag. These have the following meaning:
The context is gradually put together as the functions are executed, starting with the main entry point, which adds :completion: and the function element if necessary. The completer then adds the completer element. The contextual completion adds the command and argument options. Finally, the tag is added when the types of completion are known. For example, the context name
:completion::complete:dvips:option-o-1:files
says that normal completion was attempted as the first argument to the option -o of the command dvips:
dvips -o ...
and the completion function will generate filenames.
Usually completion will be tried for all possible tags in an order given by the completion function. However, this can be altered by using the tag-order style. Completion is then restricted to the list of given tags in the given order.
The _complete_help bindable command shows all the contexts and tags available for completion at a particular point. This provides an easy way of finding information for tag-order and other styles. It is described in the section `Bindable Commands' below.
When looking up styles the completion system uses full context names, including the tag. Looking up the value of a style therefore consists of two things: the context, which is matched to the most specific (best fitting) style pattern, and the name of the style itself, which must be matched exactly. The following examples demonstrate that style patterns may be loosely defined for styles that apply broadly, or as tightly defined as desired for styles that apply in narrower circumstances.
For example, many completion functions can generate matches in a simple and a verbose form and use the verbose style to decide which form should be used. To make all such functions use the verbose form, put
zstyle ':completion:*' verbose yes
in a startup file (probably .zshrc). This gives the verbose style the value yes in every context inside the completion system, unless that context has a more specific definition. It is best to avoid giving the context as `*' in case the style has some meaning outside the completion system.
Many such general purpose styles can be configured simply by using the compinstall function.
A more specific example of the use of the verbose style is by the completion for the kill builtin. If the style is set, the builtin lists full job texts and process command lines; otherwise it shows the bare job numbers and PIDs. To turn the style off for this use only:
zstyle ':completion:*:*:kill:*:*' verbose no
For even more control, the style can use one of the tags `jobs' or `processes'. To turn off verbose display only for jobs:
zstyle ':completion:*:*:kill:*:jobs' verbose no
The -e option to zstyle even allows completion function code to appear as the argument to a style; this requires some understanding of the internals of completion functions (see see zshcompwid(1))). For example,
zstyle -e ':completion:*' hosts 'reply=($myhosts)'
This forces the value of the hosts style to be read from the variable myhosts each time a host name is needed; this is useful if the value of myhosts can change dynamically. For another useful example, see the example in the description of the file-list style below. This form can be slow and should be avoided for commonly examined styles such as menu and list-rows-first.
Note that the order in which styles are defined does not matter; the style mechanism uses the most specific possible match for a particular style to determine the set of values. More precisely, strings are preferred over patterns (for example, `:completion::complete:::foo' is more specific than `:completion::complete:::*'), and longer patterns are preferred over shorter patterns.
A good rule of thumb is that any completion style pattern that needs to include more than one wildcard (*) and that does not end in a tag name, should include all six colons (:), possibly surrounding additional wildcards.
Style names like those of tags are arbitrary and depend on the completion function. However, the following two sections list some of the most common tags and styles.
Some of the following are only used when looking up particular styles and do not refer to a type of match.
Note that the values of several of these styles represent boolean values. Any of the strings `true', `on', `yes', and `1' can be used for the value `true' and any of the strings `false', `off', `no', and `0' for the value `false'. The behavior for any other value is undefined except where explicitly mentioned. The default value may be either `true' or `false' if the style is not set.
Some of these styles are tested first for every possible tag corresponding to a type of match, and if no style was found, for the default tag. The most notable styles of this type are menu, list-colors and styles controlling completion listing such as list-packed and last-prompt. When tested for the default tag, only the function field of the context will be set so that a style using the default tag will normally be defined along the lines of:
zstyle ':completion:*:default' menu ...
When completing pathnames (where the tag used is `paths') this style accepts any number of patterns as the value in addition to the boolean values. Pathnames matching one of these patterns will be accepted immediately even if the command line contains some more partially typed pathname components and these match no file under the directory accepted.
This style is also used by the _expand completer to decide if words beginning with a tilde or parameter expansion should be expanded. For example, if there are parameters foo and foobar, the string `$foo' will only be expanded if accept-exact is set to `true'; otherwise the completion system will be allowed to complete $foo to $foobar. If the style is set to `continue', _expand will add the expansion as a match and the completion system will also be allowed to continue.
When this style is `true', any prefix of a path that matches an existing directory is accepted without any attempt to complete it further. Hence, in the given example, the path /usr/bin/ is accepted immediately and completion tried in that directory.
This style is also useful when completing after directories that magically appear when referenced, such as ZFS .zfs directories or NetApp .snapshot directories. When the style is set the shell does not check for the existence of the directory within the parent directory.
If you wish to inhibit this behaviour entirely, set the path-completion style (see below) to `false'.
The _prefix completer uses this style as a simple boolean value to decide if a space should be inserted before the suffix.
The default is to complete lists when the word on the line already contains a colon.
The default value for this style is `_expand _old_list _correct _approximate', i.e. it contains the completers for which a string with all matches will almost never be wanted.
As an example, the completion function for process IDs uses this style with the processes tag to generate the IDs to complete and the list of processes to display (if the verbose style is `true'). The list produced by the command should look like the output of the ps command. The first line is not displayed, but is searched for the string `PID' (or `pid') to find the position of the process IDs in the following lines. If the line does not contain `PID', the first numbers in each of the other lines are taken as the process IDs to complete.
Note that the completion function generally has to call the specified command for each attempt to generate the completion list. Hence care should be taken to specify only commands that take a short time to run, and in particular to avoid any that may never terminate.
Each string may be either the name of a completer function or a string of the form `function:name'. In the first case the completer field of the context will contain the name of the completer without the leading underscore and with all other underscores replaced by hyphens. In the second case the function is the name of the completer to call, but the context will contain the user-defined name in the completer field of the context. If the name starts with a hyphen, the string for the context will be build from the name of the completer function as in the first case with the name appended to it. For example:
zstyle ':completion:*' completer _complete _complete:-foo
Here, completion will call the _complete completer twice, once using `complete' and once using `complete-foo' in the completer field of the context. Normally, using the same completer more than once only makes sense when used with the `functions:name' form, because otherwise the context name will be the same in all calls to the completer; possible exceptions to this rule are the _ignored and _prefix completers.
The default value for this style is `_complete _ignored': only completion will be done, first using the ignored-patterns style and the $fignore array and then without ignoring matches.
zstyle ':completion:*:sudo::' environ \ PATH="/sbin:/usr/sbin:$PATH" HOME="/root"
If one of its values is the string `prefix', the partially typed word from the line will be expanded as far as possible even if trailing parts cannot be completed.
If one of its values is the string `suffix', matching names for components after the first ambiguous one will also be added. This means that the resulting string is the longest unambiguous string possible. However, menu completion can be used to cycle through all matches.
It is important to use a sufficiently restrictive context when specifying fake strings. Note that the styles fake-files and fake-parameters provide additional features when completing files or parameters.
The following shows a way of supplementing any tag with arbitrary data, but having it behave for display purposes like a separate tag. In this example we use the features of the tag-order style to divide the named-directories tag into two when performing completion with the standard completer complete for arguments of cd. The tag named-directories-normal behaves as normal, but the tag named-directories-mine contains a fixed set of directories. This has the effect of adding the match group `extra directories' with the given completions.
zstyle ':completion::complete:cd:*' tag-order \ 'named-directories:-mine:extra\ directories named-directories:-normal:named\ directories *' zstyle ':completion::complete:cd:*:named-directories-mine' \ fake-always mydir1 mydir2 zstyle ':completion::complete:cd:*:named-directories-mine' \ ignored-patterns '*'
This can be useful on systems that support special file systems whose top-level pathnames can not be listed or generated with glob patterns (but see accept-exact-dirs for a more general way of dealing with this problem). It can also be used for directories for which one does not have read permission.
The pattern form can be used to add a certain `magic' entry to all directories on a particular file system.
zmodload -i zsh/stat disable stat
The style may either be set to a `true' value (or `all'), or one of the values `insert' or `list', indicating that files are to be listed in long format in all circumstances, or when attempting to insert a file name, or when listing file names without attempting to insert one.
More generally, the value may be an array of any of the above values, optionally followed by =num. If num is present it gives the maximum number of matches for which long listing style will be used. For example,
zstyle ':completion:*' file-list list=20 insert=10
specifies that long format will be used when listing up to 20 files or inserting a file with up to 10 matches (assuming a listing is to be shown at all, for example on an ambiguous completion), else short format will be used.
zstyle -e ':completion:*' file-list \
'(( ${+NUMERIC} )) && reply=(true)'
specifies that long format will be used any time a numeric argument is supplied, else short format.
The file-patterns style provides alternatives to the default tags, which are not used. Its value consists of elements of the form `pattern:tag'; each string may contain any number of such specifications separated by spaces.
The pattern is a pattern that is to be used to generate filenames. Any occurrence of the sequence `%p' is replaced by any pattern(s) passed by the function calling _files. Colons in the pattern must be preceded by a backslash to make them distinguishable from the colon before the tag. If more than one pattern is needed, the patterns can be given inside braces, separated by commas.
The tags of all strings in the value will be offered by _files and used when looking up other styles. Any tags in the same word will be offered at the same time and before later words. If no `:tag' is given the `files' tag will be used.
The tag may also be followed by an optional second colon and a description, which will be used for the `%d' in the value of the format style (if that is set) instead of the default description supplied by the completion function. If the description given here contains itself a `%d', that is replaced with the description supplied by the completion function.
For example, to make the rm command first complete only names of object files and then the names of all files if there is no matching object file:
zstyle ':completion:*:*:rm:*:*' file-patterns \
'*.o:object-files' '%p:all-files'
To alter the default behaviour of file completion -- offer files matching a pattern and directories on the first attempt, then all files -- to offer only matching files on the first attempt, then directories, and finally all files:
zstyle ':completion:*' file-patterns \
'%p:globbed-files' '*(-/):directories' '*:all-files'
This works even where there is no special pattern: _files matches all files using the pattern `*' at the first step and stops when it sees this pattern. Note also it will never try a pattern more than once for a single completion attempt.
During the execution of completion functions, the EXTENDED_GLOB option is in effect, so the characters `#', `~' and `^' have special meanings in the patterns.
This style is tested for the default tag as well as for each tag valid for the current completion. Hence the listing can be forced only for certain types of match.
The style is tested with each tag valid for the current completion before it is tested for the descriptions tag. Hence different format strings can be defined for different types of match.
Note also that some completer functions define additional `%'-sequences. These are described for the completer functions that make use of them.
Some completion functions display messages that may be customised by setting this style for the messages tag. Here, the `%d' is replaced with a message given by the completion function.
Finally, the format string is looked up with the warnings tag, for use when no matches could be generated at all. In this case the `%d' is replaced with the descriptions for the matches that were expected separated by spaces. The sequence `%D' is replaced with the same descriptions separated by newlines.
It is possible to use printf-style field width specifiers with `%d' and similar escape sequences. This is handled by the zformat builtin command from the zsh/zutil module, see zshmodules(1).
zstyle ':completion:*:*:-command-:*:commands' \
group-name commands
zstyle ':completion:*:*:-command-:*:functions' \
group-name functions
As a consequence, any match with the same tag will be displayed in the same group.
If the name given is the empty string the name of the tag for the matches will be used as the name of the group. So, to have all different types of matches displayed separately, one can just set:
zstyle ':completion:*' group-name ''
All matches for which no group name is defined will be put in a group named -default-.
For example, to have names of builtin commands, shell functions and external commands appear in that order when completing in command position:
zstyle ':completion:*:*:-command-:*:*' group-order \
builtins functions commands
Note that the matches will still be completed; they are just not shown in the list. To avoid having matches considered as possible completions at all, the tag-order style can be modified as described below.
The values `current' and `current-shown' are a bit like the opposite of the accept-exact style: only strings with missing characters will be completed.
Note that you almost certainly don't want to set this to `true' or `other' for a general context such as `:completion:*'. This is because it would disallow completion of, for example, options multiple times even if the command in question accepts the option more than once.
In addition, the value may include one or both of:
Excluded values act in a similar fashion to values of the ignored-patterns style, so they can be restored to consideration by the _ignored completer.
Note that the EXTENDED_GLOB option is set during the execution of completion functions, so the characters `#', `~' and `^' have special meanings in the patterns.
If the value of the style is `single', the shell will wait until the user has typed enough to make the command unique before converting the name to an ID; attempts at completion will be unsuccessful until that point. If the value is any other string, menu completion will be started when the string typed by the user is longer than the common prefix to the corresponding IDs.
The value may also contain the substrings `pending' or `pending=val'. In this case, the typed character will be inserted instead of starting completion when there is unprocessed input pending. If a val is given, completion will not be done if there are at least that many characters of unprocessed input. This is often useful when pasting characters into a terminal. Note however, that it relies on the $PENDING special parameter from the zsh/zle module being set properly which is not guaranteed on all platforms.
The default value of this style is `true' except for completion within vared builtin command where it is `false'.
In the case of the _approximate completer, the completer field in the context will already have been set to one of correct-num or approximate-num, where num is the number of errors that were accepted.
In the case of the _match completer, the style may also be set to the string `pattern'. Then the pattern on the line is left unchanged if it does not match unambiguously.
The behaviour of _expand when this style is `true' is to cause _expand to give up when a single expansion with the restored prefix is the same as the original; hence any remaining completers may be called.
If this style is set for the default tag, the strings in the value are taken as specifications that are to be used everywhere. If it is set for other tags, the specifications are used only for matches of the type described by the tag. For this to work best, the group-name style must be set to an empty string.
In addition to setting styles for specific tags, it is also possible to use group names specified explicitly by the group-name tag together with the `(group)' syntax allowed by the ZLS_COLORS and ZLS_COLOURS parameters and simply using the default tag.
It is possible to use any color specifications already set up for the GNU version of the ls command:
zstyle ':completion:*:default' list-colors \
${(s.:.)LS_COLORS}
The default colors are the same as for the GNU ls command and can be obtained by setting the style to an empty string (i.e. '').
The value may contain the escape sequences: `%l' or `%L', which will be replaced by the number of the last line displayed and the total number of lines; `%m' or `%M', the number of the last match shown and the total number of matches; and `%p' and `%P', `Top' when at the beginning of the list, `Bottom' when at the end and the position shown as a percentage of the total length otherwise. In each case the form with the uppercase letter will be replaced by a string of fixed width, padded to the right with spaces, while the lowercase form will be replaced by a variable width string. As in other prompt strings, the escape sequences `%S', `%s', `%B', `%b', `%U', `%u' for entering and leaving the display modes standout, bold and underline, and `%F', `%f', `%K', `%k' for changing the foreground background colour, are also available, as is the form `%{...%}' for enclosing escape sequences which display with zero (or, with a numeric argument, some other) width.
After deleting this prompt the variable LISTPROMPT should be unset for the removal to take effect.
For example:
zstyle ':completion:*' local toast \
/var/http/public/toast public_html
Completion after `http://toast/stuff/' will look for files in the directory /var/http/public/toast/stuff, while completion after `http://toast/~yousir/' will look for files in the directory ~yousir/public_html.
For notes comparing the use of this and the matcher-list style, see under the description of the tag-order style.
zstyle ':completion:*' matcher-list '' 'm:{a-zA-Z}={A-Za-z}'
By default each specification replaces the previous one; however, if a specification is prefixed with +, it is added to the existing list. Hence it is possible to create increasingly general specifications without repetition:
zstyle ':completion:*' matcher-list \
'' '+m:{a-z}={A-Z}' '+m:{A-Z}={a-z}'
It is possible to create match specifications valid for particular completers by using the third field of the context. This applies only to completers that override the global matcher-list, which as of this writing includes only _prefix and _ignored. For example, to use the completers _complete and _prefix but allow case-insensitive completion only with _complete:
zstyle ':completion:*' completer _complete _prefix
zstyle ':completion:*:complete:*:*:*' matcher-list \
'' 'm:{a-zA-Z}={A-Za-z}'
User-defined names, as explained for the completer style, are available. This makes it possible to try the same completer more than once with different match specifications each time. For example, to try normal completion without a match specification, then normal completion with case-insensitive matching, then correction, and finally partial-word completion:
zstyle ':completion:*' completer \
_complete _correct _complete:foo
zstyle ':completion:*:complete:*:*:*' matcher-list \
'' 'm:{a-zA-Z}={A-Za-z}'
zstyle ':completion:*:foo:*:*:*' matcher-list \
'm:{a-zA-Z}={A-Za-z} r:|[-_./]=* r:|=*'
If the style is unset in any context no match specification is applied. Note also that some completers such as _correct and _approximate do not use the match specifications at all, though these completers will only ever be called once even if the matcher-list contains more than one element.
Where multiple specifications are useful, note that the entire completion is done for each element of matcher-list, which can quickly reduce the shell's performance. As a rough rule of thumb, one to three strings will give acceptable performance. On the other hand, putting multiple space-separated values into the same string does not have an appreciable impact on performance.
If there is no current matcher or it is empty, and the option NO_CASE_GLOB is in effect, the matching for files is performed case-insensitively in any case. However, any matcher must explicitly specify case-insensitive matching if that is required.
For notes comparing the use of this and the matcher style, see under the description of the tag-order style.
If the value for this style contains the string `numeric', the completer function will take any numeric argument as the maximum number of errors allowed. For example, with
zstyle ':completion:*:approximate:::' max-errors 2 numeric
two errors are allowed if no numeric argument is given, but with a numeric argument of six (as in `ESC-6 TAB'), up to six errors are accepted. Hence with a value of `0 numeric', no correcting completion will be attempted unless a numeric argument is given.
If the value contains the string `not-numeric', the completer will not try to generate corrected completions when given a numeric argument, so in this case the number given should be greater than zero. For example, `2 not-numeric' specifies that correcting completion with two errors will usually be performed, but if a numeric argument is given, correcting completion will not be performed.
The default value for this style is `2 numeric'.
This has the most impact when several matches have the same description and so will be grouped together. Increasing the style will allow more matches to be grouped together; decreasing it will allow more of the description to be visible.
If none of the values found in this way is `true' but at least one is set to `auto', the shell behaves as if the AUTO_MENU option is set.
If one of the values is explicitly set to `false', menu completion will be explicitly turned off, overriding the MENU_COMPLETE option and other settings.
In the form `yes=num', where `yes' may be any of the `true' values (`yes', `true', `on' and `1'), menu completion will be turned on if there are at least num matches. In the form `yes=long', menu completion will be turned on if the list does not fit on the screen. This does not activate menu completion if the widget normally only lists completions, but menu completion can be activated in that case with the value `yes=long-list' (Typically, the value `select=long-list' described later is more useful as it provides control over scrolling.)
Similarly, with any of the `false' values (as in `no=10'), menu completion will not be used if there are num or more matches.
The value of this widget also controls menu selection, as implemented by the zsh/complist module. The following values may appear either alongside or instead of the values above.
If the value contains the string `select', menu selection will be started unconditionally.
In the form `select=num', menu selection will only be started if there are at least num matches. If the values for more than one tag provide a number, the smallest number is taken.
Menu selection can be turned off explicitly by defining a value containing the string`no-select'.
It is also possible to start menu selection only if the list of matches does not fit on the screen by using the value `select=long'. To start menu selection even if the current widget only performs listing, use the value `select=long-list'.
To turn on menu completion or menu selection when there are a certain number of matches or the list of matches does not fit on the screen, both of `yes=' and `select=' may be given twice, once with a number and once with `long' or `long-list'.
Finally, it is possible to activate two special modes of menu selection. The word `interactive' in the value causes interactive mode to be entered immediately when menu selection is started; see the description of the zsh/complist module in zshmodules(1) for a description of interactive mode. Including the string `search' does the same for incremental search mode. To select backward incremental search, include the string `search-backward'.
For example, suppose you type ^Xc to use the _correct_word widget, which generates a list of corrections for the word under the cursor. Usually, typing ^D would generate a standard list of completions for the word on the command line, and show that. With _oldlist, it will instead show the list of corrections already generated.
As another example consider the _match completer: with the insert-unambiguous style set to `true' it inserts only a common prefix string, if there is any. However, this may remove parts of the original pattern, so that further completion could produce more matches than on the first attempt. By using the _oldlist completer and setting this style to _match, the list of matches generated on the first attempt will be used again.
If this style is set it is generally unwise to call the _all_matches completer unconditionally. One possible use is for either this style or the completer style to be defined with the -e option to zstyle to make the style conditional.
For example, suppose you type ^Xc to generate a list of corrections, and menu completion is started in one of the usual ways. Usually, or with this style set to `false', typing TAB at this point would start trying to complete the line as it now appears. With _oldlist, it instead continues to cycle through the list of corrections.
zstyle ':completion:*:complete:dpkg:option--status-1:*' \
packageset avail
causes available packages, rather than only installed packages, to be completed for `dpkg --status'.
Even with the style set to `false', it is still possible to complete multiple paths by setting the option COMPLETE_IN_WORD and moving the cursor back to the first component in the path to be completed. For example, /u/b/z can be completed to /usr/bin/zsh if the cursor is after the /u.
The default value for this style is `false'.
The style is applicable to the options, signals, jobs, functions, and parameters completion tags.
For command options, this means that the initial `-', `+', or `--' must be typed explicitly before option names will be completed.
For signals, an initial `-' is required before signal names will be completed.
For jobs, an initial `%' is required before job names will be completed.
For function and parameter names, an initial `_' or `.' is required before function or parameter names starting with those characters will be completed.
The default value for this style is `false' for function and parameter completions, and `true' otherwise.
If it is a single number, only the last N words from the history will be completed.
If it is a range of the form `max:slice', the last slice words will be completed; then if that yields no matches, the slice words before those will be tried and so on. This process stops either when at least one match has been found, or max words have been tried.
The default is to complete all words from the history at once.
For example,
zstyle ':completion:*' recursive-files '*/zsh/*'
If the current directory is /home/pws/zsh/Src, then zle_trTAB can be completed to Zle/zle_tricky.c.
It is not always possible to know if connections are in fact to a remote site, so some may be prevented unnecessarily.
If its value is `true' or `false', sorting is enabled or disabled. Additionally the values associated with the `-o' option to compadd can also be listed: match, nosort, numeric, reverse. If it is not set for the context, the standard behaviour of the calling widget is used.
The style is tested first against the full context including the tag, and if that fails to produce a value against the context without the tag.
In many cases where a calling widget explicitly selects a particular ordering in lieu of the default, a value of `true' is not honoured. An example of where this is not the case is for command history where the default of sorting matches chronologically may be overridden by setting the style to `true'.
In the _expand completer, if it is set to `true', the expansions generated will always be sorted. If it is set to `menu', then the expansions are only sorted when they are offered as single strings but not in the string containing all possible expansions.
The following example sets special-dirs to `..' when the current prefix is empty, is a single `.', or consists only of a path beginning with `../'. Otherwise the value is `false'.
zstyle -e ':completion:*' special-dirs \ '[[ $PREFIX = (../)#(|.|..) ]] && reply=(..)'
The default for this style is `false'.
The default is `true'.
The values for the style are sets of space-separated lists of tags. The tags in each value will be tried at the same time; if no match is found, the next value is used. (See the file-patterns style for an exception to this behavior.)
For example:
zstyle ':completion:*:complete:-command-:*:*' tag-order \
'commands functions'
specifies that completion in command position first offers external commands and shell functions. Remaining tags will be tried if no completions are found.
In addition to tag names, each string in the value may take one of the following forms:
If the label starts with a hyphen, the tag is prepended to the label to form the name used for lookup. This can be used to make the completion system try a certain tag more than once, supplying different style settings for each attempt; see below for an example.
In any of the forms above the tag may be a pattern or several patterns in the form `{pat1,pat2...}'. In this case all matching tags will be used except for any given explicitly in the same string.
One use of these features is to try one tag more than once, setting other styles differently on each attempt, but still to use all the other tags without having to repeat them all. For example, to make completion of function names in command position ignore all the completion functions starting with an underscore the first time completion is tried:
zstyle ':completion:*:*:-command-:*:*' tag-order \
'functions:-non-comp *' functions
zstyle ':completion:*:functions-non-comp' \
ignored-patterns '_*'
On the first attempt, all tags will be offered but the functions tag will be replaced by functions-non-comp. The ignored-patterns style is set for this tag to exclude functions starting with an underscore. If there are no matches, the second value of the tag-order style is used which completes functions using the default tag, this time presumably including all function names.
The matches for one tag can be split into different groups. For example:
zstyle ':completion:*' tag-order \
'options:-long:long\ options
options:-short:short\ options
options:-single-letter:single\ letter\ options'
zstyle ':completion:*:options-long' \
ignored-patterns '[-+](|-|[^-]*)'
zstyle ':completion:*:options-short' \
ignored-patterns '--*' '[-+]?'
zstyle ':completion:*:options-single-letter' \
ignored-patterns '???*'
With the group-names style set, options beginning with `--', options beginning with a single `-' or `+' but containing multiple characters, and single-letter options will be displayed in separate groups with different descriptions.
Another use of patterns is to try multiple match specifications one after another. The matcher-list style offers something similar, but it is tested very early in the completion system and hence can't be set for single commands nor for more specific contexts. Here is how to try normal completion without any match specification and, if that generates no matches, try again with case-insensitive matching, restricting the effect to arguments of the command foo:
zstyle ':completion:*:*:foo:*:*' tag-order '*' '*:-case'
zstyle ':completion:*-case' matcher 'm:{a-z}={A-Z}'
First, all the tags offered when completing after foo are tried using the normal tag name. If that generates no matches, the second value of tag-order is used, which tries all tags again except that this time each has -case appended to its name for lookup of styles. Hence this time the value for the matcher style from the second call to zstyle in the example is used to make completion case-insensitive.
It is possible to use the -e option of the zstyle builtin command to specify conditions for the use of particular tags. For example:
zstyle -e '*:-command-:*' tag-order '
if [[ -n $PREFIX$SUFFIX ]]; then
reply=( )
else
reply=( - )
fi'
Completion in command position will be attempted only if the string typed so far is not empty. This is tested using the PREFIX special parameter; see zshcompwid for a description of parameters which are special inside completion widgets. Setting reply to an empty array provides the default behaviour of trying all tags at once; setting it to an array containing only a hyphen disables the use of all tags and hence of all completions.
If no tag-order style has been defined for a context, the strings `(|*-)argument-* (|*-)option-* values' and `options' plus all tags offered by the completion function will be used to provide a sensible default behavior that causes arguments (whether normal command arguments or arguments of options) to be completed before option names for most commands.
If the value consists of more than one string, or if the only string does not name a file or directory, the strings are used as the URLs to complete.
If the value contains only one string which is the name of a normal file the URLs are taken from that file (where the URLs may be separated by white space or newlines).
Finally, if the only string in the value names a directory, the directory hierarchy rooted at this directory gives the completions. The top level directory should be the file access method, such as `http', `ftp', `bookmark' and so on. In many cases the next level of directories will be a filename. The directory hierarchy can descend as deep as necessary.
For example,
zstyle ':completion:*' urls ~/.urls mkdir -p ~/.urls/ftp/ftp.zsh.org/pub
allows completion of all the components of the URL ftp://ftp.zsh.org/pub after suitable commands such as `netscape' or `lynx'. Note, however, that access methods and files are completed separately, so if the hosts style is set hosts can be completed without reference to the urls style.
See the description in the function _urls itself for more information (e.g. `more $^fpath/_urls(N)').
Note that this is only intended to smooth the transition from compctl to the new completion system and may disappear in the future.
Note also that the definitions from compctl will only be used if there is no specific completion function for the command in question. For example, if there is a function _foo to complete arguments to the command foo, compctl will never be invoked for foo. However, the compctl version will be tried if foo only uses default completion.
It is possible to group values for sets of commands which allow a remote login, such as rlogin and ssh, by using the my-accounts tag. Similarly, values for sets of commands which usually refer to the accounts of other people, such as talk and finger, can be grouped by using the other-accounts tag. More ambivalent commands may use the accounts tag.
The initialization script compinit redefines all the widgets which perform completion to call the supplied widget function _main_complete. This function acts as a wrapper calling the so-called `completer' functions that generate matches. If _main_complete is called with arguments, these are taken as the names of completer functions to be called in the order given. If no arguments are given, the set of functions to try is taken from the completer style. For example, to use normal completion and correction if that doesn't generate any matches:
zstyle ':completion:*' completer _complete _correct
after calling compinit. The default value for this style is `_complete _ignored', i.e. normally only ordinary completion is tried, first with the effect of the ignored-patterns style and then without it. The _main_complete function uses the return status of the completer functions to decide if other completers should be called. If the return status is zero, no other completers are tried and the _main_complete function returns.
If the first argument to _main_complete is a single hyphen, the arguments will not be taken as names of completers. Instead, the second argument gives a name to use in the completer field of the context and the other arguments give a command name and arguments to call to generate the matches.
The following completer functions are contained in the distribution, although users may write their own. Note that in contexts the leading underscore is stripped, for example basic completion is performed in the context `:completion::complete:...'.
It may be useful to use the _generic function described below to bind _all_matches to its own keystroke, for example:
zle -C all-matches complete-word _generic bindkey '^Xa' all-matches zstyle ':completion:all-matches:*' old-matches only zstyle ':completion:all-matches::::' completer _all_matches
Note that this does not generate completions by itself: first use any of the standard ways of generating a list of completions, then use ^Xa to show all matches. It is possible instead to add a standard completer to the list and request that the list of all matches should be directly inserted:
zstyle ':completion:all-matches::::' completer \
_all_matches _complete
zstyle ':completion:all-matches:*' insert true
In this case the old-matches style should not be set.
zstyle ':completion:*' completer _complete _approximate
This will give correcting completion if and only if normal completion yields no possible completions. When corrected completions are found, the completer will normally start menu completion allowing you to cycle through these strings.
This completer uses the tags corrections and original when generating the possible corrections and the original string. The format style for the former may contain the additional sequences `%e' and `%o' which will be replaced by the number of errors accepted to generate the corrections and the original string, respectively.
The completer progressively increases the number of errors allowed up to the limit by the max-errors style, hence if a completion is found with one error, no completions with two errors will be shown, and so on. It modifies the completer name in the context to indicate the number of errors being tried: on the first try the completer field contains `approximate-1', on the second try `approximate-2', and so on.
When _approximate is called from another function, the number of errors to accept may be passed with the -a option. The argument is in the same format as the max-errors style, all in one string.
Note that this completer (and the _correct completer mentioned below) can be quite expensive to call, especially when a large number of errors are allowed. One way to avoid this is to set up the completer style using the -e option to zstyle so that some completers are only used when completion is attempted a second time on the same string, e.g.:
zstyle -e ':completion:*' completer '
if [[ $_last_try != "$HISTNO$BUFFER$CURSOR" ]]; then
_last_try="$HISTNO$BUFFER$CURSOR"
reply=(_complete _match _prefix)
else
reply=(_ignored _correct _approximate)
fi'
This uses the HISTNO parameter and the BUFFER and CURSOR special parameters that are available inside zle and completion widgets to find out if the command line hasn't changed since the last time completion was tried. Only then are the _ignored, _correct and _approximate completers called.
-A, if specified, takes the paths from the array variable specified. Paths can also be specified on the command line as shown above. -N, if specified, prevents canonicalizing the paths given before using them for completion, in case they are already so. The options -M, -J, -V, -1, -2, -n, -F, -X are passed to compadd.
See _description for a description of tag and descr.
This function takes no arguments.
To complete arguments of commands, _complete uses the utility function _normal, which is in turn responsible for finding the particular function; it is described below. Various contexts of the form -context- are handled specifically. These are all mentioned above as possible arguments to the #compdef tag.
Before trying to find a function for a specific context, _complete checks if the parameter `compcontext' is set. Setting `compcontext' allows the usual completion dispatching to be overridden which is useful in places such as a function that uses vared for input. If it is set to an array, the elements are taken to be the possible matches which will be completed using the tag `values' and the description `value'. If it is set to an associative array, the keys are used as the possible completions and the values (if non-empty) are used as descriptions for the matches. If `compcontext' is set to a string containing colons, it should be of the form `tag:descr:action'. In this case the tag and descr give the tag and description to use and the action indicates what should be completed in one of the forms accepted by the _arguments utility function described below.
Finally, if `compcontext' is set to a string without colons, the value is taken as the name of the context to use and the function defined for that context will be called. For this purpose, there is a special context named -command-line- that completes whole command lines (commands and their arguments). This is not used by the completion system itself but is nonetheless handled when explicitly called.
For example, with:
zstyle ':completion:::::' completer \
_complete _correct _approximate
zstyle ':completion:*:correct:::' max-errors 2 not-numeric
zstyle ':completion:*:approximate:::' max-errors 3 numeric
correction will accept up to two errors. If a numeric argument is given, correction will not be performed, but correcting completion will be, and will accept as many errors as given by the numeric argument. Without a numeric argument, first correction and then correcting completion will be tried, with the first one accepting two errors and the second one accepting three errors.
When _correct is called as a function, the number of errors to accept may be given following the -a option. The argument is in the same form a values to the accept style, all in one string.
This completer function is intended to be used without the _approximate completer or, as in the example, just before it. Using it after the _approximate completer is useless since _approximate will at least generate the corrected strings generated by the _correct completer -- and probably more.
The tags used when generating expansions are all-expansions for the string containing all possible expansions, expansions when adding the possible expansions as single matches and original when adding the original string from the line. The order in which these strings are generated, if at all, can be controlled by the group-order and tag-order styles, as usual.
The format string for all-expansions and for expansions may contain the sequence `%o' which will be replaced by the original string from the line.
The kind of expansion to be tried is controlled by the substitute, glob and subst-globs-only styles.
It is also possible to call _expand as a function, in which case the different modes may be selected with options: -s for substitute, -g for glob and -o for subst-globs-only.
This function is also a bindable command, see the section `Bindable Commands' below.
This is intended to be used via _generic, bound to a custom key combination. Note that pattern matching is enabled so matching is performed similar to how it works with the _match completer.
Normally completion will be performed by taking the pattern from the line, inserting a `*' at the cursor position and comparing the resulting pattern with the possible completions generated. This can be modified with the match-original style described above.
The generated matches will be offered in a menu completion unless the insert-unambiguous style is set to `true'; see the description above for other options for this style.
Note that matcher specifications defined globally or used by the completion functions (the styles matcher-list and matcher) will not be used.
The completer style is used to decide which other completers are to be called to generate matches. If this style is unset, the list of completers set for the current context is used -- except, of course, the _prefix completer itself. Furthermore, if this completer appears more than once in the list of completers only those completers not already tried by the last invocation of _prefix will be called.
For example, consider this global completer style:
zstyle ':completion:*' completer \
_complete _prefix _correct _prefix:foo
Here, the _prefix completer tries normal completion but ignoring the suffix. If that doesn't generate any matches, and neither does the call to the _correct completer after it, _prefix will be called a second time and, now only trying correction with the suffix ignored. On the second invocation the completer part of the context appears as `foo'.
To use _prefix as the last resort and try only normal completion when it is invoked:
zstyle ':completion:*' completer _complete ... _prefix zstyle ':completion::prefix:*' completer _complete
The add-space style is also respected. If it is set to `true' then _prefix will insert a space between the matches generated (if any) and the suffix.
Note that this completer is only useful if the COMPLETE_IN_WORD option is set; otherwise, the cursor will be moved to the end of the current word before the completion code is called and hence there will be no suffix.
The expansion depends on the array style user-expand being defined for the current context; remember that the context for completers is less specific than that for contextual completion as the full context has not yet been determined. Elements of the array may have one of the following forms:
hash is the name of an associative array. Note this is not a full parameter expression, merely a $, suitably quoted to prevent immediate expansion, followed by the name of an associative array. If the trial expansion word matches a key in hash, the resulting expansion is the corresponding value.
_func is the name of a shell function whose name must begin with _ but is not otherwise special to the completion system. The function is called with the trial word as an argument. If the word is to be expanded, the function should set the array reply to a list of expansions. Optionally, it can set REPLY to a word that will be used as a description for the set of expansions. The return status of the function is irrelevant.
In addition to the context-dependent completions provided, which are expected to work in an intuitively obvious way, there are a few widgets implementing special behaviour which can be bound separately to keys. The following is a list of these and their default bindings.
for key in '!' '$' '@' '/' '~'; do bindkey "\e$key" _bash_complete-word bindkey "^X$key" _bash_list-choices done
This includes the bindings for `~' in case they were already bound to something else; the completion code does not override user bindings.
When used as a bindable command there is one additional feature that can be selected by setting the complete style to `true'. In this case, if the word is not the name of an alias, _expand_alias tries to complete the word to a full alias name without expanding it. It leaves the cursor directly after the completed word so that invoking _expand_alias once more will expand the now-complete alias name.
zle -C foo complete-word _generic bindkey '...' foo zstyle ':completion:foo:*' menu yes select=1
Note in particular that the completer style may be set for the context in order to change the set of functions used to generate possible matches. If _generic is called with arguments, those are passed through to _main_complete as the list of completers in place of those defined by the completer style.
Second, any other string will be passed as a set of arguments to compadd and should hence be an expression specifying what should be completed.
A very restricted set of editing commands is available when reading the string: `DEL' and `^H' delete the last character; `^U' deletes the line, and `^C' and `^G' abort the function, while `RET' accepts the completion. Note the string is used verbatim as a command line, so arguments must be quoted in accordance with standard shell rules.
Once a string has been read, the next call to _read_comp will use the existing string instead of reading a new one. To force a new string to be read, call _read_comp with a numeric argument.
Note that the information about styles may be incomplete; it depends on the information available from the completion functions called, which in turn is determined by the user's own styles and other settings.
If this widget is created using the name of the function, as it is by default, then when executed it will read a key sequence. This is expected to be bound to a call to a completion function that uses the _generic widget. That widget will be executed, and information provided in the same format that the _complete_help widget displays for contextual completion.
If the widget's name contains debug, for example if it is created as `zle -N _complete_debug_generic _complete_help_generic', it will read and execute the keystring for a generic widget as before, but then generate debugging information as done by _complete_debug for contextual completion.
If the widget's name contains noread, it will not read a keystring but instead arrange that the next use of a generic widget run in the same shell will have the effect as described above.
The widget works by setting the shell parameter ZSH_TRACE_GENERIC_WIDGET which is read by _generic. Unsetting the parameter cancels any pending effect of the noread form.
For example, after executing the following:
zle -N _complete_debug_generic _complete_help_generic bindkey '^x:' _complete_debug_generic
typing `C-x :' followed by the key sequence for a generic widget will cause trace output for that widget to be saved to a file.
Descriptions follow for utility functions that may be useful when writing completion functions. If functions are installed in subdirectories, most of these reside in the Base subdirectory. Like the example functions for commands in the distribution, the utility functions generating matches all follow the convention of returning status zero if they generated completions and non-zero if no matching completions could be added.
For example:
local expl ... if _requested foo; then ... _all_labels foo expl '...' compadd ... - $matches fi
Will complete the strings from the matches parameter, using compadd with additional options which will take precedence over those generated by _all_labels.
The tags to use and the action to perform if a tag is requested are described using the specs which are of the form: `tag:descr:action'. The tags are offered using _tags and if the tag is requested, the action is executed with the given description descr. The actions are those accepted by the _arguments function (described below), excluding the `->state' and `=...' forms.
For example, the action may be a simple function call:
_alternative \
'users:user:_users' \
'hosts:host:_hosts'
offers usernames and hostnames as possible matches, generated by the _users and _hosts functions respectively.
Like _arguments, this function uses _all_labels to execute the actions, which will loop over all sets of tags. Special handling is only required if there is an additional valid tag, for example inside a function called from _alternative.
The option `-O name' is used in the same way as by the _arguments function. In other words, the elements of the name array will be passed to compadd when executing an action.
Like _tags this function supports the -C option to give a different name for the argument context field.
Options Overview
Options to _arguments itself must be in separate words, i.e. -s -w, not -sw. The options are followed by specs that describe options and arguments of the analyzed command. To avoid ambiguity, all options to _arguments itself may be separated from the spec forms by a single colon.
The `--' form is used to intuit spec forms from the help output of the command being analyzed, and is described in detail below. The opts for the `--' form are otherwise the same options as the first form. Note that `-s' following `--' has a distinct meaning from `-s' preceding `--', and both may appear.
The option switches -s, -S, -A, -w, and -W affect how _arguments parses the analyzed command line's options. These switches are useful for commands with standard argument parsing.
The options of _arguments have the following meanings:
Options beginning with a single hyphen or plus sign are eligible for stacking; words beginning with two hyphens are not.
Note that -s after -- has a different meaning, which is documented in the segment entitled `Deriving spec forms from the help output'.
foobar -x -- -y
the `-x' is considered an option, the `-y' is considered an argument, and the `--' is considered to be neither.
r:|[_-]=* r:|=*
specs: overview
Each of the following forms is a spec describing individual sets of options or arguments on the command line being analyzed.
With two colons before the message, the words special array and the CURRENT special parameter are modified to refer only to the normal arguments when the action is executed or evaluated. With three colons before the message they are modified to refer only to the normal arguments covered by this description.
The following forms are available for the initial optspec, whether or not the option has arguments.
Either of `-+optname' and `+-optname' can be used to specify that -optname and +optname are both valid.
In all the remaining forms, the leading `-' may be replaced by or paired with `+' in this way.
The verbose style is used to decide whether the explanation strings are displayed with the option in a completion listing.
If no bracketed explanation string is given but the auto-description style is set and only one argument is described for this optspec, the value of the style is displayed, with any appearance of the sequence `%d' in it replaced by the message of the first optarg that follows the optspec; see below.
It is possible for options with a literal `+' or `=' to appear, but that character must be quoted, for example `-\+'.
Each optarg following an optspec must take one of the following forms:
This group may be repeated for options which take multiple arguments. In other words, :message1:action1:message2:action2 specifies that the option takes two arguments.
Multiple colons are treated as for the `*:...' forms for ordinary arguments: when the message is preceded by two colons, the words special array and the CURRENT special parameter are modified during the execution or evaluation of the action to refer only to the words after the option. When preceded by three colons, they are modified to refer only to the words covered by this description.
Any literal colon in an optname, message, or action must be preceded by a backslash, `\:'.
Each of the forms above may be preceded by a list in parentheses of option names and argument numbers. If the given option is on the command line, the options and arguments indicated in parentheses will not be offered. For example, `(-two -three 1)-one:...' completes the option `-one'; if this appears on the command line, the options -two and -three and the first ordinary argument will not be completed after it. `(-foo):...' specifies an ordinary argument completion; -foo will not be completed if that argument is already present.
Other items may appear in the list of excluded options to indicate various other items that should not be applied when the current specification is matched: a single star (*) for the rest arguments (i.e. a specification of the form `*:...'); a colon (:) for all normal (non-option-) arguments; and a hyphen (-) for all options. For example, if `(*)' appears before an option and the option appears on the command line, the list of remaining arguments (those shown in the above table beginning with `*:') will not be completed.
To aid in reuse of specifications, it is possible to precede any of the forms above with `!'; then the form will no longer be completed, although if the option or argument appears on the command line they will be skipped as normal. The main use for this is when the arguments are given by an array, and _arguments is called repeatedly for more specific contexts: on the first call `_arguments $global_options' is used, and on subsequent calls `_arguments !$^global_options'.
specs: actions
In each of the forms above the action determines how completions should be generated. Except for the `->string' form below, the action will be executed by calling the _all_labels function to process all tag labels. No special handling of tags is needed unless a function call introduces a new one.
The functions called to execute actions will be called with the elements of the array named by the `-O name' option as arguments. This can be used, for example, to pass the same set of options for the compadd builtin to all actions.
The forms for action are as follows.
:foo:(foo bar baz)
:foo:((a\:bar b\:baz))
The matches will be listed together with their descriptions if the description style is set with the values tag in the context.
Where _arguments encounters action in the `->string' format, it will strip all leading and trailing whitespace from string and set the array state to the set of all strings for which an action is to be performed. The elements of the array state_descr are assigned the corresponding message field from each optarg containing such an action.
By default and in common with all other well behaved completion functions, _arguments returns status zero if it was able to add matches and non-zero otherwise. However, if the -R option is given, _arguments will instead return a status of 300 to indicate that $state is to be handled.
In addition to $state and $state_descr, _arguments also sets the global parameters `context', `line' and `opt_args' as described below, and does not reset any changes made to the special parameters such as PREFIX and words. This gives the calling function the choice of resetting these parameters or propagating changes in them.
A function calling _arguments with at least one action containing a `->string' must therefore declare appropriate local parameters:
local context state state_descr line typeset -A opt_args
to prevent _arguments from altering the global environment.
This is most useful with one of the specifiers that restrict the words on the command line on which the action is to operate (the two- and three-colon forms above). One particular use is when an action itself causes _arguments on a restricted range; it is necessary to use this trick to insert an appropriate command name into the range for the second call to _arguments to be able to parse the line.
Otherwise it will be invoked with some extra strings placed after the first word; these are to be passed down as options to the compadd builtin. They ensure that the state specified by _arguments, in particular the descriptions of options and arguments, is correctly passed to the completion command. These additional arguments are taken from the array parameter `expl'; this will be set up before executing the action and hence may be referred to inside it, typically in an expansion of the form `$expl[@]' which preserves empty elements of the array.
During the performance of the action the array `line' will be set to the normal arguments from the command line, i.e. the words from the command line after the command name excluding all options and their arguments. Options are stored in the associative array `opt_args' with option names as keys and their arguments as the values. For options that have more than one argument these are given as one string, separated by colons. All colons and backslashes in the original arguments are preceded with backslashes.
The parameter `context' is set when returning to the calling function to perform an action of the form `->string'. It is set to an array of elements corresponding to the elements of $state. Each element is a suitable name for the argument field of the context: either a string of the form `option-opt-n' for the n'th argument of the option -opt, or a string of the form `argument-n' for the n'th argument. For `rest' arguments, that is those in the list at the end not handled by position, n is the string `rest'. For example, when completing the argument of the -o option, the name is `option-o-1', while for the second normal (non-option-) argument it is `argument-2'.
Furthermore, during the evaluation of the action the context name in the curcontext parameter is altered to append the same string that is stored in the context parameter.
The option -C tells _arguments to modify the curcontext parameter for an action of the form `->state'. This is the standard parameter used to keep track of the current context. Here it (and not the context array) should be made local to the calling function to avoid passing back the modified value and should be initialised to the current value at the start of the function:
local curcontext="$curcontext"
This is useful where it is not possible for multiple states to be valid together.
Grouping Options
Options can be grouped to simplify exclusion lists. A group is introduced with `+' followed by a name for the group in the subsequent word. Whole groups can then be referenced in an exclusion list or a group name can be used to disambiguate between two forms of the same option. For example:
_arguments \
'(group2--x)-a' \
+ group1 \
-m \
'(group2)-n' \
+ group2 \
-x -y
If the name of a group is specified in the form `(name)' then only one value from that group will ever be completed; more formally, all specifications are mutually exclusive to all other specifications in that group. This is useful for defining options that are aliases for each other. For example:
_arguments \
-a -b \
+ '(operation)' \
{-c,--compress}'[compress]' \
{-d,--decompress}'[decompress]' \
{-l,--list}'[list]'
If an option in a group appears on the command line, it is stored in the associative array `opt_args' with 'group-option' as a key. In the example above, a key `operation--c' is used if the option `-c' is present on the command line.
Specifying Multiple Sets of Arguments
It is possible to specify multiple sets of options and arguments with the sets separated by single hyphens. This differs from groups in that sets are considered to be mutually exclusive of each other.
Specifications before the first set and from any group are common to all sets. For example:
_arguments \
-a \
- set1 \
-c \
- set2 \
-d \
':arg:(x2 y2)'
This defines two sets. When the command line contains the option `-c', the `-d' option and the argument will not be considered possible completions. When it contains `-d' or an argument, the option `-c' will not be considered. However, after `-a' both sets will still be considered valid.
As for groups, the name of a set may appear in exclusion lists, either alone or preceding a normal option or argument specification.
The completion code has to parse the command line separately for each set. This can be slow so sets should only be used when necessary. A useful alternative is often an option specification with rest-arguments (as in `-foo:*:...'); here the option -foo swallows up all remaining arguments as described by the optarg definitions.
Deriving spec forms from the help output
The option `--' allows _arguments to work out the names of long options that support the `--help' option which is standard in many GNU commands. The command word is called with the argument `--help' and the output examined for option names. Clearly, it can be dangerous to pass this to commands which may not support this option as the behaviour of the command is unspecified.
In addition to options, `_arguments --' will try to deduce the types of arguments available for options when the form `--opt=val' is valid. It is also possible to provide hints by examining the help text of the command and adding helpspec of the form `pattern:message:action'; note that other _arguments spec forms are not used. The pattern is matched against the help text for an option, and if it matches the message and action are used as for other argument specifiers. The special case of `*:' means both message and action are empty, which has the effect of causing options having no description in the help output to be ordered in listings ahead of options that have a description.
For example:
_arguments -- '*\*:toggle:(yes no)' \
'*=FILE*:file:_files' \
'*=DIR*:directory:_files -/' \
'*=PATH*:directory:_files -/'
Here, `yes' and `no' will be completed as the argument of options whose description ends in a star; file names will be completed for options that contain the substring `=FILE' in the description; and directories will be completed for options whose description contains `=DIR' or `=PATH'. The last three are in fact the default and so need not be given explicitly, although it is possible to override the use of these patterns. A typical help text which uses this feature is:
-C, --directory=DIR change to directory DIR
so that the above specifications will cause directories to be completed after `--directory', though not after `-C'.
Note also that _arguments tries to find out automatically if the argument for an option is optional. This can be specified explicitly by doubling the colon before the message.
If the pattern ends in `(-)', this will be removed from the pattern and the action will be used only directly after the `=', not in the next word. This is the behaviour of a normal specification defined with the form `=-'.
By default, the command (with the option `--help') is run after resetting all the locale categories (except for LC_CTYPE) to `C'. If the localized help output is known to work, the option `-l' can be specified after the `_arguments --' so that the command is run in the current locale.
The `_arguments --' can be followed by the option `-i patterns' to give patterns for options which are not to be completed. The patterns can be given as the name of an array parameter or as a literal list in parentheses. For example,
_arguments -- -i \
"(--(en|dis)able-FEATURE*)"
will cause completion to ignore the options `--enable-FEATURE' and `--disable-FEATURE' (this example is useful with GNU configure).
The `_arguments --' form can also be followed by the option `-s pair' to describe option aliases. The pair consists of a list of alternating patterns and corresponding replacements, enclosed in parens and quoted so that it forms a single argument word in the _arguments call.
For example, some configure-script help output describes options only as `--enable-foo', but the script also accepts the negated form `--disable-foo'. To allow completion of the second form:
_arguments -- -s "((#s)--enable- --disable-)"
Miscellaneous notes
Finally, note that _arguments generally expects to be the primary function handling any completion for which it is used. It may have side effects which change the treatment of any matches added by other functions called after it. To combine _arguments with other functions, those functions should be called either before _arguments, as an action within a spec, or in handlers for `->state' actions.
Here is a more general example of the use of _arguments:
_arguments '-l+:left border:' \
'-format:paper size:(letter A4)' \
'*-copy:output file:_files::resolution:(300 600)' \
':postscript file:_files -g \*.\(ps\|eps\)' \
'*:page number:'
This describes three options: `-l', `-format', and `-copy'. The first takes one argument described as `left border' for which no completion will be offered because of the empty action. Its argument may come directly after the `-l' or it may be given as the next word on the line.
The `-format' option takes one argument in the next word, described as `paper size' for which only the strings `letter' and `A4' will be completed.
The `-copy' option may appear more than once on the command line and takes two arguments. The first is mandatory and will be completed as a filename. The second is optional (because of the second colon before the description `resolution') and will be completed from the strings `300' and `600'.
The last two descriptions say what should be completed as arguments. The first describes the first argument as a `postscript file' and makes files ending in `ps' or `eps' be completed. The last description gives all other arguments the description `page numbers' but does not offer completions.
Example:
_example_caching_policy () {
# rebuild if cache is more than a week old
local -a oldp
oldp=( "$1"(Nm+7) )
(( $#oldp ))
}
The return status of _call_function itself is zero if the function name exists and was called and non-zero otherwise.
By default, the command is run in an environment where all the locale categories (except for LC_CTYPE) are reset to `C' by calling the utility function _comp_locale (see below). If the option `-l' is given, the command is run with the current locale.
If the option `-p' is supplied it indicates that the command output is influenced by the permissions it is run with. If the gain-privileges style is set to true, _call_program will make use of commands such as sudo, if present on the command-line, to match the permissions to whatever the final command is likely to run under. When looking up the gain-privileges and command styles, the command component of the zstyle context will end with a slash (`/') followed by the command that would be used to gain privileges.
The style name consists of field names separated by hyphens, for example `users-hosts-ports'. For each field for a value is already known, a spec of the form `field=pattern' is given. For example, if the command line so far specifies a user `pws', the argument `users=pws' should appear.
The next argument with no equals sign is taken as the name of the field for which completions should be generated (presumably not one of the fields for which the value is known).
The matches generated will be taken from the value of the style. These should contain the possible values for the combinations in the appropriate order (users, hosts, ports in the example above). The values for the different fields are separated by colons. This can be altered with the option -s to _combination which specifies a pattern. Typically this is a character class, as for example `-s "[:@]"' in the case of the users-hosts style. Each `field=pattern' specification restricts the completions which apply to elements of the style with appropriately matching fields.
If no style with the given name is defined for the given tag, or if none of the strings in style's value match, but a function name of the required field preceded by an underscore is defined, that function will be called to generate the matches. For example, if there is no `users-hosts-ports' or no matching hostname when a host is required, the function `_hosts' will automatically be called.
If the same name is used for more than one field, in both the `field=pattern' and the argument that gives the name of the field to be completed, the number of the field (starting with one) may be given after the fieldname, separated from it by a colon.
All arguments after the required field name are passed to compadd when generating matches from the style value, or to the functions for the fields if they are called.
This function should normally be run only in a subshell, because the new locale is exported to the environment. Typical usage would be `$(_comp_locale; command ...)'.
The descr is taken as a string to display above the matches if the format style for the descriptions tag is set. This is followed by one or two names of arrays followed by options to pass to compadd. The array name1 contains the possible completions with their descriptions in the form `completion:description'. Any literal colons in completion must be quoted with a backslash. If a name2 is given, it should have the same number of elements as name1; in this case the corresponding elements are added as possible completions instead of the completion strings from name1. The completion list will retain the descriptions from name1. Finally, a set of completion options can appear.
If the option `-o' appears before the first argument, the matches added will be treated as names of command options (N.B. not shell options), typically following a `-', `--' or `+' on the command line. In this case _describe uses the prefix-hidden, prefix-needed and verbose styles to find out if the strings should be added as completions and if the descriptions should be shown. Without the `-o' option, only the verbose style is used to decide how descriptions are shown. If `-O' is used instead of `-o', command options are completed as above but _describe will not handle the prefix-needed style.
With the -t option a tag can be specified. The default is `values' or, if the -o option is given, `options'.
The options -1, -2, -J, -V, -x are passed to _next_label.
If selected by the list-grouped style, strings with the same description will appear together in the list.
_describe uses the _all_labels function to generate the matches, so it does not need to appear inside a loop over tag labels.
The styles listed below are tested in the current context using the given tag. The resulting options for compadd are put into the array named name (this is traditionally `expl', but this convention is not enforced). The description for the corresponding set of matches is passed to the function in descr.
The styles tested are: format, hidden, matcher, ignore-line, ignored-patterns, group-name and sort. The format style is first tested for the given tag and then for the descriptions tag if no value was found, while the remainder are only tested for the tag given as the first argument. The function also calls _setup which tests some more styles.
The string returned by the format style (if any) will be modified so that the sequence `%d' is replaced by the descr given as the third argument without any leading or trailing white space. If, after removing the white space, the descr is the empty string, the format style will not be used and the options put into the name array will not contain an explanation string to be displayed above the matches.
If _description is called with more than three arguments, the additional specs should be of the form `char:str'. These supply escape sequence replacements for the format style: every appearance of `%char' will be replaced by string.
If the -x option is given, the description will be passed to compadd using the -x option instead of the default -X. This means that the description will be displayed even if there are no corresponding matches.
The options placed in the array name take account of the group-name style, so matches are placed in a separate group where necessary. The group normally has its elements sorted (by passing the option -J to compadd), but if an option starting with `-V', `-J', `-1', or `-2' is passed to _description, that option will be included in the array. Hence it is possible for the completion group to be unsorted by giving the option `-V', `-1V', or `-2V'.
In most cases, the function will be used like this:
local expl _description files expl file compadd "$expl[@]" - "$files[@]"
Note the use of the parameter expl, the hyphen, and the list of matches. Almost all calls to compadd within the completion system use a similar format; this ensures that user-specified styles are correctly passed down to the builtins which implement the internals of completion.
The function sets the parameter $service to the string being tried, and sets the context/command field (the fourth) of the $curcontext parameter to the context given as the first argument.
The following plugins are available by default: _email-ldap (see the filter style), _email-local (completes user@hostname Unix addresses), _email-mail (completes aliases from ~/.mailrc), _email-mush, _email-mutt, and _email-pine.
Addresses from the _email-foo plugin are added under the tag `email-foo'.
Writing plugins
Plugins are written as separate functions with names starting with `_email-'. They are invoked with the -c option and compadd options. They should either do their own completion or set the $reply array to a list of `alias:address' elements and return 300. New plugins will be picked up and run automatically.
This function accepts the full set of options allowed by _path_files, described below.
compdef _gnu_generic foo bar
after the call to compinit.
The completion system as supplied is conservative in its use of this function, since it is important to be sure the command understands the option `--help'.
The return status is zero if the message was displayed and the word to complete is not empty, and non-zero otherwise.
The pattern may be preceded by any of the options understood by compadd that are passed down from _description, namely -M, -J, -V, -1, -2, -n, -F and -X. All of these options will be ignored. This fits in conveniently with the argument-passing conventions of actions for _arguments.
As an example, consider a command taking the options -n and -none, where -n must be followed by a numeric value in the same word. By using:
_arguments '-n-: :_guard "[0-9]#" "numeric value"' '-none'
_arguments can be made to both display the message `numeric value' and complete options after `-n<TAB>'. If the `-n' is already followed by one or more digits (the pattern passed to _guard) only the message will be displayed; if the `-n' is followed by another character, only options are completed.
The format style is examined with the messages tag to find a message; the usual tag, descriptions, is used only if the style is not set with the former.
If the -r option is given, no style is used; the descr is taken literally as the string to display. This is most useful when the descr comes from a pre-processed argument list which already contains an expanded description. Note that this option does not disable the `%'-sequence parsing done by compadd.
The -12VJ options and the group are passed to compadd and hence determine the group the message string is added to.
The second -e form gives a description for completions with the tag tag to be shown even if there are no matches for that tag. This form is called by _arguments in the event that there is no action for an option specification. The tag can be omitted and if so the tag is taken from the parameter $curtag; this is maintained by the completion system and so is usually correct. Note that if there are no matches at the time this function is called, compstate[insert] is cleared, so additional matches generated later are not inserted on the command line.
The -i option causes _multi_parts to insert a unique match even if that requires multiple separators to be inserted. This is not usually the expected behaviour with filenames, but certain other types of completion, for example those with a fixed set of possibilities, may be more suited to this form.
Like other utility functions, this function accepts the `-V', `-J', `-1', `-2', `-n', `-f', `-X', `-M', `-P', `-S', `-r', `-R', and `-q' options and passes them to the compadd builtin.
The -x12VJ options and the first three arguments are passed to the _description function. Where appropriate the tag will be replaced by a tag label in this call. Any description given in the tag-order style is preferred to the descr passed to _next_label.
The options given after the descr are set in the parameter given by name, and hence are to be passed to compadd or whatever function is called to add the matches.
Here is a typical use of this function for the tag foo. The call to _requested determines if tag foo is required at all; the loop over _next_label handles any labels defined for the tag in the tag-order style.
local expl ret=1
...
if _requested foo; then
...
while _next_label foo expl '...'; do
compadd "$expl[@]" ... && ret=0
done
...
fi
return ret
A second use is to reexamine the command line specified by the $words array and the $CURRENT parameter after those have been modified. For example, the function _precommand, which completes after precommand specifiers such as nohup, removes the first word from the words array, decrements the CURRENT parameter, then calls `_normal -p $service'. The effect is that `nohup cmd ...' is treated in the same way as `cmd ...'.
If the command name matches one of the patterns given by one of the options -p or -P to compdef, the corresponding completion function is called and then the parameter _compskip is checked. If it is set completion is terminated at that point even if no matches have been found. This is the same effect as in the -first- context.
Note that you need to uncomment a few lines in the _main_complete function for these functions to work properly. The lines in question are used to store the option settings in effect before the completion widget locally sets the options it needs. Hence these functions are not generally used by the completion system.
The option `-g pattern' limits the completion to parameters whose type matches the pattern. The type of a parameter is that shown by `print ${(t)param}', hence judicious use of `*' in pattern is probably necessary.
All other arguments are passed to the compadd builtin.
The options accepted by both _path_files and _files are:
Both _path_files and _files also accept the following options which are passed to compadd: `-J', `-V', `-1', `-2', `-n', `-X', `-M', `-P', `-S', `-q', `-r', and `-R'.
Finally, the _path_files function uses the styles expand, ambiguous, special-dirs, list-suffixes and file-sort described above.
The command to run is taken from the first element of the array words unless this is overridden by the option -c. This command is run and its output is compared with a series of patterns. Arguments to be passed to the command can be specified at the end after all the other arguments. The patterns to try in order are given by the arguments label=pattern; if the output of `command arg ...' contains pattern, then label is selected as the label for the command variant. If none of the patterns match, the final command label is selected and status 1 is returned.
If the `-b builtin-label' is given, the command is tested to see if it is provided as a shell builtin, possibly autoloaded; if so, the label builtin-label is selected as the label for the variant.
If the `-r name' is given, the label picked is stored in the parameter named name.
The results are also cached in the _cmd_variant associative array indexed by the name of the command run.
The arguments are grouped as sets of alternatives separated by `|', which are tried one after the other until one matches. Each alternative consists of a one or more specifications which are tried left to right, with each pattern matched being stripped in turn from the command line being tested, until all of the group succeeds or until one fails; in the latter case, the next alternative is tried. This structure can be repeated to arbitrary depth by using parentheses; matching proceeds from inside to outside.
A special procedure is applied if no test succeeds but the remaining command line string contains no null character (implying the remaining word is the one for which completions are to be generated). The completion target is restricted to the remaining word and any actions for the corresponding patterns are executed. In this case, nothing is stripped from the command line string. The order of evaluation of the actions can be determined by the tag-order style; the various formats supported by _alternative can be used in action. The descr is used for setting up the array parameter expl.
Specification arguments take one of following forms, in which metacharacters such as `(', `)', `#' and `|' should be quoted.
The argument starting with : is used in the same manner as an argument to _alternative.
A component is used as follows: pattern is tested to see if the component already exists on the command line. If it does, any following specifications are examined to find something to complete. If a component is reached but no such pattern exists yet on the command line, the string containing the action is used to generate matches to insert at that point.
The function _regex_words can be used as a helper function to generate matches for a set of alternative words possibly with their own arguments as a command line argument.
Examples:
_regex_arguments _tst /$'[^\0]#\0'/ \
/$'[^\0]#\0'/ :'compadd aaa'
This generates a function _tst that completes aaa as its only argument. The tag and description for the action have been omitted for brevity (this works but is not recommended in normal use). The first component matches the command word, which is arbitrary; the second matches any argument. As the argument is also arbitrary, any following component would not depend on aaa being present.
_regex_arguments _tst /$'[^\0]#\0'/ \
/$'aaa\0'/ :'compadd aaa'
This is a more typical use; it is similar, but any following patterns would only match if aaa was present as the first argument.
_regex_arguments _tst /$'[^\0]#\0'/ \( \
/$'aaa\0'/ :'compadd aaa' \
/$'bbb\0'/ :'compadd bbb' \) \#
In this example, an indefinite number of command arguments may be completed. Odd arguments are completed as aaa and even arguments as bbb. Completion fails unless the set of aaa and bbb arguments before the current one is matched correctly.
_regex_arguments _tst /$'[^\0]#\0'/ \
\( /$'aaa\0'/ :'compadd aaa' \| \
/$'bbb\0'/ :'compadd bbb' \) \#
This is similar, but either aaa or bbb may be completed for any argument. In this case _regex_words could be used to generate a suitable expression for the arguments.
Each spec gives one of a set of words that may be completed at this point, together with arguments. It is thus roughly equivalent to the _arguments function when used in normal (non-regex) completion.
The part of the spec before the first colon is the word to be completed. This may contain a *; the entire word, before and after the * is completed, but only the text before the * is required for the context to be matched, so that further arguments may be completed after the abbreviated form.
The second part of spec is a description for the word being completed.
The optional third part of the spec describes how words following the one being completed are themselves to be completed. It will be evaluated in order to avoid problems with quoting. This means that typically it contains a reference to an array containing previously generated regex arguments.
The option -t term specifies a terminator for the word instead of the usual space. This is handled as an auto-removable suffix in the manner of the option -s sep to _values.
The result of the processing by _regex_words is placed in the array reply, which should be made local to the calling function. If the set of words and arguments may be matched repeatedly, a # should be appended to the generated array at that point.
For example:
local -a reply _regex_words mydb-commands 'mydb commands' \ 'add:add an entry to mydb:$mydb_add_cmds' \ 'show:show entries in mydb' _regex_arguments _mydb "$reply[@]" _mydb "$@"
This shows a completion function for a command mydb which takes two command arguments, add and show. show takes no arguments, while the arguments for add have already been prepared in an array mydb_add_cmds, quite possibly by a previous call to _regex_words.
_tags foo bar baz
while _tags; do
if _requested foo; then
... # perform completion for foo
fi
... # test the tags bar and baz in the same way
... # exit loop if matches were generated
done
Note that the test for whether matches were generated is not performed until the end of the _tags loop. This is so that the user can set the tag-order style to specify a set of tags to be completed at the same time.
If name and descr are given, _requested calls the _description function with these arguments together with the options passed to _requested.
If command is given, the _all_labels function will be called immediately with the same arguments. In simple cases this makes it possible to perform the test for the tag and the matching in one go. For example:
local expl ret=1
_tags foo bar baz
while _tags; do
_requested foo expl 'description' \
compadd foobar foobaz && ret=0
...
(( ret )) || break
done
If the command is not compadd, it must nevertheless be prepared to handle the same options.
See _store_cache below for more details.
This function accepts the compadd options `-V', `-J', `-1', `-2', `-n', `-X', `-M', `-P', `-S', `-r', `-R', and `-q' and passes them on to the compadd builtin used to add the matches.
Common compadd options are passed on to the function. It is possible to use compadd directly with _sequence, though _values may be more appropriate in this situation.
The optional group supplies the name of the group in which the matches will be placed. If it is not given, the tag is used as the group name.
This function is called automatically from _description and hence is not normally called explicitly.
The cache_identifier specifies the file which the data should be dumped to. The file is stored in a directory specified by the cache-path style which defaults to ~/.zcompcache. The remaining params arguments are the parameters to dump to the file.
The return status is zero if storage was successful. The function will only attempt storage if the use-cache style is set, so you can call this function without worrying about whether the user wanted to use the caching layer.
The completion function may avoid calling _retrieve_cache when it already has the completion data available as parameters. However, in that case it should call _cache_invalid to check whether the data in the parameters and in the cache are still valid.
See the _perl_modules completion function for a simple example of the usage of the caching layer.
Next, _tags is called repeatedly without arguments from the same completion function. This successively selects the first, second, etc. set of tags requested by the user. The return status is zero if at least one of the tags is requested and non-zero otherwise. To test if a particular tag is to be tried, the _requested function should be called (see above).
If `-C name' is given, name is temporarily stored in the argument field (the fifth) of the context in the curcontext parameter during the call to _tags; the field is restored on exit. This allows _tags to use a more specific context without having to change and reset the curcontext parameter (which has the same effect).
If the first argument is the option `-O name', it will be used in the same way as by the _arguments function. In other words, the elements of the name array will be passed to compadd when executing an action.
If the first argument (or the first argument after `-O name') is `-s', the next argument is used as the character that separates multiple values. This character is automatically added after each value in an auto-removable fashion (see below); all values completed by `_values -s' appear in the same word on the command line, unlike completion using _arguments. If this option is not present, only a single value will be completed per word.
Normally, _values will only use the current word to determine which values are already present on the command line and hence are not to be completed again. If the -w option is given, other arguments are examined as well.
The first non-option argument, desc, is used as a string to print as a description before listing the values.
All other arguments describe the possible values and their arguments in the same format used for the description of options by the _arguments function (see above). The only differences are that no minus or plus sign is required at the beginning, values can have only one argument, and the forms of action beginning with an equal sign are not supported.
The character separating a value from its argument can be set using the option -S (like -s, followed by the character to use as the separator in the next argument). By default the equals sign will be used as the separator between values and arguments.
Example:
_values -s , 'description' \
'*foo[bar]' \
'(two)*one[number]:first count:' \
'two[another number]::second count:(1 2 3)'
This describes three possible values: `foo', `one', and `two'. The first is described as `bar', takes no argument and may appear more than once. The second is described as `number', may appear more than once, and takes one mandatory argument described as `first count'; no action is specified, so it will not be completed. The `(two)' at the beginning says that if the value `one' is on the line, the value `two' will no longer be considered a possible completion. Finally, the last value (`two') is described as `another number' and takes an optional argument described as `second count' for which the completions (to appear after an `=') are `1', `2', and `3'. The _values function will complete lists of these values separated by commas.
Like _arguments, this function temporarily adds another context name component to the arguments element (the fifth) of the current context while executing the action. Here this name is just the name of the value for which the argument is completed.
The style verbose is used to decide if the descriptions for the values (but not those for the arguments) should be printed.
The associative array val_args is used to report values and their arguments; this works similarly to the opt_args associative array used by _arguments. Hence the function calling _values should declare the local parameters state, state_descr, line, context and val_args:
local context state state_descr line typeset -A val_args
when using an action of the form `->string'. With this function the context parameter will be set to the name of the value whose argument is to be completed. Note that for _values, the state and state_descr are scalars rather than arrays. Only a single matching state is returned.
Note also that _values normally adds the character used as the separator between values as an auto-removable suffix (similar to a `/' after a directory). However, this is not possible for a `->string' action as the matches for the argument are generated by the calling function. To get the usual behaviour, the calling function can add the separator x as a suffix by passing the options `-qS x' either directly or indirectly to compadd.
The option -C is treated in the same way as it is by _arguments. In that case the parameter curcontext should be made local instead of context (as described above).
The arguments to _wanted are the same as those to _requested, i.e. arguments to be passed to _description. However, in this case the command is not optional; all the processing of tags, including the loop over both tags and tag labels and the generation of matches, is carried out automatically by _wanted.
Hence to offer only one tag and immediately add the corresponding matches with the given description:
local expl
_wanted tag expl 'description' \
compadd matches...
Note that, as for _requested, the command must be able to accept options to be passed down to compadd.
Like _tags this function supports the -C option to give a different name for the argument context field. The -x option has the same meaning as for _description.
There are some standard variables, initialised by the _main_complete function and then used from other functions.
The standard variables are:
Two more features are offered by the _main_complete function. The arrays compprefuncs and comppostfuncs may contain names of functions that are to be called immediately before or after completion has been tried. A function will only be called once unless it explicitly reinserts itself into the array.
In the source distribution, the files are contained in various subdirectories of the Completion directory. They may have been installed in the same structure, or into one single function directory. The following is a description of the files found in the original directory structure. If you wish to alter an installed file, you will need to copy it to some directory which appears earlier in your fpath than the standard directory where it appears.
Control the editor's completion behavior according to the supplied set of options. Various editing commands, notably expand-or-complete-word, usually bound to tab, will attempt to complete a word typed by the user, while others, notably delete-char-or-list, usually bound to ^D in EMACS editing mode, list the possibilities; compctl controls what those possibilities are. They may for example be filenames (the most common case, and hence the default), shell variables, or words from a user-specified list.
Any of the command strings may be patterns of the form normally used for filename generation. These should be quoted to protect them from immediate expansion; for example the command string 'foo*' arranges for completion of the words of any command beginning with foo. When completion is attempted, all pattern completions are tried in the reverse order of their definition until one matches. By default, completion then proceeds as normal, i.e. the shell will try to generate more matches for the specific command on the command line; this can be overridden by including -tn in the flags for the pattern completion.
Note that aliases are expanded before the command name is determined unless the COMPLETE_ALIASES option is set. Commands may not be combined with the -C, -D or -T flags.
compctl -T -x 's[~] C[0,[^/]#]' -k friends -S/ -tn
to complete the strings in the array friends after a `~'. The C[...] argument is necessary so that this form of ~-completion is not tried after the directory name is finished.
If the + flag is alone and followed immediately by the command list, the completion behavior for all the commands in the list is reset to the default. In other words, completion will subsequently use the options specified by the -D flag.
The form with -M as the first and only option defines global matching specifications (see zshcompwid). The match specifications given will be used for every completion attempt (only when using compctl, not with the new completion system) and are tried in the order in which they are defined until one generates at least one match. E.g.:
compctl -M '' 'm:{a-zA-Z}={A-Za-z}'
This will first try completion without any global match specifications (the empty string) and, if that generates no matches, will try case insensitive completion.
The remaining options specify the type of command arguments to look for during completion. Any combination of these flags may be specified; the result is a sorted list of all the possibilities. The options are as follows.
compctl -k "(cputime filesize datasize stacksize
coredumpsize resident descriptors)" limit
function whoson { reply=(`users`); }
compctl -K whoson talk
completes only logged-on users after `talk'. Note that `whoson' must return an array, so `reply=`users`' would be incorrect.
compctl -D -f + -H 0 ''
which forces completion to look back in the history list for a word if no filename matches.
compctl -j -P "%" kill
inserts a `%' after the kill command and then completes job names.
compctl -/ -W ~/Mail maildirs
completes any subdirectories to any depth beneath the directory ~/Mail, although that prefix does not appear on the command line. The file-prefix may also be of the form accepted by the -k flag, i.e. the name of an array or a literal list in parenthesis. In this case all the directories in the list will be searched for possible completions.
compctl -x 'r[-exec,;]' -l '' -- find
completes arguments between `-exec' and the following `;' (or the end of the command line if there is no such string) as if they were a separate command line.
Note that the returned list does not have to correspond, even in length, to the original set of matches, and may be passed as a scalar instead of an array. No special formatting of characters is performed on the output in this case; in particular, newlines are printed literally and if they appear output in columns is suppressed.
The sequences %B, %b, %S, %s, %U, and %u specify output attributes (bold, standout, and underline), %F, %f, %K, %k specify foreground and background colours, and %{...%} can be used to include literal escape sequences as in prompts.
(i) With -T, or when trying a list of pattern completions, when compctl would usually continue with ordinary processing after finding matches; this can be suppressed with `-tn'.
(ii) With a list of alternatives separated by +, when compctl would normally stop when one of the alternatives generates matches. It can be forced to consider the next set of completions by adding `-t+' to the flags of the alternative before the `+'.
(iii) In an extended completion list (see below), when compctl would normally continue until a set of conditions succeeded, then use only the immediately following flags. With `-t-', compctl will continue trying extended completions after the next `-'; with `-tx' it will attempt completion with the default flags, in other words those before the `-x'.
This can be useful with non-exclusive alternative completions. For example, in
compctl -f -J files -t+ + -v -J variables foo
both files and variables are possible completions, as the -t+ forces both sets of alternatives before and after the + to be considered at once. Because of the -J options, however, all files are listed before all variables.
The form with `+' specifies alternative options. Completion is tried with the options before the first `+'. If this produces no matches completion is tried with the flags after the `+' and so on. If there are no flags after the last `+' and a match has not been found up to that point, default completion is tried. If the list of flags contains a -t with a + character, the next list of flags is used even if the current list produced matches.
Additional options are available that restrict completion to some part of the command line; this is referred to as `extended completion'.
The form with `-x' specifies extended completion for the commands given; as shown, it may be combined with alternative completion using `+'. Each pattern is examined in turn; when a match is found, the corresponding options, as described in the section `Option Flags' above, are used to generate possible completions. If no pattern matches, the options given before the -x are used.
Note that each pattern should be supplied as a single argument and should be quoted to prevent expansion of metacharacters by the shell.
A pattern is built of sub-patterns separated by commas; it matches if at least one of these sub-patterns matches (they are `or'ed). These sub-patterns are in turn composed of other sub-patterns separated by white spaces which match if all of the sub-patterns match (they are `and'ed). An element of the sub-patterns is of the form `c[...][...]', where the pairs of brackets may be repeated as often as necessary, and matches if any of the sets of brackets match (an `or'). The example below makes this clearer.
The elements may be any of the following:
compctl -s '`users`' -x 'n[1,@]' -k hosts -- talk
will usually complete usernames, but if you insert an @ after the name, names from the array hosts (assumed to contain hostnames, though you must make the array yourself) will be completed. Other commands such as rcp can be handled similarly.
compctl -u -x 's[+] c[-1,-f],s[-f+]' \ -g '~/Mail/*(:t)' - 's[-f],c[-1,-f]' -f -- mail
This is to be interpreted as follows:
If the current command is mail, then
if ((the current word begins with + and the previous word is -f) or (the current word begins with -f+)), then complete the non-directory part (the `:t' glob modifier) of files in the directory ~/Mail; else if the current word begins with -f or the previous word was -f, then complete any file; else complete user names.
The modules that are bundled with the zsh distribution are:
The zsh/attr module is used for manipulating extended attributes. The -h option causes all commands to operate on symbolic links instead of their targets. The builtins in this module are:
zgetattr and zlistattr allocate memory dynamically. If the attribute or list of attributes grows between the allocation and the call to get them, they return 2. On all other errors, 1 is returned. This allows the calling function to check for this case and retry.
The zsh/cap module is used for manipulating POSIX.1e (POSIX.6) capability sets. If the operating system does not support this interface, the builtins defined by this module will do nothing. The builtins in this module are:
The zsh/clone module makes available one builtin command:
The return status of the builtin is zero in both shells if successful, and non-zero on error.
The target of clone should be an unused terminal, such as an unused virtual console or a virtual terminal created by
xterm -e sh -c 'trap : INT QUIT TSTP; tty;
while :; do sleep 100000000; done'
Some words of explanation are warranted about this long xterm command line: when doing clone on a pseudo-terminal, some other session ("session" meant as a unix session group, or SID) is already owning the terminal. Hence the cloned zsh cannot acquire the pseudo-terminal as a controlling tty. That means two things:
This does not apply when cloning to an unused vc.
Cloning to a used (and unprepared) terminal will result in two processes reading simultaneously from the same terminal, with input bytes going randomly to either process.
clone is mostly useful as a shell built-in replacement for openvt.
The zsh/compctl module makes available two builtin commands. compctl, is the old, deprecated way to control completions for ZLE. See zshcompctl(1). The other builtin command, compcall can be used in user-defined completion widgets, see zshcompwid(1).
The zsh/complete module makes available several builtin commands which can be used in user-defined completion widgets, see zshcompwid(1).
The zsh/complist module offers three extensions to completion listings: the ability to highlight matches in such a list, the ability to scroll through long lists and a different style of menu completion.
The parameters ZLS_COLORS and ZLS_COLOURS describe how matches are highlighted. To turn on highlighting an empty value suffices, in which case all the default values given below will be used. The format of the value of these parameters is the same as used by the GNU version of the ls command: a colon-separated list of specifications of the form `name=value'. The name may be one of the following strings, most of which specify file types for which the value will be used. The strings and their default values are:
Apart from these strings, the name may also be an asterisk (`*') followed by any string. The value given for such a string will be used for all files whose name ends with the string. The name may also be an equals sign (`=') followed by a pattern; the EXTENDED_GLOB option will be turned on for evaluation of the pattern. The value given for this pattern will be used for all matches (not just filenames) whose display string are matched by the pattern. Definitions for the form with the leading equal sign take precedence over the values defined for file types, which in turn take precedence over the form with the leading asterisk (file extensions).
The leading-equals form also allows different parts of the displayed strings to be colored differently. For this, the pattern has to use the `(#b)' globbing flag and pairs of parentheses surrounding the parts of the strings that are to be colored differently. In this case the value may consist of more than one color code separated by equal signs. The first code will be used for all parts for which no explicit code is specified and the following codes will be used for the parts matched by the sub-patterns in parentheses. For example, the specification `=(#b)(?)*(?)=0=3=7' will be used for all matches which are at least two characters long and will use the code `3' for the first character, `7' for the last character and `0' for the rest.
All three forms of name may be preceded by a pattern in parentheses. If this is given, the value will be used only for matches in groups whose names are matched by the pattern given in the parentheses. For example, `(g*)m*=43' highlights all matches beginning with `m' in groups whose names begin with `g' using the color code `43'. In case of the `lc', `rc', and `ec' codes, the group pattern is ignored.
Note also that all patterns are tried in the order in which they appear in the parameter value until the first one matches which is then used. Patterns may be matched against completions, descriptions (possibly with spaces appended for padding), or lines consisting of a completion followed by a description. For consistent coloring it may be necessary to use more than one pattern or a pattern with backreferences.
When printing a match, the code prints the value of lc, the value for the file-type or the last matching specification with a `*', the value of rc, the string to display for the match itself, and then the value of ec if that is defined or the values of lc, no, and rc if ec is not defined.
The default values are ISO 6429 (ANSI) compliant and can be used on vt100 compatible terminals such as xterms. On monochrome terminals the default values will have no visible effect. The colors function from the contribution can be used to get associative arrays containing the codes for ANSI terminals (see the section `Other Functions' in zshcontrib(1)). For example, after loading colors, one could use `$color[red]' to get the code for foreground color red and `$color[bg-green]' for the code for background color green.
If the completion system invoked by compinit is used, these parameters should not be set directly because the system controls them itself. Instead, the list-colors style should be used (see the section `Completion System Configuration' in zshcompsys(1)).
If the parameter LISTPROMPT is set, the completion code will not ask if the list should be shown. Instead it immediately starts displaying the list, stopping after the first screenful, showing the prompt at the bottom, waiting for a keypress after temporarily switching to the listscroll keymap. Some of the zle functions have a special meaning while scrolling lists:
Every other character stops listing and immediately processes the key as usual. Any key that is not bound in the listscroll keymap or that is bound to undefined-key is looked up in the keymap currently selected.
As for the ZLS_COLORS and ZLS_COLOURS parameters, LISTPROMPT should not be set directly when using the shell function based completion system. Instead, the list-prompt style should be used.
Menu selection can be invoked directly by the widget menu-select defined by this module. This is a standard ZLE widget that can be bound to a key in the usual way as described in zshzle(1).
Alternatively, the parameter MENUSELECT can be set to an integer, which gives the minimum number of matches that must be present before menu selection is automatically turned on. This second method requires that menu completion be started, either directly from a widget such as menu-complete, or due to one of the options MENU_COMPLETE or AUTO_MENU being set. If MENUSELECT is set, but is 0, 1 or empty, menu selection will always be started during an ambiguous menu completion.
When using the completion system based on shell functions, the MENUSELECT parameter should not be used (like the ZLS_COLORS and ZLS_COLOURS parameters described above). Instead, the menu style should be used with the select=... keyword.
After menu selection is started, the matches will be listed. If there are more matches than fit on the screen, only the first screenful is shown. The matches to insert into the command line can be selected from this list. In the list one match is highlighted using the value for ma from the ZLS_COLORS or ZLS_COLOURS parameter. The default value for this is `7' which forces the selected match to be highlighted using standout mode on a vt100-compatible terminal. If neither ZLS_COLORS nor ZLS_COLOURS is set, the same terminal control sequence as for the `%S' escape in prompts is used.
If there are more matches than fit on the screen and the parameter MENUPROMPT is set, its value will be shown below the matches. It supports the same escape sequences as LISTPROMPT, but the number of the match or line shown will be that of the one where the mark is placed. If its value is the empty string, a default prompt will be used.
The MENUSCROLL parameter can be used to specify how the list is scrolled. If the parameter is unset, this is done line by line, if it is set to `0' (zero), the list will scroll half the number of lines of the screen. If the value is positive, it gives the number of lines to scroll and if it is negative, the list will be scrolled the number of lines of the screen minus the (absolute) value.
As for the ZLS_COLORS, ZLS_COLOURS and LISTPROMPT parameters, neither MENUPROMPT nor MENUSCROLL should be set directly when using the shell function based completion system. Instead, the select-prompt and select-scroll styles should be used.
The completion code sometimes decides not to show all of the matches in the list. These hidden matches are either matches for which the completion function which added them explicitly requested that they not appear in the list (using the -n option of the compadd builtin command) or they are matches which duplicate a string already in the list (because they differ only in things like prefixes or suffixes that are not displayed). In the list used for menu selection, however, even these matches are shown so that it is possible to select them. To highlight such matches the hi and du capabilities in the ZLS_COLORS and ZLS_COLOURS parameters are supported for hidden matches of the first and second kind, respectively.
Selecting matches is done by moving the mark around using the zle movement functions. When not all matches can be shown on the screen at the same time, the list will scroll up and down when crossing the top or bottom line. The following zle functions have special meaning during menu selection. Note that the following always perform the same task within the menu selection map and cannot be replaced by user defined widgets, nor can the set of functions be extended:
All movement functions wrap around at the edges; any other zle function not listed leaves menu selection and executes that function. It is possible to make widgets in the above list do the same by using the form of the widget with a `.' in front. For example, the widget `.accept-line' has the effect of leaving menu selection and accepting the entire command line.
During this selection the widget uses the keymap menuselect. Any key that is not defined in this keymap or that is bound to undefined-key is looked up in the keymap currently selected. This is used to ensure that the most important keys used during selection (namely the cursor keys, return, and TAB) have sensible defaults. However, keys in the menuselect keymap can be modified directly using the bindkey builtin command (see zshmodules(1)). For example, to make the return key leave menu selection without accepting the match currently selected one could call
bindkey -M menuselect '^M' send-break
after loading the zsh/complist module.
The zsh/computil module adds several builtin commands that are used by some of the completion functions in the completion system based on shell functions (see zshcompsys(1) ). Except for compquote these builtin commands are very specialised and thus not very interesting when writing your own completion functions. In summary, these builtin commands are:
Once compdescribe has been called with either the -i or the -I option, it can be repeatedly called with the -g option and the names of four parameters as its arguments. This will step through the different sets of matches and store the value of compstate[list] in the first scalar, the options for compadd in the second array, the matches in the third array, and the strings to be displayed in the completion listing in the fourth array. The arrays may then be directly given to compadd to register the matches with the completion code.
The return status is non-zero in case of an error and zero otherwise.
The zsh/curses module makes available one builtin command and various parameters.
The subcommand addwin creates a window with nlines lines and ncols columns. Its upper left corner will be placed at row begin_y and column begin_x of the screen. targetwin is a string and refers to the name of a window that is not currently assigned. Note in particular the curses convention that vertical values appear before horizontal values.
If addwin is given an existing window as the final argument, the new window is created as a subwindow of parentwin. This differs from an ordinary new window in that the memory of the window contents is shared with the parent's memory. Subwindows must be deleted before their parent. Note that the coordinates of subwindows are relative to the screen, not the parent, as with other windows.
Use the subcommand delwin to delete a window created with addwin. Note that end does not implicitly delete windows, and that delwin does not erase the screen image of the window.
The window corresponding to the full visible screen is called stdscr; it always exists after `zcurses init' and cannot be delete with delwin.
The subcommand refresh will refresh window targetwin; this is necessary to make any pending changes (such as characters you have prepared for output with char) visible on the screen. refresh without an argument causes the screen to be cleared and redrawn. If multiple windows are given, the screen is updated once at the end.
The subcommand touch marks the targetwins listed as changed. This is necessary before refreshing windows if a window that was in front of another window (which may be stdscr) is deleted.
The subcommand move moves the cursor position in targetwin to new coordinates new_y and new_x. Note that the subcommand string (but not the subcommand char) advances the cursor position over the characters added.
The subcommand clear erases the contents of targetwin. One (and no more than one) of three options may be specified. With the option redraw, in addition the next refresh of targetwin will cause the screen to be cleared and repainted. With the option eol, targetwin is only cleared to the end of the current cursor line. With the option bot, targetwin is cleared to the end of the window, i.e everything to the right and below the cursor is cleared.
The subcommand position writes various positions associated with targetwin into the array named array. These are, in order:
Outputting characters and strings are achieved by char and string respectively.
To draw a border around window targetwin, use border. Note that the border is not subsequently handled specially: in other words, the border is simply a set of characters output at the edge of the window. Hence it can be overwritten, can scroll off the window, etc.
The subcommand attr will set targetwin's attributes or foreground/background color pair for any successive character output. Each attribute given on the line may be prepended by a + to set or a - to unset that attribute; + is assumed if absent. The attributes supported are blink, bold, dim, reverse, standout, and underline.
Each fg_col/bg_col attribute (to be read as `fg_col on bg_col') sets the foreground and background color for character output. The color default is sometimes available (in particular if the library is ncurses), specifying the foreground or background color with which the terminal started. The color pair default/default is always available. To use more than the 8 named colors (red, green, etc.) construct the fg_col/bg_col pairs where fg_col and bg_col are decimal integers, e.g 128/200. The maximum color value is 254 if the terminal supports 256 colors.
bg overrides the color and other attributes of all characters in the window. Its usual use is to set the background initially, but it will overwrite the attributes of any characters at the time when it is called. In addition to the arguments allowed with attr, an argument @char specifies a character to be shown in otherwise blank areas of the window. Owing to limitations of curses this cannot be a multibyte character (use of ASCII characters only is recommended). As the specified set of attributes override the existing background, turning attributes off in the arguments is not useful, though this does not cause an error.
The subcommand scroll can be used with on or off to enabled or disable scrolling of a window when the cursor would otherwise move below the window due to typing or output. It can also be used with a positive or negative integer to scroll the window up or down the given number of lines without changing the current cursor position (which therefore appears to move in the opposite direction relative to the window). In the second case, if scrolling is off it is temporarily turned on to allow the window to be scrolled.
The subcommand input reads a single character from the window without echoing it back. If param is supplied the character is assigned to the parameter param, else it is assigned to the parameter REPLY.
If both param and kparam are supplied, the key is read in `keypad' mode. In this mode special keys such as function keys and arrow keys return the name of the key in the parameter kparam. The key names are the macros defined in the curses.h or ncurses.h with the prefix `KEY_' removed; see also the description of the parameter zcurses_keycodes below. Other keys cause a value to be set in param as before. On a successful return only one of param or kparam contains a non-empty string; the other is set to an empty string.
If mparam is also supplied, input attempts to handle mouse input. This is only available with the ncurses library; mouse handling can be detected by checking for the exit status of `zcurses mouse' with no arguments. If a mouse button is clicked (or double- or triple-clicked, or pressed or released with a configurable delay from being clicked) then kparam is set to the string MOUSE, and mparam is set to an array consisting of the following elements:
Not all mouse events may be passed through to the terminal window; most terminal emulators handle some mouse events themselves. Note that the ncurses manual implies that using input both with and without mouse handling may cause the mouse cursor to appear and disappear.
The subcommand mouse can be used to configure the use of the mouse. There is no window argument; mouse options are global. `zcurses mouse' with no arguments returns status 0 if mouse handling is possible, else status 1. Otherwise, the possible arguments (which may be combined on the same command line) are as follows. delay num sets the maximum delay in milliseconds between press and release events to be considered as a click; the value 0 disables click resolution, and the default is one sixth of a second. motion proceeded by an optional `+' (the default) or - turns on or off reporting of mouse motion in addition to clicks, presses and releases, which are always reported. However, it appears reports for mouse motion are not currently implemented.
The subcommand timeout specifies a timeout value for input from targetwin. If intval is negative, `zcurses input' waits indefinitely for a character to be typed; this is the default. If intval is zero, `zcurses input' returns immediately; if there is typeahead it is returned, else no input is done and status 1 is returned. If intval is positive, `zcurses input' waits intval milliseconds for input and if there is none at the end of that period returns status 1.
The subcommand querychar queries the character at the current cursor position. The return values are stored in the array named param if supplied, else in the array reply. The first value is the character (which may be a multibyte character if the system supports them); the second is the color pair in the usual fg_col/bg_col notation, or 0 if color is not supported. Any attributes other than color that apply to the character, as set with the subcommand attr, appear as additional elements.
The subcommand resize resizes stdscr and all windows to given dimensions (windows that stick out from the new dimensions are resized down). The underlying curses extension (resize_term call) can be unavailable. To verify, zeroes can be used for height and width. If the result of the subcommand is 0, resize_term is available (2 otherwise). Tests show that resizing can be normally accomplished by calling zcurses end and zcurses refresh. The resize subcommand is provided for versatility. Multiple system configurations have been checked and zcurses end and zcurses refresh are still needed for correct terminal state after resize. To invoke them with resize, use endwin argument. Using nosave argument will cause new terminal state to not be saved internally by zcurses. This is also provided for versatility and should normally be not needed.
The zsh/datetime module makes available one builtin command:
In most implementations of strftime any timezone in the timestring is ignored and the local timezone declared by the TZ environment variable is used; other parameters are set to zero if not present.
If timestring does not match format the command returns status 1 and prints an error message. If timestring matches format but not all characters in timestring were used, the conversion succeeds but also prints an error message.
If either of the system functions strptime or mktime is not available, status 2 is returned and an error message is printed.
Note that depending on the system's declared integral time type, strftime may produce incorrect results for epoch times greater than 2147483647 which corresponds to 2038-01-19 03:14:07 +0000.
The zsh/datetime module makes available several parameters; all are readonly:
for secs nsecs in $epochtime; do ... done
The zsh/db/gdbm module is used to create "tied" associative arrays that interface to database files. If the GDBM interface is not available, the builtins defined by this module will report an error. This module is also intended as a prototype for creating additional database interfaces, so the ztie builtin may move to a more generic module in the future.
The builtins in this module are:
local -A sampledb ztie -d db/gdbm -f sample.gdbm sampledb
The -r option opens the database file for reading only, creating a parameter with the readonly attribute. Without this option, using `ztie' on a file for which the user does not have write permission is an error. If writable, the database is opened synchronously so fields changed in arrayname are immediately written to filename.
Changes to the file modes filename after it has been opened do not alter the state of arrayname, but `typeset -r arrayname' works as expected.
This happens automatically if the parameter is explicitly unset or its local scope (function) ends. Note that a readonly parameter may not be explicitly unset, so the only way to unset a global parameter created with `ztie -r' is to use `zuntie -u'.
The fields of an associative array tied to GDBM are neither cached nor otherwise stored in memory, they are read from or written to the database on each reference. Thus, for example, the values in a readonly array may be changed by a second writer of the same database file.
The zsh/deltochar module makes available two ZLE functions:
The zsh/example module makes available one builtin command:
The purpose of the module is to serve as an example of how to write a module.
The zsh/files module makes available some common commands for file manipulation as builtins; these commands are probably not needed for many normal situations but can be useful in emergency recovery situations with constrained resources. The commands do not implement all features now required by relevant standards committees.
For all commands, a variant beginning zf_ is also available and loaded automatically. Using the features capability of zmodload will let you load only those names you want. Note that it's possible to load only the builtins with zsh-specific names using the following command:
zmodload -m -F zsh/files b:zf_\*
The commands loaded by default are:
The specified mode must be in octal.
The -R option causes chmod to recursively descend into directories, changing the mode of all files in the directory after changing the mode of the directory itself.
The -s option is a zsh extension to chmod functionality. It enables paranoid behaviour, intended to avoid security problems involving a chmod being tricked into affecting files other than the ones intended. It will refuse to follow symbolic links, so that (for example) ``chmod 600 /tmp/foo/passwd'' can't accidentally chmod /etc/passwd if /tmp/foo happens to be a link to /etc. It will also check where it is after leaving directories, so that a recursive chmod of a deep directory tree can't end up recursively chmoding /usr as a result of directories being moved up the tree.
The user-spec can be in four forms:
In each case, the `:' may instead be a `.'. The rule is that if there is a `:' then the separator is `:', otherwise if there is a `.' then the separator is `.', otherwise there is no separator.
Each of user and group may be either a username (or group name, as appropriate) or a decimal user ID (group ID). Interpretation as a name takes precedence, if there is an all-numeric username (or group name).
If the target is a symbolic link, the -h option causes chown to set the ownership of the link instead of its target.
The -R option causes chown to recursively descend into directories, changing the ownership of all files in the directory after changing the ownership of the directory itself.
The -s option is a zsh extension to chown functionality. It enables paranoid behaviour, intended to avoid security problems involving a chown being tricked into affecting files other than the ones intended. It will refuse to follow symbolic links, so that (for example) ``chown luser /tmp/foo/passwd'' can't accidentally chown /etc/passwd if /tmp/foo happens to be a link to /etc. It will also check where it is after leaving directories, so that a recursive chown of a deep directory tree can't end up recursively chowning /usr as a result of directories being moved up the tree.
Normally, ln will not attempt to create hard links to directories. This check can be overridden using the -d option. Typically only the super-user can actually succeed in creating hard links to directories. This does not apply to symbolic links in any case.
By default, existing files cannot be replaced by links. The -i option causes the user to be queried about replacing existing files. The -f option causes existing files to be silently deleted, without querying. -f takes precedence.
The -h and -n options are identical and both exist for compatibility; either one indicates that if the target is a symlink then it should not be dereferenced. Typically this is used in combination with -sf so that if an existing link points to a directory then it will be removed, instead of followed. If this option is used with multiple filenames and the target is a symbolic link pointing to a directory then the result is an error.
By default, the user will be queried before replacing any file that the user cannot write to, but writable files will be silently removed. The -i option causes the user to be queried about replacing any existing files. The -f option causes any existing files to be silently deleted, without querying. -f takes precedence.
Note that this mv will not move files across devices. Historical versions of mv, when actual renaming is impossible, fall back on copying and removing files; if this behaviour is desired, use cp and rm manually. This may change in a future version.
Normally, rm will not remove directories (except with the -R or -r options). The -d option causes rm to try removing directories with unlink (see unlink(2)), the same method used for files. Typically only the super-user can actually succeed in unlinking directories in this way. -d takes precedence over -R and -r.
By default, the user will be queried before removing any file that the user cannot write to, but writable files will be silently removed. The -i option causes the user to be queried about removing any files. The -f option causes files to be silently deleted, without querying, and suppresses all error indications. -f takes precedence.
The -R and -r options cause rm to recursively descend into directories, deleting all files in the directory before removing the directory with the rmdir system call (see rmdir(2)).
The -s option is a zsh extension to rm functionality. It enables paranoid behaviour, intended to avoid common security problems involving a root-run rm being tricked into removing files other than the ones intended. It will refuse to follow symbolic links, so that (for example) ``rm /tmp/foo/passwd'' can't accidentally remove /etc/passwd if /tmp/foo happens to be a link to /etc. It will also check where it is after leaving directories, so that a recursive removal of a deep directory tree can't end up recursively removing /usr as a result of directories being moved up the tree.
The zsh/langinfo module makes available one parameter:
Your implementation may support a number of the following keys:
CODESET, D_T_FMT, D_FMT, T_FMT, RADIXCHAR, THOUSEP, YESEXPR, NOEXPR, CRNCYSTR, ABDAY_{1..7}, DAY_{1..7}, ABMON_{1..12}, MON_{1..12}, T_FMT_AMPM, AM_STR, PM_STR, ERA, ERA_D_FMT, ERA_D_T_FMT, ERA_T_FMT, ALT_DIGITS
The zsh/mapfile module provides one special associative array parameter of the same name.
When the array is accessed as a whole, the keys are the names of files in the current directory, and the values are empty (to save a huge overhead in memory). Thus ${(k)mapfile} has the same effect as the glob operator *(D), since files beginning with a dot are not special. Care must be taken with expressions such as rm ${(k)mapfile}, which will delete every file in the current directory without the usual `rm *' test.
The parameter mapfile may be made read-only; in that case, files referenced may not be written or deleted.
A file may conveniently be read into an array as one line per element with the form `array=("${(f@)mapfile[filename]}")'. The double quotes and the `@' are necessary to prevent empty lines from being removed. Note that if the file ends with a newline, the shell will split on the final newline, generating an additional empty field; this can be suppressed by using `array=("${(f@)${mapfile[filename]%$'\n'}}")'.
Although reading and writing of the file in question is efficiently handled, zsh's internal memory management may be arbitrarily baroque; however, mapfile is usually very much more efficient than anything involving a loop. Note in particular that the whole contents of the file will always reside physically in memory when accessed (possibly multiple times, due to standard parameter substitution operations). In particular, this means handling of sufficiently long files (greater than the machine's swap space, or than the range of the pointer type) will be incorrect.
No errors are printed or flagged for non-existent, unreadable, or unwritable files, as the parameter mechanism is too low in the shell execution hierarchy to make this convenient.
It is unfortunate that the mechanism for loading modules does not yet allow the user to specify the name of the shell parameter to be given the special behaviour.
The zsh/mathfunc module provides standard mathematical functions for use when evaluating mathematical formulae. The syntax agrees with normal C and FORTRAN conventions, for example,
(( f = sin(0.3) ))
assigns the sine of 0.3 to the parameter f.
Most functions take floating point arguments and return a floating point value. However, any necessary conversions from or to integer type will be performed automatically by the shell. Apart from atan with a second argument and the abs, int and float functions, all functions behave as noted in the manual page for the corresponding C function, except that any arguments out of range for the function in question will be detected by the shell and an error reported.
The following functions take a single floating point argument: acos, acosh, asin, asinh, atan, atanh, cbrt, ceil, cos, cosh, erf, erfc, exp, expm1, fabs, floor, gamma, j0, j1, lgamma, log, log10, log1p, log2, logb, sin, sinh, sqrt, tan, tanh, y0, y1. The atan function can optionally take a second argument, in which case it behaves like the C function atan2. The ilogb function takes a single floating point argument, but returns an integer.
The function signgam takes no arguments, and returns an integer, which is the C variable of the same name, as described in gamma(3). Note that it is therefore only useful immediately after a call to gamma or lgamma. Note also that `signgam()' and `signgam' are distinct expressions.
The functions min, max, and sum are defined not in this module but in the zmathfunc autoloadable function, described in the section `Mathematical Functions' in zshcontrib(1).
The following functions take two floating point arguments: copysign, fmod, hypot, nextafter.
The following take an integer first argument and a floating point second argument: jn, yn.
The following take a floating point first argument and an integer second argument: ldexp, scalb.
The function abs does not convert the type of its single argument; it returns the absolute value of either a floating point number or an integer. The functions float and int convert their arguments into a floating point or integer value (by truncation) respectively.
Note that the C pow function is available in ordinary math evaluation as the `**' operator and is not provided here.
The function rand48 is available if your system's mathematical library has the function erand48(3). It returns a pseudo-random floating point number between 0 and 1. It takes a single string optional argument.
If the argument is not present, the random number seed is initialised by three calls to the rand(3) function --- this produces the same random numbers as the next three values of $RANDOM.
If the argument is present, it gives the name of a scalar parameter where the current random number seed will be stored. On the first call, the value must contain at least twelve hexadecimal digits (the remainder of the string is ignored), or the seed will be initialised in the same manner as for a call to rand48 with no argument. Subsequent calls to rand48(param) will then maintain the seed in the parameter param as a string of twelve hexadecimal digits, with no base signifier. The random number sequences for different parameters are completely independent, and are also independent from that used by calls to rand48 with no argument.
For example, consider
print $(( rand48(seed) )) print $(( rand48() )) print $(( rand48(seed) ))
Assuming $seed does not exist, it will be initialised by the first call. In the second call, the default seed is initialised; note, however, that because of the properties of rand() there is a correlation between the seeds used for the two initialisations, so for more secure uses, you should generate your own 12-byte seed. The third call returns to the same sequence of random numbers used in the first call, unaffected by the intervening rand48().
The zsh/nearcolor module replaces colours specified as hex triplets with the nearest colour in the 88 or 256 colour palettes that are widely used by terminal emulators. By default, 24-bit true colour escape codes are generated when colours are specified using hex triplets. These are not supported by all terminals. The purpose of this module is to make it easier to define colour preferences in a form that can work across a range of terminal emulators.
Aside from the default colour, the ANSI standard for terminal escape codes provides for eight colours. The bright attribute brings this to sixteen. These basic colours are commonly used in terminal applications due to being widely supported. Expanded 88 and 256 colour palettes are also common and, while the first sixteen colours vary somewhat between terminals and configurations, these add a generally consistent and predictable set of colours.
In order to use the zsh/nearcolor module, it only needs to be loaded. Thereafter, whenever a colour is specified using a hex triplet, it will be compared against each of the available colours and the closest will be selected. The first sixteen colours are never matched in this process due to being unpredictable.
It isn't possible to reliably detect support for true colour in the terminal emulator. It is therefore recommended to be selective in loading the zsh/nearcolor module. For example, the following checks the COLORTERM environment variable:
[[ $COLORTERM = *(24bit|truecolor)* ]] || zmodload zsh/nearcolor
Note that some terminals accept the true color escape codes but map them internally to a more limited palette in a similar manner to the zsh/nearcolor module.
The zsh/newuser module is loaded at boot if it is available, the RCS option is set, and the PRIVILEGED option is not set (all three are true by default). This takes place immediately after commands in the global zshenv file (typically /etc/zshenv), if any, have been executed. If the module is not available it is silently ignored by the shell; the module may safely be removed from $MODULE_PATH by the administrator if it is not required.
On loading, the module tests if any of the start-up files .zshenv, .zprofile, .zshrc or .zlogin exist in the directory given by the environment variable ZDOTDIR, or the user's home directory if that is not set. The test is not performed and the module halts processing if the shell was in an emulation mode (i.e. had been invoked as some other shell than zsh).
If none of the start-up files were found, the module then looks for the file newuser first in a sitewide directory, usually the parent directory of the site-functions directory, and if that is not found the module searches in a version-specific directory, usually the parent of the functions directory containing version-specific functions. (These directories can be configured when zsh is built using the --enable-site-scriptdir=dir and --enable-scriptdir=dir flags to configure, respectively; the defaults are prefix/share/zsh and prefix/share/zsh/$ZSH_VERSION where the default prefix is /usr/local.)
If the file newuser is found, it is then sourced in the same manner as a start-up file. The file is expected to contain code to install start-up files for the user, however any valid shell code will be executed.
The zsh/newuser module is then unconditionally unloaded.
Note that it is possible to achieve exactly the same effect as the zsh/newuser module by adding code to /etc/zshenv. The module exists simply to allow the shell to make arrangements for new users without the need for intervention by package maintainers and system administrators.
The script supplied with the module invokes the shell function zsh-newuser-install. This may be invoked directly by the user even if the zsh/newuser module is disabled. Note, however, that if the module is not installed the function will not be installed either. The function is documented in the section User Configuration Functions in zshcontrib(1).
The zsh/parameter module gives access to some of the internal hash tables used by the shell by defining some special parameters.
For an autoloaded function that has already been loaded, or marked for autoload with an absolute path, or that has had its path resolved with `functions -r', this is the file found for autoloading, resolved to an absolute path.
For a function defined within the body of a script or sourced file, this is the name of that file. In this case, this is the exact path originally used to that file, which may be a relative path.
For any other function, including any defined at an interactive prompt or an autoload function whose path has not yet been resolved, this is the empty string. However, the hash element is reported as defined just so long as the function is present: the keys to this hash are the same as those to $functions.
Setting or unsetting keys in this array is not possible.
The keys of the associative arrays are usually valid job numbers, and these are the values output with, for example, ${(k)jobdirs}. Non-numeric job references may be used when looking up a value; for example, ${jobdirs[%+]} refers to the current job.
Handling of the keys of the associative array is as described for jobdirs above.
Handling of the keys of the associative array is as described for jobdirs above.
For functions autoloaded from a file in native zsh format, where only the body of the function occurs in the file, or for files that have been executed by the source or `.' builtins, the trace information is shown as filename:0, since the entire file is the definition. The source file name is resolved to an absolute path when the function is loaded or the path to it otherwise resolved.
Most users will be interested in the information in the funcfiletrace array instead.
The standard shell array zsh_eval_context can be used to determine the type of shell construct being executed at each depth: note, however, that is in the opposite order, with the most recent item last, and it is more detailed, for example including an entry for toplevel, the main shell code being executed either interactively or from a script, which is not present in $funcstack.
The zsh/pcre module makes some commands available as builtins:
Option -a will force the pattern to be anchored. Option -i will compile a case-insensitive pattern. Option -m will compile a multi-line pattern; that is, ^ and $ will match newlines within the pattern. Option -x will compile an extended pattern, wherein whitespace and # comments are ignored. Option -s makes the dot metacharacter match all characters, including those that indicate newline.
Upon successful match, if the expression captures substrings within parentheses, pcre_match will set the array match to those substrings, unless the -a option is given, in which case it will set the array arr. Similarly, the variable MATCH will be set to the entire matched portion of the string, unless the -v option is given, in which case the variable var will be set. No variables are altered if there is no successful match. A -n option starts searching for a match from the byte offset position in string. If the -b option is given, the variable ZPCRE_OP will be set to an offset pair string, representing the byte offset positions of the entire matched portion within the string. For example, a ZPCRE_OP set to "32 45" indicates that the matched portion began on byte offset 32 and ended on byte offset 44. Here, byte offset position 45 is the position directly after the matched portion. Keep in mind that the byte position isn't necessarily the same as the character position when UTF-8 characters are involved. Consequently, the byte offset positions are only to be relied on in the context of using them for subsequent searches on string, using an offset position as an argument to the -n option. This is mostly used to implement the "find all non-overlapping matches" functionality.
A simple example of "find all non-overlapping matches":
string="The following zip codes: 78884 90210 99513"
pcre_compile -m "\d{5}"
accum=()
pcre_match -b -- $string
while [[ $? -eq 0 ]] do
b=($=ZPCRE_OP)
accum+=$MATCH
pcre_match -b -n $b[2] -- $string
done
print -l $accum
The zsh/pcre module makes available the following test condition:
For example,
[[ "$text" -pcre-match ^d+$ ]] && print text variable contains only "d's".
If the REMATCH_PCRE option is set, the =~ operator is equivalent to -pcre-match, and the NO_CASE_MATCH option may be used. Note that NO_CASE_MATCH never applies to the pcre_match builtin, instead use the -i switch of pcre_compile.
The zsh/param/private module is used to create parameters whose scope is limited to the current function body, and not to other functions called by the current function.
This module provides a single autoloaded builtin:
If used at the top level (outside a function scope), private creates a normal parameter in the same manner as declare or typeset. A warning about this is printed if WARN_CREATE_GLOBAL is set (zshoptions(1)). Used inside a function scope, private creates a local parameter similar to one declared with local, except having special properties noted below.
Special parameters which expose or manipulate internal shell state, such as ARGC, argv, COLUMNS, LINES, UID, EUID, IFS, PROMPT, RANDOM, SECONDS, etc., cannot be made private unless the `-h' option is used to hide the special meaning of the parameter. This may change in the future.
As with other typeset equivalents, private is both a builtin and a reserved word, so arrays may be assigned with parenthesized word list name=(value...) syntax. However, the reserved word `private' is not available until zsh/param/private is loaded, so care must be taken with order of execution and parsing for function definitions which use private. To compensate for this, the module also adds the option `-P' to the `local' builtin to declare private parameters.
For example, this construction fails if zsh/param/private has not yet been loaded when `bad_declaration' is defined:
bad_declaration() {
zmodload zsh/param/private
private array=( one two three )
}
This construction works because local is already a keyword, and the module is loaded before the statement is executed:
good_declaration() {
zmodload zsh/param/private
local -P array=( one two three )
}
The following is usable in scripts but may have trouble with autoload:
zmodload zsh/param/private
iffy_declaration() {
private array=( one two three )
}
The private builtin may always be used with scalar assignments and for declarations without assignments.
Parameters declared with private have the following properties:
Note that this differs from the static scope defined by compiled languages derived from C, in that the a new call to the same function creates a new scope, i.e., the parameter is still associated with the call stack rather than with the function definition. It differs from ksh `typeset -S' because the syntax used to define the function has no bearing on whether the parameter scope is respected.
The zsh/regex module makes available the following test condition:
For example,
[[ alphabetical -regex-match ^a([^a]+)a([^a]+)a ]] && print -l $MATCH X $match
If the option REMATCH_PCRE is not set, then the =~ operator will automatically load this module as needed and will invoke the -regex-match operator.
If BASH_REMATCH is set, then the array BASH_REMATCH will be set instead of MATCH and match.
The zsh/sched module makes available one builtin command and one parameter.
With no arguments, prints the list of scheduled commands. If the scheduled command has the -o flag set, this is shown at the start of the command.
With the argument `-item', removes the given item from the list. The numbering of the list is continuous and entries are in time order, so the numbering can change when entries are added or deleted.
Commands are executed either immediately before a prompt, or while the shell's line editor is waiting for input. In the latter case it is useful to be able to produce output that does not interfere with the line being edited. Providing the option -o causes the shell to clear the command line before the event and redraw it afterwards. This should be used with any scheduled event that produces visible output to the terminal; it is not needed, for example, with output that updates a terminal emulator's title bar.
To effect changes to the editor buffer when an event executes, use the `zle' command with no arguments to test whether the editor is active, and if it is, then use `zle widget' to access the editor via the named widget.
The sched builtin is not made available by default when the shell starts in a mode emulating another shell. It can be made available with the command `zmodload -F zsh/sched b:sched'.
The sched builtin should be used for manipulating the events. Note that this will have an immediate effect on the contents of the array, so that indices may become invalid.
The zsh/net/socket module makes available one builtin command:
If -d is specified, its argument will be taken as the target file descriptor for the connection.
In order to elicit more verbose output, use -v.
File descriptors can be closed with normal shell syntax when no longer needed, for example:
exec {REPLY}>&-
If -d is specified, its argument will be taken as the target file descriptor for the connection.
In order to elicit more verbose output, use -v.
If -d is specified, its argument will be taken as the target file descriptor for the connection.
If -t is specified, zsocket will return if no incoming connection is pending. Otherwise it will wait for one.
In order to elicit more verbose output, use -v.
The zsh/stat module makes available one builtin command under two possible names:
If the stat call fails, the appropriate system error message printed and status 1 is returned. The fields of struct stat give information about the files provided as arguments to the command. In addition to those available from the stat call, an extra element `link' is provided. These elements are:
A particular element may be selected by including its name preceded by a `+' in the option list; only one element is allowed. The element may be shortened to any unique set of leading characters. Otherwise, all elements will be shown for all files.
Options:
The zsh/system module makes available various builtin commands and parameters.
Instead of the error number, a name errname, for example ENOENT, may be used. The set of names is the same as the contents of the array errnos, see below.
If the string prefix is given, it is printed in front of the error message, with no intervening space.
If errvar is supplied, the entire message, without a newline, is assigned to the parameter names errvar and nothing is output.
A return status of 0 indicates the message was successfully printed (although it may not be useful if the error number was out of the system's range), a return status of 1 indicates an error in the parameters, and a return status of 2 indicates the error name was not recognised (no message is printed for this).
The -o option allows various system specific options to be specified as a comma-separated list. The following is a list of possible options. Note that, depending on the system, some may not be available.
To close the file, use one of the following:
exec {fd}<&-
exec {fd}>&-
The maximum number of bytes read is bufsize or 8192 if that is not given, however the command returns as soon as any number of bytes was successfully read.
If timeout is given, it specifies a timeout in seconds, which may be zero to poll the file descriptor. This is handled by the poll system call if available, otherwise the select system call if available.
If outfd is given, an attempt is made to write all the bytes just read to the file descriptor outfd. If this fails, because of a system error other than EINTR or because of an internal zsh error during an interrupt, the bytes read but not written are stored in the parameter named by param if supplied (no default is used in this case), and the number of bytes read but not written is stored in the parameter named by countvar if that is supplied. If it was successful, countvar contains the full number of bytes transferred, as usual, and param is not set.
The error EINTR (interrupted system call) is handled internally so that shell interrupts are transparent to the caller. Any other error causes a return.
The possible return statuses are
If countvar is given, the number of byte written is stored in the parameter named by countvar; this may not be the full length of data if an error occurred.
The error EINTR (interrupted system call) is handled internally by retrying; otherwise an error causes the command to return. For example, if the file descriptor is set to non-blocking output, an error EAGAIN (on some systems, EWOULDBLOCK) may result in the command returning early.
The return status may be 0 for success, 1 for an error in the parameters to the command, or 2 for an error on the write; no error message is printed in the last case, but the parameter ERRNO will reflect the error that occurred.
In the first form the named file, which must already exist, is locked by opening a file descriptor to the file and applying a lock to the file descriptor. The lock terminates when the shell process that created the lock exits; it is therefore often convenient to create file locks within subshells, since the lock is automatically released when the subshell exits. Note that use of the print builtin with the -u option will, as a side effect, release the lock, as will redirection to the file in the shell holding the lock. To work around this use a subshell, e.g. `(print message) >> file'. Status 0 is returned if the lock succeeds, else status 1.
In the second form the file descriptor given by the arithmetic expression fd_expr is closed, releasing a lock. The file descriptor can be queried by using the `-f var' form during the lock; on a successful lock, the shell variable var is set to the file descriptor used for locking. The lock will be released if the file descriptor is closed by any other means, for example using `exec {var}>&-'; however, the form described here performs a safety check that the file descriptor is in use for file locking.
By default the shell waits indefinitely for the lock to succeed. The option -t timeout specifies a timeout for the lock in seconds; currently this must be an integer. The shell will attempt to lock the file once a second during this period. If the attempt times out, status 2 is returned.
If the option -e is given, the file descriptor for the lock is preserved when the shell uses exec to start a new process; otherwise it is closed at that point and the lock released.
If the option -r is given, the lock is only for reading, otherwise it is for reading and writing. The file descriptor is opened accordingly.
Note that aliases for errors are not handled; only the canonical name is used.
The zsh/net/tcp module makes available one builtin command:
If ztcp is run with no options, it will output the contents of its session table.
If it is run with only the option -L, it will output the contents of the session table in a format suitable for automatic parsing. The option is ignored if given with a command to open or close a session. The output consists of a set of lines, one per session, each containing the following elements separated by spaces:
If -d is specified, its argument will be taken as the target file descriptor for the connection.
In order to elicit more verbose output, use -v.
If -d is specified, its argument will be taken as the target file descriptor for the connection.
In order to elicit more verbose output, use -v.
If -d is specified, its argument will be taken as the target file descriptor for the connection.
If -t is specified, ztcp will return if no incoming connection is pending. Otherwise it will wait for one.
In order to elicit more verbose output, use -v.
Normally, sockets registered by zftp (see zshmodules(1) ) cannot be closed this way. In order to force such a socket closed, use -f.
In order to elicit more verbose output, use -v.
On host1,
zmodload zsh/net/tcp ztcp -l 5123 listenfd=$REPLY ztcp -a $listenfd fd=$REPLY
Now create a connection from host2 (which may, of course, be the same machine):
zmodload zsh/net/tcp ztcp host1 5123 fd=$REPLY
Now on each host, $fd contains a file descriptor for talking to the other. For example, on host1:
print This is a message >&$fd
read -r line <&$fd; print -r - $line
To tidy up, on host1:
ztcp -c $listenfd ztcp -c $fd
ztcp -c $fd
The zsh/termcap module makes available one builtin command:
The zsh/termcap module makes available one parameter:
The zsh/terminfo module makes available one builtin command:
The zsh/terminfo module makes available one parameter:
The zsh/zftp module makes available one builtin command:
All commands consist of the command name zftp followed by the name of a subcommand. These are listed below. The return status of each subcommand is supposed to reflect the success or failure of the remote operation. See a description of the variable ZFTP_VERBOSE for more information on how responses from the server may be printed.
If IPv6 addresses in colon format are used, the host should be surrounded by quoted square brackets to distinguish it from the port, for example '[fe80::203:baff:fe02:8b56]'. For consistency this is allowed with all forms of host.
Remaining arguments are passed to the login subcommand. Note that if no arguments beyond host are supplied, open will not automatically call login. If no arguments at all are supplied, open will use the parameters set by the params subcommand.
After a successful open, the shell variables ZFTP_HOST, ZFTP_PORT, ZFTP_IP and ZFTP_SYSTEM are available; see `Variables' below.
After a successful login, the shell variables ZFTP_USER, ZFTP_ACCOUNT and ZFTP_PWD are available; see `Variables' below.
This command may be re-issued when a user is already logged in, and the server will first be reinitialized for a new user.
Any of the parameters may be specified as a `?', which may need to be quoted to protect it from shell expansion. In this case, the appropriate parameter will be read from stdin as with the login subcommand, including special handling of password. If the `?' is followed by a string, that is used as the prompt for reading the parameter instead of the default message (any necessary punctuation and whitespace should be included at the end of the prompt). The first letter of the parameter (only) may be quoted with a `\'; hence an argument "\\$word" guarantees that the string from the shell parameter $word will be treated literally, whether or not it begins with a `?'.
If instead a single `-' is given, the existing parameters, if any, are deleted. In that case, calling open with no arguments will cause an error.
The list of parameters is not deleted after a close, however it will be deleted if the zsh/zftp module is unloaded.
For example,
zftp params ftp.elsewhere.xx juser '?Password for juser: '
will store the host ftp.elsewhere.xx and the user juser and then prompt the user for the corresponding password with the given prompt.
The FTP default for a transfer is ASCII. However, if zftp finds that the remote host is a UNIX machine with 8-bit byes, it will automatically switch to using binary for file transfers upon open. This can subsequently be overridden.
The transfer type is only passed to the remote host when a data connection is established; this command involves no network overhead.
Not all servers support the commands for retrieving this information. In that case, the remote command will print nothing and return status 2, compared with status 1 for a file not found.
The local command (but not remote) may be used with no arguments, in which case the information comes from examining file descriptor zero. This is the same file as seen by a put command with no further redirection.
Each session remembers the status of the connection, the set of connection-specific shell parameters (the same set as are unset when a connection closes, as given in the description of close), and any user parameters specified with the params subcommand. Changing to a previous session restores those values; changing to a new session initialises them in the same way as if zftp had just been loaded. The name of the current session is given by the parameter ZFTP_SESSION.
In the unlikely event that your system does not have the appropriate conversion functions, this appears in network byte order. If your system is little-endian, the port then consists of two swapped bytes and the standard port will be reported as 5376. In that case, numeric ports passed to zftp open will also need to be in this format.
If ZFTP_PREFS is not set when zftp is loaded, it will be set to a default of `PS', i.e. use passive mode if available, otherwise fall back to sendport mode.
It should be noted that, for unknown reasons, the reply `Service not available', which forces termination of a connection, is classified as 421, i.e. `transient negative', an interesting interpretation of the word `transient'.
The code 0 is special: it indicates that all but the last line of multiline replies read from the server will be printed to standard error in a processed format. By convention, servers use this mechanism for sending information for the user to read. The appropriate reply code, if it matches the same response, takes priority.
If ZFTP_VERBOSE is not set when zftp is loaded, it will be set to the default value 450, i.e., messages destined for the user and all errors will be printed. A null string is valid and specifies that no messages should be printed.
When it is called with a transfer in progress, the following additional shell parameters are set:
The function is initially called with ZFTP_TRANSFER set appropriately and ZFTP_COUNT set to zero. After the transfer is finished, the function will be called one more time with ZFTP_TRANSFER set to GF or PF, in case it wishes to tidy up. It is otherwise never called twice with the same value of ZFTP_COUNT.
Sometimes the progress meter may cause disruption. It is up to the user to decide whether the function should be defined and to use unfunction when necessary.
A connection may not be opened in the left hand side of a pipe as this occurs in a subshell and the file information is not updated in the main shell. In the case of type or mode changes or closing the connection in a subshell, the information is returned but variables are not updated until the next call to zftp. Other status changes in subshells will not be reflected by changes to the variables (but should be otherwise harmless).
Deleting sessions while a zftp command is active in the background can have unexpected effects, even if it does not use the session being deleted. This is because all shell subprocesses share information on the state of all connections, and deleting a session changes the ordering of that information.
On some operating systems, the control connection is not valid after a fork(), so that operations in subshells, on the left hand side of a pipeline, or in the background are not possible, as they should be. This is presumably a bug in the operating system.
The zsh/zle module contains the Zsh Line Editor. See zshzle(1).
The zsh/zleparameter module defines two special parameters that can be used to access internal information of the Zsh Line Editor (see zshzle(1)).
When loaded, the zsh/zprof causes shell functions to be profiled. The profiling results can be obtained with the zprof builtin command made available by this module. There is no way to turn profiling off other than unloading the module.
At the top there is a summary listing all functions that were called at least once. This summary is sorted in decreasing order of the amount of time spent in each. The lines contain the number of the function in order, which is used in other parts of the list in suffixes of the form `[num]', then the number of calls made to the function. The next three columns list the time in milliseconds spent in the function and its descendants, the average time in milliseconds spent in the function and its descendants per call and the percentage of time spent in all shell functions used in this function and its descendants. The following three columns give the same information, but counting only the time spent in the function itself. The final column shows the name of the function.
After the summary, detailed information about every function that was invoked is listed, sorted in decreasing order of the amount of time spent in each function and its descendants. Each of these entries consists of descriptions for the functions that called the function described, the function itself, and the functions that were called from it. The description for the function itself has the same format as in the summary (and shows the same information). The other lines don't show the number of the function at the beginning and have their function named indented to make it easier to distinguish the line showing the function described in the section from the surrounding lines.
The information shown in this case is almost the same as in the summary, but only refers to the call hierarchy being displayed. For example, for a calling function the column showing the total running time lists the time spent in the described function and its descendants only for the times when it was called from that particular calling function. Likewise, for a called function, this columns lists the total time spent in the called function and its descendants only for the times when it was called from the function described.
Also in this case, the column showing the number of calls to a function also shows a slash and then the total number of invocations made to the called function.
As long as the zsh/zprof module is loaded, profiling will be done and multiple invocations of the zprof builtin command will show the times and numbers of calls since the module was loaded. With the -c option, the zprof builtin command will reset its internal counters and will not show the listing.
The zsh/zpty module offers one builtin:
With the -e option, the pseudo-terminal is set up so that input characters are echoed.
With the -b option, input to and output from the pseudo-terminal are made non-blocking.
The shell parameter REPLY is set to the file descriptor assigned to the master side of the pseudo-terminal. This allows the terminal to be monitored with ZLE descriptor handlers (see zshzle(1)) or manipulated with sysread and syswrite (see THE ZSH/SYSTEM MODULE in zshmodules(1)). Warning: Use of sysread and syswrite is not recommended; use zpty -r and zpty -w unless you know exactly what you are doing.
If no string is provided, the standard input is copied to the pseudo-terminal; this may stop before copying the full input if the pseudo-terminal is non-blocking. The exact input is always copied: the -n option is not applied.
Note that the command under the pseudo-terminal sees this input as if it were typed, so beware when sending special tty driver characters such as word-erase, line-kill, and end-of-file.
When also given a param argument, at most one line is read and stored in the parameter named param. Less than a full line may be read if the pseudo-terminal is non-blocking. The return status is zero if at least one character is stored in param.
If a pattern is given as well, output is read until the whole string read matches the pattern, even in the non-blocking case. The return status is zero if the string read matches the pattern, or if the command has exited but at least one character could still be read. If the option -m is present, the return status is zero only if the pattern matches. As of this writing, a maximum of one megabyte of output can be consumed this way; if a full megabyte is read without matching the pattern, the return status is non-zero.
In all cases, the return status is non-zero if nothing could be read, and is 2 if this is because the command has finished.
If the -r option is combined with the -t option, zpty tests whether output is available before trying to read. If no output is available, zpty immediately returns the status 1. When used with a pattern, the behaviour on a failed poll is similar to when the command has exited: the return value is zero if at least one character could still be read even if the pattern failed to match.
The zsh/zselect module makes available one builtin command:
Arguments and options may be intermingled in any order. Non-option arguments are file descriptors, which must be decimal integers. By default, file descriptors are to be tested for reading, i.e. zselect will return when data is available to be read from the file descriptor, or more precisely, when a read operation from the file descriptor will not block. After a -r, -w and -e, the given file descriptors are to be tested for reading, writing, or error conditions. These options and an arbitrary list of file descriptors may be given in any order.
(The presence of an `error condition' is not well defined in the documentation for many implementations of the select system call. According to recent versions of the POSIX specification, it is really an exception condition, of which the only standard example is out-of-band data received on a socket. So zsh users are unlikely to find the -e option useful.)
The option `-t timeout' specifies a timeout in hundredths of a second. This may be zero, in which case the file descriptors will simply be polled and zselect will return immediately. It is possible to call zselect with no file descriptors and a non-zero timeout for use as a finer-grained replacement for `sleep'; note, however, the return status is always 1 for a timeout.
The option `-a array' indicates that array should be set to indicate the file descriptor(s) which are ready. If the option is not given, the array reply will be used for this purpose. The array will contain a string similar to the arguments for zselect. For example,
zselect -t 0 -r 0 -w 1
might return immediately with status 0 and $reply containing `-r 0 -w 1' to show that both file descriptors are ready for the requested operations.
The option `-A assoc' indicates that the associative array assoc should be set to indicate the file descriptor(s) which are ready. This option overrides the option -a, nor will reply be modified. The keys of assoc are the file descriptors, and the corresponding values are any of the characters `rwe' to indicate the condition.
The command returns status 0 if some file descriptors are ready for reading. If the operation timed out, or a timeout of 0 was given and no file descriptors were ready, or there was an error, it returns status 1 and the array will not be set (nor modified in any way). If there was an error in the select operation the appropriate error message is printed.
The zsh/zutil module only adds some builtins:
A pattern is considered to be more specific than another if it contains more components (substrings separated by colons) or if the patterns for the components are more specific, where simple strings are considered to be more specific than patterns and complex patterns are considered to be more specific than the pattern `*'. A `*' in the pattern will match zero or more characters in the context; colons are not treated specially in this regard. If two patterns are equally specific, the tie is broken in favour of the pattern that was defined first.
Example
For example, to define your preferred form of precipitation depending on which city you're in, you might set the following in your zshrc:
zstyle ':weather:europe:*' preferred-precipitation rain zstyle ':weather:europe:germany:* preferred-precipitation none zstyle ':weather:europe:germany:*:munich' preferred-precipitation snow
Then, the fictional `weather' plugin might run under the hood a command such as
zstyle -s ":weather:${continent}:${country}:${county}:${city}" preferred-precipitation REPLY
in order to retrieve your preference into the scalar variable $REPLY.
Usage
The forms that operate on patterns are the following.
If the -L option is given, listing is done in the form of calls to zstyle. The optional first argument, metapattern, is a pattern which will be matched against the string supplied as pattern when the style was defined. Note: this means, for example, `zstyle -L ":completion:*"' will match any supplied pattern beginning `:completion:', not just ":completion:*": use ':completion:\*' to match that. The optional second argument limits the output to a specific style (not a pattern). -L is not compatible with any other options.
The other forms can be used to look up or test styles for a given context.
Return 0 if the style is set, 1 otherwise.
Return 0 if name is set to `yes', 1 otherwise.
Return 0 if the style is set, 1 otherwise.
The -T option tests the values of the style like -t, but it returns status zero (rather than 2) if the style is not defined for any matching pattern.
The %-escapes also understand ternary expressions in the form used by prompts. The % is followed by a `(' and then an ordinary format specifier character as described above. There may be a set of digits either before or after the `('; these specify a test number, which defaults to zero. Negative numbers are also allowed. An arbitrary delimiter character follows the format specifier, which is followed by a piece of `true' text, the delimiter character again, a piece of `false' text, and a closing parenthesis. The complete expression (without the digits) thus looks like `%(X.text1.text2)', except that the `.' character is arbitrary. The value given for the format specifier in the char:string expressions is evaluated as a mathematical expression, and compared with the test number. If they are the same, text1 is output, else text2 is output. A parenthesis may be escaped in text2 as %). Either of text1 or text2 may contain nested %-escapes.
For example:
zformat -f REPLY "The answer is '%3(c.yes.no)'." c:3
outputs "The answer is 'yes'." to REPLY since the value for the format specifier c is 3, agreeing with the digit argument to the ternary expression.
The second form, using the -a option, can be used for aligning strings. Here, the specs are of the form `left:right' where `left' and `right' are arbitrary strings. These strings are modified by replacing the colons by the sep string and padding the left strings with spaces to the right so that the sep strings in the result (and hence the right strings after them) are all aligned if the strings are printed below each other. All strings without a colon are left unchanged and all strings with an empty right string have the trailing colon removed. In both cases the lengths of the strings are not used to determine how the other strings are to be aligned. A colon in the left string can be escaped with a backslash. The resulting strings are stored in the array.
Note that it is an error to give any spec without an `=array' unless one of the -a or -A options is used.
Unless the -E option is given, parsing stops at the first string that isn't described by one of the specs. Even with -E, parsing always stops at a positional parameter equal to `-' or `--'. See also -F.
The opt description must be one of the following. Any of the special characters can appear in the option name provided it is preceded by a backslash.
If a `+' appears after name, the option is appended to array each time it is found in the positional parameters; without the `+' only the last occurrence of the option is preserved.
If one of these forms is used, the option takes no argument, so parsing stops if the next positional parameter does not also begin with `-' (unless the -E option is used).
An optional argument is put into the same array element as the option name (note that this makes empty strings as arguments indistinguishable). A mandatory argument is added as a separate element unless the `:-' form is used, in which case the argument is put into the same element.
A `+' as described above may appear between the name and the first colon.
In all cases, option-arguments must appear either immediately following the option in the same positional parameter or in the next one. Even an optional argument may appear in the next parameter, unless it begins with a `-'. There is no special handling of `=' as with GNU-style argument parsers; given the spec `-foo:', the positional parameter `--foo=bar' is parsed as `--foo' with an argument of `=bar'.
When the names of two options that take no arguments overlap, the longest one wins, so that parsing for the specs `-foo -foobar' (for example) is unambiguous. However, due to the aforementioned handling of option-arguments, ambiguities may arise when at least one overlapping spec takes an argument, as in `-foo: -foobar'. In that case, the last matching spec wins.
The options of zparseopts itself cannot be stacked because, for example, the stack `-DEK' is indistinguishable from a spec for the GNU-style long option `--DEK'. The options of zparseopts itself are:
Note that the appearance in the positional parameters of an option without its required argument always aborts parsing and returns an error as described above regardless of whether this option is used.
For example,
set -- -a -bx -c y -cz baz -cend zparseopts a=foo b:=bar c+:=bar
will have the effect of
foo=(-a) bar=(-b x -c y -c z)
The arguments from `baz' on will not be used.
As an example for the -E option, consider:
set -- -a x -b y -c z arg1 arg2 zparseopts -E -D b:=bar
will have the effect of
bar=(-b y) set -- -a x -c z arg1 arg2
I.e., the option -b and its arguments are taken from the positional parameters and put into the array bar.
The -M option can be used like this:
set -- -a -bx -c y -cz baz -cend zparseopts -A bar -M a=foo b+: c:=b
to have the effect of
foo=(-a) bar=(-a '' -b xyz)
The shell is supplied with a series of functions to replace and enhance the traditional Unix calendar programme, which warns the user of imminent or future events, details of which are stored in a text file (typically calendar in the user's home directory). The version provided here includes a mechanism for alerting the user when an event is due.
In addition functions age, before and after are provided that can be used in a glob qualifier; they allow files to be selected based on their modification times.
The format of the calendar file and the dates used there in and in the age function are described first, then the functions that can be called to examine and modify the calendar file.
The functions here depend on the availability of the zsh/datetime module which is usually installed with the shell. The library function strptime() must be available; it is present on most recent operating systems.
The calendar file is by default ~/calendar. This can be configured by the calendar-file style, see the section STYLES below. The basic format consists of a series of separate lines, with no indentation, each including a date and time specification followed by a description of the event.
Various enhancements to this format are supported, based on the syntax of Emacs calendar mode. An indented line indicates a continuation line that continues the description of the event from the preceding line (note the date may not be continued in this way). An initial ampersand (&) is ignored for compatibility.
An indented line on which the first non-whitespace character is # is not displayed with the calendar entry, but is still scanned for information. This can be used to hide information useful to the calendar system but not to the user, such as the unique identifier used by calendar_add.
The Emacs extension that a date with no description may refer to a number of succeeding events at different times is not supported.
Unless the done-file style has been altered, any events which have been processed are appended to the file with the same name as the calendar file with the suffix .done, hence ~/calendar.done by default.
An example is shown below.
The format of the date and time is designed to allow flexibility without admitting ambiguity. (The words `date' and `time' are both used in the documentation below; except where specifically noted this implies a string that may include both a date and a time specification.) Note that there is no localization support; month and day names must be in English and separator characters are fixed. Matching is case insensitive, and only the first three letters of the names are significant, although as a special case a form beginning "month" does not match "Monday". Furthermore, time zones are not handled; all times are assumed to be local.
It is recommended that, rather than exploring the intricacies of the system, users find a date format that is natural to them and stick to it. This will avoid unexpected effects. Various key facts should be noted.
The following give some obvious examples; users finding here a format they like and not subject to vagaries of style may skip the full description. As dates and times are matched separately (even though the time may be embedded in the date), any date format may be mixed with any format for the time of day provide the separators are clear (whitespace, colons, commas).
2007/04/03 13:13 2007/04/03:13:13 2007/04/03 1:13 pm 3rd April 2007, 13:13 April 3rd 2007 1:13 p.m. Apr 3, 2007 13:13 Tue Apr 03 13:13:00 2007 13:13 2007/apr/3
More detailed rules follow.
Times are parsed and extracted before dates. They must use colons to separate hours and minutes, though a dot is allowed before seconds if they are present. This limits time formats to the following:
Here, square brackets indicate optional elements, possibly with alternatives. Fractions of a second are recognised but ignored. For absolute times (the normal format require by the calendar file and the age, before and after functions) a date is mandatory but a time of day is not; the time returned is at the start of the date. One variation is allowed: if a.m. or p.m. or one of their variants is present, an hour without a minute is allowed, e.g. 3 p.m..
Time zones are not handled, though if one is matched following a time specification it will be removed to allow a surrounding date to be parsed. This only happens if the format of the timezone is not too unusual. The following are examples of forms that are understood:
+0100 GMT GMT-7 CET+1CDT
Any part of the timezone that is not numeric must have exactly three capital letters in the name.
Dates suffer from the ambiguity between DD/MM/YYYY and MM/DD/YYYY. It is recommended this form is avoided with purely numeric dates, but use of ordinals, eg. 3rd/04/2007, will resolve the ambiguity as the ordinal is always parsed as the day of the month. Years must be four digits (and the first two must be 19 or 20); 03/04/08 is not recognised. Other numbers may have leading zeroes, but they are not required. The following are handled:
Here, MNM is at least the first three letters of a month name, matched case-insensitively. The remainder of the month name may appear but its contents are irrelevant, so janissary, febrile, martial, apricot, maybe, junta, etc. are happily handled.
Where the year is shown as optional, the current year is assumed. There are only two such cases, the form Jun 20 or 14 September (the only two commonly occurring forms, apart from a "the" in some forms of English, which isn't currently supported). Such dates will of course become ambiguous in the future, so should ideally be avoided.
Times may follow dates with a colon, e.g. 1965/07/12:09:45; this is in order to provide a format with no whitespace. A comma and whitespace are allowed, e.g. 1965/07/12, 09:45. Currently the order of these separators is not checked, so illogical formats such as 1965/07/12, : ,09:45 will also be matched. For simplicity such variations are not shown in the list above. Otherwise, a time is only recognised as being associated with a date if there is only whitespace in between, or if the time was embedded in the date.
Days of the week are not normally scanned, but will be ignored if they occur at the start of the date pattern only. However, in contexts where it is useful to specify dates relative to today, days of the week with no other date specification may be given. The day is assumed to be either today or within the past week. Likewise, the words yesterday, today and tomorrow are handled. All matches are case-insensitive. Hence if today is Monday, then Sunday is equivalent to yesterday, Monday is equivalent to today, but Tuesday gives a date six days ago. This is not generally useful within the calendar file. Dates in this format may be combined with a time specification; for example Tomorrow, 8 p.m..
For example, the standard date format:
Fri Aug 18 17:00:48 BST 2006
is handled by matching HH:MM:SS and removing it together with the matched (but unused) time zone. This leaves the following:
Fri Aug 18 2006
Fri is ignored and the rest is matched according to the standard rules.
In certain places relative times are handled. Here, a date is not allowed; instead a combination of various supported periods are allowed, together with an optional time. The periods must be in order from most to least significant.
In some cases, a more accurate calculation is possible when there is an anchor date: offsets of months or years pick the correct day, rather than being rounded, and it is possible to pick a particular day in a month as `(1st Friday)', etc., as described in more detail below.
Anchors are available in the following cases. If one or two times are passed to the function calendar, the start time acts an anchor for the end time when the end time is relative (even if the start time is implicit). When examining calendar files, the scheduled event being examined anchors the warning time when it is given explicitly by means of the WARN keyword; likewise, the scheduled event anchors a repetition period when given by the RPT keyword, so that specifications such as RPT 2 months, 3rd Thursday are handled properly. Finally, the -R argument to calendar_scandate directly provides an anchor for relative calculations.
The periods handled, with possible abbreviations are:
Spaces between the numbers are optional, but are required between items, although a comma may be used (with or without spaces).
The forms yearly to hourly allow the number to be omitted; it is assumed to be 1. For example, 1 d and daily are equivalent. Note that using those forms with plurals is confusing; 2 yearly is the same as 2 years, not twice yearly, so it is recommended they only be used without numbers.
When an anchor time is present, there is an extension to handle regular events in the form of the nth someday of the month. Such a specification must occur immediately after any year and month specification, but before any time of day, and must be in the form n(th|st|rd) day, for example 1st Tuesday or 3rd Monday. As in other places, days are matched case insensitively, must be in English, and only the first three letters are significant except that a form beginning `month' does not match `Monday'. No attempt is made to sanitize the resulting date; attempts to squeeze too many occurrences into a month will push the day into the next month (but in the obvious fashion, retaining the correct day of the week).
Here are some examples:
30 years 3 months 4 days 3:42:41 14 days 5 hours Monthly, 3rd Thursday 4d,10hr
Here is an example calendar file. It uses a consistent date format, as recommended above.
Feb 1, 2006 14:30 Pointless bureaucratic meeting Mar 27, 2006 11:00 Mutual recrimination and finger pointing Bring water pistol and waterproofs Mar 31, 2006 14:00 Very serious managerial pontification # UID 12C7878A9A50 Apr 10, 2006 13:30 Even more pointless blame assignment exercise WARN 30 mins May 18, 2006 16:00 Regular moaning session RPT monthly, 3rd Thursday
The second entry has a continuation line. The third entry has a continuation line that will not be shown when the entry is displayed, but the unique identifier will be used by the calendar_add function when updating the event. The fourth entry will produce a warning 30 minutes before the event (to allow you to equip yourself appropriately). The fifth entry repeats after a month on the 3rd Thursday, i.e. June 15, 2006, at the same time.
This section describes functions that are designed to be called directly by the user. The first part describes those functions associated with the user's calendar; the second part describes the use in glob qualifiers.
With no arguments, show events from the start of today until the end of the next working day after today. In other words, if today is Friday, Saturday, or Sunday, show up to the end of the following Monday, otherwise show today and tomorrow.
If end is given, show events from the start of today up to the time and date given, which is in the format described in the previous section. Note that if this is a date the time is assumed to be midnight at the start of the date, so that effectively this shows all events before the given date.
end may start with a +, in which case the remainder of the specification is a relative time format as described in the previous section indicating the range of time from the start time that is to be included.
If start is also given, show events starting from that time and date. The word now can be used to indicate the current time.
To implement an alert when events are due, include calendar -s in your ~/.zshrc file.
Options:
The timed event usually runs the programme calendar_show to show the event, as described in the section UTILITY FUNCTIONS below.
By default, a warning of the event is shown five minutes before it is due. The warning period can be configured by the style warn-time or for a single calendar entry by including WARN reltime in the first line of the entry, where reltime is one of the usual relative time formats.
A repeated event may be indicated by including RPT reldate in the first line of the entry. After the scheduled event has been displayed it will be re-entered into the calendar file at a time reldate after the existing event. Note that this is currently the only use made of the repeat count, so that it is not possible to query the schedule for a recurrence of an event in the calendar until the previous event has passed.
If RPT is used, it is also possible to specify that certain recurrences of an event are rescheduled or cancelled. This is done with the OCCURRENCE keyword, followed by whitespace and the date and time of the occurrence in the regular sequence, followed by whitespace and either the date and time of the rescheduled event or the exact string CANCELLED. In this case the date and time must be in exactly the "date with local time" format used by the text/calendar MIME type (RFC 2445), <YYYY><MM><DD>T<hh><mm><ss> (note the presence of the literal character T). The first word (the regular recurrence) may be something other than a proper date/time to indicate that the event is additional to the normal sequence; a convention that retains the formatting appearance is XXXXXXXXTXXXXXX.
Furthermore, it is useful to record the next regular recurrence (as then the displayed date may be for a rescheduled event so cannot be used for calculating the regular sequence). This is specified by RECURRENCE and a time or date in the same format. calendar_add adds such an indication when it encounters a recurring event that does not include one, based on the headline date/time.
If calendar_add is used to update occurrences the UID keyword described there should be present in both the existing entry and the added occurrence in order to identify recurring event sequences.
For example,
Thu May 6, 2010 11:00 Informal chat RPT 1 week # RECURRENCE 20100506T110000 # OCCURRENCE 20100513T110000 20100513T120000 # OCCURRENCE 20100520T110000 CANCELLED
The event that occurs at 11:00 on 13th May 2010 is rescheduled an hour later. The event that occurs a week later is cancelled. The occurrences are given on a continuation line starting with a # character so will not usually be displayed as part of the event. As elsewhere, no account of time zones is taken with the times. After the next event occurs the headline date/time will be `Thu May 13, 2010 12:00' while the RECURRENCE date/time will be `20100513T110000' (note that cancelled and moved events are not taken account of in the RECURRENCE, which records what the next regular recurrence is, but they are accounted for in the headline date/time).
It is safe to run calendar -s to reschedule an existing event (if the calendar file has changed, for example), and also to have it running in multiples instances of the shell since the calendar file is locked when in use.
By default, expired events are moved to the "done" file; see the -d option. Use -D to prevent this.
The option -B indicates that backing up the calendar file will be handled by the caller and should not be performed by calendar_add. The option -L indicates that calendar_add does not need to lock the calendar file as it is already locked. These options will not usually be needed by users.
If the style reformat-date is true, the date and time of the new entry will be rewritten into the standard date format: see the descriptions of this style and the style date-format.
The function can use a unique identifier stored with each event to ensure that updates to existing events are treated correctly. The entry should contain the word UID, followed by whitespace, followed by a word consisting entirely of hexadecimal digits of arbitrary length (all digits are significant, including leading zeroes). As the UID is not directly useful to the user, it is convenient to hide it on an indented continuation line starting with a #, for example:
Aug 31, 2007 09:30 Celebrate the end of the holidays # UID 045B78A0
The second line will not be shown by the calendar function.
It is possible to specify the RPT keyword followed by CANCELLED instead of a relative time. This causes any matched event or series of events to be cancelled (the original event does not have to be marked as recurring in order to be cancelled by this method). A UID is required in order to match an existing event in the calendar.
calendar_add will attempt to manage recurrences and occurrences of repeating events as described for event scheduling by calendar -s above. To reschedule or cancel a single event calendar_add should be called with an entry that includes the correct UID but does not include the RPT keyword as this is taken to mean the entry applies to a series of repeating events and hence replaces all existing information. Each rescheduled or cancelled occurrence must have an OCCURRENCE keyword in the entry passed to calendar_add which will be merged into the calendar file. Any existing reference to the occurrence is replaced. An occurrence that does not refer to a valid existing event is added as a one-off occurrence to the same calendar entry.
If the calendar scheduler was running, then after editing the file calendar -s is called to update it.
This function locks out the calendar system during the edit. Hence it should be used to edit the calendar file if there is any possibility of a calendar event occurring meanwhile. Note this can lead to another shell with calendar functions enabled hanging waiting for a lock, so it is necessary to quit the editor as soon as possible.
With the option -r nothing is printed but the value of the date and time in seconds since the epoch is stored in the parameter REPLY.
With the option -f fmt the given date/time conversion format is passed to strftime; see notes on the date-format style below.
In order to avoid ambiguity with negative relative date specifications, options must occur in separate words; in other words, -r and -f should not be combined in the same word.
age selects files having a given modification time for use as a glob qualifier. The format of the date is the same as that understood by the calendar system, described in the section FILE AND DATE FORMATS above.
The function can take one or two arguments, which can be supplied either directly as command or arguments, or separately as shell parameters.
print *(e:age 2006/10/04 2006/10/09:)
The example above matches all files modified between the start of those dates. The second argument may alternatively be a relative time introduced by a +:
print *(e:age 2006/10/04 +5d:)
The example above is equivalent to the previous example.
In addition to the special use of days of the week, today and yesterday, times with no date may be specified; these apply to today. Obviously such uses become problematic around midnight.
print *(e-age 12:00 13:30-)
The example above shows files modified between 12:00 and 13:00 today.
print *(e:age 2006/10/04:)
The example above matches all files modified on that date. If the second argument is omitted it is taken to be exactly 24 hours after the first argument (even if the first argument contains a time).
print *(e-age 2006/10/04:10:15 2006/10/04:10:45-)
The example above supplies times. Note that whitespace within the time and date specification must be quoted to ensure age receives the correct arguments, hence the use of the additional colon to separate the date and time.
AGEREF=2006/10/04:10:15 AGEREF2=2006/10/04:10:45 print *(+age)
This shows the same example before using another form of argument passing. The dates and times in the parameters AGEREF and AGEREF2 stay in effect until unset, but will be overridden if any argument is passed as an explicit argument to age. Any explicit argument causes both parameters to be ignored.
Instead of an explicit date and time, it's possible to use the modification time of a file as the date and time for either argument by introducing the file name with a colon:
print *(e-age :file1-)
matches all files created on the same day (24 hours starting from midnight) as file1.
print *(e-age :file1 :file2-)
matches all files modified no earlier than file1 and no later than file2; precision here is to the nearest second.
The two following examples are therefore equivalent:
print *(e-after 12:00-) print *(e-after today:12:00-)
The zsh style mechanism using the zstyle command is describe in zshmodules(1). This is the same mechanism used in the completion system.
The styles below are all examined in the context :datetime:function:, for example :datetime:calendar:.
This is used for outputting dates in calendar, both to support the -v option and when adding recurring events back to the calendar file, and in calendar_showdate as the final output format.
If the style is not set, the default used is similar the standard system format as output by the date command (also known as `ctime format'): `%a %b %d %H:%M:%S %Z %Y'.
The default is the function calendar_show.
Three attempts are made to lock the file before giving up. If the module zsh/zselect is available, the times of the attempts are jittered so that multiple instances of the calling function are unlikely to retry at the same time.
The files locked are appended to the array lockfiles, which should be local to the caller.
If all files were successfully locked, status zero is returned, else status one.
This function may be used as a general file locking function, although this will only work if only this mechanism is used to lock files.
Returns status zero if the date and time were successfully parsed, else one.
Options:
The event is always printed to standard output. If the command line editor is active (which will usually be the case) the command line will be redisplayed after the output.
If the parameter DISPLAY is set and the start and end times are the same (indicating a scheduled event), the function uses the command xmessage to display a window with the event details.
As the system is based entirely on shell functions (with a little support from the zsh/datetime module) the mechanisms used are not as robust as those provided by a dedicated calendar utility. Consequently the user should not rely on the shell for vital alerts.
There is no calendar_delete function.
There is no localization support for dates and times, nor any support for the use of time zones.
Relative periods of months and years do not take into account the variable number of days.
The calendar_show function is currently hardwired to use xmessage for displaying alerts on X Window System displays. This should be configurable and ideally integrate better with the desktop.
calendar_lockfiles hangs the shell while waiting for a lock on a file. If called from a scheduled task, it should instead reschedule the event that caused it.
A module zsh/net/tcp is provided to provide network I/O over TCP/IP from within the shell; see its description in zshmodules(1). This manual page describes a function suite based on the module. If the module is installed, the functions are usually installed at the same time, in which case they will be available for autoloading in the default function search path. In addition to the zsh/net/tcp module, the zsh/zselect module is used to implement timeouts on read operations. For troubleshooting tips, consult the corresponding advice for the zftp functions described in zshzftpsys(1).
There are functions corresponding to the basic I/O operations open, close, read and send, named tcp_open etc., as well as a function tcp_expect for pattern match analysis of data read as input. The system makes it easy to receive data from and send data to multiple named sessions at once. In addition, it can be linked with the shell's line editor in such a way that input data is automatically shown at the terminal. Other facilities available including logging, filtering and configurable output prompts.
To use the system where it is available, it should be enough to `autoload -U tcp_open' and run tcp_open as documented below to start a session. The tcp_open function will autoload the remaining functions.
If sess is given, this becomes the name of the session which can be used to refer to multiple different TCP connections. If sess is not given, the function will invent a numeric name value (note this is not the same as the file descriptor to which the session is attached). It is recommended that session names not include `funny' characters, where funny characters are not well-defined but certainly do not include alphanumerics or underscores, and certainly do include whitespace.
In the second case, one or more sessions to be opened are given by name. A single session name is given after -s and a comma-separated list after -l; both options may be repeated as many times as necessary. A failure to open any session causes tcp_open to abort. The host and port are read from the file .ztcp_sessions in the same directory as the user's zsh initialisation files, i.e. usually the home directory, but $ZDOTDIR if that is set. The file consists of lines each giving a session name and the corresponding host and port, in that order (note the session name comes first, not last), separated by whitespace.
The third form allows passive and fake TCP connections. If the option -a is used, its argument is a file descriptor open for listening for connections. No function front-end is provided to open such a file descriptor, but a call to `ztcp -l port' will create one with the file descriptor stored in the parameter $REPLY. The listening port can be closed with `ztcp -c fd'. A call to `tcp_open -a fd' will block until a remote TCP connection is made to port on the local machine. At this point, a session is created in the usual way and is largely indistinguishable from an active connection created with one of the first two forms.
If the option -f is used, its argument is a file descriptor which is used directly as if it were a TCP session. How well the remainder of the TCP function system copes with this depends on what actually underlies this file descriptor. A regular file is likely to be unusable; a FIFO (pipe) of some sort will work better, but note that it is not a good idea for two different sessions to attempt to read from the same FIFO at once.
If the option -q is given with any of the three forms, tcp_open will not print informational messages, although it will in any case exit with an appropriate status.
If the line editor (zle) is in use, which is typically the case if the shell is interactive, tcp_open installs a handler inside zle which will check for new data at the same time as it checks for keyboard input. This is convenient as the shell consumes no CPU time while waiting; the test is performed by the operating system. Giving the option -z to any of the forms of tcp_open prevents the handler from being installed, so data must be read explicitly. Note, however, this is not necessary for executing complete sets of send and read commands from a function, as zle is not active at this point. Generally speaking, the handler is only active when the shell is waiting for input at a command prompt or in the vared builtin. The option has no effect if zle is not active; `[[ -o zle]]' will test for this.
The first session to be opened becomes the current session and subsequent calls to tcp_open do not change it. The current session is stored in the parameter $TCP_SESS; see below for more detail about the parameters used by the system.
The function tcp_on_open, if defined, is called when a session is opened. See the description below.
If the session being closed is the current one, $TCP_SESS is unset, leaving no current session, even if there are other sessions still open.
If the session was opened with tcp_open -f, the file descriptor is closed so long as it is in the range 0 to 9 accessible directly from the command line. If the option -n is given, no attempt will be made to close file descriptors in this case. The -n option is not used for genuine ztcp session; the file descriptors are always closed with the session.
If the option -q is given, no informational messages will be printed.
The function checks for new data available on all the sessions listed. Unless the -b option is given, it will not block waiting for new data. Any one line of data from any of the available sessions will be read, stored in the parameter $TCP_LINE, and displayed to standard output unless $TCP_SILENT contains a non-empty string. When printed to standard output the string $TCP_PROMPT will be shown at the start of the line; the default form for this includes the name of the session being read. See below for more information on these parameters. In this mode, tcp_read can be called repeatedly until it returns status 2 which indicates all pending input from all specified sessions has been handled.
With the option -b, equivalent to an infinite timeout, the function will block until a line is available to read from one of the specified sessions. However, only a single line is returned.
The option -d indicates that all pending input should be drained. In this case tcp_read may process multiple lines in the manner given above; only the last is stored in $TCP_LINE, but the complete set is stored in the array $tcp_lines. This is cleared at the start of each call to tcp_read.
The options -t and -T specify a timeout in seconds, which may be a floating point number for increased accuracy. With -t the timeout is applied before each line read. With -T, the timeout applies to the overall operation, possibly including multiple read operations if the option -d is present; without this option, there is no distinction between -t and -T.
The function does not print informational messages, but if the option -q is given, no error message is printed for a non-existent session.
A return status of 2 indicates a timeout or no data to read. Any other non-zero return status indicates some error condition.
See tcp_log for how to control where data is sent by tcp_read.
The option -c causes tcp_send to behave like cat. It reads lines from standard input until end of input and sends them in turn to the specified session(s) exactly as if they were given as data arguments to individual tcp_send commands.
The option -n prevents tcp_send from putting a newline at the end of the data strings.
The remaining options all behave as for tcp_read.
The data arguments are not further processed once they have been passed to tcp_send; they are simply passed down to print -r.
If the parameter $TCP_OUTPUT is a non-empty string and logging is enabled then the data sent to each session will be echoed to the log file(s) with $TCP_OUTPUT in front where appropriate, much in the manner of $TCP_PROMPT.
The first form creates an alias for a session name; alias can then be used to refer to the existing session sess. As many aliases may be listed as required.
The second form lists any aliases specified, or all aliases if none.
The third form deletes all the aliases listed. The underlying sessions are not affected.
The option -q suppresses an inconsistently chosen subset of error messages.
With the option -s, per-session logging is enabled. Input from tcp_read is output to the file logfile.sess. As the session is automatically discriminated by the filename, the contents are raw (no $TCP_PROMPT). The option -a applies as above. Per-session logging and logging of all data in one file are not mutually exclusive.
The option -c closes all logging, both complete and per-session logs.
The options -n and -N respectively turn off or restore output of data read by tcp_read to standard output; hence `tcp_log -cn' turns off all output by tcp_read.
The function is purely a convenient front end to setting the parameters $TCP_LOG, $TCP_LOG_SESS, $TCP_SILENT, which are described below.
With a sess argument, set the current session to sess. This is equivalent to changing $TCP_SESS directly.
With additional arguments, temporarily set the current session while executing `command arg ...'. command is re-evaluated so as to expand aliases etc., but the remaining args are passed through as that appear to tcp_sess. The original session is restored when tcp_sess exits.
This is generally only useful for interactive use, to prevent the display becoming fragmented by output returned from the connection. Within a programme or function it is generally better to handle reading data by a more explicit method.
Sessions are specified in the same way as tcp_read: the default is to use the current session, otherwise the sessions specified by -a, -s, or -l are used.
Each pattern is a standard zsh extended-globbing pattern; note that it needs to be quoted to avoid it being expanded immediately by filename generation. It must match the full line, so to match a substring there must be a `*' at the start and end. The line matched against includes the $TCP_PROMPT added by tcp_read. It is possible to include the globbing flags `#b' or `#m' in the patterns to make backreferences available in the parameters $MATCH, $match, etc., as described in the base zsh documentation on pattern matching.
Unlike tcp_read, the default behaviour of tcp_expect is to block indefinitely until the required input is found. This can be modified by specifying a timeout with -t or -T; these function as in tcp_read, specifying a per-read or overall timeout, respectively, in seconds, as an integer or floating-point number. As tcp_read, the function returns status 2 if a timeout occurs.
The function returns as soon as any one of the patterns given match. If the caller needs to know which of the patterns matched, the option -p var can be used; on return, $var is set to the number of the pattern using ordinary zsh indexing, i.e. the first is 1, and so on. Note the absence of a `$' in front of var. To avoid clashes, the parameter cannot begin with `_expect'. The index -1 is used if there is a timeout and 0 if there is no match.
The option -P var works similarly to -p, but instead of numerical indexes the regular arguments must begin with a prefix followed by a colon: that prefix is then used as a tag to which var is set when the argument matches. The tag timeout is used if there is a timeout and the empty string if there is no match. Note it is acceptable for different arguments to start with the same prefix if the matches do not need to be distinguished.
The option -q is passed directly down to tcp_read.
As all input is done via tcp_read, all the usual rules about output of lines read apply. One exception is that the parameter $tcp_lines will only reflect the line actually matched by tcp_expect; use $tcp_expect_lines for the full set of lines read during the function call.
The first argument is a TCP port on which the function will listen.
The remaining arguments give a command and its arguments to execute with standard input, standard output and standard error redirected to the file descriptor on which the TCP session has been accepted. If no command is given, a new zsh is started. This gives everyone on your network direct access to your account, which in many cases will be a bad thing.
The command is run in the background, so tcp_proxy can then accept new connections. It continues to accept new connections until interrupted.
The sessions may be selected explicitly with the standard -a, -s or -l options, or may be chosen implicitly. If none of the three options is given the rules are: first, if the array $tcp_spam_list is set, this is taken as the list of sessions, otherwise all sessions are taken. Second, any sessions given in the array $tcp_no_spam_list are removed from the list of sessions.
Normally, any sessions added by the `-a' flag or when all sessions are chosen implicitly are spammed in alphabetic order; sessions given by the $tcp_spam_list array or on the command line are spammed in the order given. The -r flag reverses the order however it was arrived it.
The -v flag specifies that a $TCP_PROMPT will be output before each session. This is output after any modification to TCP_SESS by the user-defined tcp_on_spam function described below. (Obviously that function is able to generate its own output.)
If the option -e is present, the line given as `cmd [ arg ... ]' is executed using eval, otherwise it is executed without any further processing.
An escape string, $TCP_TALK_ESCAPE, default `:', is used to allow access to normal shell operation. If it is on its own at the start of the line, or followed only by whitespace, the line editor returns to normal operation. Otherwise, the string and any following whitespace are skipped and the remainder of the line executed as shell input without any change of the line editor's operating mode.
The current implementation is somewhat deficient in terms of use of the command history. For this reason, many users will prefer to use some form of alternative approach for sending data easily to the current session. One simple approach is to alias some special character (such as `%') to `tcp_command --'.
To transfer a file from host woodcock to host springes, on springes:
tcp_point 8091 >output_file
and on woodcock:
tcp_shoot springes 8091 <input_file
As these two functions do not require tcp_open to set up a TCP connection first, they may need to be autoloaded separately.
Certain functions, if defined by the user, will be called by the function system in certain contexts. This facility depends on the module zsh/parameter, which is usually available in interactive shells as the completion system depends on it. None of the functions need be defined; they simply provide convenient hooks when necessary.
Typically, these are called after the requested action has been taken, so that the various parameters will reflect the new state.
The variable TCP_INVALIDATE_ZLE will be a non-empty string if it is necessary to invalidate the line editor display using `zle -I' before printing output from the function.
(`AWOL' is military jargon for `absent without leave' or some variation. It has no pre-existing technical meaning known to the author.)
This function is called after $TCP_SESS is set to reflect the session to be spammed, but before any use of it is made. Hence it is possible to alter the value of $TCP_SESS within this function. For example, the session arguments to tcp_spam could include extra information to be stripped off and processed in tcp_on_spam.
If the function sets the parameter $REPLY to `done', the command line is not executed; in addition, no prompt is printed for the -v option to tcp_spam.
The following functions are used by the TCP function system but will rarely if ever need to be called directly.
While active, the function sets the parameter TCP_HANDLER_ACTIVE to 1. This allows shell code called internally (for example, by setting tcp_on_read) to tell if is being called when the shell is otherwise idle at the editor prompt.
The prompt to use is specified by -P; the default is the empty string. It can contain:
The option -q suppresses output to standard output, but not to any log files which are configured.
The -S and -F options are used to pass in the session name and file descriptor for possible replacement in the prompt.
Parameters follow the usual convention that uppercase is used for scalars and integers, while lowercase is used for normal and associative array. It is always safe for user code to read these parameters. Some parameters may also be set; these are noted explicitly. Others are included in this group as they are set by the function system for the user's benefit, i.e. setting them is typically not useful but is benign.
It is often also useful to make settable parameters local to a function. For example, `local TCP_SILENT=1' specifies that data read during the function call will not be printed to standard output, regardless of the setting outside the function. Likewise, `local TCP_SESS=sess' sets a session for the duration of a function, and `local TCP_PROMPT=' specifies that no prompt is used for input during the function.
If the prompt starts with %P, this is stripped and the complete result of the previous stage is passed through standard prompt %-style formatting before being output.
The functions tcp_read and tcp_expect use the shell's SECONDS parameter for their own timing purposes. If that parameter is not of floating point type on entry to one of the functions, it will create a local parameter SECONDS which is floating point and set the parameter TCP_SECONDS_START to the previous value of $SECONDS. If the parameter is already floating point, it is used without a local copy being created and TCP_SECONDS_START is not set. As the global value is zero, the shell elapsed time is guaranteed to be the sum of $SECONDS and $TCP_SECONDS_START.
This can be avoided by setting SECONDS globally to a floating point value using `typeset -F SECONDS'; then the TCP functions will never make a local copy and never set TCP_SECONDS_START to a non-zero value.
The following parameters are not set by the function system, but have a special effect if set by the user.
If any function called to handle a line returns a non-zero status, the line is not output. Thus a tcp_on_read handler containing only the instruction `return 1' can be used to suppress output of particular lines (see, however, tcp_filter above). However, the line is still stored in TCP_LINE and tcp_lines; this occurs after all tcp_on_read processing.
These parameters are controlled by the function system; they may be read directly, but should not usually be set by user code.
Here is a trivial example using a remote calculator.
To create a calculator server on port 7337 (see the dc manual page for quite how infuriating the underlying command is):
tcp_proxy 7337 dc
To connect to this from the same host with a session also named `dc':
tcp_open localhost 7337 dc
To send a command to the remote session and wait a short while for output (assuming dc is the current session):
tcp_command 2 4 + p
To close the session:
tcp_close
The tcp_proxy needs to be killed to be stopped. Note this will not usually kill any connections which have already been accepted, and also that the port is not immediately available for reuse.
The following chunk of code puts a list of sessions into an xterm header, with the current session followed by a star.
print -n "\033]2;TCP:" ${(k)tcp_by_name:/$TCP_SESS/$TCP_SESS\*} "\a"
The function tcp_read uses the shell's normal read builtin. As this reads a complete line at once, data arriving without a terminating newline can cause the function to block indefinitely.
Though the function suite works well for interactive use and for data arriving in small amounts, the performance when large amounts of data are being exchanged is likely to be extremely poor.
This describes the set of shell functions supplied with the source distribution as an interface to the zftp builtin command, allowing you to perform FTP operations from the shell command line or within functions or scripts. The interface is similar to a traditional FTP client (e.g. the ftp command itself, see ftp(1)), but as it is entirely done within the shell all the familiar completion, editing and globbing features, and so on, are present, and macros are particularly simple to write as they are just ordinary shell functions.
The prerequisite is that the zftp command, as described in zshmodules(1) , must be available in the version of zsh installed at your site. If the shell is configured to load new commands at run time, it probably is: typing `zmodload zsh/zftp' will make sure (if that runs silently, it has worked). If this is not the case, it is possible zftp was linked into the shell anyway: to test this, type `which zftp' and if zftp is available you will get the message `zftp: shell built-in command'.
Commands given directly with zftp builtin may be interspersed between the functions in this suite; in a few cases, using zftp directly may cause some of the status information stored in shell parameters to become invalid. Note in particular the description of the variables $ZFTP_TMOUT, $ZFTP_PREFS and $ZFTP_VERBOSE for zftp.
You should make sure all the functions from the Functions/Zftp directory of the source distribution are available; they all begin with the two letters `zf'. They may already have been installed on your system; otherwise, you will need to find them and copy them. The directory should appear as one of the elements of the $fpath array (this should already be the case if they were installed), and at least the function zfinit should be autoloaded; it will autoload the rest. Finally, to initialize the use of the system you need to call the zfinit function. The following code in your .zshrc will arrange for this; assume the functions are stored in the directory ~/myfns:
fpath=(~/myfns $fpath) autoload -U zfinit zfinit
Note that zfinit assumes you are using the zmodload method to load the zftp command. If it is already built into the shell, change zfinit to zfinit -n. It is helpful (though not essential) if the call to zfinit appears after any code to initialize the new completion system, else unnecessary compctl commands will be given.
The sequence of operations in performing a file transfer is essentially the same as that in a standard FTP client. Note that, due to a quirk of the shell's getopts builtin, for those functions that handle options you must use `--' rather than `-' to ensure the remaining arguments are treated literally (a single `-' is treated as an argument).
A single argument `-' will delete the stored parameters. This will also cause the memory of the last directory (and so on) on the other host to be deleted.
If the command was successful, and the terminal is compatible with xterm or is sun-cmd, a summary will appear in the title bar, giving the local host:directory and the remote host:directory; this is handled by the function zftp_chpwd, described below.
Normally, the host, user and password are internally recorded for later re-opening, either by a zfopen with no arguments, or automatically (see below). With the option `-1', no information is stored. Also, if an open command with arguments failed, the parameters will not be retained (and any previous parameters will also be deleted). A zfopen on its own, or a zfopen -1, never alters the stored parameters.
Both zfopen and zfanon (but not zfparams) understand URLs of the form ftp://host/path... as meaning to connect to the host, then change directory to path (which must be a directory, not a file). The `ftp://' can be omitted; the trailing `/' is enough to trigger recognition of the path. Note prefixes other than `ftp:' are not recognized, and that all characters after the first slash beyond host are significant in path.
In the first form with dir present, change to the directory dir. The command `zfcd ..' is treated specially, so is guaranteed to work on non-UNIX servers (note this is handled internally by zftp). If dir is omitted, has the effect of `zfcd ~'.
The second form changes to the directory previously current.
The third form attempts to change the current directory by replacing the first occurrence of the string old with the string new in the current directory.
Note that in this command, and indeed anywhere a remote filename is expected, the string which on the local host corresponds to `~' is converted back to a `~' before being passed to the remote machine. This is convenient because of the way expansion is performed on the command line before zfcd receives a string. For example, suppose the command is `zfcd ~/foo'. The shell will expand this to a full path such as `zfcd /home/user2/pws/foo'. At this stage, zfcd recognises the initial path as corresponding to `~' and will send the directory to the remote host as ~/foo, so that the `~' will be expanded by the server to the correct remote host directory. Other named directories of the form `~name' are not treated in this fashion.
The directory is usually cached for re-use. In fact, two caches are maintained. One is for use when there is no dir-options or dir, i.e. a full listing of the current remote directory; it is flushed when the current remote directory changes. The other is kept for repeated use of zfdir with the same arguments; for example, repeated use of `zfdir /pub/gnu' will only require the directory to be retrieved on the first call. Alternatively, this cache can be re-viewed with the -r option. As relative directories will confuse zfdir, the -f option can be used to force the cache to be flushed before the directory is listed. The option -d will delete both caches without showing a directory listing; it will also delete the cache of file names in the current remote directory, if any.
Note that this requires a commonly implemented, but non-standard, version of the FTP protocol, so is not guaranteed to work on all servers.
In the first form, copy remote-file from the server to the local file local-file.
In the second form, copy all the remote files rfile1 ... into the local directory ldir retaining the same basenames. This assumes UNIX directory semantics.
With the option -r, if any of the files are directories they are sent recursively with all their subdirectories, including files beginning with `.'. This requires that the remote machine understand UNIX file semantics, since `/' is used as a directory separator.
With two arguments, copy local-file to the server as remote-file.
With more than two arguments, copy all the local files lfile1 ... into the existing remote directory rdir retaining the same basenames. This assumes UNIX directory semantics.
A problem arises if you attempt to use zfpcp lfile1 rdir, i.e. the second form of copying but with two arguments, as the command has no simple way of knowing if rdir corresponds to a directory or a filename. It attempts to resolve this in various ways. First, if the rdir argument is `.' or `..' or ends in a slash, it is assumed to be a directory. Secondly, if the operation of copying to a remote file in the first form failed, and the remote server sends back the expected failure code 553 and a reply including the string `Is a directory', then zfpcp will retry using the second form.
With no arguments, zfsession prints the name of the current session; with the option -l it lists all sessions which currently exist, and with the option -v it gives a verbose list showing the host and directory for each session, where the current session is marked with an asterisk. With -o, it will switch to the most recent previous session.
With -d, the given session (or else the current one) is removed; everything to do with it is completely forgotten. If it was the only session, a new session called `default' is created and made current. It is safest not to delete sessions while background commands using zftp are active.
The operation is performed using pipes, so it is required that the connections still be valid in a subshell, which is not the case under versions of some operating systems, presumably due to a system bug.
If not given an argument, list the existing bookmarks and the points to which they refer in the form user@host:directory; this is the format in which they are stored, and the file may be edited directly.
With the option -n, the bookmark is taken to be a nickname stored by the ncftp program in its bookmark file, which is assumed to be ~/.ncftp/bookmarks. The function works identically in other ways. Note that there is no mechanism for adding or modifying ncftp bookmarks from the zftp functions.
chpwd() {
if [[ -n $ZFTP_USER ]]; then
zftp_chpwd
else
# usual chpwd e.g put host:directory in title bar
fi
}
fits in well.
Various styles are available using the standard shell style mechanism, described in zshmodules(1). Briefly, the command `zstyle ':zftp:*' style value ...'. defines the style to have value value; more than one value may be given, although that is not useful in the cases described here. These values will then be used throughout the zftp function system. For more precise control, the first argument, which gives a context in which the style applies, can be modified to include a particular function, as for example `:zftp:zfget': the style will then have the given value only in the zfget function. Values for the same style in different contexts may be set; the most specific function will be used, where strings are held to be more specific than patterns, and longer patterns and shorter patterns. Note that only the top level function name, as called by the user, is used; calling of lower level functions is transparent to the user. Hence modifications to the title bar in zftp_chpwd use the contexts :zftp:zfopen, :zftp:zfcd, etc., depending where it was called from. The following styles are understood:
When zfinit is run, if this style is not defined for the context :zftp:*, it will be set to `bar'.
As described for progress, zfinit will force this to default to 1.
As described for progress, zfinit will force this to default to 1.
When zfinit is run, it will determine whether chpwd exists and if so it will set the default value for the style to 1 if none exists already.
Note that there is also an associative array zfconfig which contains values used by the function system. This should not be modified or overwritten.
The commands for retrieving files usually perform filename generation (globbing) on their arguments; this can be turned off by passing the option -G to each of the commands. Normally this operates by retrieving a complete list of files for the directory in question, then matching these locally against the pattern supplied. This has the advantage that the full range of zsh patterns (respecting the setting of the option EXTENDED_GLOB) can be used. However, it means that the directory part of a filename will not be expanded and must be given exactly. If the remote server does not support the UNIX directory semantics, directory handling is problematic and it is recommended that globbing only be used within the current directory. The list of files in the current directory, if retrieved, will be cached, so that subsequent globs in the same directory without an intervening zfcd are much faster.
If the remote-glob style (see above) is set, globbing is instead performed on the remote host: the server is asked for a list of matching files. This is highly dependent on how the server is implemented, though typically UNIX servers will provide support for basic glob patterns. This may in some cases be faster, as it avoids retrieving the entire list of directory contents.
As described for the zfopen command, a subsequent zfopen with no parameters will reopen the connection to the last host (this includes connections made with the zfanon command). Opened in this fashion, the connection starts in the default remote directory and will remain open until explicitly closed.
Automatic re-opening is also available. If a connection is not currently open and a command requiring a connection is given, the last connection is implicitly reopened. In this case the directory which was current when the connection was closed again becomes the current directory (unless, of course, the command given changes it). Automatic reopening will also take place if the connection was close by the remote server for whatever reason (e.g. a timeout). It is not available if the -1 option to zfopen or zfanon was used.
Furthermore, if the command issued is a file transfer, the connection will be closed after the transfer is finished, hence providing a one-shot mode for transfers. This does not apply to directory changing or listing commands; for example a zfdir may reopen a connection but will leave it open. Also, automatic closure will only ever happen in the same command as automatic opening, i.e a zfdir directly followed by a zfget will never close the connection automatically.
Information about the previous connection is given by the zfstat function. So, for example, if that reports:
Session: default Not connected. Last session: ftp.bar.com:/pub/textfiles
then the command zfget file.txt will attempt to reopen a connection to ftp.bar.com, retrieve the file /pub/textfiles/file.txt, and immediately close the connection again. On the other hand, zfcd .. will open the connection in the directory /pub and leave it open.
Note that all the above is local to each session; if you return to a previous session, the connection for that session is the one which will be reopened.
Completion of local and remote files, directories, sessions and bookmarks is supported. The older, compctl-style completion is defined when zfinit is called; support for the new widget-based completion system is provided in the function Completion/Zsh/Command/_zftp, which should be installed with the other functions of the completion system and hence should automatically be available.
The Zsh source distribution includes a number of items contributed by the user community. These are not inherently a part of the shell, and some may not be available in every zsh installation. The most significant of these are documented here. For documentation on other contributed items such as shell functions, look for comments in the function source files.
The key sequence ESC h is normally bound by ZLE to execute the run-help widget (see zshzle(1)). This invokes the run-help command with the command word from the current input line as its argument. By default, run-help is an alias for the man command, so this often fails when the command word is a shell builtin or a user-defined function. By redefining the run-help alias, one can improve the on-line help provided by the shell.
The helpfiles utility, found in the Util directory of the distribution, is a Perl program that can be used to process the zsh manual to produce a separate help file for each shell builtin and for many other shell features as well. The autoloadable run-help function, found in Functions/Misc, searches for these helpfiles and performs several other tests to produce the most complete help possible for the command.
Help files are installed by default to a subdirectory of /usr/share/zsh or /usr/local/share/zsh.
To create your own help files with helpfiles, choose or create a directory where the individual command help files will reside. For example, you might choose ~/zsh_help. If you unpacked the zsh distribution in your home directory, you would use the commands:
mkdir ~/zsh_help perl ~/zsh-5.8/Util/helpfiles ~/zsh_help
The HELPDIR parameter tells run-help where to look for the help files. When unset, it uses the default installation path. To use your own set of help files, set this to the appropriate path in one of your startup files:
HELPDIR=~/zsh_help
To use the run-help function, you need to add lines something like the following to your .zshrc or equivalent startup file:
unalias run-help autoload run-help
Note that in order for `autoload run-help' to work, the run-help file must be in one of the directories named in your fpath array (see zshparam(1)). This should already be the case if you have a standard zsh installation; if it is not, copy Functions/Misc/run-help to an appropriate directory.
If you frequently edit your zsh functions, or periodically update your zsh installation to track the latest developments, you may find that function digests compiled with the zcompile builtin are frequently out of date with respect to the function source files. This is not usually a problem, because zsh always looks for the newest file when loading a function, but it may cause slower shell startup and function loading. Also, if a digest file is explicitly used as an element of fpath, zsh won't check whether any of its source files has changed.
The zrecompile autoloadable function, found in Functions/Misc, can be used to keep function digests up to date.
In the first form, each name is the name of a compiled file or a directory containing *.zwc files that should be checked. If no arguments are given, the directories and *.zwc files in fpath are used.
When -t is given, no compilation is performed, but a return status of zero (true) is set if there are files that need to be re-compiled and non-zero (false) otherwise. The -q option quiets the chatty output that describes what zrecompile is doing.
Without the -t option, the return status is zero if all files that needed re-compilation could be compiled and non-zero if compilation for at least one of the files failed.
If the -p option is given, the args are interpreted as one or more sets of arguments for zcompile, separated by `--'. For example:
zrecompile -p \
-R ~/.zshrc -- \
-M ~/.zcompdump -- \
~/zsh/comp.zwc ~/zsh/Completion/*/_*
This compiles ~/.zshrc into ~/.zshrc.zwc if that doesn't exist or if it is older than ~/.zshrc. The compiled file will be marked for reading instead of mapping. The same is done for ~/.zcompdump and ~/.zcompdump.zwc, but this compiled file is marked for mapping. The last line re-creates the file ~/zsh/comp.zwc if any of the files matching the given pattern is newer than it.
Without the -p option, zrecompile does not create function digests that do not already exist, nor does it add new functions to the digest.
The following shell loop is an example of a method for creating function digests for all functions in your fpath, assuming that you have write permission to the directories:
for ((i=1; i <= $#fpath; ++i)); do
dir=$fpath[i]
zwc=${dir:t}.zwc
if [[ $dir == (.|..) || $dir == (.|..)/* ]]; then
continue
fi
files=($dir/*(N-.))
if [[ -w $dir:h && -n $files ]]; then
files=(${${(M)files%/*/*}#/})
if ( cd $dir:h &&
zrecompile -p -U -z $zwc $files ); then
fpath[i]=$fpath[i].zwc
fi
fi
done
The -U and -z options are appropriate for functions in the default zsh installation fpath; you may need to use different options for your personal function directories.
Once the digests have been created and your fpath modified to refer to them, you can keep them up to date by running zrecompile with no arguments.
The large number of possible combinations of keyboards, workstations, terminals, emulators, and window systems makes it impossible for zsh to have built-in key bindings for every situation. The zkbd utility, found in Functions/Misc, can help you quickly create key bindings for your configuration.
Run zkbd either as an autoloaded function, or as a shell script:
zsh -f ~/zsh-5.8/Functions/Misc/zkbd
When you run zkbd, it first asks you to enter your terminal type; if the default it offers is correct, just press return. It then asks you to press a number of different keys to determine characteristics of your keyboard and terminal; zkbd warns you if it finds anything out of the ordinary, such as a Delete key that sends neither ^H nor ^?.
The keystrokes read by zkbd are recorded as a definition for an associative array named key, written to a file in the subdirectory .zkbd within either your HOME or ZDOTDIR directory. The name of the file is composed from the TERM, VENDOR and OSTYPE parameters, joined by hyphens.
You may read this file into your .zshrc or another startup file with the `source' or `.' commands, then reference the key parameter in bindkey commands, like this:
source ${ZDOTDIR:-$HOME}/.zkbd/$TERM-$VENDOR-$OSTYPE
[[ -n ${key[Left]} ]] && bindkey "${key[Left]}" backward-char
[[ -n ${key[Right]} ]] && bindkey "${key[Right]}" forward-char
# etc.
Note that in order for `autoload zkbd' to work, the zkdb file must be in one of the directories named in your fpath array (see zshparam(1)). This should already be the case if you have a standard zsh installation; if it is not, copy Functions/Misc/zkbd to an appropriate directory.
Occasionally you may encounter what appears to be a bug in the shell, particularly if you are using a beta version of zsh or a development release. Usually it is sufficient to send a description of the problem to one of the zsh mailing lists (see zsh(1)), but sometimes one of the zsh developers will need to recreate your environment in order to track the problem down.
The script named reporter, found in the Util directory of the distribution, is provided for this purpose. (It is also possible to autoload reporter, but reporter is not installed in fpath by default.) This script outputs a detailed dump of the shell state, in the form of another script that can be read with `zsh -f' to recreate that state.
To use reporter, read the script into your shell with the `.' command and redirect the output into a file:
. ~/zsh-5.8/Util/reporter > zsh.report
You should check the zsh.report file for any sensitive information such as passwords and delete them by hand before sending the script to the developers. Also, as the output can be voluminous, it's best to wait for the developers to ask for this information before sending it.
You can also use reporter to dump only a subset of the shell state. This is sometimes useful for creating startup files for the first time. Most of the output from reporter is far more detailed than usually is necessary for a startup file, but the aliases, options, and zstyles states may be useful because they include only changes from the defaults. The bindings state may be useful if you have created any of your own keymaps, because reporter arranges to dump the keymap creation commands as well as the bindings for every keymap.
As is usual with automated tools, if you create a startup file with reporter, you should edit the results to remove unnecessary commands. Note that if you're using the new completion system, you should not dump the functions state to your startup files with reporter; use the compdump function instead (see zshcompsys(1)).
If the state is omitted, all is assumed.
With the exception of `all', every state can be abbreviated by any prefix, even a single letter; thus a is the same as aliases, z is the same as zstyles, etc.
hook is one of chpwd, periodic, precmd, preexec, zshaddhistory, zshexit, or zsh_directory_name, the special functions in question. Note that zsh_directory_name is called in a different way from the other functions, but may still be manipulated as a hook.
function is name of an ordinary shell function. If no options are given this will be added to the array of functions to be executed in the given context. Functions are invoked in the order they were added.
If the option -L is given, the current values for the hook arrays are listed with typeset.
If the option -d is given, the function is removed from the array of functions to be executed.
If the option -D is given, the function is treated as a pattern and any matching names of functions are removed from the array of functions to be executed.
The options -U, -z and -k are passed as arguments to autoload for function. For functions contributed with zsh, the options -Uz are appropriate.
hook is one of isearch-exit, isearch-update, line-pre-redraw, line-init, line-finish, history-line-set, or keymap-select, corresponding to each of the special widgets zle-isearch-exit, etc. The special widget names are also accepted as the hook argument.
widgetname is the name of a ZLE widget. If no options are given this is added to the array of widgets to be invoked in the given hook context. Widgets are invoked in the order they were added, with
zle widgetname -Nw -- "$@"
Note that this means that the `WIDGET' special parameter tracks the widgetname when the widget function is called, rather than tracking the name of the corresponding special hook widget.
If the option -d is given, the widgetname is removed from the array of widgets to be executed.
If the option -D is given, the widgetname is treated as a pattern and any matching names of widgets are removed from the array.
If widgetname does not name an existing widget when added to the array, it is assumed that a shell function also named widgetname is meant to provide the implementation of the widget. This name is therefore marked for autoloading, and the options -U, -z and -k are passed as arguments to autoload as with add-zsh-hook. The widget is also created with `zle -N widgetname' to cause the corresponding function to be loaded the first time the hook is called.
The arrays of widgetname are currently maintained in zstyle contexts, one for each hook context, with a style of `widgets'. If the -L option is given, this set of styles is listed with `zstyle -L'. This implementation may change, and the special widgets that refer to the styles are created only if add-zle-hook-widget is called to add at least one widget, so if this function is used for any hooks, then all hooks should be managed only via this function.
The function cdr allows you to change the working directory to a previous working directory from a list maintained automatically. It is similar in concept to the directory stack controlled by the pushd, popd and dirs builtins, but is more configurable, and as it stores all entries in files it is maintained across sessions and (by default) between terminal emulators in the current session. Duplicates are automatically removed, so that the list reflects the single most recent use of each directory.
Note that the pushd directory stack is not actually modified or used by cdr unless you configure it to do so as described in the configuration section below.
The system works by means of a hook function that is called every time the directory changes. To install the system, autoload the required functions and use the add-zsh-hook function described above:
autoload -Uz chpwd_recent_dirs cdr add-zsh-hook add-zsh-hook chpwd chpwd_recent_dirs
Now every time you change directly interactively, no matter which command you use, the directory to which you change will be remembered in most-recent-first order.
All direct user interaction is via the cdr function.
The argument to cdr is a number N corresponding to the Nth most recently changed-to directory. 1 is the immediately preceding directory; the current directory is remembered but is not offered as a destination. Note that if you have multiple windows open 1 may refer to a directory changed to in another window; you can avoid this by having per-terminal files for storing directory as described for the recent-dirs-file style below.
If you set the recent-dirs-default style described below cdr will behave the same as cd if given a non-numeric argument, or more than one argument. The recent directory list is updated just the same however you change directory.
If the argument is omitted, 1 is assumed. This is similar to pushd's behaviour of swapping the two most recent directories on the stack.
Completion for the argument to cdr is available if compinit has been run; menu selection is recommended, using:
zstyle ':completion:*:*:cdr:*:*' menu selection
to allow you to cycle through recent directories; the order is preserved, so the first choice is the most recent directory before the current one. The verbose style is also recommended to ensure the directory is shown; this style is on by default so no action is required unless you have changed it.
The behaviour of cdr may be modified by the following options.
If output is to a terminal, then the function will print the new list after pruning and prompt for confirmation by the user. This output and confirmation step can be skipped by using -P instead of -p.
Configuration is by means of the styles mechanism that should be familiar from completion; if not, see the description of the zstyle command in see zshmodules(1). The context for setting styles should be ':chpwd:*' in case the meaning of the context is extended in future, for example:
zstyle ':chpwd:*' recent-dirs-max 0
sets the value of the recent-dirs-max style to 0. In practice the style name is specific enough that a context of '*' should be fine.
An exception is recent-dirs-insert, which is used exclusively by the completion system and so has the usual completion system context (':completion:*' if nothing more specific is needed), though again '*' should be fine in practice.
The value of this style may be an array. In this case, the first file in the list will always be used for saving directories while any other files are left untouched. When reading the recent directory list, if there are fewer than the maximum number of entries in the first file, the contents of later files in the array will be appended with duplicates removed from the list shown. The contents of the two files are not sorted together, i.e. all the entries in the first file are shown first. The special value + can appear in the list to indicate the default file should be read at that point. This allows effects like the following:
zstyle ':chpwd:*' recent-dirs-file \
~/.chpwd-recent-dirs-${TTY##*/} +
Recent directories are read from a file numbered according to the terminal. If there are insufficient entries the list is supplemented from the default file.
It is possible to use zstyle -e to make the directory configurable at run time:
zstyle -e ':chpwd:*' recent-dirs-file pick-recent-dirs-file
pick-recent-dirs-file() {
if [[ $PWD = ~/text/writing(|/*) ]]; then
reply=(~/.chpwd-recent-dirs-writing)
else
reply=(+)
fi
}
In this example, if the current directory is ~/text/writing or a directory under it, then use a special file for saving recent directories, else use the default.
If the value of the style is always, then only recent directories will be completed; in that case, use the cd command when you want to complete other directories.
If the value is fallback, recent directories will be tried first, then normal directory completion is performed if recent directory completion failed to find a match.
Finally, if the value is both then both sets of completions are presented; the usual tag mechanism can be used to distinguish results, with recent directories tagged as recent-dirs. Note that the recent directories inserted are abbreviated with directory names where appropriate.
It is possible to refer to recent directories using the dynamic directory name syntax by using the supplied function zsh_directory_name_cdr a hook:
autoload -Uz add-zsh-hook add-zsh-hook -Uz zsh_directory_name zsh_directory_name_cdr
When this is done, ~[1] will refer to the most recent directory other than $PWD, and so on. Completion after ~[... also works.
This section is for the curious or confused; most users will not need to know this information.
Recent directories are saved to a file immediately and hence are preserved across sessions. Note currently no file locking is applied: the list is updated immediately on interactive commands and nowhere else (unlike history), and it is assumed you are only going to change directory in one window at once. This is not safe on shared accounts, but in any case the system has limited utility when someone else is changing to a different set of directories behind your back.
To make this a little safer, only directory changes instituted from the command line, either directly or indirectly through shell function calls (but not through subshells, evals, traps, completion functions and the like) are saved. Shell functions should use cd -q or pushd -q to avoid side effects if the change to the directory is to be invisible at the command line. See the contents of the function chpwd_recent_dirs for more details.
The dynamic directory naming system is described in the subsection Dynamic named directories of the section Filename Expansion in expn(1). In this, a reference to ~[...] is expanded by a function found by the hooks mechanism.
The contributed function zsh_directory_name_generic provides a system allowing the user to refer to directories with only a limited amount of new code. It supports all three of the standard interfaces for directory naming: converting from a name to a directory, converting in the reverse direction to find a short name, and completion of names.
The main feature of this function is a path-like syntax, combining abbreviations at multiple levels separated by ":". As an example, ~[g:p:s] might specify:
Name components may be completed: if a description is shown at the top of the list of completions, it includes the path to which previous components expand, while the description for an individual completion shows the path segment it would add. No additional configuration is needed for this as the completion system is aware of the dynamic directory name mechanism.
To use the function, first define a wrapper function for your specific case. We'll assume it's to be autoloaded. This can have any name but we'll refer to it as zdn_mywrapper. This wrapper function will define various variables and then call this function with the same arguments that the wrapper function gets. This configuration is described below.
Then arrange for the wrapper to be run as a zsh_directory_name hook:
autoload -Uz add-zsh-hook zsh_diretory_name_generic zdn_mywrapper add-zsh-hook -U zsh_directory_name zdn_mywrapper
The wrapper function should define a local associative array zdn_top. Alternatively, this can be set with a style called mapping. The context for the style is :zdn:wrapper-name where wrapper-name is the function calling zsh_directory_name_generic; for example:
zstyle :zdn:zdn_mywrapper: mapping zdn_mywrapper_top
The keys in this associative array correspond to the first component of the name. The values are matching directories. They may have an optional suffix with a slash followed by a colon and the name of a variable in the same format to give the next component. (The slash before the colon is to disambiguate the case where a colon is needed in the path for a drive. There is otherwise no syntax for escaping this, so path components whose names start with a colon are not supported.) A special component :default: specifies a variable in the form /:var (the path section is ignored and so is usually empty) that will be used for the next component if no variable is given for the path. Variables referred to within zdn_top have the same format as zdn_top itself, but contain relative paths.
For example,
local -A zdn_top=( g ~/git ga ~/alternate/git gs /scratch/$USER/git/:second2 :default: /:second1 )
This specifies the behaviour of a directory referred to as ~[g:...] or ~[ga:...] or ~[gs:...]. Later path components are optional; in that case ~[g] expands to ~/git, and so on. gs expands to /scratch/$USER/git and uses the associative array second2 to match the second component; g and ga use the associative array second1 to match the second component.
When expanding a name to a directory, if the first component is not g or ga or gs, it is not an error; the function simply returns 1 so that a later hook function can be tried. However, matching the first component commits the function, so if a later component does not match, an error is printed (though this still does not stop later hooks from being executed).
For components after the first, a relative path is expected, but note that multiple levels may still appear. Here is an example of second1:
local -A second1=( p myproject s somproject os otherproject/subproject/:third )
The path as found from zdn_top is extended with the matching directory, so ~[g:p] becomes ~/git/myproject. The slash between is added automatically (it's not possible to have a later component modify the name of a directory already matched). Only os specifies a variable for a third component, and there's no :default:, so it's an error to use a name like ~[g:p:x] or ~[ga:s:y] because there's nowhere to look up the x or y.
The associative arrays need to be visible within this function; the generic function therefore uses internal variable names beginning _zdn_ in order to avoid clashes. Note that the variable reply needs to be passed back to the shell, so should not be local in the calling function.
The function does not test whether directories assembled by component actually exist; this allows the system to work across automounted file systems. The error from the command trying to use a non-existent directory should be sufficient to indicate the problem.
Here is a full fictitious but usable autoloadable definition of the example function defined by the code above. So ~[gs:p:s] expands to /scratch/$USER/git/myscratchproject/top/srcdir (with $USER also expanded).
local -A zdn_top=( g ~/git ga ~/alternate/git gs /scratch/$USER/git/:second2 :default: /:second1 ) local -A second1=( p myproject s somproject os otherproject/subproject/:third ) local -A second2=( p myscratchproject s somescratchproject ) local -A third=( s top/srcdir d top/documentation ) # autoload not needed if you did this at initialisation... autoload -Uz zsh_directory_name_generic zsh_directory_name_generic "$@
It is also possible to use global associative arrays, suitably named, and set the style for the context of your wrapper function to refer to this. Then your set up code would contain the following:
typeset -A zdn_mywrapper_top=(...) # ... and so on for other associative arrays ... zstyle ':zdn:zdn_mywrapper:' mapping zdn_mywrapper_top autoload -Uz add-zsh-hook zsh_directory_name_generic zdn_mywrapper add-zsh-hook -U zsh_directory_name zdn_mywrapper
and the function zdn_mywrapper would contain only the following:
zsh_directory_name_generic "$@"
In a lot of cases, it is nice to automatically retrieve information from version control systems (VCSs), such as subversion, CVS or git, to be able to provide it to the user; possibly in the user's prompt. So that you can instantly tell which branch you are currently on, for example.
In order to do that, you may use the vcs_info function.
The following VCSs are supported, showing the abbreviated name by which they are referred to within the system:
There is also support for the patch management system quilt (https://savannah.nongnu.org/projects/quilt). See Quilt Support below for details.
To load vcs_info:
autoload -Uz vcs_info
It can be used in any existing prompt, because it does not require any specific $psvar entries to be available.
To get this feature working quickly (including colors), you can do the following (assuming, you loaded vcs_info properly - see above):
zstyle ':vcs_info:*' actionformats \
'%F{5}(%f%s%F{5})%F{3}-%F{5}[%F{2}%b%F{3}|%F{1}%a%F{5}]%f '
zstyle ':vcs_info:*' formats \
'%F{5}(%f%s%F{5})%F{3}-%F{5}[%F{2}%b%F{5}]%f '
zstyle ':vcs_info:(sv[nk]|bzr):*' branchformat '%b%F{1}:%F{3}%r'
precmd () { vcs_info }
PS1='%F{5}[%F{2}%n%F{5}] %F{3}%3~ ${vcs_info_msg_0_}%f%# '
Obviously, the last two lines are there for demonstration. You need to call vcs_info from your precmd function. Once that is done you need a single quoted '${vcs_info_msg_0_}' in your prompt.
To be able to use '${vcs_info_msg_0_}' directly in your prompt like this, you will need to have the PROMPT_SUBST option enabled.
Now call the vcs_info_printsys utility from the command line:
% vcs_info_printsys ## list of supported version control backends: ## disabled systems are prefixed by a hash sign (#) bzr cdv cvs darcs fossil git hg mtn p4 svk svn tla ## flavours (cannot be used in the enable or disable styles; they ## are enabled and disabled with their master [git-svn -> git]) ## they *can* be used in contexts: ':vcs_info:git-svn:*'. git-p4 git-svn hg-git hg-hgsubversion hg-hgsvn
You may not want all of these because there is no point in running the code to detect systems you do not use. So there is a way to disable some backends altogether:
zstyle ':vcs_info:*' disable bzr cdv darcs mtn svk tla
You may also pick a few from that list and enable only those:
zstyle ':vcs_info:*' enable git cvs svn
If you rerun vcs_info_printsys after one of these commands, you will see the backends listed in the disable style (or backends not in the enable style - if you used that) marked as disabled by a hash sign. That means the detection of these systems is skipped completely. No wasted time there.
The vcs_info feature can be configured via zstyle.
First, the context in which we are working:
:vcs_info:vcs-string:user-context:repo-root-name
Currently, hooks in quilt mode don't add the `.quilt-quilt-mode' information. This may change in the future.
There are three special values for vcs-string: The first is named -init-, that is in effect as long as there was no decision what VCS backend to use. The second is -preinit-; it is used before vcs_info is run, when initializing the data exporting variables. The third special value is formats and is used by the vcs_info_lastmsg for looking up its styles.
The initial value of repo-root-name is -all- and it is replaced with the actual name, as soon as it is known. Only use this part of the context for defining the formats, actionformats or branchformat styles, as it is guaranteed that repo-root-name is set up correctly for these only. For all other styles, just use '*' instead.
There are two pre-defined values for user-context:
You can of course use ':vcs_info:*' to match all VCSs in all user-contexts at once.
This is a description of all styles that are looked up.
Say, ~/.zsh is a directory under version control, in which you do not want vcs_info to be active, do:
zstyle ':vcs_info:*' disable-patterns "${(b)HOME}/.zsh(|/*)"
For this style to be evaluated with the hg backend, the get-revision style needs to be set and the use-simple style needs to be unset. The latter is the default; the former is not.
With the bzr backend, lightweight checkouts only honor this style if the use-server style is set.
Note, the actions taken if this style is enabled are potentially expensive (read: they may be slow, depending on how big the current repository is). Therefore, it is disabled by default.
In the git backend, this style checks for changes in the index. Other backends do not currently implement this style.
This style is disabled by default.
If the value of this style is empty (which is the default), the used binary name is the name of the backend in use (e.g. svn is used in an svn repository).
The repo-root-name part in the context is always the default -all- when this style is looked up.
For example, this style can be used to use binaries from non-default installation directories. Assume, git is installed in /usr/bin but your sysadmin installed a newer version in /usr/local/bin. Instead of changing the order of your $PATH parameter, you can do this:
zstyle ':vcs_info:git:*:-all-' command /usr/local/bin/git
The Bazaar backend (bzr) uses this to permit contacting the server about lightweight checkouts, see the check-for-changes style.
The default is to generate a comma-separated list of all bookmark names that refer to the currently checked out revision. If a bookmark is active, its name is suffixed an asterisk and placed first in the list.
Used by the quilt and hg backends.
The default values for these styles in all contexts are:
In normal formats and actionformats the following replacements are done:
The hg and git backends use this expando to display patch information. hg sources patch information from the mq extensions; git from in-progress rebase and cherry-pick operations and from the stgit extension. The patch-format and nopatch-format styles control the generated string. The former is used when at least one patch from the patch queue has been applied, and the latter otherwise.
The hg backend displays bookmark information in this expando (in addition to mq information). See the get-mq and get-bookmarks styles. Both of these styles may be enabled at the same time. If both are enabled, both resulting strings will be shown separated by a semicolon (that cannot currently be customized).
The quilt `standalone' backend sets this expando to the same value as the %Q expando.
See Quilt Support below for details.
In branchformat these replacements are done:
In hgrevformat these replacements are done:
In patch-format and nopatch-format these replacements are done:
Not all VCS backends have to support all replacements. For nvcsformats no replacements are performed at all, it is just a string.
If you want to use the %b (bold off) prompt expansion in formats, which expands %b itself, use %%b. That will cause the vcs_info expansion to replace %%b with %b, so that zsh's prompt expansion mechanism can handle it. Similarly, to hand down %b from branchformat, use %%%%b. Sorry for this inconvenience, but it cannot be easily avoided. Luckily we do not clash with a lot of prompt expansions and this only needs to be done for those.
When one of the gen-applied-string, gen-unapplied-string, and set-patch-format hooks is defined, applying %-escaping (`foo=${foo//'%'/%%}') to the interpolated values for use in the prompt is the responsibility of those hooks (jointly); when neither of those hooks is defined, vcs_info handles escaping by itself. We regret this coupling, but it was required for backwards compatibility.
Quilt is not a version control system, therefore this is not implemented as a backend. It can help keeping track of a series of patches. People use it to keep a set of changes they want to use on top of software packages (which is tightly integrated into the package build process - the Debian project does this for a large number of packages). Quilt can also help individual developers keep track of their own patches on top of real version control systems.
The vcs_info integration tries to support both ways of using quilt by having two slightly different modes of operation: `addon' mode and `standalone' mode).
Quilt integration is off by default; to enable it, set the use-quilt style, and add %Q to your formats or actionformats style:
zstyle ':vcs_info:*' use-quilt true
Styles looked up from the Quilt support code include `.quilt-quilt-mode' in the vcs-string part of the context, where quilt-mode is either addon or standalone. Example: :vcs_info:git.quilt-addon:default:repo-root-name.
For `addon' mode to become active vcs_info must have already detected a real version control system controlling the directory. If that is the case, a directory that holds quilt's patches needs to be found. That directory is configurable via the `QUILT_PATCHES' environment variable. If that variable exists its value is used, otherwise the value `patches' is assumed. The value from $QUILT_PATCHES can be overwritten using the `quilt-patches' style. (Note: you can use vcs_info to keep the value of $QUILT_PATCHES correct all the time via the post-quilt hook).
When the directory in question is found, quilt is assumed to be active. To gather more information, vcs_info looks for a directory called `.pc'; Quilt uses that directory to track its current state. If this directory does not exist we know that quilt has not done anything to the working directory (read: no patches have been applied yet).
If patches are applied, vcs_info will try to find out which. If you want to know which patches of a series are not yet applied, you need to activate the get-unapplied style in the appropriate context.
vcs_info allows for very detailed control over how the gathered information is presented (see the Configuration and Hooks in vcs_info sections), all of which are documented below. Note there are a number of other patch tracking systems that work on top of a certain version control system (like stgit for git, or mq for hg); the configuration for systems like that are generally configured the same way as the quilt support.
If the quilt support is working in `addon' mode, the produced string is available as a simple format replacement (%Q to be precise), which can be used in formats and actionformats; see below for details).
If, on the other hand, the support code is working in `standalone' mode, vcs_info will pretend as if quilt were an actual version control system. That means that the version control system identifier (which otherwise would be something like `svn' or `cvs') will be set to `-quilt-'. This has implications on the used style context where this identifier is the second element. vcs_info will have filled in a proper value for the "repository's" root directory and the string containing the information about quilt's state will be available as the `misc' replacement (and %Q for compatibility with `addon' mode).
What is left to discuss is how `standalone' mode is detected. The detection itself is a series of searches for directories. You can have this detection enabled all the time in every directory that is not otherwise under version control. If you know there is only a limited set of trees where you would like vcs_info to try and look for Quilt in `standalone' mode to minimise the amount of searching on every call to vcs_info, there are a number of ways to do that:
Essentially, `standalone' mode detection is controlled by a style called `quilt-standalone'. It is a string style and its value can have different effects. The simplest values are: `always' to run detection every time vcs_info is run, and `never' to turn the detection off entirely.
If the value of quilt-standalone is something else, it is interpreted differently. If the value is the name of a scalar variable the value of that variable is checked and that value is used in the same `always'/`never' way as described above.
If the value of quilt-standalone is an array, the elements of that array are used as directory names under which you want the detection to be active.
If quilt-standalone is an associative array, the keys are taken as directory names under which you want the detection to be active, but only if the corresponding value is the string `true'.
Last, but not least, if the value of quilt-standalone is the name of a function, the function is called without arguments and the return value decides whether detection should be active. A `0' return value is true; a non-zero return value is interpreted as false.
Note, if there is both a function and a variable by the name of quilt-standalone, the function will take precedence.
All functions named VCS_INFO_* are for internal use only.
All variables named VCS_INFO_* are for internal use only.
Hooks are places in vcs_info where you can run your own code. That code can communicate with the code that called it and through that, change the system's behaviour.
For configuration, hooks change the style context:
:vcs_info:vcs-string+hook-name:user-context:repo-root-name
To register functions to a hook, you need to list them in the hooks style in the appropriate context.
Example:
zstyle ':vcs_info:*+foo:*' hooks bar baz
This registers functions to the hook `foo' for all backends. In order to avoid namespace problems, all registered function names are prepended by a `+vi-', so the actual functions called for the `foo' hook are `+vi-bar' and `+vi-baz'.
If you would like to register a function to a hook regardless of the current context, you may use the vcs_info_hookadd function. To remove a function that was added like that, the vcs_info_hookdel function can be used.
If something seems weird, you can enable the `debug' boolean style in the proper context and the hook-calling code will print what it tried to execute and whether the function in question existed.
When you register more than one function to a hook, all functions are executed one after another until one function returns non-zero or until all functions have been called. Context-sensitive hook functions are executed before statically registered ones (the ones added by vcs_info_hookadd).
You may pass data between functions via an associative array, user_data. For example:
+vi-git-myfirsthook(){
user_data[myval]=$myval
}
+vi-git-mysecondhook(){
# do something with ${user_data[myval]}
}
There are a number of variables that are special in hook contexts:
Finally, the full list of currently available hooks:
This hook gets the names of the Mercurial bookmarks that vcs_info collected from `hg'.
If a bookmark is active, the key ${hook_com[hg-active-bookmark]} is set to its name. The key is otherwise unset.
When setting ret to non-zero, the string in ${hook_com[hg-bookmark-string]} will be used in the %m escape in formats and actionformats and will be available in the global backend_misc array as ${backend_misc[bookmarks]}.
This hook gets the names of all applied patches which vcs_info collected so far in the opposite order, which means that the first argument is the top-most patch and so forth.
When setting ret to non-zero, the string in ${hook_com[applied-string]} will be available as %p in the patch-format and nopatch-format styles. This hook is, in concert with set-patch-format, responsible for %-escaping that value for use in the prompt. (See the Oddities section.)
This hook gets the names of all unapplied patches which vcs_info collected so far in order, which means that the first argument is the patch next-in-line to be applied and so forth.
When setting ret to non-zero, the string in ${hook_com[unapplied-string]} will be available as %u in the patch-format and nopatch-format styles. This hook is, in concert with set-patch-format, responsible for %-escaping that value for use in the prompt. (See the Oddities section.)
This hook gets the names of any active mq guards.
When setting ret to non-zero, the string in ${hook_com[guards-string]} will be used in the %g escape in the patch-format and nopatch-format styles.
The `hook_com' parameter is not used.
The `hook_com' keys available are as for the set-message hook.
The `hook_com' parameter is not used.
The `hook_com' keys considered are `branch' and `revision'. They are set to the values figured out so far by vcs_info and any change will be used directly when the actual replacement is done.
If ret is set to non-zero, the string in ${hook_com[branch-replace]} will be used unchanged as the `%b' replacement in the variables set by vcs_info.
The `hook_com' keys considered are `hash' and `localrev'. They are set to the values figured out so far by vcs_info and any change will be used directly when the actual replacement is done.
If ret is set to non-zero, the string in ${hook_com[rev-replace]} will be used unchanged as the `%i' replacement in the variables set by vcs_info.
Setting the `ret' variable in this hook to a non-zero value avoids any quilt specific actions from being run at all.
This hook is used in the git, hg and quilt backends.
The hook allows the control of the %p (${hook_com[applied]}) and %u (${hook_com[unapplied]}) expansion in all backends that use the hook. With the mercurial backend, the %g (${hook_com[guards]}) expansion is controllable in addition to that.
If ret is set to non-zero, the string in ${hook_com[patch-replace]} will be used unchanged instead of an expanded format from patch-format or nopatch-format.
This hook is, in concert with the gen-applied-string or gen-unapplied-string hooks if they are defined, responsible for %-escaping the final patch-format value for use in the prompt. (See the Oddities section.)
There are a number of `hook_com' keys, that are used here: `action', `branch', `base', `base-name', `subdir', `staged', `unstaged', `revision', `misc', `vcs' and one `miscN' entry for each backend-specific data field (N starting at zero). They are set to the values figured out so far by vcs_info and any change will be used directly when the actual replacement is done.
Since this hook is triggered multiple times (once for each configured formats or actionformats), each of the `hook_com' keys mentioned above (except for the miscN entries) has an `_orig' counterpart, so even if you changed a value to your liking you can still get the original value in the next run. Changing the `_orig' values is probably not a good idea.
If ret is set to non-zero, the string in ${hook_com[message]} will be used unchanged as the message by vcs_info.
If all of this sounds rather confusing, take a look at the Examples section below and also in the Misc/vcs_info-examples file in the Zsh source. They contain some explanatory code.
Don't use vcs_info at all (even though it's in your prompt):
zstyle ':vcs_info:*' enable NONE
Disable the backends for bzr and svk:
zstyle ':vcs_info:*' disable bzr svk
Disable everything but bzr and svk:
zstyle ':vcs_info:*' enable bzr svk
Provide a special formats for git:
zstyle ':vcs_info:git:*' formats ' GIT, BABY! [%b]' zstyle ':vcs_info:git:*' actionformats ' GIT ACTION! [%b|%a]'
All %x expansion in all sorts of formats (formats, actionformats, branchformat, you name it) are done using the `zformat' builtin from the `zsh/zutil' module. That means you can do everything with these %x items what zformat supports. In particular, if you want something that is really long to have a fixed width, like a hash in a mercurial branchformat, you can do this: %12.12i. That'll shrink the 40 character hash to its 12 leading characters. The form is actually `%min.maxx'. More is possible. See the section `The zsh/zutil Module' in zshmodules(1) for details.
Use the quicker bzr backend
zstyle ':vcs_info:bzr:*' use-simple true
If you do use use-simple, please report if it does `the-right-thing[tm]'.
Display the revision number in yellow for bzr and svn:
zstyle ':vcs_info:(svn|bzr):*' \
branchformat '%b%{'${fg[yellow]}'%}:%r'
If you want colors, make sure you enclose the color codes in %{...%} if you want to use the string provided by vcs_info in prompts.
Here is how to print the VCS information as a command (not in a prompt):
alias vcsi='vcs_info command; vcs_info_lastmsg'
This way, you can even define different formats for output via vcs_info_lastmsg in the ':vcs_info:*:command:*' namespace.
Now as promised, some code that uses hooks: say, you'd like to replace the string `svn' by `subversion' in vcs_info's %s formats replacement.
First, we will tell vcs_info to call a function when populating the message variables with the gathered information:
zstyle ':vcs_info:*+set-message:*' hooks svn2subversion
Nothing happens. Which is reasonable, since we didn't define the actual function yet. To see what the hooks subsystem is trying to do, enable the `debug' style:
zstyle ':vcs_info:*+*:*' debug true
That should give you an idea what is going on. Specifically, the function that we are looking for is `+vi-svn2subversion'. Note, the `+vi-' prefix. So, everything is in order, just as documented. When you are done checking out the debugging output, disable it again:
zstyle ':vcs_info:*+*:*' debug false
Now, let's define the function:
function +vi-svn2subversion() {
[[ ${hook_com[vcs_orig]} == svn ]] && hook_com[vcs]=subversion
}
Simple enough. And it could have even been simpler, if only we had registered our function in a less generic context. If we do it only in the `svn' backend's context, we don't need to test which the active backend is:
zstyle ':vcs_info:svn+set-message:*' hooks svn2subversion
function +vi-svn2subversion() {
hook_com[vcs]=subversion
}
And finally a little more elaborate example, that uses a hook to create a customised bookmark string for the hg backend.
Again, we start off by registering a function:
zstyle ':vcs_info:hg+gen-hg-bookmark-string:*' hooks hgbookmarks
And then we define the `+vi-hgbookmarks' function:
function +vi-hgbookmarks() {
# The default is to connect all bookmark names by
# commas. This mixes things up a little.
# Imagine, there's one type of bookmarks that is
# special to you. Say, because it's *your* work.
# Those bookmarks look always like this: "sh/*"
# (because your initials are sh, for example).
# This makes the bookmarks string use only those
# bookmarks. If there's more than one, it
# concatenates them using commas.
# The bookmarks returned by `hg' are available in
# the function's positional parameters.
local s="${(Mj:,:)@:#sh/*}"
# Now, the communication with the code that calls
# the hook functions is done via the hook_com[]
# hash. The key at which the `gen-hg-bookmark-string'
# hook looks is `hg-bookmark-string'. So:
hook_com[hg-bookmark-string]=$s
# And to signal that we want to use the string we
# just generated, set the special variable `ret' to
# something other than the default zero:
ret=1
return 0
}
Some longer examples and code snippets which might be useful are available in the examples file located at Misc/vcs_info-examples in the Zsh source directory.
This concludes our guided tour through zsh's vcs_info.
You should make sure all the functions from the Functions/Prompts directory of the source distribution are available; they all begin with the string `prompt_' except for the special function`promptinit'. You also need the `colors' and `add-zsh-hook' functions from Functions/Misc. All these functions may already be installed on your system; if not, you will need to find them and copy them. The directory should appear as one of the elements of the fpath array (this should already be the case if they were installed), and at least the function promptinit should be autoloaded; it will autoload the rest. Finally, to initialize the use of the system you need to call the promptinit function. The following code in your .zshrc will arrange for this; assume the functions are stored in the directory ~/myfns:
fpath=(~/myfns $fpath) autoload -U promptinit promptinit
Use the prompt command to select your preferred theme. This command may be added to your .zshrc following the call to promptinit in order to start zsh with a theme already selected.
In some cases the theme may be modified by one or more arguments, which should be given after the theme name. See the help for each theme for descriptions of these arguments.
Options are:
Note that you can undo `prompt off' and `prompt default' with `prompt restore', but a second restore does not undo the first.
The first step for adding your own theme is to choose a name for it, and create a file `prompt_name_setup' in a directory in your fpath, such as ~/myfns in the example above. The file should at minimum contain assignments for the prompt variables that your theme wishes to modify. By convention, themes use PS1, PS2, RPS1, etc., rather than the longer PROMPT and RPROMPT.
The file is autoloaded as a function in the current shell context, so it may contain any necessary commands to customize your theme, including defining additional functions. To make some complex tasks easier, your setup function may also do any of the following:
prompt_cleanup command
These functions all implement user-defined ZLE widgets (see zshzle(1)) which can be bound to keystrokes in interactive shells. To use them, your .zshrc should contain lines of the form
autoload function zle -N function
followed by an appropriate bindkey command to associate the function with a key sequence. Suggested bindings are described below.
autoload -U select-word-style select-word-style bash
The simplest way of configuring the functions is to use select-word-style, which can either be called as a normal function with the appropriate argument, or invoked as a user-defined widget that will prompt for the first character of the word style to be used. The first time it is invoked, the first eight -match functions will automatically replace the builtin versions, so they do not need to be loaded explicitly.
The word styles available are as follows. Only the first character is examined.
All but `default' can be input as an upper case character, which has the same effect but with subword matching turned on. In this case, words with upper case characters are treated specially: each separate run of upper case characters, or an upper case character followed by any number of other characters, is considered a word. The style subword-range can supply an alternative character range to the default `[:upper:]'; the value of the style is treated as the contents of a `[...]' pattern (note that the outer brackets should not be supplied, only those surrounding named ranges).
More control can be obtained using the zstyle command, as described in zshmodules(1). Each style is looked up in the context :zle:widget where widget is the name of the user-defined widget, not the name of the function implementing it, so in the case of the definitions supplied by select-word-style the appropriate contexts are :zle:forward-word, and so on. The function select-word-style itself always defines styles for the context `:zle:*' which can be overridden by more specific (longer) patterns as well as explicit contexts.
The style word-style specifies the rules to use. This may have the following values.
The first three of those rules usually use $WORDCHARS, but the value in the parameter can be overridden by the style word-chars, which works in exactly the same way as $WORDCHARS. In addition, the style word-class uses character class syntax to group characters and takes precedence over word-chars if both are set. The word-class style does not include the surrounding brackets of the character class; for example, `-:[:alnum:]' is a valid word-class to include all alphanumerics plus the characters `-' and `:'. Be careful including `]', `^' and `-' as these are special inside character classes.
word-style may also have `-subword' appended to its value to turn on subword matching, as described above.
The style skip-chars is mostly useful for transpose-words and similar functions. If set, it gives a count of characters starting at the cursor position which will not be considered part of the word and are treated as space, regardless of what they actually are. For example, if
zstyle ':zle:transpose-words' skip-chars 1
has been set, and transpose-words-match is called with the cursor on the X of fooXbar, where X can be any character, then the resulting expression is barXfoo.
Finer grained control can be obtained by setting the style word-context to an array of pairs of entries. Each pair of entries consists of a pattern and a subcontext. The shell argument the cursor is on is matched against each pattern in turn until one matches; if it does, the context is extended by a colon and the corresponding subcontext. Note that the test is made against the original word on the line, with no stripping of quotes. Special handling is done between words: the current context is examined and if it contains the string between the word is set to a single space; else if it is contains the string back, the word before the cursor is considered, else the word after cursor is considered. Some examples are given below.
The style skip-whitespace-first is only used with the forward-word widget. If it is set to true, then forward-word skips any non-word-characters, followed by any non-word-characters: this is similar to the behaviour of other word-orientated widgets, and also that used by other editors, however it differs from the standard zsh behaviour. When using select-word-style the widget is set in the context :zle:* to true if the word style is bash and false otherwise. It may be overridden by setting it in the more specific context :zle:forward-word*.
It is possible to create widgets with specific behaviour by defining a new widget implemented by the appropriate generic function, then setting a style for the context of the specific widget. For example, the following defines a widget backward-kill-space-word using backward-kill-word-match, the generic widget implementing backward-kill-word behaviour, and ensures that the new widget always implements space-delimited behaviour.
zle -N backward-kill-space-word backward-kill-word-match zstyle :zle:backward-kill-space-word word-style space
The widget backward-kill-space-word can now be bound to a key.
Here are some further examples of use of the styles, actually taken from the simplified interface in select-word-style:
zstyle ':zle:*' word-style standard zstyle ':zle:*' word-chars ''
Implements bash-style word handling for all widgets, i.e. only alphanumerics are word characters; equivalent to setting the parameter WORDCHARS empty for the given context.
style ':zle:*kill*' word-style space
Uses space-delimited words for widgets with the word `kill' in the name. Neither of the styles word-chars nor word-class is used in this case.
Here are some examples of use of the word-context style to extend the context.
zstyle ':zle:*' word-context \
"*/*" filename "[[:space:]]" whitespace
zstyle ':zle:transpose-words:whitespace' word-style shell
zstyle ':zle:transpose-words:filename' word-style normal
zstyle ':zle:transpose-words:filename' word-chars ''
This provides two different ways of using transpose-words depending on whether the cursor is on whitespace between words or on a filename, here any word containing a /. On whitespace, complete arguments as defined by standard shell rules will be transposed. In a filename, only alphanumerics will be transposed. Elsewhere, words will be transposed using the default style for :zle:transpose-words.
The word matching and all the handling of zstyle settings is actually implemented by the function match-words-by-style. This can be used to create new user-defined widgets. The calling function should set the local parameter curcontext to :zle:widget, create the local parameter matched_words and call match-words-by-style with no arguments. On return, matched_words will be set to an array with the elements: (1) the start of the line (2) the word before the cursor (3) any non-word characters between that word and the cursor (4) any non-word character at the cursor position plus any remaining non-word characters before the next word, including all characters specified by the skip-chars style, (5) the word at or following the cursor (6) any non-word characters following that word (7) the remainder of the line. Any of the elements may be an empty string; the calling function should test for this to decide whether it can perform its function.
If the variable matched_words is defined by the caller to match-words-by-style as an associative array (local -A matched_words), then the seven values given above should be retrieved from it as elements named start, word-before-cursor, ws-before-cursor, ws-after-cursor, word-after-cursor, ws-after-word, and end. In addition the element is-word-start is 1 if the cursor is on the start of a word or subword, or on white space before it (the cases can be distinguished by testing the ws-after-cursor element) and 0 otherwise. This form is recommended for future compatibility.
It is possible to pass options with arguments to match-words-by-style to override the use of styles. The options are:
For example, match-words-by-style -w shell -c 0 may be used to extract the command argument around the cursor.
The word-context style is implemented by the function match-word-context. This should not usually need to be called directly.
The bracketed-paste-magic widget is meant to replace bracketed-paste with a wrapper that re-enables these self-insert actions, and other actions as selected by zstyles. Therefore this widget is installed with
autoload -Uz bracketed-paste-magic zle -N bracketed-paste bracketed-paste-magic
Other than enabling some widget processing, bracketed-paste-magic attempts to replicate bracketed-paste as faithfully as possible.
The following zstyles may be set to control processing of pasted text. All are looked up in the context `:bracketed-paste-magic'.
If this style is not set (explicitly deleted) or set to an empty value, no widgets are active and the pasted text is inserted literally. If the value includes `undefined-key', any unknown sequences are discarded from the pasted text.
A non-zero return from a paste-init function does not prevent the paste itself from proceeding.
Loading bracketed-paste-magic defines backward-extend-paste, a helper function for use in paste-init.
zstyle :bracketed-paste-magic paste-init \
backward-extend-paste
When a paste would insert into the middle of a word or append text to a word already on the line, backward-extend-paste moves the prefix from LBUFFER into PASTED so that the active-widgets see the full word so far. This may be useful with url-quote-magic.
A non-zero return from a paste-finish function does not prevent the paste itself from proceeding.
Loading bracketed-paste-magic also defines quote-paste, a helper function for use in paste-finish.
zstyle :bracketed-paste-magic paste-finish \
quote-paste
zstyle :bracketed-paste-magic:finish quote-style \
qqq
When the pasted text is inserted into BUFFER, it is quoted per the quote-style value. To forcibly turn off the built-in numeric prefix quoting of bracketed-paste, use:
zstyle :bracketed-paste-magic:finish quote-style \
none
Important: During active-widgets processing of the paste (after paste-init and before paste-finish), BUFFER starts empty and history is restricted, so cursor motions, etc., may not pass outside of the pasted content. Text assigned to BUFFER by the active widgets is copied back into PASTED before paste-finish.
If insert-last-word has been used to retrieve the last word on a previous history line, repeated invocations will replace that word with earlier words from the same line.
Otherwise, the widget applies to words on the line currently being edited. The widget style can be set to the name of another widget that should be called to retrieve words. This widget must accept the same three arguments as insert-last-word.
This widget allows the cursor to be easily moved to the other interesting spots. It can be invoked repeatedly to cycle between all positions reported by the completion system.
The basic behaviour is to delete the word around the cursor. There is no numeric argument handling; only the single word around the cursor is considered. If the widget contains the string kill, the removed text will be placed in the cutbuffer for future yanking. This can be obtained by defining kill-whole-word-match as follows:
zle -N kill-whole-word-match delete-whole-word-match
and then binding the widget kill-whole-word-match.
bindkey -M vicmd v edit-command-line
Although you autoload only one function, the commands to use it are slightly different because it implements two widgets.
zle -N history-beginning-search-backward-end \
history-search-end
zle -N history-beginning-search-forward-end \
history-search-end
bindkey '\e^P' history-beginning-search-backward-end
bindkey '\e^N' history-beginning-search-forward-end
With a numeric argument, the search is not anchored to the start of the line; the string typed by the use may appear anywhere in the line in the history.
If the widget name contains `-end' the cursor is moved to the end of the line inserted. If the widget name contains `-space' any space in the text typed is treated as a wildcard and can match anything (hence a leading space is equivalent to giving a numeric argument). Both forms can be combined, for example:
zle -N history-beginning-search-menu-space-end \
history-beginning-search-menu
The commands to create bindable widgets are similar to those in the example immediately above:
autoload -U history-pattern-search zle -N history-pattern-search-backward history-pattern-search zle -N history-pattern-search-forward history-pattern-search
bindkey '^X+' incarg
This works only with the new function based completion system.
bindkey '^Xi' incremental-complete-word
The function may optionally be followed by up to two characters which replace one or both of the characters read from the keyboard; if both characters are supplied, no input is read. For example, insert-composed-char a: can be used within a widget to insert an a with umlaut into the command line. This has the advantages over use of a literal character that it is more portable.
For best results zsh should have been built with support for multibyte characters (configured with --enable-multibyte); however, the function works for the limited range of characters available in single-byte character sets such as ISO-8859-1.
The character is converted into the local representation and inserted into the command line at the cursor position. (The conversion is done within the shell, using whatever facilities the C library provides.) With a numeric argument, the character and its code are previewed in the status line
The function may be run outside zle in which case it prints the character (together with a newline) to standard output. Input is still read from keystrokes.
See insert-unicode-char for an alternative way of inserting Unicode characters using their hexadecimal character number.
The set of accented characters is reasonably complete up to Unicode character U+0180, the set of special characters less so. However, it is very sporadic from that point. Adding new characters is easy, however; see the function define-composed-chars. Please send any additions to zsh-workers@zsh.org.
The codes for the second character when used to accent the first are as follows. Note that not every character can take every accent.
The most common characters from the Arabic, Cyrillic, Greek and Hebrew alphabets are available; consult RFC 1345 for the appropriate sequences. In addition, a set of two letter codes not in RFC 1345 are available for the double-width characters corresponding to ASCII characters from ! to ~ (0x21 to 0x7e) by preceding the character with ^, for example ^A for a double-width A.
The following other two-character sequences are understood.
bindkey '^Xf' insert-files
See insert-composed-char for a way of inserting characters using a two-character mnemonic.
narrow-to-region may be used as a widget or called as a function from a user-defined widget; by default, the text outside the editable area remains visible. A recursive-edit is performed and the original widening status is then restored. Various options and arguments are available when it is called as a function.
The options -p pretext and -P posttext may be used to replace the text before and after the display for the duration of the function; either or both may be an empty string.
If the option -n is also given, pretext or posttext will only be inserted if there is text before or after the region respectively which will be made invisible.
Two numeric arguments may be given which will be used instead of the cursor and mark positions.
The option -S statepm is used to narrow according to the other options while saving the original state in the parameter with name statepm, while the option -R statepm is used to restore the state from the parameter; note in both cases the name of the parameter is required. In the second case, other options and arguments are irrelevant. When this method is used, no recursive-edit is performed; the calling widget should call this function with the option -S, perform its own editing on the command line or pass control to the user via `zle recursive-edit', then call this function with the option -R. The argument statepm must be a suitable name for an ordinary parameter, except that parameters beginning with the prefix _ntr_ are reserved for use within narrow-to-region. Typically the parameter will be local to the calling function.
The options -l lbufvar and -r rbufvar may be used to specify parameters where the widget will store the resulting text from the operation. The parameter lbufvar will contain LBUFFER and rbufvar will contain RBUFFER. Neither of these two options may be used with -S or -R.
narrow-to-region-invisible is a simple widget which calls narrow-to-region with arguments which replace any text outside the region with `...'. It does not take any arguments.
The display is restored (and the widget returns) upon any zle command which would usually cause the line to be accepted or aborted. Hence an additional such command is required to accept or abort the current line.
The return status of both widgets is zero if the line was accepted, else non-zero.
Here is a trivial example of a widget using this feature.
local state narrow-to-region -p $'Editing restricted region\n' \ -P '' -S state zle recursive-edit narrow-to-region -R state
With the function based completion system (which is needed for this), you should be able to type TAB at almost any point to advance the cursor to the next ``interesting'' character position (usually the end of the current word, but sometimes somewhere in the middle of the word). And of course as soon as the entire line is what you want, you can accept with return, without needing to move the cursor to the end first.
The first time predict-on is used, it creates several additional widget functions:
Although you autoload only the predict-on function, it is necessary to create a keybinding for predict-off as well.
zle -N predict-on zle -N predict-off bindkey '^X^Z' predict-on bindkey '^Z' predict-off
If one argument is supplied to the function it is taken as a prompt, otherwise `? ' is used. If two arguments are supplied, they are the prompt and the initial value of $LBUFFER, and if a third argument is given it is the initial value of $RBUFFER. This provides a default value and starting cursor placement. Upon return the entire buffer is the value of $REPLY.
One option is available: `-k num' specifies that num characters are to be read instead of a whole line. The line editor is not invoked recursively in this case, so depending on the terminal settings the input may not be visible, and only the input keys are placed in $REPLY, not the entire buffer. Note that unlike the read builtin num must be given; there is no default.
The name is a slight misnomer, as in fact the shell's own minibuffer is not used. Hence it is still possible to call executed-named-cmd and similar functions while reading a value.
If a numeric argument is given, that specifies the argument to be replaced. 0 means the command name, as in history expansion. A negative numeric argument counts backward from the last word.
If no numeric argument is given, the current argument is replaced; this is the last argument if the previous history line is being used.
The function prompts for a replacement argument.
If the widget contains the string edit, for example is defined as
zle -N replace-argument-edit replace-argument
then the function presents the current value of the argument for editing, otherwise the editing buffer for the replacement is initially empty.
If the widget name contains the word `pattern', for example by defining the widget using the command `zle -N replace-pattern replace-string', then the matching is performed using zsh patterns. All zsh extended globbing patterns can be used in the source string; note that unlike filename generation the pattern does not need to match an entire word, nor do glob qualifiers have any effect. In addition, the replacement string can contain parameter or command substitutions. Furthermore, a `&' in the replacement string will be replaced with the matched source string, and a backquoted digit `\N' will be replaced by the Nth parenthesised expression matched. The form `\{N}' may be used to protect the digit from following digits.
If the widget instead contains the word `regex' (or `regexp'), then the matching is performed using regular expressions, respecting the setting of the option RE_MATCH_PCRE (see the description of the function regexp-replace below). The special replacement facilities described above for pattern matching are available.
By default the previous source or replacement string will not be offered for editing. However, this feature can be activated by setting the style edit-previous in the context :zle:widget (for example, :zle:replace-string) to true. In addition, a positive numeric argument forces the previous values to be offered, a negative or zero argument forces them not to be.
The function replace-string-again can be used to repeat the previous replacement; no prompting is done. As with replace-string, if the name of the widget contains the word `pattern' or `regex', pattern or regular expression matching is performed, else a literal string replacement. Note that the previous source and replacement text are the same whether pattern, regular expression or string matching is used.
In addition, replace-string shows the previous replacement above the prompt, so long as there was one during the current session; if the source string is empty, that replacement will be repeated without the widget prompting for a replacement string.
For example, starting from the line:
print This line contains fan and fond
and invoking replace-pattern with the source string `f(?)n' and the replacement string `c\1r' produces the not very useful line:
print This line contains car and cord
The range of the replacement string can be limited by using the narrow-to-region-invisible widget. One limitation of the current version is that undo will cycle through changes to the replacement and source strings before undoing the replacement itself.
If one argument is supplied to the function it is taken as a prompt, otherwise `Non-echoed text: ' is used (as in emacs). If a second and third argument are supplied they are used to begin and end the reference to $INVISIBLE that is inserted into the buffer. The default is to open with ${, then INVISIBLE, and close with }, but many other effects are possible.
zle -N insert-last-word smart-insert-last-word
With a numeric argument, or when passed command line arguments in a call from another widget, it behaves like insert-last-word, except that words in comments are ignored when INTERACTIVE_COMMENTS is set.
Otherwise, the rightmost ``interesting'' word from the previous command is found and inserted. The default definition of ``interesting'' is that the word contains at least one alphabetic character, slash, or backslash. This definition may be overridden by use of the match style. The context used to look up the style is the widget name, so usually the context is :insert-last-word. However, you can bind this function to different widgets to use different patterns:
zle -N insert-last-assignment smart-insert-last-word zstyle :insert-last-assignment match '[[:alpha:]][][[:alnum:]]#=*' bindkey '\e=' insert-last-assignment
If no interesting word is found and the auto-previous style is set to a true value, the search continues upward through the history. When auto-previous is unset or false (the default), the widget must be invoked repeatedly in order to search earlier history lines.
Transpose the current line with the previous line and move the cursor to the start of the next line. Repeating this (which can be done by providing a positive numeric argument) has the effect of moving the line above the cursor down by a number of lines.
With a negative numeric argument, requires two lines above the cursor. These two lines are transposed and the cursor moved to the start of the previous line. Using a numeric argument less than -1 has the effect of moving the line above the cursor up by minus that number of lines.
Styles to control quoting behavior:
Loading url-quote-magic also defines a helper function `urlglobber' and aliases `globurl' to `noglob urlglobber'. This function takes a local URL apart, attempts to pattern-match the local file portion of the URL path, and then puts the results back into URL format again.
The function serves as an example for reading a vi movement command from within a user-defined widget.
The style whence is available in the context :zle:$WIDGET; this may be set to an array to give the command and options that will be used to investigate the command word found. The default is whence -c.
Hence, for example, typing `+12' followed by return adds 12 to the previous result.
If zcalc is in RPN mode (-r option) the effect of this binding is automatically suppressed as operators alone on a line are meaningful.
When not in zcalc, the key simply inserts the symbol itself.
These functions are useful in constructing widgets. They should be loaded with `autoload -U function' and called as indicated from user-defined widgets.
The parameter REPLY is set to the index of the word in reply which contains the character after the cursor, where the first element has index 1. The parameter REPLY2 is set to the index of the character under the cursor in that word, where the first character has index 1.
Hence reply, REPLY and REPLY2 should all be made local to the enclosing function.
See the function modify-current-argument, described below, for an example of how to call this function.
The argument can be an expression which when evaluated operates on the shell parameter ARG, which will have been set to the command line argument under the cursor. The expression should be suitably quoted to prevent it being evaluated too early.
Alternatively, if the argument does not contain the string ARG, it is assumed to be a shell function, to which the current command line argument is passed as the only argument. The function should set the variable REPLY to the new value for the command line argument. If the function returns non-zero status, so does the calling function.
For example, a user-defined widget containing the following code converts the characters in the argument under the cursor into all upper case:
modify-current-argument '${(U)ARG}'
The following strips any quoting from the current word (whether backslashes or one of the styles of quotes), and replaces it with single quoting throughout:
modify-current-argument '${(qq)${(Q)ARG}}'
The following performs directory expansion on the command line argument and replaces it by the absolute path:
expand-dir() {
REPLY=${~1}
REPLY=${REPLY:a}
}
modify-current-argument expand-dir
In practice the function expand-dir would probably not be defined within the widget where modify-current-argument is called.
The behavior of several of the above widgets can be controlled by the use of the zstyle mechanism. In particular, widgets that interact with the completion system pass along their context to any completions that they invoke.
zstyle ':completion:*' completer \
_complete _correct _approximate
zstyle ':completion:incremental:*' completer \
_complete _correct
zstyle ':completion:predict:*' completer \
_complete
It is a good idea to restrict the completers used in prediction, because they may be automatically invoked as you type. The _list and _menu completers should never be used with prediction. The _approximate, _correct, _expand, and _match completers may be used, but be aware that they may change characters anywhere in the word behind the cursor, so you need to watch carefully that the result is what you intended.
Any other value for this style unconditionally leaves the cursor at the position where the completion code left it.
The insert-and-predict widget uses this style to decide if the completion should be shown even if there is only one possible completion. This is done if the value of this style is the string always. In this case the context is `:predict' (not `:completion:predict').
zstyle :insert-last-word match '*[[:alpha:]/\\]*'
However, you might want to include words that contain spaces:
zstyle :insert-last-word match '*[[:alpha:][:space:]/\\]*'
Or include numbers as long as the word is at least two characters long:
zstyle :insert-last-word match '*([[:digit:]]?|[[:alpha:]/\\])*'
The above example causes redirections like "2>" to be included.
Like `break-keys', this uses the `:incremental' context.
zstyle :copy-earlier-word widget smart-insert-last-word
Check the documentation for the calling widget or function to determine whether the widget style is used.
Two functions are provided to enable zsh to provide exception handling in a form that should be familiar from other languages.
The functions are designed to be used together with the always construct described in zshmisc(1). This is important as only this construct provides the required support for exceptions. A typical example is as follows.
{
# "try" block
# ... nested code here calls "throw MyExcept"
} always {
# "always" block
if catch MyExcept; then
print "Caught exception MyExcept"
elif catch ''; then
print "Caught a shell error. Propagating..."
throw ''
fi
# Other exceptions are not handled but may be caught further
# up the call stack.
}
If all exceptions should be caught, the following idiom might be preferable.
{
# ... nested code here throws an exception
} always {
if catch *; then
case $CAUGHT in
(MyExcept)
print "Caught my own exception"
;;
(*)
print "Caught some other exception"
;;
esac
fi
}
In common with exception handling in other languages, the exception may be thrown by code deeply nested inside the `try' block. However, note that it must be thrown inside the current shell, not in a subshell forked for a pipeline, parenthesised current-shell construct, or some form of command or process substitution.
The system internally uses the shell variable EXCEPTION to record the name of the exception between throwing and catching. One drawback of this scheme is that if the exception is not handled the variable EXCEPTION remains set and may be incorrectly recognised as the name of an exception if a shell error subsequently occurs. Adding unset EXCEPTION at the start of the outermost layer of any code that uses exception handling will eliminate this problem.
Three functions are available to provide handling of files recognised by extension, for example to dispatch a file text.ps when executed as a command to an appropriate viewer.
To enable the system, the function zsh-mime-setup should be autoloaded and run. This allows files with extensions to be treated as executable; such files be completed by the function completion system. The function zsh-mime-handler should not need to be called by the user.
The system works by setting up suffix aliases with `alias -s'. Suffix aliases already installed by the user will not be overwritten.
For suffixes defined in lower case, upper case variants will also automatically be handled (e.g. PDF is automatically handled if handling for the suffix pdf is defined), but not vice versa.
Repeated calls to zsh-mime-setup do not override the existing mapping between suffixes and executable files unless the option -f is given. Note, however, that this does not override existing suffix aliases assigned to handlers other than zsh-mime-handler.
Calling zsh-mime-setup with the option -l lists the existing mappings without altering them. Suffixes to list (which may contain pattern characters that should be quoted from immediate interpretation on the command line) may be given as additional arguments, otherwise all suffixes are listed.
Calling zsh-mime-setup with the option -v causes verbose output to be shown during the setup operation.
The system respects the mailcap flags needsterminal and copiousoutput, see mailcap(4).
The functions use the following styles, which are defined with the zstyle builtin command (see zshmodules(1)). They should be defined before zsh-mime-setup is run. The contexts used all start with :mime:, with additional components in some cases. It is recommended that a trailing * (suitably quoted) be appended to style patterns in case the system is extended in future. Some examples are given below.
For files that have multiple suffixes, e.g. .pdf.gz, where the context includes the suffix it will be looked up starting with the longest possible suffix until a match for the style is found. For example, if .pdf.gz produces a match for the handler, that will be used; otherwise the handler for .gz will be used. Note that, owing to the way suffix aliases work, it is always required that there be a handler for the shortest possible suffix, so in this example .pdf.gz can only be handled if .gz is also handled (though not necessarily in the same way). Alternatively, if no handling for .gz on its own is needed, simply adding the command
alias -s gz=zsh-mime-handler
to the initialisation code is sufficient; .gz will not be handled on its own, but may be in combination with other suffixes.
zstyle ':mime:*' execute-never '/mnt/windows/*'
will ensure that any files found in that area will be executed as MIME types even if they are executable. As this example shows, the complete file name is matched against the pattern, regardless of how the file was passed to the handler. The file is resolved to a full path using the :P modifier described in the subsection Modifiers in zshexpn(1); this means that symbolic links are resolved where possible, so that links into other file systems behave in the correct fashion.
Note that as this style is handled during initialisation, the context is always :mime:, with no discrimination by suffix.
Examples:
zstyle ':mime:*' mailcap ~/.mailcap /usr/local/etc/mailcap zstyle ':mime:.txt:' handler less %s zstyle ':mime:.txt:' flags needsterminal
When zsh-mime-setup is subsequently run, it will look for mailcap entries in the two files given. Files of suffix .txt will be handled by running `less file.txt'. The flag needsterminal is set to show that this program must run attached to a terminal.
As there are several steps to dispatching a command, the following should be checked if attempting to execute a file by extension .ext does not have the expected effect.
The command `alias -s ext' should show `ps=zsh-mime-handler'. If it shows something else, another suffix alias was already installed and was not overwritten. If it shows nothing, no handler was installed: this is most likely because no handler was found in the .mime.types and mailcap combination for .ext files. In that case, appropriate handling should be added to ~/.mime.types and mailcap.
If the extension is handled by zsh-mime-handler but the file is not opened correctly, either the handler defined for the type is incorrect, or the flags associated with it are in appropriate. Running zsh-mime-setup -l will show the handler and, if there are any, the flags. A %s in the handler is replaced by the file (suitably quoted if necessary). Check that the handler program listed lists and can be run in the way shown. Also check that the flags needsterminal or copiousoutput are set if the handler needs to be run under a terminal; the second flag is used if the output should be sent to a pager. An example of a suitable mailcap entry for such a program is:
text/html; /usr/bin/lynx '%s'; needsterminal
Running `zsh-mime-handler -l command line' prints the command line that would be executed, simplified to remove the effect of any flags, and quoted so that the output can be run as a complete zsh command line. This is used by the completion system to decide how to complete after a file handled by zsh-mime-setup.
autoload -U pick-web-browser alias -s html=pick-web-browser
It is provided as an intelligent front end to dispatch a web browser. It may be run as either a function or a shell script. The status 255 is returned if no browser could be started.
Various styles are available to customize the choice of browsers:
If the style is not set the default running x tty is used.
zstyle ':mime:*' x-browsers opera konqueror firefox
specifies that pick-web-browser should first look for a running instance of Opera, Konqueror or Firefox, in that order, and if it fails to find any should attempt to start Opera. The default is firefox mozilla netscape opera konqueror.
Non-programmers should note that, as in many other programming languages, expressions involving only integers (whether constants without a `.', variables containing such constants as strings, or variables declared to be integers) are by default evaluated using integer arithmetic, which is not how an ordinary desk calculator operates. To force floating point operation, pass the option -f; see further notes below.
If the file ~/.zcalcrc exists it will be sourced inside the function once it is set up and about to process the command line. This can be used, for example, to set shell options; emulate -L zsh and setopt extendedglob are in effect at this point. Any failure to source the file if it exists is treated as fatal. As with other initialisation files, the directory $ZDOTDIR is used instead of $HOME if it is set.
The mathematical library zsh/mathfunc will be loaded if it is available; see the section `The zsh/mathfunc Module' in zshmodules(1). The mathematical functions correspond to the raw system libraries, so trigonometric functions are evaluated using radians, and so on.
Each line typed is evaluated as an expression. The prompt shows a number, which corresponds to a positional parameter where the result of that calculation is stored. For example, the result of the calculation on the line preceded by `4> ' is available as $4. The last value calculated is available as ans. Full command line editing, including the history of previous calculations, is available; the history is saved in the file ~/.zcalc_history. To exit, enter a blank line or type `:q' on its own (`q' is allowed for historical compatibility).
A line ending with a single backslash is treated in the same fashion as it is in command line editing: the backslash is removed, the function prompts for more input (the prompt is preceded by `...' to indicate this), and the lines are combined into one to get the final result. In addition, if the input so far contains more open than close parentheses zcalc will prompt for more input.
If arguments are given to zcalc on start up, they are used to prime the first few positional parameters. A visual indication of this is given when the calculator starts.
The constants PI (3.14159...) and E (2.71828...) are provided. Parameter assignment is possible, but note that all parameters will be put into the global namespace unless the :local special command is used. The function creates local variables whose names start with _, so users should avoid doing so. The variables ans (the last answer) and stack (the stack in RPN mode) may be referred to directly; stack is an array but elements of it are numeric. Various other special variables are used locally with their standard meaning, for example compcontext, match, mbegin, mend, psvar.
The output base can be initialised by passing the option `-#base', for example `zcalc -#16' (the `#' may have to be quoted, depending on the globbing options set).
If the option `-e' is set, the function runs non-interactively: the arguments are treated as expressions to be evaluated as if entered interactively line by line.
If the option `-f' is set, all numbers are treated as floating point, hence for example the expression `3/4' evaluates to 0.75 rather than 0. Options must appear in separate words.
If the option `-r' is set, RPN (Reverse Polish Notation) mode is entered. This has various additional properties:
The prompt is configurable via the parameter ZCALCPROMPT, which undergoes standard prompt expansion. The index of the current entry is stored locally in the first element of the array psvar, which can be referred to in ZCALCPROMPT as `%1v'. The default prompt is `%1v> '.
The variable ZCALC_ACTIVE is set within the function and can be tested by nested functions; it has the value rpn if RPN mode is active, else 1.
A few special commands are available; these are introduced by a colon. For backward compatibility, the colon may be omitted for certain commands. Completion is available if compinit has been run.
The output precision may be specified within zcalc by special commands familiar from many calculators.
Other special commands:
Note that zcalc takes care of all quoting. Hence for example:
:f cube $1 * $1 * $1
defines a function to cube the sole argument. Functions so defined, or indeed any functions defined directly or indirectly using functions -M, are available to execute by typing only the name on the line in RPN mode; this pops the appropriate number of arguments off the stack to pass to the function, i.e. 1 in the case of the example cube function. If there are optional arguments only the mandatory arguments are supplied by this means.
See the comments in the function for a few extra tips.
Not to be confused with the zsh/mathfunc module, described in the section `The zsh/mathfunc Module' in zshmodules(1).
With two arguments, define a mathematical function named mathfunc which can be used in any form of arithmetic evaluation. body is a mathematical expression to implement the function. It may contain references to position parameters $1, $2, ... to refer to mandatory parameters and ${1:-defvalue} ... to refer to optional parameters. Note that the forms must be strictly adhered to for the function to calculate the correct number of arguments. The implementation is held in a shell function named zsh_math_func_mathfunc; usually the user will not need to refer to the shell function directly. Any existing function of the same name is silently replaced.
With one argument, remove the mathematical function mathfunc as well as the shell function implementation.
With no arguments, list all mathfunc functions in a form suitable for restoring the definition. The functions have not necessarily been defined by zmathfuncdef.
The zsh/newuser module comes with a function to aid in configuring shell options for new users. If the module is installed, this function can also be run by hand. It is available even if the module's default behaviour, namely running the function for a new user logging in without startup files, is inhibited.
By default the function exits immediately if it finds any of the files .zshenv, .zprofile, .zshrc, or .zlogin in the appropriate directory. The option -f is required in order to force the function to continue. Note this may happen even if .zshrc itself does not exist.
As currently configured, the function will exit immediately if the user has root privileges; this behaviour cannot be overridden.
Once activated, the function's behaviour is supposed to be self-explanatory. Menus are present allowing the user to alter the value of options and parameters. Suggestions for improvements are always welcome.
When the script exits, the user is given the opportunity to save the new file or not; changes are not irreversible until this point. However, the script is careful to restrict changes to the file only to a group marked by the lines `# Lines configured by zsh-newuser-install' and `# End of lines configured by zsh-newuser-install'. In addition, the old version of .zshrc is saved to a file with the suffix .zni appended.
If the function edits an existing .zshrc, it is up to the user to ensure that the changes made will take effect. For example, if control usually returns early from the existing .zshrc the lines will not be executed; or a later initialization file may override options or parameters, and so on. The function itself does not attempt to detect any such conflicts.
There are a large number of helpful functions in the Functions/Misc directory of the zsh distribution. Most are very simple and do not require documentation here, but a few are worthy of special mention.
The eight base colors are: black, red, green, yellow, blue, magenta, cyan, and white. Each of these has codes for foreground and background. In addition there are seven intensity attributes: bold, faint, standout, underline, blink, reverse, and conceal. Finally, there are seven codes used to negate attributes: none (reset all attributes to the defaults), normal (neither bold nor faint), no-standout, no-underline, no-blink, no-reverse, and no-conceal.
Some terminals do not support all combinations of colors and intensities.
The associative arrays are:
Although it is a misnomer to call them `colors', these arrays also map the other fourteen attributes from names to codes and codes to names.
In addition, the scalar parameters reset_color and bold_color are set to the ANSI terminal escapes that turn off all attributes and turn on bold intensity, respectively.
This is useful in startup files to set options and other state that are not available in all versions of zsh.
is-at-least 3.1.6-15 && setopt NO_GLOBAL_RCS is-at-least 3.1.0 && setopt HIST_REDUCE_BLANKS is-at-least 2.6-17 || print "You can't use is-at-least here."
See also the pager, prompt and rprompt styles below.
var is the name of the variable containing the string to be matched. The variable will be modified directly by the function. The variables MATCH, MBEGIN, MEND, match, mbegin, mend should be avoided as these are used by the regular expression code.
regexp is the regular expression to match against the string.
replace is the replacement text. This can contain parameter, command and arithmetic expressions which will be replaced: in particular, a reference to $MATCH will be replaced by the text matched by the pattern.
The return status is 0 if at least one match was performed, else 1.
In the discussion which follows, if cmd is a file system path, it is first reduced to its rightmost component (the file name).
Help is first sought by looking for a file named cmd in the directory named by the HELPDIR parameter. If no file is found, an assistant function, alias, or command named run-help-cmd is sought. If found, the assistant is executed with the rest of the current command line (everything after the command name cmd) as its arguments. When neither file nor assistant is found, the external command `man cmd' is run.
An example assistant for the "ssh" command:
run-help-ssh() {
emulate -LR zsh
local -a args
# Delete the "-l username" option
zparseopts -D -E -a args l:
# Delete other options, leaving: host command
args=(${@:#-*})
if [[ ${#args} -lt 2 ]]; then
man ssh
else
run-help $args[2]
fi
}
Several of these assistants are provided in the Functions/Misc directory. These must be autoloaded, or placed as executable scripts in your search path, in order to be found and used by run-help.
This function must be used as a ZLE widget:
autoload -U tetris zle -N tetris bindkey keys tetris
To start a game, execute the widget by typing the keys. Whatever command line you were editing disappears temporarily, and your keymap is also temporarily replaced by the Tetris control keys. The previous editor state is restored when you quit the game (by pressing `q') or when you lose.
If you quit in the middle of a game, the next invocation of the tetris widget will continue where you left off. If you lost, it will start a new game.
This version does not save the game state between invocations, and is not invoked as a widget, but rather as:
autoload -U tetriscurses tetriscurses
The option list represents options of the zargs command itself, which are the same as those of xargs. The input list is the collection of strings (often file names) that become the arguments of the command, analogous to the standard input of xargs. Finally, the arg list consists of those arguments (usually options) that are passed to the command each time it runs. The arg list precedes the elements from the input list in each run. If no command is provided, then no arg list may be provided, and in that event the default command is `print' with arguments `-r --'.
For example, to get a long ls listing of all non-hidden plain files in the current directory or its subdirectories:
autoload -U zargs zargs -- **/*(.) -- ls -ld --
The first and third occurrences of `--' are used to mark the end of options for zargs and ls respectively to guard against filenames starting with `-', while the second is used to separate the list of files from the command to run (`ls -ld --').
The first `--' would also be needed if there was a chance the list might be empty as in:
zargs -r -- ./*.back(#qN) -- rm -f
In the event that the string `--' is or may be an input, the -e option may be used to change the end-of-inputs marker. Note that this does not change the end-of-options marker. For example, to use `..' as the marker:
zargs -e.. -- **/*(.) .. ls -ld --
This is a good choice in that example because no plain file can be named `..', but the best end-marker depends on the circumstances.
The options -i, -I, -l, -L, and -n differ slightly from their usage in xargs. There are no input lines for zargs to count, so -l and -L count through the input list, and -n counts the number of arguments passed to each execution of command, including any arg list. Also, any time -i or -I is used, each input is processed separately as if by `-L 1'.
For details of the other zargs options, see xargs(1) (but note the difference in function between zargs and xargs) or run zargs with the --help option.
Only one name argument is allowed. If the -f option is given, the name is taken to be that of a function; if the function is marked for autoloading, zed searches for it in the fpath and loads it. Note that functions edited this way are installed into the current shell, but not written back to the autoload file. In this case the -x option specifies that leading tabs indenting the function according to syntax should be converted into the given number of spaces; `-x 2' is consistent with the layout of functions distributed with the shell.
Without -f, name is the path name of the file to edit, which need not exist; it is created on write, if necessary.
While editing, the function sets the main keymap to zed and the vi command keymap to zed-vicmd. These will be copied from the existing main and vicmd keymaps if they do not exist the first time zed is run. They can be used to provide special key bindings used only in zed.
If it creates the keymap, zed rebinds the return key to insert a line break and `^X^W' to accept the edit in the zed keymap, and binds `ZZ' to accept the edit in the zed-vicmd keymap.
The bindings alone can be installed by running `zed -b'. This is suitable for putting into a startup file. Note that, if rerun, this will overwrite the existing zed and zed-vicmd keymaps.
Completion is available, and styles may be set with the context prefix `:completion:zed'.
A zle widget zed-set-file-name is available. This can be called by name from within zed using `\ex zed-set-file-name' (note, however, that because of zed's rebindings you will have to type ^j at the end instead of the return key), or can be bound to a key in either of the zed or zed-vicmd keymaps after `zed -b' has been run. When the widget is called, it prompts for a new name for the file being edited. When zed exits the file will be written under that name and the original file will be left alone. The widget has no effect with `zed -f'.
While zed-set-file-name is running, zed uses the keymap zed-normal-keymap, which is linked from the main keymap in effect at the time zed initialised its bindings. (This is to make the return key operate normally.) The result is that if the main keymap has been changed, the widget won't notice. This is not a concern for most users.
zmv '(*).lis' '$1.txt'
renames `foo.lis' to `foo.txt', `my.old.stuff.lis' to `my.old.stuff.txt', and so on.
The pattern is always treated as an EXTENDED_GLOB pattern. Any file whose name is not changed by the substitution is simply ignored. Any error (a substitution resulted in an empty string, two substitutions gave the same result, the destination was an existing regular file and -f was not given) causes the entire function to abort without doing anything.
In addition to pattern replacement, the variable $f can be referrred to in the second (replacement) argument. This makes it possible to use variable substitution to alter the argument; see examples below.
Options:
Further examples:
zmv -v '(* *)' '${1// /_}'
For any file in the current directory with at least one space in the name, replace every space by an underscore and display the commands executed.
zmv -v '* *' '${f// /_}'
This does exactly the same by referring to the file name stored in $f.
For more complete examples and other implementation details, see the zmv source file, usually located in one of the directories named in your fpath, or in Functions/Misc/zmv in the zsh distribution.
zstyle+ ':foo:bar' style1 value1 \
+':baz' style2 value2 \
+':frob' style3 value3
This defines style1 with value1 for the context :foo:bar as usual, but it also defines style2 with value2 for the context :foo:bar:baz and style3 with value3 for :foo:bar:frob. Any subcontext may be the empty string to re-use the first context unchanged.
IEEE Standard for information Technology - Portable Operating System Interface (POSIX) - Part 2: Shell and Utilities, IEEE Inc, 1993, ISBN 1-55937-255-9.