path: root/doc/extractor.texi

\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 extractor.info
@include version.texi
@settitle The GNU libextractor Reference Manual
@c Unify some of the indices.
@c %**end of header
@copying
This manual is for GNU libextractor
(version @value{VERSION}, @value{UPDATED}).

GNU libextractor is a GNU package.


Copyright @copyright{} 2007, 2010 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 GNU Libraries
@direntry
* libextractor: (extractor).  Meta data extraction library.
@end direntry



@c
@c Titlepage
@c
@setchapternewpage odd
@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
@page


@macro gnu{}
@acronym{GNU}
@end macro

@macro gpl{}
@acronym{GPL}
@end macro

@macro api{}
@acronym{API}
@end macro

@macro cfunction{arg}
@code{\arg\()}
@end macro

@macro mynull{}
@code{NULL}
@end macro

@macro gnule{}
@acronym{GNU libextractor}
@end macro


@ifnottex
@node Top
@top The GNU libextractor Reference Manual
@insertcopying
@end ifnottex

@menu
* Introduction::                 What is @gnule{}.
* Preparation::                  What you should do before using the library.
* Generalities::                 General library functions and data types.
* Extracting meta data::         How to use @gnule{} to obtain meta data.
* Language bindings::            How to use @gnule{} from languages other than C.
* Utility functions::            Utility functions of @gnule{}.
* Existing Plugins::             What plugins are available.
* Writing new Plugins::          How to write new plugins for @gnule{}.
* Internal utility functions::   Utility functions of @gnule{} for writing plugins.
* Reporting bugs::               How to report bugs or request new features.

Appendices

* Copying::                     The GNU General Public License says how you
                                can copy and share some parts of @gnule{}.

Indices

* Concept Index::               Index of concepts and programs.
* Function and Data Index::     Index of functions, variables and data types.
* Type Index::                  Index of data types.

@end menu



@c **********************************************************
@c *******************  Introduction  ***********************
@c **********************************************************
@node Introduction
@chapter Introduction

@cindex error handling
@gnule{} 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, @gnule{} deliberately
avoids to ever report any errors.  Unexpected file contents simply
result in less or possibly no meta data being extracted.  

@cindex plugin
@gnule{} 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,
@gnule{} 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.

@gnule{} 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
@gnule{}.  

This manual focuses on providing documentation for writing software
with @gnule{}.  The only relevant parts for end-users
are the chapter on compiling and installing @gnule{}
(@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
@gnule{} is licensed under the GNU General Public License.  The
developers have frequently received requests to license GNU
libextractor under alternative terms.  However, @gnule{}
borrows plenty of GPL-licensed code from various other projects.
Hence we cannot change the license (even if we wanted to).@footnote{It
maybe possible to switch to GPLv3 in the future.  For this, an audit
of the license status of our dependencies would be required.  The new
code that was developed specifically for @gnule{} has
always been licensed under GPLv2 @emph{or any later version}.}

@node Preparation
@chapter Preparation

Compiling @gnule{} follows the standard GNU autotools
build process using @command{configure} and @command{make}.  For
details, read the @file{INSTALL} file and query 
@verb{|./configure --help|} for additional options.

@gnule{} has various dependencies, some 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 (unstable) packages that should be installed.

You absolutely need:

@itemize @bullet
@item
libtool
@item
gcc
@item
make
@item
g++ 
@item
libltdl7-dev
@item
zlib1g-dev
@item
libbz2-dev
@end itemize

Recommended dependencies are:
@itemize @bullet
@item
libgtk2.0-dev
@item
libvorbis-dev
@item
libflac-dev
@item
libgsf-1-dev
@item
libmpeg2-4-dev
@item
libqt4-dev
@item
librpm-dev
@item
libpoppler-dev
@item
libexiv2-dev
@end itemize

Optional dependencies (you would need to additionally specify 
the configure option @code{--enable-ffmpeg}) to make use of these
are:
@itemize @bullet
@item
libavformat-dev
@item
libswscale-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 @gnule{}, 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
@gnule{}.  The build process also installs the @file{extract} binary and
man pages for @file{extract} and @gnule{}.  The @file{extract} man page
documents the @file{extract} tool.  The @gnule{} man page gives a brief
summary of the C API for @gnule{}.

@cindex packageing
@cindex directory structure
@cindex plugin
@cindex environment variables
@vindex LIBEXTRACTOR_PREFIX
When you install @gnule{}, various plugins will be
installed in the @file{lib/libextractor/} directory.  The main library
will be installed as @file{lib/libextractor.so}.  Note that
@gnule{} 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
@gnule{} cannot locate a plugin, it will look in
@verb{|LIBEXTRACTOR_PREFIX/lib/libextractor/|}.

@section Note to package maintainers

The suggested way to package GNU libextractor is to split it into
roughly the following binary packages:@footnote{Debian policy
furthermore requires a @file{-dev} (meta) package that would depend on
all of the above 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)
@item
libextractor-dev (extractor.h header and man page)
@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)
@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, QT and X11) for systems that have limited
resources.


@node Generalities
@chapter Generalities

Each public symbol exported by @gnule{} has the prefix
@verb{|EXTRACTOR_|}.  All-caps names are used for constants.  For the
impatient, the minimal C code for using @gnule{} (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


@node Extracting meta data
@chapter Extracting meta data

In order to extract meta data with @gnule{} 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

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.  Creating multiple
plugin lists and using them concurrently is supported as long as
the @code{EXTRACTOR_OPTION_IN_PROCESS} option is not used. 

Generally, @gnule{} is fully thread-safe and mostly reentrant.
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.  Hence plugins are executed (by default) out of
process.  This also ensures that plugins that crash do not cause
the main application to fail as well.  

Plugins can be executed in-process by giving the option
@code{EXTRACTOR_OPTION_IN_PROCESS} when loading the plugin.  This
option is only recommended when debugging plugins and not for
production use.  Due to the use of shared-memory IPC the
out-of-process execution of plugins should not be a concern for
performance.


@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 NULL.  The second argument specifies the name of the plugin (i.e. ``ogg'').  The third argument can be 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 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 @gnule{} 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 @gnule{} 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 @gnule{} (@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 @gnule{} (@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 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 @gnule{}.  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 NULL, then the @samp{data} argument must point to the in-memory data to extract meta data from.  If @samp{filename} is non-NULL, @samp{data} can be 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, @gnule{} should not call @samp{proc} with any meta data. By design, @gnule{} should never crash or leak memory, even given corrupt files as input.  Note however, that running @gnule{} 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.  While @gnule{} 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.   @gnule{} 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 @gnule{} 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

@gnule{} works immediately with C and C++ code. Bindings for Java, Mono, Ruby, Perl, PHP and Python are available for download from the main @gnule{} 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

This 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 @gnule{} 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
APPLEFILE
@item
ASF
@item
DEB
@item
DVI
@item
ELF
@item
EXIV2
@item
FLAC
@item
FLV
@item
GIF
@item
HTML
@item
ID3 (v2.0, v2.3, v2.4)
@item
IT
@item
JPEG
@item
OLE2
@item
thumbnail (GTK, QT or FFMPEG-based)
@item
MAN
@item
MIME
@item
MP3 (ID3v1)
@item
MPEG
@item
NSF and NSFE
@item
ODF
@item
PNG
@item
PS (PostScript)
@item
QT (QuickTime)
@item
REAL
@item
RIFF
@item
RPM
@item 
S3M
@item
SID
@item
TAR
@item
TIFF
@item
WAV
@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).

@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|}, with the following 
signature:
@verbatim
int
EXTRACTOR_XXX_extract
   (const char *data,
    size_t data_size,
    EXTRACTOR_MetaDataProcessor proc,
    void *proc_cls,
    const char * options);
@end verbatim

@samp{data} is a pointer to the typically memory mapped contents of
the file.  Note that plugins cannot ignore the @verb{|const|}
annotation since the memory mapping may have been done read-only (and
thus writes to this page will result in an error).  The @samp{data_size}
argument specifies the size of the @samp{data} buffer in bytes.

@samp{proc} should be called on each meta data item found.  If @samp{proc} 
returns non-zero, processing should be aborted and the @code{extract}
function must return 1.  Otherwise @code{extract} should always return zero.


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

int
EXTRACTOR_mymime_extract
   (const char *data,
    size_t data_size,
    EXTRACTOR_MetaDataProcessor proc,
    void *proc_cls,
    const char * options)
{
  if (data_size < 4)
    return 0;
  if (0 != memcmp (data, "\177ELF", 4))
    return 0;
  if (0 != proc (proc_cls, 
                 "mymime",
                 EXTRACTOR_METATYPE_MIMETYPE,
                 EXTRACTOR_METAFORMAT_UTF8,
                 "text/plain",
                 "application/x-executable",
                 1 + strlen("application/x-executable")))
    return 1;
  /* more calls to 'proc' here as needed */
  return 0;
}

@end example

@section Plugin execution options

Plugins can request that their execution be done in a particular way.
For this, the plugin defines a function with the following signature:

@verbatim
const char *
EXTRACTOR_XXX_options (void);
@end verbatim

The function should return a string with the execution options.
Individual options in this string should be separated by semicolons.
Options that are included in the string but not known to the library
are ignored.  The following options are supported:

@itemize @bullet
@item
@code{oop-only} ensures that the plugin is only run out-of-process; if
this is not possible, the plugin will not be executed at all if this
option is set.

@item
@code{close-stderr} ensures that @code{stderr} is closed during the
execution of the plugin.  This is useful if the plugin uses libraries
that write (error) messages to @code{stderr} and where this behavior cannot be 
turned off.  This option only works if the plugin is executed out-of-process.

@item
@code{close-stdout} ensures that @code{stdout} is closed during the
execution of the plugin.  This is useful if the plugin uses libraries
that write messages to @code{stdout} and where this behavior cannot be 
turned off.  This option only works if the plugin is executed out-of-process.

@item
@code{force-kill} kills and restarts the plugin process for each
file that is being analyzed.  This is useful if the plugin uses
libraries that keep global state between runs that is problematic or
if the plugin uses libraries that are known to have serious resource
leaks (such as memory leaks).

@item
@code{want-tail} 
In order to limit memory consumption, limit the amount if reading from
disk and to keep the API simple, the @samp{data} argument passed to
the @code{EXTRACTOR_XXX_extract} method bounded (to 32 MB of normal
data; for compressed data, a limit of 16 MB is imposed).@footnote{If
@gnule{} was given a pointer to an existing, uncompressed block of
data in memory, no bound is imposed for plugins executing in-process;
for out-of-process plugins, a 32 MB limit is still imposed.}  Since
some file formats contain meta data at the end of the file, this option
provides a way for plugins to access not the first 16--32 MB of a file
but instead the last (roughly) 32 MB. 

Note that even for files larger than 32 MB, @samp{size} is not
guaranteed to be 32 MB since @samp{data} will be aligned to the page
size of the operating system.  However, the last byte of @samp{data}
is guaranteed to be the last byte of the file.  Furthermore, if the
file was large and compressed, unlike in the case of meta data
extraction from the header, the end of the file will not be
automatically decompressed by @gnule{}.  

@end itemize

Note that using options other than @code{want-tail} is pretty much
always a kludge and should thus be avoided.

@section Example for an options method

The following example shows how a plugin can set some of the options listed above:
@example
const char *
EXTRACTOR_id3_options ()
{
  return "close-stderr;want-tail";
}
@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 @gnule{} 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 @gnule{} 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
@gnule{} uses the @url{http://gnunet.org/bugs/,Mantis bugtracking
system}.  If possible, please report bugs there.  You can also e-mail
the @gnule{} mailinglist at @url{libextractor@@gnu.org}.



@c **********************************************************
@c *******************  Appendices  *************************
@c **********************************************************

@include gpl.texi

@node Concept Index
@unnumbered Concept Index

@printindex cp

@node Function and Data Index
@unnumbered Function and Data Index

@printindex fn

@node Type Index
@unnumbered Type Index

@printindex tp

@bye