tta/README

tta/README

  Copyright 2011-2026 Free Software Foundation, Inc.

  Copying and distribution of this file, with or without modification,
  are permitted in any medium without royalty provided the copyright
  notice and this notice are preserved.

See also README-hacking file at top level.

texi2any is a Perl or C program (experimental) for converting Texinfo
input to diverse output formats.  All the code is available in pure Perl,
the C code coverage is partial.  The Perl code can call native code
extensions for speed, while the C code can call Perl modules and embed
a Perl interpreter.  The Perl code, Perl extension interfaces (XS) code
and embedded Perl initialization code are in the perl/ subdirectory.
The C code, including code calling Perl modules, code called by the
extension interfaces and code loading embedded Perl is in the C/
subdirectory.

Specific installation instructions

* The `texi2any' (makeinfo) program is a Perl program in the default case.
  If you prefer an experimental C implementation of the texi2any program,
  you can give the --enable-using-c-texi2any flag to `configure'.  The C
  implementation will only be actually used if all the prerequisites are
  found, which includes a working iconv library, the possibility to embed
  a Perl interpreter and enabled Perl extension modules, known as XS modules.

* The C texi2any implementation uses native code for the parsing and the
  structure step.  For the conversion step, however, Perl code is often
  needed, therefore the C texi2any implementation embeds a Perl interpreter
  to call Perl code.  In that case, XS modules are always used.


Module requirements

These modules and libraries are required (all have been standard
parts of Perl for years, at least since 5.7.3):
  Carp, Config, Data::Dumper, Encode, File::Basename, File::Spec
  Getopt::Long, Unicode::Normalize, Storable

Unicode::Collate (also a standard module since 5.7.3) is required for
correct index sorting.

texi2any also uses the less widely-available modules:
  Locale::Messages, Unicode::EastAsianWidth, Text::Unidecode
For these, internal versions are included, and are installed and used as
part of Texinfo (not disturbing the Perl installation at all) (although
see the output of 'configure --help' for how to use externally installed
modules).

Archive::Zip is needed for the EPUB output format final file creation.
It is not included in Texinfo; it is detected at runtime and an error is
issued if it cannot be loaded.


About running the Texinfo programs from a development source tree:

Regarding texi2any (aka makeinfo), you can run tta/perl/texi2any.pl
directly.  This is the original source file for the program, so it's
convenient to be able to make changes and then run it.

To run the output "tta/perl/texi2any" instead, you can set the environment
variable TEXINFO_DEV_SOURCE to 1.  Otherwise, it will try to use
Texinfo's Perl modules in the installed locations.  "tta/perl/texi2any"
uses the Perl interpreter found by configure, so you might want to run
that script instead of texi2any.pl if it's different from the default
interpreter in your environment.

To directly run programs out of source, you should set the t2a_builddir
variable to the tta/ build directory, in order to have compiled modules
and translated in-document strings found.  If you use scripts with
names ending in .pl or .t test files, the source directory obtained
from the script name might be used to determine the source directories
and you may not need to set them explicitely.  Otherwise, you could
need to set the srcdir environment variable, and/or set the t2a_srcdir
environment variables to the in-source tta/ directory.


Running tests:

Tests in perl/t/ test the Perl modules used by the texi2any command.
To run those tests, you also need:
  Test::More, Data::Compare, Test::Deep
On Debian-based distros, Test::More is part of perl-modules and thus
installed with perl, the packages corresponding to the other modules
are named:
  libdata-compare-perl libtest-deep-perl

Devel::Refcount is needed for some checks.  It is detected at configure
time.  If checks using Devel::Refcount fail, Devel::FindRef, detected at
runtime, gives additional interesting debugging information.

For tests in perl/t/, if the Text::Diff module is installed, the difference
between reference and obtained test results are shown as a diff, which
should be more readable.

The tests in tests/ test the command itself.

Tests of cycles in Perl data require the Devel::Cycle module and uncommenting
the find_cycle calls in Perl modules codes manually.  This intervention
is needed because the run time increases significantly with these calls.
Cycles being found do not cause the tests to fail, so the test logs should
be checked for this information.

Cycles in data are relevant, because they prevent Perl from releasing
memory.  Note, however, that cycles are not removed by default, since it
takes more time to do so than to let the memory be released at the end of a
run.  Removing cycles requires to specifically remove some data that creates
the cycles and is only activated in some cases (more than one manual
converted, tests).


Internals:

The conversion of Texinfo to output formats is done in three steps:
1) input Texinfo code parsing into a Texinfo tree representing the
   Texinfo code structure.  Using the Texinfo::ParserNonXS Perl module
   or C code in the C/parsetexi directory.
2) information gathering on the document structure and Texinfo tree
   transformations.  Using the Texinfo::Structuring and
   Texinfo::Transformation Perl modules or C code in the C/structuring_transfo
   directory.
3) conversion of the Texinfo tree to output format by a converter
   backend.  Using Texinfo::Converter::**** Perl modules or C code in
   C/convert.

The C converage of steps 1) and 2) is complete.  Many converter backends
are only available in Perl.

Much of the code is available in two versions, both Perl and C.  Please
check if any code you are updating needs to be updated in both languages.
Affected files should have a comment near the top stating if this is the
case but you can grep the code base for symbol names if that fails.
You can check for an "ALTIMP" comment giving the name of a source code
file that is part of an alternative implementation - paths are relative
either to tta/ or the directory containing the file.

Some, but not all of the C code is also available to Perl code as
"XSUBs" (external subroutines).  These interfaces are defined in files
with a .xs file extension.


If you want to delve into making a new Perl converter backend, the
documentation in perl/Texinfo/Convert/Converter.pm is a good starting
point, as it describes the existing backends and other places to look.

The Texinfo parser builds a complicated parse tree.  It can output a lot of
debug information about the tree, and what it's doing generally.  For example,
these commands output the tree (in different forms):
  texi2any -c TEXINFO_OUTPUT_FORMAT=debugtree document.texi
Or:
  texi2any -c DUMP_TREE=1 -c TEXINFO_OUTPUT_FORMAT=parse document.texi
In addition (or instead) setting the DEBUG configuration variable will
dump more information about what it's doing.


Sometime the messages on syntax errors can be incomplete because of the
system used to set the paths.  It is possible in that case to use something
along the following line to test the validity of one source file from
the perl/ subdirectory:

perl -c -I '.' -I '../maintain/lib/libintl-perl/lib/' -I '../maintain/lib/Unicode-EastAsianWidth/lib/' -I '../maintain/lib/Text-Unidecode/lib' Texinfo/Convert/HTML.pm


The modules in perl/Texinfo/Example are not developped anymore.  The Docbook
conversion module in this directory was developped using Texinfo::Reader.
The Reader interface is not used from Perl, as it proved to be too slow
in Perl and difficult to implement with XS code.  Going forward, the SWIG
interface Reader should be used.  The SWIG interface is in the swig directory.