Section: btparse (1)
Updated: 2021-01-27
Page Index


bt_split_names - splitting up BibTeX names and lists of names  


   bt_stringlist * bt_split_list (char *   string,
                                  char *   delim,
                                  char *   filename,
                                  int      line,
                                  char *   description);
   void bt_free_list (bt_stringlist *list);
   bt_name * bt_split_name (char *  name,
                            char *  filename, 
                            int     line,
                            int     name_num);
   void bt_free_name (bt_name * name);



When BibTeX files are used for their original purpose---bibliographic entries describing scholarly publications---processing lists of names (authors and editors mostly) becomes important. Although such name-processing is outside the general-purpose database domain of most of the btparse library, these splitting functions are provided as a concession to reality: most BibTeX data files use the BibTeX conventions for author names, and a library to process that data ought to be capable of processing the names.

Name-processing comes in two stages: first, split up a list of names into individual strings; second, split up each name into ``parts'' (first, von, last, and jr). The first is actually quite general: you could pick a delimiter (such as 'and', used for lists of names) and use it to divide any string into substrings. "bt_split_list()" could then be called to break up the original string and extract the substrings. "bt_split_name()", however, is quite specific to four-part author names written using BibTeX conventions. (These conventions are described informally in any BibTeX documentation; the description you will find here is more formal and algorithmic---and thus harder to understand.)

See bt_format_names for information on turning split-up names back into strings in a variety of ways.  


   bt_stringlist * bt_split_list (char *   string,
                                  char *   delim,
                                  char *   filename,
                                  int      line,
                                  char *   description)

Splits "string" into substrings delimited by "delim" (a fixed string). The splitting is done according to the rules used by BibTeX for splitting up a list of names, in particular:

delimiters at beginning or end of string are ignored
delimiters must be surrounded by whitespace
matching of delimiters is case insensitive
delimiters at non-zero brace depth are ignored

For instance, if the delimiter is "and", then the string

   Candy and Apples AnD {Green Eggs and Ham}

splits into three substrings: "Candy", "Apples", and "{Green Eggs and Ham}".

If there are extra delimiters at the extremities of the string---say, an "and" at the beginning of the string---then they are included in the first/last string; no warning is currently printed, but this may change. Successive delimiters ("and and") result in a warning and a NULL string being added to the list of substrings. For instance, the string

   and Joe Q. Blow and and Smith, Jr., John

would split into three substrings: "and Joe Q. Blow", "NULL", and "Smith, Jr., John".

(If these rules seem somewhat odd, don't blame me: I just implemented BibTeX's observed behaviour and added warning messages for one of the more obvious and easily-detected mistakes.)

The substrings are returned as a "bt_stringlist" structure:

   typedef struct
      char *  string;
      int     num_items;
      char ** items;
   } bt_stringlist;

There is currently no elegant interface to this structure: you just have to poke around in it yourself. The fields are:

a copy of the "string" parameter passed to "bt_split_list()", but with NUL characters replacing the space after each substring. (This is safe because delimiters must be surrounded by whitespace, which means that each substring is followed by whitespace which is not part of the substring.) You probably shouldn't fiddle with "string"; it's just there so that "bt_free_list()" has something to "free()".
the number of substrings found in the string passed to "bt_split_list()".
an array of "num_items" pointers into "string". For instance, "items[1]" points to the second substring. Since "string" has been mangled with NUL characters, it is safe to treat "items[i]" as a regular C string.

"filename", "line", and "description" are all used for generating warning messages. "filename" and "line" simply describe where the string came from, and "description" is a brief (one word) description of the substrings. For instance, if you are splitting a list of names, supply "name" for "description"---that way, warnings will refer to ``name X'' rather than ``substring x''.

   void bt_free_list (bt_stringlist *list)

Frees a "bt_stringlist" structure as returned by "bt_split_list()". That is, it frees the copy of the string you passed to "bt_split_list()", and then frees the structure itself.

   bt_name * bt_split_name (char *  name,
                            char *  filename, 
                            int     line,
                            int     name_num)

Splits a single BibTeX-style author name into four parts: first, von, last, and jr. This can handle almost all names in the style of the major Western European languages, but not quite. (Alas!)

A name is split by first dividing into tokens; tokens are separated by whitespace or commas at brace-level zero. Thus the name

   van der Graaf, Horace Q.

has five tokens, whereas the name

   {Foo, Bar, and Sons}

consists of a single token.

How tokens are divided into parts depends on the form of the name. If the name has no commas at brace-level zero (as in the second example), then it is assumed to be in either ``first last'' or ``first von last'' form. If there are no tokens that start with a lower-case letter, then ``first last'' form is assumed: the final token is the last name, and all other tokens form the first name. Otherwise, the earliest contiguous sequence of tokens with initial lower-case letters is taken as the `von' part; if this sequence includes the final token, then a warning is printed and the final token is forced to be the `last' part.

If a name has a single comma, then it is assumed to be in ``von last, first'' form. A leading sequence of tokens with initial lower-case letters, if any, forms the `von' part; tokens between the `von' and the comma form the `last' part; tokens following the comma form the `first' part. Again, if there are no token following a leading sequence of lowercase tokens, a warning is printed and the token immediately preceding the comma is taken to be the `last' part.

If a name has more than two commas, a warning is printed and the name is treated as though only the first two commas were present.

Finally, if a name has two commas, it is assumed to be in ``von last, jr, first'' form. (This is the only way to represent a name with a `jr' part.) The parsing of the name is the same as for a one-comma name, except that tokens between the two commas are taken to be the `jr' part.

The one case not properly handled by BibTeX name conventions is a name with a 'jr' part not separated from the last name by a comma; for example:

   Henry Ford Jr.
   George Herbert Walker Bush III

Both of these would be incorrectly interpreted by both BibTeX and bt_split_name(): the "Jr." or "III" token would be taken as the last name, and the other tokekens as a two- or four-part first name. The workaround is to shoehorn the 'jr' into the last name:

   Henry {Ford Jr.}
   George Herbert Walker {Bush III}

but this will make it impossible to extract the last name on its own, e.g. to generate ``author-year'' style citations. This design flaw may be fixed in a future version of btparse.

The split-up name is returned as a "bt_name" structure:

   typedef struct
      bt_stringlist * tokens;
      char ** parts[BT_MAX_NAMEPARTS];
      int     part_len[BT_MAX_NAMEPARTS];
   } bt_name;

Again, there's no nice interface to this structure; you'll just have to access the fields individually. They are:

the name, broken down into a flat list of tokens. See above for a description of the "bt_stringlist" structure.
an array of arrays of pointers into the token list. The major dimension of this beast is the ``name part;'' you should index this dimension using the "bt_namepart" enum. For instance, "parts[BTN_LAST]" is an array of pointers to the tokens comprising the last name; "parts[BTN_LAST][1]" is a "char *": the second token of the 'last' part; and "parts[BTN_LAST][1][0]" is the first character of the second token of the 'last' part.
the length, in tokens, of each part. For instance, you might loop over all tokens in the 'first' part as follows (assuming "name" is a "bt_name *" returned by "bt_split_name()"):

   for (i = 0; i < name->part_len[BTN_FIRST]; i++)
      printf ("token %d of first name: %s\n",
              i, name->parts[BTN_FIRST][i]);

   void bt_free_name (bt_name * name)

Frees the "bt_name" structure created by "bt_split_name()" (including the "bt_stringlist" structure inside the "bt_name").



btparse, bt_format_names  


Greg Ward <>