Changes for version 1.00
- Rewrite of the btparse C library. See "UPGRADING FROM 0.x" in Text::BibTeX for the incompatible changes.
- The PCCTS-generated parser is replaced by a hand-written, reentrant lexer and parser. All state (lexer state, error counts, macro tables) lives in explicit objects, so several files can be parsed at once and Text::BibTeX is thread-safe. The fixed-size parser stacks (and their limits on the number of fields in an entry) are gone.
- The C library never prints, exits or aborts. Errors are collected and reported as Text::BibTeX::Error objects through Perl's warn/die: they can be caught with $SIG{__WARN__}, redirected with STDERR, controlled with the 'Text::BibTeX' warnings category, or handled by $Text::BibTeX::Error::Handler. Entries record their errors (Text::BibTeX::Entry::errors). Usage errors throw exceptions.
- Input is read through PerlIO: any filehandle works, including in-memory handles, tied handles and handles with :encoding layers. Text::BibTeX::File::open accepts a filehandle or a scalar reference, and new 'encoding', 'private_macros' and 'filename' options.
- Full Unicode support: binmode 'utf-8' is now the default; input is validated (invalid bytes become U+FFFD with a warning) and normalised once when read, rather than on every access. Field names, entry types and macro names use full Unicode case mappings; name splitting ("von" detection), initials (grapheme clusters), purify_string and change_case are Unicode-aware, using tables generated at build time from the building perl's Unicode database. TeX markup is passed through unchanged. Files opened for writing in utf-8 mode get an :encoding(UTF-8) layer.
- Better error recovery: a missing closing brace no longer loses the following entry, and a syntax error no longer stops the rest of the file being read.
- Fixed crashes and undefined behaviour in the old library: consecutive or leading commas in names, "and and" with doubled spaces, empty delimiters, preambles containing macros in preserve mode, stray bytes at the end of change_case results, and many uses of uninitialised memory found by fuzzing.
- The C library is compiled straight into the XS module: libbtparse is no longer built or installed (no more library path problems), and the biblex, bibparse and dumpnames programs are gone. The build no longer needs Config::AutoConf or ExtUtils::LibBuilder.
- Inputs which BibTeX accepts are no longer syntax errors: an entry with a key but no fields (@misc{key}), and @comment not followed by a braced or parenthesised comment, whose text is then skipped like other text between entries (an unterminated @comment{ is still an error).
- New: Text::BibTeX::unicode_version(); a stand-alone fuzzing harness for the C library (fuzz/).
- Text::BibTeX::Error objects record the entry key (key) and the calling location (caller_file, caller_line); name warnings from Text::BibTeX::Entry::names number names from 1.
- Requires Perl 5.26.
Documentation
the BibTeX data language, as recognized by Text::BibTeX
Frequently-Asked Questions about Text::BibTeX
formatting BibTeX names for consistent output
input/parsing functions in btparse library
accessing and manipulating the btparse macro table
miscellaneous BibTeX-like string-processing utilities
post-processing of BibTeX strings, values, and entries
splitting up BibTeX names and lists of names
AST traversal/query functions in btparse library
C library for parsing and processing BibTeX data files
Modules
interface to read and parse BibTeX files
defines the "Bib" database structure
formats bibliography entries
generate sort keys for bibliographic entries
read and parse BibTeX files
errors and warnings from the BibTeX parser
interface to whole BibTeX files
interface to BibTeX-style author names
format BibTeX-style author names
provides base classes for user structure modules
interfaces to BibTeX values and simple values
Provides
in lib/Text/BibTeX/Bib.pm
in lib/Text/BibTeX/Value.pm
in lib/Text/BibTeX/Structure.pm