NAME
btparse - C library for parsing and processing BibTeX data files
SYNOPSIS
#include <btparse.h>
/* Basic library initialization / cleanup */
void bt_initialize (void);
void bt_free_ast (AST *ast);
void bt_cleanup (void);
/* Input / interface to parser */
void bt_set_stringopts (bt_metatype_t metatype, ushort options);
AST * bt_parse_entry_s (char * entry_text,
char * filename,
int line,
ushort options,
boolean * status);
AST * bt_parse_entry (FILE * infile,
char * filename,
ushort options,
boolean * status);
AST * bt_parse_file (char * filename,
ushort options,
boolean * overall_status);
/* AST traversal/query */
AST * bt_next_entry (AST * entry_list,
AST * prev_entry)
AST * bt_next_field (AST *entry, AST *prev, char **name);
AST * bt_next_value (AST *head,
AST *prev,
bt_nodetype_t *nodetype,
char **text);
bt_metatype_t bt_entry_metatype (AST *entry);
char *bt_entry_type (AST *entry);
char *bt_entry_key (AST *entry);
char *bt_get_text (AST *node);
/* Splitting 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);
/* Formatting names */
bt_name_format * bt_create_name_format (char * parts, boolean abbrev_first);
void bt_free_name_format (bt_name_format * format);
void bt_set_format_text (bt_name_format * format,
bt_namepart part,
char * pre_part,
char * post_part,
char * pre_token,
char * post_token);
void bt_set_format_options (bt_name_format * format,
bt_namepart part,
boolean abbrev,
bt_joinmethod join_tokens,
bt_joinmethod join_part);
char * bt_format_name (bt_name * name, bt_name_format * format);
/* Construct tree from TeX groups */
bt_tex_tree * bt_build_tex_tree (char * string);
void bt_free_tex_tree (bt_tex_tree **top);
void bt_dump_tex_tree (bt_tex_tree *node, int depth, FILE *stream);
char * bt_flatten_tex_tree (bt_tex_tree *top);
/* Miscellaneous string utilities */
void bt_purify_string (char * string, ushort options);
void bt_change_case (char transform, char * string, ushort options);
DESCRIPTION
btparse is a C library for parsing and processing BibTeX files. It
provides a lexical scanner and LR parser (constructed by PCCTS), both
of which are efficient and offer good error detection and recovery; a
set of functions for traversing the AST (abstract syntax tree)
generated by the parser; and utility functions for manipulating strings
according to BibTeX conventions. (Note that nothing in the library
assumes that you’re using BibTeX files for their original purpose of
bibliographic data for scholarly publications; you could use the file
format for any conceivable purpose that fits it. However, there is
some code in the library that is really only appropriate for use with
strings meant to be processed in the same way that BibTeX itself does.
This is all entirely optional, though.)
Note that the interface provided by btparse, while complete, is fairly
low-level. If you have more sophisticated needs, you might be
interested my "Text::BibTeX" module for Perl 5 (available on CPAN).
CONCEPTS AND TERMINOLOGY
To understand this document and use btparse, you should already be
familiar with the BibTeX language---more specifically, the BibTeX data
description language. (BibTeX being the complex beast that it is, one
can conceive of the term applying to the program, the data language,
the particular database structure described in the original BibTeX
documentation, the ".bst" formatting language, and the set of
conventions embodied in the standard styles included with the BibTeX
distribution. In this document, I’ll stick to the first two
meanings---the data language because that’s what btparse deals with,
and the program because it’s occasionally necessary to explain
differences between my parser and BibTeX’s.)
In particular, you should have a good idea what’s going on in the
following:
@string{and = { and },
joe = "Blow, Joe",
john = "John Smith"}
@book(ourbook,
author = joe # and # john,
title = {Our Little Book})
If this looks like something you want to parse, but don’t want to have
to write your own parser for, you’ve come to the right place.
Before going much further, though, you’re going to have to learn some
of the terminology I use for describing BibTeX data. Most of it’s the
same as you’ll find in any BibTeX documentation, but it’s important to
be sure that we’re talking about the same things here. So, some
definitions:
top-level
All text in a BibTeX file from the start of the file to the start
of the first entry, and between entries thereafter.
name
A string of letters, digits, and the following characters:
! $ & * + - . / : ; < > ? [ ] ^ _ ‘ │
A "name" is a catch-all used for entry types, entry keys, and field
and macro names. For BibTeX compatibility, there are slightly
different rules for these four entities; currently, the only such
rule actually implemented is that field and macro names may not
begin with a digit. Some names in the above example: "string",
"and".
entry
A chunk of text starting with an "at" sign ("@") at top-level,
followed by a name (the entry type), an entry delimiter ("{" or
"("), and proceeding to the matching closing delimiter. Also, the
data structure that results from parsing this chunk of text. There
are two entries in the above example.
entry type
The name that comes right after an "@" at top-level. Examples from
above: "string", "book".
entry metatype
A classification of entry types that allows us to group one or more
entry types under the same heading. With the standard BibTeX
database structure, "article", "book", "inbook", etc. all fall
under the "regular entry" metatype. Other metatypes are "macro
definition" (for "string" entries), "preamble" (for "preamble")
entries, and "comment" ("comment" entries). In fact, any entry
whose type is not one of "string", "preamble", or "comment" is
called a "regular" entry.
entry delimiters
"{" and "}", or "(" and ")": the pair of characters that (almost)
mark the boundaries of an entry. "Almost" because the start of an
entry is marked by an "@", not by the "entry open" delimiter.
entry key
(Or just key when it’s clear what we’re speaking of.) The name
immediately following the entry open delimiter in a regular entry,
which uniquely identifies the entry. Example from above:
"ourbook". Only regular entries have keys.
field
A name to the left of an equals sign in a regular or macro-
definition entry. In the latter context, might also be called a
macro name. Examples from above: "joe", "author".
field list
In a regular entry, everything between the entry delimiters except
for the entry key. In a macro definition entry, everything between
the entry delimiters (possibly also called a macro list).
compound value
(Usually just "value".) The text that follows an equals sign ("=")
in a regular or macro definition entry, up to a comma or the entry
close delimiter; a list of one or more simple values joined by hash
signs ("#").
simple value
A string, macro, or number.
string
(Or, sometimes, "quoted string.") A chunk of text between quotes
(""") or braces ("{" and "}"). Braces must balance: "{this is a
{string}" is not a BibTeX string, but "{this is a {string}}" is.
("this is a {string" is also illegal, mainly to avoid the
possibility of generating bogus TeX code--which BibTeX will do in
certain cases.)
macro
A name that appears on the right-hand side of an equals sign (i.e.
as one simple value in a compound value). Implies that this name
was defined as a macro in an earlier macro definition entry, but
this is only checked if btparse is being asked to expand macros to
their full definitions.
number
An unquoted string of digits.
Working with btparse generally consists of passing the library some
BibTeX data (or a source for some BibTeX data, such as a filename or a
file pointer), which it then lexically scans, parses, and constructs an
abstract syntax tree (AST) from. It returns this AST to you, and you
call other btparse functions to traverse and query the tree.
The contents of AST nodes are the private domain of the library, and
you shouldn’t go poking into them. This being C, though, there’s
nothing to prevent you from doing so except good manners and the
possibility that I might change the AST structure in future releases,
breaking any badly-behaved code. Also, it’s not necessary to know the
structural relationships between nodes in the AST---that’s taken care
of by the query/traversal functions.
However, it’s useful to know some of the things that btparse deposits
in the AST and returns to you through those query/traversal functions.
First off, each node has a "node type," which records the syntactic
element corresponding to each node. For instance, the entry
@book{mybook, author = "Joe Blow", title = "My Little Book"}
is rooted by an "entry" node; under this would be found a "key" node
(for the entry key), two "field" nodes (for the "author" and "title"
fields); and associated with each field node would be a "string" node.
The only time this concerns you is when you ask the library for a
simple value; just looking at the text is not enough to distinguish
quoted strings, numbers, and macro names, so btparse returns the
nodetype as well.
In addition to the nodetype, btparse records the metatype of each
"entry" node. This allows you (and the library) to distinguish, say,
regular entries from comment entries. Not only do they have very
different structures and must therefore be traversed differently by the
library, but certain traversal functions make no sense on certain entry
metatypes---thus it’s necessary for you to be able to make the
distinction as well.
That said, everything you need to know to work with the AST is
explained in bt_traversal.
DATA TYPES AND MACROS
btparse defines several types required for the external interface.
First, it trivially defines a "boolean" type (along with "TRUE" and
"FALSE" macros). This might affect you when including the btparse.h
header in your own code---since it’s not possible for the code to
detect if there is already a "boolean" type defined, you might have to
define the "HAVE_BOOLEAN" pre-processor token to deactivate btparse.h’s
"typedef" of "boolean".
Next, two enumeration types are defined: "bt_metatype" and
"bt_nodetype". Both of these are used extensively in the library
itself, and are made available to users of the library because they can
be found in nodes of the "btparse" AST (abstract syntax tree). (I.e.,
querying the AST can give you "bt_metatype" and "bt_nodetype" values,
so the "typedef"s must be available to your code.)
Entry metatype enum
"bt_metatype_t" has the following values:
· "BTE_UNKNOWN"
· "BTE_REGULAR"
· "BTE_COMMENT"
· "BTE_PREAMBLE"
· "BTE_MACRODEF"
which are determined by the "entry type" token. (@string entries have
the "BTE_MACRODEF" metatype; @comment and @preamble correspond to
"BTE_COMMENT" and "BTE_PREAMBLE"; and any other entry type has the
"BTE_REGULAR" metatype.)
AST nodetype enum
"bt_nodetype" has the following values:
· "BTAST_UNKNOWN"
· "BTAST_ENTRY"
· "BTAST_KEY"
· "BTAST_FIELD"
· "BTAST_STRING"
· "BTAST_NUMBER"
· "BTAST_MACRO"
Of these, you’ll only ever deal with the last three. They are returned
when you query the AST for a simple value---just seeing the text isn’t
enough to distinguish between a quoted string, a number, and a macro,
so the AST nodetype is supplied along with the text.
String processing option macros
Since BibTeX is essentially a system for glueing strings together in a
wide variety of ways, the processing done to its strings is fairly
important. Most of the string transformations are done outside of the
lexer/parser; this reduces their complexity, and makes it easier to
switch different transformations on and off. This switching is done
with an "options" bitmap which can be specified on a per-entry-metatype
basis. (That is, you can have one set of transformations done to the
strings in all regular entries, another set done to the strings in all
macro definition entries, and so on.) If you need finer control than
that, it’s currently unavailable outside of the library (but it’s just
a matter of making a couple functions available and documenting
them---so bug me if you need this feature).
There are three basic macros for constructing this bitmap:
"BTO_CONVERT"
Convert "number" values to strings. (The conversion is trivial,
involving changing the type of the AST node representing the number
from "BTAST_NUMBER" to "BTAST_STRING". "Number" values are stored
as strings of digits, just as they are in the input data.)
"BTO_EXPAND"
Expand macro invocations to the full macro text.
"BTO_PASTE"
Paste simple values together.
"BTO_COLLAPSE"
Collapse whitespace according to the BibTeX rules.
For instance, supplying "BTO_CONVERT │ BTO_EXPAND" as the string
options bitmap for the "BTE_REGULAR" metatype means that all simple
values in "regular" entries will be converted to strings: numbers will
simply have their "nodetype" changed, and macros will be expanded.
Nothing else will be done to the simple values, though---they will not
be concatenated, nor will whitespace be collapsed. See the
"bt_set_stringopts()" and "bt_parse_*()" functions in bt_input for more
information on the various options for parsing; see bt_postprocess for
details on the post-processing.
USING THE LIBRARY
The following code is a skeletal example of using the btparse library:
#include <btparse.h>
int main (void)
{
bt_initialize ();
/* process some data */
bt_cleanup ();
exit (0);
}
Please note the call to "bt_initialize()"; this is very important!
Without it, the library may crash or fail mysteriously. You must call
"bt_initialize()" before calling any other btparse functions.
"bt_cleanup()" just frees the memory allocated by "bt_initialize()"; if
you are careful to call it before exiting, and "bt_free_ast()" on any
abstract syntax trees generated by btparse when you are done with them,
then your program shouldn’t have any memory leaks. (Unless they’re due
to your own code, of course!)
BUGS AND LIMITATIONS
btparse has several inherent limitations that are due to the lexical
scanner and parser generated by PCCTS 1.x. In short, the scanner and
parser are both heavily dependent on global variables, meaning that
thread safety -- or even the ability to have two files open and being
parsed at the same time -- is well-nigh impossible. This will not
change until I get with the times and adopt ANTLR 2.0, the successor to
PCCTS -- presuming of course that it can generate more modular C
scanners and parsers.
Another limitation that is due to PCCTS: entries with a large number of
fields (more than about 90, if each field value is just a single
string) will cause the parser to crash. This is unavoidable due to the
parser using statically-allocated stacks for attributes and abstract-
syntax tree nodes. I could increase the static allocation, but that
would just decrease the likelihood of encountering the problem, not
make it go away. Again, the chances of this changing as long as I’m
using PCCTS 1.x are nil.
Apart from those inherent limitations, there are no known bugs in
btparse. Any segmentation faults or bus errors from the library should
be considered bugs. They probably result from using the library
incorrectly (eg. attempting to interleave the parsing of two files),
but I do make an attempt to catch all such mistakes, and if I’ve missed
any I’d like to know about it.
Any memory leaks from the library are also a concern; as long as you
are conscientious about calling the cleanup functions ("bt_free_ast()"
and "bt_cleanup()"), then the library shouldn’t leak.
SEE ALSO
To read and parse BibTeX data files, see bt_input.
To traverse the syntax tree that results, see bt_traversal.
To learn what is done to values in parsed entries, and how to customize
that munging, see bt_postprocess.
To learn how btparse deals with strings, see bt_strings (oops, I
haven’t written this one yet!).
To manipulate and access the btparse macro table, see bt_macros.
For splitting author names and lists "the BibTeX way" using btparse,
bt_split_names.
To put author names back together again, see bt_format_names.
Miscellaneous functions for processing strings "the BibTeX way":
bt_misc.
A semi-formal language definition is in bt_language.
AUTHOR
Greg Ward <gward@python.net>
COPYRIGHT
Copyright (c) 1996-97 by Gregory P. Ward.
This library is free software; you can redistribute it and/or modify it
under the terms of the GNU Library General Public License as published
by the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This library is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
AVAILABILITY
The btOOL home page, where you can get up-to-date information about
btparse (and download the latest version) is
http://starship.python.net/~gward/btOOL/
You will also find the latest version of Text::BibTeX, the Perl library
that provides a high-level front-end to btparse, there. btparse is
needed to build "Text::BibTeX", and must be downloaded separately.
Both libraries are also available on CTAN (the Comprehensive TeX
Archive Network, "http://www.ctan.org/tex-archive/") and CPAN (the
Comprehensive Perl Archive Network, "http://www.cpan.org/"). Look in
biblio/bibtex/utils/btOOL/ on CTAN, and authors/Greg_Ward/ on CPAN.
For example,
http://www.ctan.org/tex-archive/biblio/bibtex/utils/btOOL/
http://www.cpan.org/authors/Greg_Ward
will both get you to the latest version of "Text::BibTeX" and btparse
-- but of course, you should always access busy sites like CTAN and
CPAN through a mirror.