libextractor

libextractor.texi (46742B)

---

 \input texinfo                  @c -*- Texinfo -*-
 @c % The structure of this document is based on the
 @c % Texinfo manual from libgcrypt by Werner Koch and 
 @c % and Moritz Schulte.
 @c %**start of header
 @setfilename libextractor.info
 @include version.texi
 @settitle The GNU libextractor Reference Manual
 @c Unify all the indices into concept index.
 @syncodeindex fn cp
 @syncodeindex vr cp
 @syncodeindex ky cp
 @syncodeindex pg cp
 @syncodeindex tp cp
 @c %**end of header
 @copying
 This manual is for GNU libextractor
 (version @value{VERSION}, @value{UPDATED}), a library for metadata
 extraction.
 
 Copyright @copyright{} 2007, 2010, 2012 Christian Grothoff
 
 @quotation
 Permission is granted to copy, distribute and/or modify this document
 under the terms of the GNU Free Documentation License, Version 1.3
 or any later version published by the Free Software Foundation;
 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
 Texts.  A copy of the license is included in the section entitled ``GNU
 Free Documentation License''.
 @end quotation
 @end copying
 
 @dircategory Software libraries
 @direntry
 * Libextractor: (libextractor).    Metadata extraction library.
 @end direntry
 
 
 
 @c
 @c Titlepage
 @c
 @titlepage
 @title The GNU libextractor Reference Manual
 @subtitle Version @value{VERSION}
 @subtitle @value{UPDATED}
 @author Christian Grothoff (@email{christian@@grothoff.org})
 
 @page
 @vskip 0pt plus 1filll
 @insertcopying
 @end titlepage
 
 @summarycontents
 @contents
 
 
 @ifnottex
 @node Top
 @top The GNU libextractor Reference Manual
 @insertcopying
 @end ifnottex
 
 @menu
 * Introduction::                 What is GNU libextractor.
 * Preparation::                  What you should do before using the library.
 * Generalities::                 General library functions and data types.
 * Extracting meta data::         How to use GNU libextractor to obtain meta data.
 * Language bindings::            How to use GNU libextractor from languages other than C.
 * Utility functions::            Utility functions of GNU libextractor.
 * Existing Plugins::             What plugins are available.
 * Writing new Plugins::          How to write new plugins for GNU libextractor.
 * Internal utility functions::   Utility functions of GNU libextractor for writing plugins.
 * Reporting bugs::               How to report bugs or request new features.
 
 Appendices
 
 * GNU Free Documentation License::  Copying this manual.
 
 Indices
 
 * Index::                       Index
 @c * Function and Data Index::     Index of functions, variables and data types.
 @c * Type Index::                  Index of data types.
 
 @end menu
 
 
 
 @c **********************************************************
 @c *******************  Introduction  ***********************
 @c **********************************************************
 @node Introduction
 @chapter Introduction
 
 @cindex error handling
 GNU libextractor is GNU's library for extracting meta data from
 files.  Meta data includes format information (such as mime type,
 image dimensions, color depth, recording frequency), content
 descriptions (such as document title or document description) and
 copyright information (such as license, author and contributors).
 Meta data extraction is an inherently uncertain business --- a parse
 error can be a corrupt file, an incompatibility in the file format
 version, an entirely different file format or a bug in the parser.  As
 a result of this uncertainty, GNU libextractor deliberately
 avoids to ever report any errors.  Unexpected file contents simply
 result in less or possibly no meta data being extracted.  
 
 @cindex plugin
 GNU libextractor uses plugins to handle various file formats.
 Technically a plugin can support multiple file formats; however, most
 plugins only support one particular format.  By default,
 GNU libextractor will use all plugins that are available and found
 in the plugin installation directory.  Applications can
 request the use of only specific plugins or the exclusion of
 certain plugins.
 
 GNU libextractor is distributed with the @command{extract} 
 command@footnote{Some distributions ship @command{extract} in a
 seperate package.} which is a command-line tool for extracting
 meta data.  @command{extract} is given a list of filenames and 
 prints the resulting meta data to the console.  The @command{extract}
 source code also serves as an advanced example for how to use
 GNU libextractor.  
 
 This manual focuses on providing documentation for writing software
 with GNU libextractor.  The only relevant parts for end-users
 are the chapter on compiling and installing GNU libextractor
 (@xref{Preparation}.).  Also, the chapter on existing plugins maybe of
 interest (@xref{Existing Plugins}.).  Additional documentation for
 end-users can be find in the man page on @command{extract} (using
 @verb{|man extract|}).
 
 @cindex license
 GNU libextractor is licensed under the GNU General Public License,
 specifically, since version 0.7, GNU libextractor is licensed under GPLv3
 @emph{or any later version}.
 
 @node Preparation
 @chapter Preparation
 
 This chapter first describes the general build instructions that
 should apply to all systems.  Specific instructions for known problems
 for particular platforms are then described in individual sections
 afterwards.
 
 Compiling GNU libextractor follows the standard GNU autotools build process
 using @command{configure} and @command{make}.  For details on the GNU
 autotools build process, read the @file{INSTALL} file and query
 @verb{|./configure --help|} for additional options.  
 
 GNU libextractor has various dependencies, most of which are optional. 
 Instead of specifying the names of the software packages, we
 will give the list in terms of the names of the respective
 Debian (wheezy) packages that should be installed.
 
 You absolutely need:
 
 @itemize @bullet
 @item
 libtool
 @item
 gcc
 @item
 make
 @item
 g++ 
 @item
 libltdl7-dev
 @end itemize
 
 Recommended dependencies are:
 @itemize @bullet
 @item
 zlib1g-dev
 @item
 libbz2-dev
 @item
 libgif-dev
 @item
 libvorbis-dev
 @item
 libflac-dev
 @item
 libmpeg2-4-dev
 @item
 librpm-dev
 @item
 libgtk2.0-dev or libgtk3.0-dev
 @item
 libgsf-1-dev
 @item
 libqt4-dev
 @item
 libpoppler-dev
 @item
 libexiv2-dev
 @item
 libavformat-dev
 @item
 libswscale-dev
 @item
 libswresample-dev
 @item
 libgstreamer1.0-dev
 @item
 libgstreamer-plugins-base1.0-dev
 @end itemize
 
 For Subversion access and compilation one also needs:
 @itemize @bullet
 @item
 subversion
 @item
 autoconf
 @item
 automake
 @end itemize
 
 Please notify us if we missed some dependencies (note that the list is
 supposed to only list direct dependencies, not transitive
 dependencies).
 
 Once you have compiled and installed GNU libextractor, you should have a file
 @file{extractor.h} installed in your @file{include/} directory.  This
 file should be the starting point for your C and C++ development with
 GNU libextractor.  The build process also installs the @file{extract} binary and
 man pages for @file{extract} and GNU libextractor.  The @file{extract} man page
 documents the @file{extract} tool.  The GNU libextractor man page gives a brief
 summary of the C API for GNU libextractor.
 
 @cindex packageing
 @cindex directory structure
 @cindex plugin
 @cindex environment variables
 @vindex LIBEXTRACTOR_PREFIX
 When you install GNU libextractor, various plugins will be
 installed in the @file{lib/libextractor/} directory.  The main library
 will be installed as @file{lib/libextractor.so}.  Note that
 GNU libextractor will attempt to find the plugins relative to the
 path of the main library.  Consequently, a package manager can move
 the library and its plugins to a different location later --- as long
 as the relative path between the main library and the plugins is
 preserved.  As a method of last resort, the user can specify an
 environment variable @verb{|LIBEXTRACTOR_PREFIX|}.  If
 GNU libextractor cannot locate a plugin, it will look in
 @verb{|LIBEXTRACTOR_PREFIX/lib/libextractor/|}.
 
 
 @section Installation on GNU/Linux
 
 Should work using the standard instructions without problems.
 
 
 @section Installation on FreeBSD
 
 Should work using the standard instructions without problems.
 
 
 @section Installation on OpenBSD
 
 OpenBSD 3.8 also doesn't have CODESET in @file{langinfo.h}.  CODESET
 is used in GNU libextractor in about three places.  This causes problems
 during compilation.
 
 
 @section Installation on NetBSD
 
 No reports so far.
 
 
 @section Installation using MinGW
 
 Linking -lstdc++ with the provided libtool fails on Cygwin, this
 is a problem with libtool, there is unfortunately no flag to tell
 libtool how to do its job on Cygwin and it seems that it cannot be the
 default to set the library check to 'pass_all'.  Patching libtool may
 help.
 
 Note: this is a rather dated report and may no longer apply.
 
 
 @section Installation on OS X
 
 libextractor has two installation methods on Mac OS X: it can be
 installed as a Mac OS X framework or with the standard
 @command{./configure; make; make install} shell commands. The
 framework package is self-contained, but currently omits some of the
 extractor plugins that can be compiled in if libextractor is installed
 with @command{./configure; make; make install} (provided that the
 required dependencies exist.)
 
 @subsection Installing and uninstalling the framework
 
 The binary framework is distributed as a disk image (@file{Extractor-x.x.xx.dmg}).
 Installation is done by opening the disk image and clicking @file{Extractor.pkg}
 inside it. The Mac OS X installer application will then run. The framework
 is installed to the root volume's @file{/Library/Frameworks} folder and installing
 will require admin privileges.
 
 The framework can be uninstalled by dragging @*
 @file{/Library/Frameworks/Extractor.framework} to the @file{Trash}.
 
 
 @subsection Using the framework
 
 In the framework, the @command{extract} command line tool can be found at @*
 @file{/Library/Frameworks/Extractor.framework/Versions/Current/bin/extract}
 
 The framework can be used in software projects as a framework or as a dynamic
 library. 
 
 When using the framework as a dynamic library in projects using autotools,
 one would most likely want to add  @*
 "-I/Library/Frameworks/Extractor.framework/Versions/Current/include"
 to CPPFLAGS and @*
 "-L/Library/Frameworks/Extractor.framework/Versions/Current/lib"
 to LDFLAGS.
 
 
 @subsection Example for using the framework
 
 @example
 @verbatim
 // hello.c
 #include <Extractor/extractor.h>
 
 int
 main (int argc, char **argv)
 {
   struct EXTRACTOR_PluginList *el;
   el = EXTRACTOR_plugin_load_defaults (EXTRACTOR_OPTION_DEFAULT_POLICY);
   // ...
   EXTRACTOR_plugin_remove_all (el);
   return 0;
 }
 @end verbatim
 @end example
 
 You can then compile the example using
 
 @verbatim
 $ gcc -o hello hello.c -framework Extractor
 @end verbatim
 
 @subsection Example for using the dynamic library
 
 @example
 @verbatim
 // hello.c
 #include <extractor.h>
 int main()
 {
   struct EXTRACTOR_PluginList *el;
   el = EXTRACTOR_plugin_load_defaults (EXTRACTOR_OPTION_DEFAULT_POLICY);
   // ...
   EXTRACTOR_plugin_remove_all (el);
   return 0;
 }
 @end verbatim
 @end example
 
 You can then compile the example using
 
 @verbatim
 $ gcc -I/Library/Frameworks/Extractor.framework/Versions/Current/include \
   -o hello hello.c \
   -L/Library/Frameworks/Extractor.framework/Versions/Current/lib \
   -lextractor
 @end verbatim
 
 Notice the difference in the @code{#include} line.
 
 
 
 
 
 
 @section Note to package maintainers
 
 The suggested way to package GNU libextractor is to split it into
 roughly the following binary packages:
 
 @itemize @bullet
 @item
 libextractor (main library only, only hard dependency for other packages depending on GNU libextractor)
 @item
 extract (command-line tool and man page extract.1)
 @item
 libextractor-dev (extractor.h header and man page libextractor.3)
 @item
 libextractor-doc (this manual)
 @item
 libextractor-plugins (plugins without external dependencies; recommended but not required by extract and libextractor package)
 @item
 libextractor-plugin-XXX (plugin with dependency on libXXX, for example for XXX=mpeg this would be @file{libextractor_mpeg.so})
 @item
 libextractor-plugins-all (meta package that requires all plugins except experimental plugins)
 @end itemize
 
 This would enable minimal installations (i.e. for embedded systems) to
 not include any plugins, as well as moderate-size installations (that
 do not trigger GTK and X11) for systems that have limited resources.
 Right now, the MP4 plugin is experimental and does nothing and should
 thus never be included at all; QuickTime, MP4, M4A and 3GPP files are
 instead handled by the @file{libextractor_qt.so} plugin, which only
 depends on zlib and is part of libextractor-plugins.  The gstreamer plugin is experimental
 but largely works with the correct version of gstreamer and can thus
 be packaged (especially if the dependency is available on the target
 system) but should probably not be part of libextractor-plugins-all.
 
 
 @node Generalities
 @chapter Generalities
 
 @section Introduction to the ``extract'' command
 
 The @command{extract} command takes a list of file names as arguments,
 extracts meta data from each of those files and prints the result to
 the console.  By default, @command{extract} will use all available
 plugins and print all (non-binary) meta data that is found.
 
 The set of plugins used by @command{extract} can be controlled using
 the ``-l'' and ``-n'' options.  Use ``-n'' to not load all of the
 default plugins.  Use ``-l NAME'' to specifically load a certain
 plugin.  For example, specify ``-n -l mime'' to only use the MIME
 plugin.
 
 The argument of ``-l'' is a colon-separated list of plugin names, each
 optionally preceded by a minus sign to remove the plugin instead of
 adding it and optionally followed by a plugin configuration in
 parentheses, as in ``-l "-pdf:pdf(fulltext)"''.  Only the last ``-l''
 option on the command line is used, so several plugins have to be
 named in a single list.  The options understood by the individual
 plugins are described in the chapter @emph{Existing Plugins}, under
 @emph{Video thumbnails and audio previews} and @emph{Full text}.
 
 Using the ``-p'' option the output of @command{extract} can be limited
 to only certain keyword types.  Similarly, using the ``-x'' option,
 certain keyword types can be excluded.  A list of all known keyword
 types can be obtained using the ``-L'' option.
 
 The output format of @command{extract} can be influenced with the
 ``-V'' (more verbose, lists filenames), ``-g'' (grep-friendly, all
 meta data on a single line per file) and ``-b'' (bibTeX style)
 options.
 
 @section Common usage examples for ``extract''
 
 The sample files used here are those of the test suite, in
 @file{src/plugins/testdata/}.  Note that the argument of @code{-l} is
 the short name of the plugin (@code{jpeg}, not @file{jpeg.so} or
 @file{libextractor_jpeg.so}).
 
 @example
 $ extract -n -l jpeg jpeg_image.jpg
 Keywords for file jpeg_image.jpg:
 mimetype - image/jpeg
 image dimensions - 3x3
 comment - (C) 2001 by Christian Grothoff, using gimp 1.2 1
 
 $ extract -n -l jpeg -x comment jpeg_image.jpg
 Keywords for file jpeg_image.jpg:
 mimetype - image/jpeg
 image dimensions - 3x3
 
 $ extract -n -l jpeg -p comment jpeg_image.jpg
 Keywords for file jpeg_image.jpg:
 comment - (C) 2001 by Christian Grothoff, using gimp 1.2 1
 
 $ extract -n -l gif -p comment jpeg_image.jpg gif_image.gif
 Keywords for file jpeg_image.jpg:
 Keywords for file gif_image.gif:
 comment - Testing keyword extraction
 @end example
 
 
 @section Introduction to the libextractor library
 
 Each public symbol exported by GNU libextractor has the prefix
 @verb{|EXTRACTOR_|}.  All-caps names are used for constants.  For the
 impatient, the minimal C code for using GNU libextractor (on the
 executing binary itself) looks like this:
 
 @verbatim
 #include <extractor.h>
 
 int 
 main (int argc, char ** argv) 
 {
   struct EXTRACTOR_PluginList *plugins
     = EXTRACTOR_plugin_add_defaults (EXTRACTOR_OPTION_DEFAULT_POLICY);
   EXTRACTOR_extract (plugins, argv[1],
                      NULL, 0, 
                      &EXTRACTOR_meta_data_print, stdout);
   EXTRACTOR_plugin_remove_all (plugins);
   return 0;
 }
 @end verbatim
 
 The minimal API illustrated by this example is actually sufficient for
 many applications.  The full external C API of GNU libextractor is described
 in chapter @xref{Extracting meta data}.  Bindings for other languages
 are described in chapter @xref{Language bindings}.  The API for
 writing new plugins is described in chapter @xref{Writing new Plugins}.
 
 Note that it is possible for GNU libextractor to encounter a @code{SIGPIPE}
 during its execution.  GNU libextractor --- as it is a library and as such
 should not interfere with your main application --- does NOT install a 
 signal handler for @code{SIGPIPE}.  You thus need to install a signal
 handler (or at least tell your system to ignore @code{SIGPIPE}) if you
 want to avoid unexpected problems during calls to GNU libextractor.  
 @cindex SIGPIPE
 
 @node Extracting meta data
 @chapter Extracting meta data
 
 In order to extract meta data with GNU libextractor you first need to
 load the respective plugins and then call the extraction API
 with the plugins and the data to process.  This section
 documents how to load and unload plugins, the various types
 and formats in which meta data is returned to the application
 and finally the extraction API itself.
 
 @menu
 * Plugin management::   How to load and unload plugins
 * Meta types::          About meta types
 * Meta formats::        About meta formats
 * Extracting::          How to use the extraction API
 @end menu
 
 
 @node Plugin management
 @section Plugin management
 
 @cindex reentrant
 @cindex concurrency
 @cindex threads
 @cindex thread-safety
 @tindex enum EXTRACTOR_Options
 
 Using GNU libextractor from a multi-threaded parent process requires some
 care.  The problem is that on most platforms GNU libextractor starts
 sub-processes for the actual extraction work.  This is useful to
 isolate the parent process from potential bugs; however, it can cause
 problems if the parent process is multi-threaded.  The issue is that
 at the time of the fork, another thread of the application may hold a
 lock (i.e. in gettext or libc).  That lock would then never be
 released in the child process (as the other thread is not present in
 the child process).  As a result, the child process would then
 deadlock on trying to acquire the lock and never terminate.  This has
 actually been observed with a lock in GNU gettext that is triggered by
 the plugin startup code when it interacts with libltdl.
 
 The problem can be solved by loading the plugins using the
 @code{EXTRACTOR_OPTION_IN_PROCESS} option, which will run GNU libextractor
 in-process and thus avoid the locking issue.  In this case, all of the
 functions for loading and unloading plugins, including
 @verb{|EXTRACTOR_plugin_add_defaults|} and
 @verb{|EXTRACTOR_plugin_remove_all|}, are thread-safe and reentrant.
 However, using the same plugin list from multiple threads at the same
 time is not safe.  
 
 All plugin code is expected required to be reentrant and state-less,
 but due to the extensive use of 3rd party libraries this cannot
 be guaranteed.
 
 
 @deftp {C Struct} EXTRACTOR_PluginList
 @tindex struct EXTRACTOR_PluginList
 
 A plugin list represents a set of GNU libextractor plugins.  Most of
 the GNU libextractor API is concerned with either constructing a
 plugin list or using it to extract meta data.  The internal representation
 of the plugin list is of no concern to users or plugin developers.
 @end deftp
 
 
 @deftypefun void EXTRACTOR_plugin_remove_all (struct EXTRACTOR_PluginList *plugins)
 @findex EXTRACTOR_plugin_remove_all
 
 Unload all of the plugins in the given list.
 @end deftypefun
 
 @deftypefun {struct EXTRACTOR_PluginList *} EXTRACTOR_plugin_remove (struct EXTRACTOR_PluginList *plugins, const char*name)
 @findex EXTRACTOR_plugin_remove
 
 Unloads a particular plugin.  The given name should be the short name of the plugin, for example ``mime'' for the mime-type extractor or ``mpeg'' for the MPEG extractor.
 @end deftypefun
 
 @deftypefun {struct EXTRACTOR_PluginList *} EXTRACTOR_plugin_add (struct EXTRACTOR_PluginList *plugins, const char* name,const char* options, enum EXTRACTOR_Options flags)
 @findex EXTRACTOR_plugin_add
 
 Loads a particular plugin.  The plugin is added to the existing list, which can be @code{NULL}.  The second argument specifies the name of the plugin (i.e. ``ogg'').  The third argument can be @code{NULL} and specifies plugin-specific options.  Finally, the last argument specifies if the plugin should be executed out-of-process (@code{EXTRACTOR_OPTION_DEFAULT_POLICY}) or not.
 @end deftypefun
 
 @deftypefun {struct EXTRACTOR_PluginList *} EXTRACTOR_plugin_add_config (struct EXTRACTOR_PluginList *plugins, const char* config, enum EXTRACTOR_Options flags)
 @findex EXTRACTOR_plugin_add_config
 
 Loads and unloads plugins based on a configuration string, modifying the existing list, which can be @code{NULL}.  The string has the format ``[-]NAME(OPTIONS)@{:[-]NAME(OPTIONS)@}*''.  Prefixing the plugin name with a ``-'' means that the plugin should be unloaded.
 @end deftypefun
 
 @deftypefun {struct EXTRACTOR_PluginList *} EXTRACTOR_plugin_add_defaults (enum EXTRACTOR_Options flags)
 @findex EXTRACTOR_plugin_add_defaults
 
 Loads all of the plugins in the plugin directory.  This function is what most GNU libextractor applications should use to setup the plugins.
 @end deftypefun
 
 
 
 @node Meta types
 @section Meta types
 
 
 @tindex enum EXTRACTOR_MetaType
 @findex EXTRACTOR_metatype_get_max
 
 @verb{|enum EXTRACTOR_MetaType|} is a C enum which defines a list of over 100 different types of meta data.  The total number can differ between different GNU libextractor releases; the maximum value for the current release can be obtained using the @verb{|EXTRACTOR_metatype_get_max|} function.  All values in this enumeration are of the form @verb{|EXTRACTOR_METATYPE_XXX|}.
 
 @deftypefun {const char *} EXTRACTOR_metatype_to_string (enum EXTRACTOR_MetaType type)
 @findex EXTRACTOR_metatype_to_string
 @cindex gettext
 @cindex internationalization
 
 The function @verb{|EXTRACTOR_metatype_to_string|} can be used to obtain a short English string @samp{s} describing the meta data type.  The string can be translated into other languages using GNU gettext with the domain set to GNU libextractor (@verb{|dgettext("libextractor", s)|}).  
 @end deftypefun
 
 @deftypefun {const char *} EXTRACTOR_metatype_to_description (enum EXTRACTOR_MetaType type)
 @findex EXTRACTOR_metatype_to_description
 @cindex gettext
 @cindex internationalization
 
 The function @verb{|EXTRACTOR_metatype_to_description|} can be used to obtain a longer English string @samp{s} describing the meta data type.  The description may be empty if the short description returned by @code{EXTRACTOR_metatype_to_string} is already comprehensive.  The string can be translated into other languages using GNU gettext with the domain set to GNU libextractor (@verb{|dgettext("libextractor", s)|}).  
 @end deftypefun
 
 
 
 @node Meta formats
 @section Meta formats
 
 @tindex enum EXTRACTOR_MetaFormat
 
 @verb{|enum EXTRACTOR_MetaFormat|} is a C enum which defines on a high level how the extracted meta data is represented.  Currently, the library uses three formats: UTF-8 strings, C strings and binary data.  A fourth value, @code{EXTRACTOR_METAFORMAT_UNKNOWN} is defined but not used.  UTF-8 strings are 0-terminated strings that have been converted to UTF-8.  The format code is @code{EXTRACTOR_METAFORMAT_UTF8}. Ideally, most text meta data will be of this format.  Some file formats fail to specify the encoding used for the text.  In this case, the text cannot be converted to UTF-8.  However, the meta data is still known to be 0-terminated and presumably human-readable.  In this case, the format code used is @code{EXTRACTOR_METAFORMAT_C_STRING}; however, this should not be understood to mean that the encoding is the same as that used by the C compiler.  Finally, for binary data (mostly images), the format @code{EXTRACTOR_METAFORMAT_BINARY} is used.
 
 Naturally this is not a precise description of the meta format. Plugins can provide a more precise description (if known) by providing the respective mime type of the meta data.  For example, binary image meta data could be also tagged as ``image/png'' and normal text would typically be tagged as ``text/plain''.  
 
 
 
 @node Extracting
 @section Extracting
 
 @deftypefn {Function Pointer} int (*EXTRACTOR_MetaDataProcessor)(void *cls, const char *plugin_name, enum EXTRACTOR_MetaType type, enum EXTRACTOR_MetaFormat format, const char *data_mime_type, const char *data, size_t data_len)
 @tindex EXTRACTOR_MetaDataProcessor
 
 Type of a function that libextractor calls for each meta data item found.
 
 @table @var
 
 @item cls 
 closure (user-defined)
 
 @item plugin_name 
 name of the plugin that produced this value; special values can be used (i.e. '<zlib>' for zlib being used in the main libextractor library and yielding meta data);
 
 @item type 
 libextractor-type describing the meta data;
 
 @item format basic 
 format information about data
 
 @item data_mime_type 
 mime-type of data (not of the original file); can be @code{NULL} (if mime-type is not known);
 
 @item data 
 actual meta-data found
 
 @item data_len 
 number of bytes in data
 
 @end table
 
 Return 0 to continue extracting, 1 to abort.
 @end deftypefn
 
 
 
 @deftypefun void EXTRACTOR_extract (struct EXTRACTOR_PluginList *plugins, const char *filename, const void *data, size_t size, EXTRACTOR_MetaDataProcessor proc, void *proc_cls)
 @findex EXTRACTOR_extract
 @cindex reentrant
 @cindex concurrency
 @cindex threads
 @cindex thread-safety
 
 This is the main function for extracting keywords with GNU libextractor.  The first argument is a plugin list which specifies the set of plugins that should be used for extracting meta data.  The @samp{filename} argument is optional and can be used to specify the name of a file to process.  If @samp{filename} is @code{NULL}, then the @samp{data} argument must point to the in-memory data to extract meta data from.  If @samp{filename} is non-@code{NULL}, @samp{data} can be @code{NULL}.  If @samp{data} is non-null, then @samp{size} is the size of @samp{data} in bytes.  Otherwise @samp{size} should be zero.  For each meta data item found, GNU libextractor will call the @samp{proc} function, passing @samp{proc_cls} as the first argument to @samp{proc}.  The other arguments to @samp{proc} depend on the specific meta data found.  
 
 @cindex SIGBUS
 @cindex bus error
 Meta data extraction should never really fail --- at worst, GNU libextractor should not call @samp{proc} with any meta data. By design, GNU libextractor should never crash or leak memory, even given corrupt files as input.  Note however, that running GNU libextractor on a corrupt file system (or incorrectly @verb{|mmap|}ed files) can result in the operating system sending a SIGBUS (bus error) to the process.  As GNU libextractor typically runs plugins out-of-process, it first maps the file into memory and then attempts to decompress it.  During decompression it is possible to encounter a SIGBUS.   GNU libextractor will @emph{not} attempt to catch this signal and your application is likely to crash.  Note again that this should only happen if the file @emph{system} is corrupt (not if individual files are corrupt).  If this is not acceptable, you might want to consider running GNU libextractor itself also out-of-process (as done, for example, by @url{http://grothoff.org/christian/doodle/,doodle}).
 
 @end deftypefun
 
 
 @node Language bindings
 @chapter Language bindings
 @cindex Java
 @cindex Mono
 @cindex Perl
 @cindex Python
 @cindex PHP
 @cindex Ruby
 
 GNU libextractor works immediately with C and C++ code. Bindings for Java, Mono, Ruby, Perl, PHP and Python are available for download from the main GNU libextractor website.  Documentation for these bindings (if available) is part of the downloads for the respective binding.  In all cases, a full installation of the C library is required before the binding can be installed.
 
 @section Java
 
 Compiling the GNU libextractor Java binding follows the usual process of
 running @command{configure} and @command{make}.  The result will be a
 shared C library @file{libextractor_java.so} with the native code and
 a JAR file (installed to @file{$PREFIX/share/java/libextractor.java}).
 
 A minimal example for using GNU libextractor's Java binding would look
 like this:
 @verbatim
 import org.gnu.libextractor.*;
 import java.util.ArrayList;
 
 public static void main(String[] args) {
   Extractor ex = Extractor.getDefault();
   for (int i=0;i<args.length;i++) {
     ArrayList keywords = ex.extract(args[i]);
     System.out.println("Keywords for " + args[i] + ":");
     for (int j=0;j<keywords.size();j++)
       System.out.println(keywords.get(j));
   }
 }
 @end verbatim
 
 The GNU libextractor library and the @file{libextractor_java.so} JNI binding
 have to be in the library search path for this to work.  Furthermore, the
 @file{libextractor.jar} file should be on the classpath.  
 
 Note that the API does not use Java 5 style generics in order to work
 with older versions of Java.
 
 @section Mono
 
 his binding is undocumented at this point.
 
 @section Perl
 
 This binding is undocumented at this point.
 
 @section Python
 
 This binding is undocumented at this point.
 
 @section PHP
 
 This binding is undocumented at this point.
 
 @section Ruby
 
 This binding is undocumented at this point.
 
 
 
 @node Utility functions
 @chapter Utility functions
 
 @cindex reentrant
 @cindex concurrency
 @cindex threads
 @cindex thread-safety
 This chapter describes various utility functions for GNU libextractor usage. All of the functions are reentrant.
 
 @menu
 * Utility Constants::
 * Meta data printing::
 @end menu
 
 @node Utility Constants
 @section Utility Constants
 
 @findex EXTRACTOR_VERSION
 The constant @verb{|EXTRACTOR_VERSION|} is a hexadecimal
 representation of the version number of the installed libextractor
 header.  The hexadecimal format is 0xAABBCCDD where AA is the major
 version (so far always 0), BB is the minor version, CC is the revision
 and DD the patch number.  For example, for version 0.5.18, we would
 have AA=0, BB=5, CC=18 and DD=0.  Minor releases such as 0.5.18a or
 significant changes in unreleased versions would be marked with DD=1
 or higher.
 
 
 @node Meta data printing
 @section Meta data printing
 
 
 @findex EXTRACTOR_meta_data_print
 The @verb{|EXTRACTOR_meta_data_print|} is a simple function which prints the meta data found with libextractor to a file.  The function is mostly useful for debugging and as an example for how to manipulate the keyword list and can be passed as the @samp{proc} argument to @code{EXTRACTOR_extract}.  The file to print to should be passed as @samp{proc_cls} (which must be of type @code{FILE *}), for example @code{stdout}.
 
 
 
 @node Existing Plugins
 @chapter Existing Plugins
 
 @itemize @bullet
 @item
 APK (Android packages and Java archives, using zlib)
 @item
 ARCHIVE (using libarchive)
 @item
 DISKIMAGE (QCOW2, VMDK, VHD and VHDX virtual disk images)
 @item
 DVI
 @item
 EBOOK (EPUB and MOBI/AZW)
 @item
 EXIV2 (using libexiv2, 0.23 or later preferred)
 @item 
 FLAC (using libFLAC)
 @item
 GEOTIFF (georeferencing tags in TIFF images)
 @item
 GPX (GPS tracks)
 @item
 GIF (using libgif)
 @item
 GSTREAMER (using libgstreamer v1.0 or later)
 @item
 HEIF (HEIC, HEIF and AVIF images)
 @item
 ID3 (MPEG audio with ID3v1/ID3v2 tags, Xing and LAME headers)
 @item
 ISO9660 (CD/DVD images)
 @item
 HTML (using libtidy)
 @item
 IT 
 @item
 JPEG (using libjpeg v8 or later)
 @item
 KML (Keyhole Markup Language)
 @item
 LNK (Windows shell links)
 @item
 MAN
 @item
 MIDI (using libsmf)
 @item
 MIME (using libmagic)
 @item
 MBOX (Unix mailboxes and RFC 5322 messages)
 @item
 MPEG (using libmpeg2)
 @item
 MSOFFICE (OOXML: @file{.docx}, @file{.xlsx}, @file{.pptx}; and
 Excel 2-4 BIFF streams)
 @item
 NSF
 @item
 PECOFF (Windows executables, DLLs and drivers)
 @item
 PLIST (Apple property lists, binary and XML)
 @item
 NSFE
 @item
 ODF
 @item
 OLE2 (with libgsf)
 @item
 OGG (with libogg)
 @item
 PNG
 @item
 PS
 @item
 RIFF
 @item
 RPM (using librpm)
 @item
 RTF (@file{.rtf}, including the revision table, comment authors and
 custom document properties)
 @item
 S3M
 @item
 SQLITE (SQLite 3 databases, write-ahead logs and journals)
 @item
 SID
 @item
 PreviewGst (audio preview, using GStreamer)
 @item
 PreviewOpus (audio preview, using FFmpeg 5.1 or later)
 @item
 ThumbnailFFMPEG (video thumbnail, using libavformat, libavcodec and
 libswscale of FFmpeg 5.1 or later)
 @item
 ThumbnailGst (video thumbnail, using GStreamer 1.8 or later)
 @item
 ThumbnailGtk (still image thumbnail, using gdk-pixbuf)
 @item
 TIFF (with libtiff, tested with v4)
 @item
 TAR (tar archives)
 @item
 WAV
 @item
 WEBP
 @item
 XM
 @item
 ZIP
 @end itemize
 
 @file{gzip} and @file{bzip2} compressed versions of these formats are 
 also supported (as well as meta data embedded by @file{gzip} itself)
 if zlib or libbz2 are available.
 
 @section Video thumbnails and audio previews
 
 Four of the plugins in that list do not report meta data that was found
 in the file, but meta data computed @emph{from} the file: a small image
 taken from a video (@code{EXTRACTOR_METATYPE_THUMBNAIL}) and a short
 audio excerpt re-encoded as Opus in an Ogg container
 (@code{EXTRACTOR_METATYPE_AUDIO_PREVIEW}).  Both are reported as
 @code{EXTRACTOR_METAFORMAT_BINARY}.
 
 Each feature comes in two implementations, one on top of GStreamer and
 one on top of FFmpeg, and which of them are built depends on what
 @command{configure} found:
 
 @table @code
 @item thumbnailgst
 @itemx thumbnailffmpeg
 Decode the video and emit a frame from about 30 seconds in, scaled to
 fit a 128x128 box, as JPEG.  Frames that are a perfectly flat surface
 (a black leader, a fade-in) are skipped.
 @item previewgst
 @itemx previewopus
 Decode about 15 seconds of audio from about 30 seconds in, fade it in
 and out, and emit it as a mono 24 kbit/s Ogg Opus stream with the MIME
 type @code{audio/ogg}.
 @end table
 
 In both cases the position is limited to a third of the duration, so a
 short file is sampled proportionally earlier.
 
 If both backends are installed, both plugins run and you get two
 thumbnails (or two previews) for the same file.  Which one you want is
 an application decision; disable the other one in the configuration
 string, as in
 
 @example
 pl = EXTRACTOR_plugin_add_defaults (EXTRACTOR_OPTION_DEFAULT_POLICY);
 pl = EXTRACTOR_plugin_add_config (pl,
                                   "-thumbnailffmpeg:-previewopus",
                                   EXTRACTOR_OPTION_DEFAULT_POLICY);
 @end example
 
 All four accept a comma separated list of options in the plugin
 configuration string:
 
 @table @code
 @item size=@var{pixels}
 Edge length of the box the thumbnail must fit into (default: 128).
 @item format=jpeg|png
 Image format of the thumbnail (default: @code{jpeg}).
 @item offset=@var{seconds}
 How far into the media to sample (default: 30).
 @item length=@var{seconds}
 Length of the audio preview (default: 15).
 @item bitrate=@var{bits-per-second}
 Opus bitrate of the audio preview (default: 24000).
 @item deadline=@var{milliseconds}
 Wall clock budget for the whole extraction; when it runs out, the
 plugin reports nothing rather than keeping the application waiting
 (default: 10000, or 60000 if @code{srt} or @code{gallery} is given).
 @end table
 
 The two thumbnail plugins also understand:
 
 @table @code
 @item srt:@var{lang}
 Additionally dump the complete subtitle track in language @var{lang}
 as an SRT file (@code{EXTRACTOR_METATYPE_SUBTITLE_TEXT}, UTF-8, MIME
 type @code{application/x-subrip}).  @var{lang} may be given as ISO
 639-1 (@code{de}), ISO 639-2/B (@code{ger}) or ISO 639-2/T
 (@code{deu}) code; only the first matching track is dumped, and only
 text subtitles (SubRip, ASS/SSA, WebVTT, timed text) can be converted.
 Without this option the plugins merely list the language of every
 subtitle track as @code{EXTRACTOR_METATYPE_SUBTITLE_LANGUAGE}.
 @item gallery:@var{n}
 Instead of a single thumbnail, emit one image showing @var{n} (at most
 100) frames taken at roughly even intervals across the whole video,
 arranged in a grid chosen so that the image is as close to 16:9 as
 possible.  @code{size} then applies to each tile.  A frame that shows
 basically nothing (99% of its pixels within a narrow colour range, such
 as a black or white screen) is replaced by one taken a little earlier
 or later.
 @end table
 
 For example, @code{"thumbnailgst(size=256,format=png)"} asks for a
 256x256 PNG, and @code{"thumbnailffmpeg(gallery:12,srt:en)"} for a
 contact sheet of twelve frames plus the English subtitles.
 
 Two caveats apply to the GStreamer variants.  Their decoders and
 encoders are themselves GStreamer plugins, resolved at run time, so a
 system that lacks the @code{jpeg} or @code{opus} GStreamer plugin
 builds them happily and then gets no results; and because GStreamer
 does not survive a @code{fork()} of a process that has already
 initialized it, an application that runs these plugins
 @code{EXTRACTOR_OPTION_IN_PROCESS} should not also use out-of-process
 plugins in the same process afterwards.
 
 @section Full text
 
 The @code{html}, @code{pdf}, @code{ebook}, @code{odf}, @code{ole2}
 and @code{dvi} plugins can report the entire text of a document as a
 single @code{EXTRACTOR_METATYPE_FULL_TEXT} item in
 @code{EXTRACTOR_METAFORMAT_UTF8} with MIME type @code{text/plain}.  As
 the text can be large, this is only done if the plugin is given the
 option @code{fulltext}, as in
 
 @example
 pl = EXTRACTOR_plugin_add_config (NULL,
                                   "pdf(fulltext):odf(fulltext)",
                                   EXTRACTOR_OPTION_DEFAULT_POLICY);
 @end example
 
 @noindent
 or, on the command line, @code{extract -n -l "pdf(fulltext)" file.pdf}.
 Note the @code{-n}: @code{EXTRACTOR_plugin_add} leaves the list
 unchanged if a plugin of that name is already in it, so adding
 @code{pdf(fulltext)} on top of the default plugin set --- which
 already contains @code{pdf}, without the option --- has no effect at
 all.  Either drop the default set with @code{-n}, or remove the plugin
 before adding it back, as in
 @code{extract -l "-pdf:pdf(fulltext)" file.pdf}.
 
 Within a paragraph, runs of whitespace (including line breaks) are
 reported as a single space; paragraphs are separated by exactly two
 newlines (@code{"\n\n"}).  Where the format makes it possible, text
 that is not part of the running text is left out: page numbers,
 running headers and footers, navigation bars (HTML) and the like.  The
 text is truncated at 16 MiB.
 
 @section Forensic and bulk-scan plugins
 
 Sixteen of the plugins in that list were added for a different job from
 the rest: identifying, and saying something useful about, every file on
 a large volume in a single fast pass.  That purpose shapes how they
 behave, in ways worth knowing about before you use them.
 
 They answer questions about @emph{provenance} rather than about
 content --- which machine, user, tool, device or account a file came
 from.  A Windows executable reports the compiler that built it, the
 path its debug symbols were written to (which routinely discloses the
 build machine's user name) and the exploit mitigations it opts into; a
 shell link reports the volume serial number, the label and the NetBIOS
 name of the machine the target lived on, and often that machine's MAC
 address; a tar archive reports the user and group names of the system
 it was packed on; a mailbox reports the relay chain a message
 travelled and the addresses it passed through.
 
 To keep a bulk scan fast they read headers, not files.  Each rejects a
 file that is not its format within the first few bytes --- which is the
 overwhelmingly common case --- and none of them decodes payload data.
 The @code{sqlite} plugin in particular reads only the 100-byte header
 and never opens the database: opening an untrusted database would be
 both slow and unsafe.  For the same reason none of these plugins links
 a third-party parsing library, so all but @code{apk} and @code{ebook},
 which must inflate members of a zip container, are built
 unconditionally.
 
 Because they characterise rather than enumerate, plugins that meet
 repeating structures --- archive members, mail messages, requested
 permissions, map placemarks --- report a bounded number of them
 together with an @code{EXTRACTOR_METATYPE_ENTRY_COUNT} giving the true
 total.  Do not read the absence of a further item as the absence of
 further items.
 
 Several of the meta types these plugins report were introduced with
 them, and an application that switches on
 @code{enum EXTRACTOR_MetaType} will not know them.  Values are only
 ever appended, so existing values keep their meaning; use
 @code{EXTRACTOR_metatype_get_max()} rather than a hard-coded bound, and
 @code{EXTRACTOR_metatype_to_string()} to render a type you do not
 recognise.
 
 Two habits of these formats are worth remembering when reading their
 output.  Timestamps are reported as ISO 8601 in UTC, but several
 formats store local time with no zone at all --- ISO 9660 and the
 ISO base media format both do --- so a value may be displaced by the
 writer's offset from UTC.  And a field being present does not make it
 true: every value here was written by whatever produced the file, an
 executable's build timestamp is trivially forged, and a disagreement
 between two fields is frequently more informative than either.
 
 @node Writing new Plugins
 @chapter Writing new Plugins
 
 Writing a new plugin for libextractor usually requires writing of or
 interfacing with an actual parser for a specific format.  How this is
 can be accomplished depends on the format and cannot be specified in
 general.  However, care should be taken for the code to be reentrant
 and highly fault-tolerant, especially with respect to malformed
 inputs.
 
 Plugins should start by verifying that the header of the data matches
 the specific format and immediately return if that is not the case.
 Even if the header matches the expected file format, plugins must not
 assume that the remainder of the file is well formed.
 
 The plugin library must be called libextractor_XXX.so, where XXX 
 denotes the file format of the plugin. The library must export a 
 method @verb{|libextractor_XXX_extract_method|}, with the following 
 signature:
 @verbatim
 void
 EXTRACTOR_XXX_extract_method (struct EXTRACTOR_ExtractContext *ec);
 @end verbatim
 
 @samp{ec} contains various information the plugin may need for its
 execution.  Most importantly, it contains functions for reading
 (``read'') and seeking (``seek'') the input data and for returning
 extracted data (``proc'').  The ``config'' member can contain
 additional configuration options.  ``proc'' should be called on
 each meta data item found.  If ``proc'' returns non-zero,
 processing should be aborted (if possible).
 
 In order to test new plugins, the @file{extract} command can be run
 with the options ``-ni'' and ``-l XXX'' .  This will run the plugin
 in-process (making it easier to debug) and without any of the other
 plugins.
 
 
 @section Example for a minimal extract method
 
 The following example shows how a plugin can return the mime type of
 a file.
 @example
 @verbatim
 void
 EXTRACTOR_mymime_extract (struct EXTRACTOR_ExtractContext *ec)
 {
   void *data;
   ssize_t data_size,
 
   if (-1 == (data_size = ec->read (ec->cls, &data, 4)))
     return; /* read error */
   if (data_size < 4)
     return; /* file too small */
   if (0 != memcmp (data, "\177ELF", 4))
     return; /* not ELF */
   if (0 != ec->proc (ec->cls, 
                      "mymime",
                      EXTRACTOR_METATYPE_MIMETYPE,
                      EXTRACTOR_METAFORMAT_UTF8,
                      "text/plain",
                      "application/x-executable",
                      1 + strlen("application/x-executable")))
     return;
   /* more calls to 'proc' here as needed */
 }
 @end verbatim
 @end example
 
 
 @node Internal utility functions
 @chapter Internal utility functions
 
 Some plugins link against the @code{libextractor_common} library which
 provides common abstractions needed by many plugins.  This section
 documents this internal API for plugin developers.  Note that the headers
 for this library are (intentionally) not installed: we do not consider
 this API stable and it should hence only be used by plugins that are 
 build and shipped with GNU libextractor.  Third-party plugins should
 not use it.
 
 @file{convert_numeric.h} defines various conversion functions for
 numbers (in particular, byte-order conversion for floating point
 numbers).  
 
 @file{unzip.h} defines an API for accessing compressed files.
 
 @file{pack.h} provides an interpreter for unpacking structs of integer
 numbers from streams and converting from big or little endian to host
 byte order at the same time.
 
 @file{convert.h} provides a function for character set conversion described
 below.
 
 @deftypefun {char *} EXTRACTOR_common_convert_to_utf8 (const char *input, size_t len, const char *charset)
 @cindex UTF-8
 @cindex character set
 @findex EXTRACTOR_common_convert_to_utf8
 Various GNU libextractor plugins make use of the internal
 @file{convert.h} header which defines a function
 
 @verb{|EXTRACTOR_common_convert_to_utf8|} which can be used to easily convert text from
 any character set to UTF-8.  This conversion is important since the
 linked list of keywords that is returned by GNU libextractor is
 expected to contain only UTF-8 strings.  Naturally, proper conversion
 may not always be possible since some file formats fail to specify the
 character set.  In that case, it is often better to not convert at
 all.
 
 The arguments to @verb{|EXTRACTOR_common_convert_to_utf8|} are the input string (which
 does @emph{not} have to be zero-terminated), the length of the input
 string, and the character set (which @emph{must} be zero-terminated).
 Which character sets are supported depends on the platform, a list can
 generally be obtained using the @command{iconv -l} command.  The
 return value from @verb{|EXTRACTOR_common_convert_to_utf8|} is a zero-terminated string
 in UTF-8 format.  The responsibility to free the string is with the
 caller, so storing the string in the keyword list is acceptable.
 @end deftypefun
 
 
 
 
 
 @node Reporting bugs
 @chapter Reporting bugs
 
 @cindex bug
 GNU libextractor uses the @url{https://gnunet.org/bugs/,Mantis bugtracking
 system}.  If possible, please report bugs there.  You can also e-mail
 the GNU libextractor mailinglist at @url{libextractor@@gnu.org}.
 
 
 
 @c **********************************************************
 @c *******************  Appendices  *************************
 @c **********************************************************
 
 @node GNU Free Documentation License
 @appendix GNU Free Documentation License
 
 @include fdl-1.3.texi
 
 
 @node Index
 @unnumbered Index
 
 @printindex cp
 
 @c @node Function and Data Index
 @c @unnumbered Function and Data Index
 @c @printindex fn
 
 @c @node Type Index
 @c @unnumbered Type Index
 @c @printindex tp
 
 @bye