//C-  -*- C++ -*-

//C- -------------------------------------------------------------------

//C- DjVuLibre-3.5

//C- Copyright (c) 2002  Leon Bottou and Yann Le Cun.

//C- Copyright (c) 2001  AT&T

//C-

//C- This software is subject to, and may be distributed under, the

//C- GNU General Public License, either Version 2 of the license,

//C- or (at your option) any later version. The license should have

//C- accompanied the software or you may obtain a copy of the license

//C- from the Free Software Foundation at http://www.fsf.org .

//C-

//C- This program is distributed in the hope that it will be useful,

//C- but WITHOUT ANY WARRANTY; without even the implied warranty of

//C- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

//C- GNU General Public License for more details.

//C- 

//C- DjVuLibre-3.5 is derived from the DjVu(r) Reference Library from

//C- Lizardtech Software.  Lizardtech Software has authorized us to

//C- replace the original DjVu(r) Reference Library notice by the following

//C- text (see doc/lizard2002.djvu and doc/lizardtech2007.djvu):

//C-

//C-  ------------------------------------------------------------------

//C- | DjVu (r) Reference Library (v. 3.5)

//C- | Copyright (c) 1999-2001 LizardTech, Inc. All Rights Reserved.

//C- | The DjVu Reference Library is protected by U.S. Pat. No.

//C- | 6,058,214 and patents pending.

//C- |

//C- | This software is subject to, and may be distributed under, the

//C- | GNU General Public License, either Version 2 of the license,

//C- | or (at your option) any later version. The license should have

//C- | accompanied the software or you may obtain a copy of the license

//C- | from the Free Software Foundation at http://www.fsf.org .

//C- |

//C- | The computer code originally released by LizardTech under this

//C- | license and unmodified by other parties is deemed "the LIZARDTECH

//C- | ORIGINAL CODE."  Subject to any third party intellectual property

//C- | claims, LizardTech grants recipient a worldwide, royalty-free, 

//C- | non-exclusive license to make, use, sell, or otherwise dispose of 

//C- | the LIZARDTECH ORIGINAL CODE or of programs derived from the 

//C- | LIZARDTECH ORIGINAL CODE in compliance with the terms of the GNU 

//C- | General Public License.   This grant only confers the right to 

//C- | infringe patent claims underlying the LIZARDTECH ORIGINAL CODE to 

//C- | the extent such infringement is reasonably necessary to enable 

//C- | recipient to make, have made, practice, sell, or otherwise dispose 

//C- | of the LIZARDTECH ORIGINAL CODE (or portions thereof) and not to 

//C- | any greater extent that may be necessary to utilize further 

//C- | modifications or combinations.

//C- |

//C- | The LIZARDTECH ORIGINAL CODE is provided "AS IS" WITHOUT WARRANTY

//C- | OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED

//C- | TO ANY WARRANTY OF NON-INFRINGEMENT, OR ANY IMPLIED WARRANTY OF

//C- | MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

//C- +------------------------------------------------------------------

#ifndef _DJVUDOCUMENT_H

#define _DJVUDOCUMENT_H

#ifdef HAVE_CONFIG_H

#include "config.h"

#endif

#if NEED_GNUG_PRAGMAS

# pragma interface

#endif

#include "DjVuPort.h"

#ifdef HAVE_NAMESPACES

namespace DJVU {

# ifdef NOT_DEFINED // Just to fool emacs c++ mode

}

#endif

#endif

class DjVmDoc;

class DjVmDir;

class DjVmDir0;

class DjVmNav;

class DjVuImage;

class DjVuFile;

class DjVuFileCache;

class DjVuNavDir;

class ByteStream;

/** @name DjVuDocument.h

    Files #"DjVuDocument.h"# and #"DjVuDocument.cpp"# contain implementation

    of the \Ref{DjVuDocument} class - the ideal tool for opening, decoding

    and saving DjVu single page and multi page documents.

    @memo DjVu document class.

    @author Andrei Erofeev <eaf@geocities.com>

*/

//@{

/** #DjVuDocument# provides convenient interface for opening, decoding

    and saving back DjVu documents in single page and multi page formats.

    {\bf Input formats}

    It can read multi page DjVu documents in either of the 4 formats: 2

    obsolete ({\em old bundled} and {\em old indexed}) and two new

    ({\em new bundled} and {\em new indirect}).

    {\bf Output formats}

    To encourage users to switch to the new formats, the #DjVuDocument# can

    save documents back only in the new formats: {\em bundled} and

    {\em indirect}.

    {\bf Conversion.} Since #DjVuDocument# can open DjVu documents in

    an obsolete format and save it in any of the two new formats

    ({\em new bundled} and {\em new indirect}), this class can be used for

    conversion from obsolete formats to the new ones. Although it can also

    do conversion between the new two formats, it's not the best way to

    do it. Please refer to \Ref{DjVmDoc} for details.

    {\bf Decoding.} #DjVuDocument# provides convenient interface for obtaining

    \Ref{DjVuImage} corresponding to any page of the document. It uses

    \Ref{DjVuFileCache} to do caching thus avoiding unnecessary multiple decoding of

    the same page. The real decoding though is accomplished by \Ref{DjVuFile}.

    {\bf Messenging.} Being derived from \Ref{DjVuPort}, #DjVuDocument#

    takes an active part in exchanging messages (requests and notifications)

    between different parties involved in decoding. It reports (relays)

    errors, progress information and even handles some requests for data (when

    these requests deal with local files).

    Typical usage of #DjVuDocument# class in a threadless command line

    program would be the following:

    \begin{verbatim}

    static const char file_name[]="/tmp/document.djvu";

    GP<DjVuDocument> doc=DjVuDocument::create_wait(file_name);

    const int pages=doc->get_pages_num();

    for(int page=0;page

    {

       GP<DjVuImage> dimg=doc->get_page(page);

       // Do something

    };

    \end{verbatim}

    {\bf Comments for the code above}

    \begin{enumerate}

       \item Since the document is assumed to be stored on the hard drive,

             we don't have to cope with \Ref{DjVuPort}s and can pass

	     #ZERO# pointer to the \Ref{init}() function. #DjVuDocument#

	     can access local data itself. In the case of a plugin though,

	     one would have to implement his own \Ref{DjVuPort}, which

	     would handle requests for data arising when the document

	     is being decoded.

       \item In a threaded program instead of calling the \Ref{init}()

             function one can call \Ref{start_init}() and \Ref{stop_init}()

	     to initiate and interrupt initialization carried out in

	     another thread. This possibility of initializing the document

	     in another thread has been added specially for the plugin

	     because the initialization itself requires data, which is

	     not immediately available in the plugin. Thus, to prevent the

	     main thread from blocking, we perform initialization in a

	     separate thread. To check if the class is completely and

	     successfully initialized, use \Ref{is_init_ok}(). To see if

	     there was an error, use \Ref{is_init_failed}(). To

	     know when initialization is over (whether successfully or not),

	     use \Ref{is_init_complete}(). To wait for this to happen use

	     \Ref{wait_for_complete_init}(). Once again, all these things are

	     not required for single-threaded program.

	     Another difference between single-threaded and multi-threaded

	     environments is that in a single-threaded program, the image is

	     fully decoded before it's returned. In a multi-threaded

	     application decoding starts in a separate thread, and the pointer

	     to the \Ref{DjVuImage} being decoded is returned immediately.

	     This has been done to enable progressive redisplay

	     in the DjVu plugin. Use communication mechanism provided by

	     \Ref{DjVuPort} and \Ref{DjVuPortcaster} to learn about progress

	     of decoding.  Or try #dimg->wait_for_complete_decode()# to wait

	     until the decoding ends.

       \item See Also: \Ref{DjVuFile}, \Ref{DjVuImage}, \Ref{GOS}.

    \end{enumerate}

    {\bf Initialization}

    As mentioned above, the #DjVuDocument# can go through several stages

    of initialization. The functionality is gradually added while it passes

    one stage after another:

    \begin{enumerate}

       \item First of all, immediately after the object is created \Ref{init}()

             or \Ref{start_init}() functions must be called. {\bf Nothing}

	     will work until this is done. \Ref{init}() function will not

	     return until the initialization is complete. You need to make

	     sure, that enough data is available. {\bf Do not call \Ref{init}()

	     in the plugin}. \Ref{start_init}() will start initialization

	     in another thread. Use \Ref{stop_init}() to interrupt it.

	     Use \Ref{is_init_complete}() to check the initialization progress.

	     Use \Ref{wait_for_complete_init}() to wait for init to finish.

       \item The first thing the initializing code learns about the document

	     is its type (#BUNDLED#, #INDIRECT#, #OLD_BUNDLED# or #OLD_INDEXED#).

	     As soon as it happens, document flags are changed and

	     #notify_doc_flags_changed()# request is sent through the

	     communication mechanism provided by \Ref{DjVuPortcaster}.

       \item After the document type becomes known, the initializing code

             proceeds with learning the document structure. Gradually the

	     flags are updated with values:

	     \begin{itemize}

	        \item #DOC_DIR_KNOWN#: Contents of the document became known.

		      This is meaningful for #BUNDLED#, #OLD_BUNDLED# and

		      #INDIRECT# documents only.

		\item #DOC_NDIR_KNOWN#: Contents of the document navigation

		      directory became known. This is meaningful for old-style

		      documents (#OLD_BUNDLED# and #OLD_INDEXED#) only

		\item #DOC_INIT_OK# or #DOC_INIT_FAILED#:

		      The initializating code finished.

	     \end{itemize}

    \end{enumerate} */

class DJVUAPI DjVuDocument : public DjVuPort

{

public:

      /** Flags describing the document initialization state.

	  \begin{itemize}

	     \item #DOC_TYPE_KNOWN#: The type of the document has been learnt.

	     \item #DOC_DIR_KNOWN#: Contents of the document became known.

		   This is meaningful for #BUNDLED#, #OLD_BUNDLED# and

		   #INDIRECT# documents only.

	     \item #DOC_NDIR_KNOWN#: Contents of the document navigation

		   directory became known. This is meaningful for old-style

		   documents (#OLD_BUNDLED# and #OLD_INDEXED#) only

	     \item #DOC_INIT_OK#: The initialization has completed successfully.

	     \item #DOC_INIT_FAILED#: The initialization failed.

	  \end{itemize} */

   enum DOC_FLAGS { DOC_TYPE_KNOWN=1, DOC_DIR_KNOWN=2,

		    DOC_NDIR_KNOWN=4, DOC_INIT_OK=8,

		    DOC_INIT_FAILED=16 };

      /** Specifies the format of #DjVuDocument#. There are currently 4 DjVu

	  multipage formats recognized by the library. Two of them are obsolete

	  and should not be used.

	  \begin{enumerate}

	     \item #OLD_BUNDLED# - Obsolete bundled format

	     \item #OLD_INDEXED# - Obsolete multipage format where every page

	           is stored in a separate file and "includes" (by means

		   of an #INCL# chunk) the file with the document directory.

	     \item #SINGLE_PAGE# - Single page document. Basically a file

	     	   with either #FORM:DJVU# or #FORM:IW44# and no multipage

		   information. For example, #OLD_INDEXED# documents with

		   document directory do not qualify even if they contain only

		   one page.

	     \item #BUNDLED# - Currently supported bundled format

	     \item #INDIRECT# - Currently supported "expanded" format, where

	           every page and component is stored in a separate file. There

		   is also a {\em top-level} file with the document directory.

          \end{enumerate} */

   enum DOC_TYPE { OLD_BUNDLED=1, OLD_INDEXED, BUNDLED, INDIRECT,

		   SINGLE_PAGE, UNKNOWN_TYPE };

   enum THREAD_FLAGS { STARTED=1, FINISHED=2 };

protected:

      /** Default creator. Please call functions \Ref{init}() or

	  \Ref{start_init}() before you start working with the #DjVuDocument#.

        */

   DjVuDocument(void);

public:

     /// Virtual Destructor

   virtual ~DjVuDocument(void);

      /** Initializes the #DjVuDocument# object using an existing document.

          This function should be called once after creating the object.

	  The #url# should point to the real data, and the creator of the

	  document should be ready to return this data to the document

	  if it's not stored locally (in which case #DjVuDocument# can

	  access it itself).

	  {\bf Initializing thread}

	  In a single-threaded application, the #start_init()# function performs

	  the complete initialization of the #DjVuDocument# before it returns.

	  In a multi-threaded application, though, it initializes some internal

	  variables, requests data for the document and starts a new

	  {\em initializing} thread, which is responsible for determining the

	  document type and structure and completing the initialization

	  process. This additional complication is justified in the case of

	  the DjVu plugin because performing initialization requires data and

	  in the plugin the data can be supplied by the main thread only.

	  Thus, if the initialization was completed by the main thread, the

	  plugin would run out of data and block.

	  {\bf Stages of initialization}

	  Immediately after the #start_init()# function terminates, the

	  #DjVuDocument# object is ready for use. Its functionality will

	  not be complete (until the initializing thread finishes), but

	  the object is still very useful. Such functions as \Ref{get_page}()

	  or \Ref{get_djvu_file}() or \Ref{id_to_url}() may be called

	  before the initializing thread completes. This allows the DjVu

	  plugin start decoding as soon as possible without waiting for

	  all data to arrive.

	  To query the current stage of initialization you can use

	  \Ref{get_doc_flags}() function or listen to the

	  #notify_doc_flags_changed()# notifications distributed with the help

	  of \Ref{DjVuPortcaster}. To wait for the initialization to

	  complete use \Ref{wait_for_complete_init}(). To stop initialization

	  call \Ref{stop_init}().

	  {\bf Querying data}

	  The query for data is done using the communication mechanism

	  provided by \Ref{DjVuPort} and \Ref{DjVuPortcaster}. If #port#

	  is not #ZERO#, then the request for data will be forwarded to it.

	  If it {\bf is} #ZERO# then #DjVuDocument# will create an internal

	  instance of \Ref{DjVuSimplePort} and will use it to access local

	  files and report errors to #stderr#. In short, if the document

	  file is stored on the local hard disk, and you're OK about reporting

	  errors to #stderr#, you may pass #ZERO# pointer to \Ref{DjVuPort}

	  as #DjVuDocument# can take care of this situation by itself.

	  {\bf The URL}

	  Depending on the document type the #url# should point to:

	  \begin{itemize}

	     \item {\bf Old bundled} and {\bf New bundled} formats: to the

	           document itself.

	     \item {\bf Old indexed} format: to any page of the document.

	     \item {\bf New indirect} format: to the top-level file of the

	           document. If (like in the {\em old indexed} format) you

		   point the #url# to a page, the page {\em will} be decoded,

		   but it will {\em not} be recognized to be part of the

		   document.

	  \end{itemize}

	  @param url The URL pointing to the document. If the document is

	         in a {\em bundled} format then the URL should point to it.

		 If the document is in the {\em old indexed} format then

		 URL may point to any page of this document. For {\em new

		 indirect} format the URL should point to the top-level

		 file of the document.

	  @param port If not #ZERO#, all requests and notifications will

	         be sent to it. Otherwise #DjVuDocument# will create an internal

		 instance of \Ref{DjVuSimplePort} for these purposes.

		 It's OK to make it #ZERO# if you're writing a command line

		 tool, which should work with files on the hard disk only

		 because #DjVuDocument# can access such files itself.

	  @param cache It's used to cache decoded \Ref{DjVuFile}s and

	         is actually useful in the plugin only.  */

   void		start_init(const GURL & url, GP<DjVuPort> port=0,

			   DjVuFileCache * cache=0);

   /** This creates a DjVuDocument without initializing it. */

   static GP<DjVuDocument> create_noinit(void) {return new DjVuDocument;}

   /** Create a version of DjVuDocument which has finished initializing. */

   static GP<DjVuDocument> create_wait(

     const GURL &url, GP<DjVuPort> xport=0, DjVuFileCache * const xcache=0);

   /** Create a version of DjVuDocument which has begun initializing. */

   static GP<DjVuDocument> create(

     const GURL &url, GP<DjVuPort> xport=0, DjVuFileCache * const xcache=0);

   /** Create a version of DjVuDocument which has begun initializing. */

   static GP<DjVuDocument> create(

     GP<DataPool> pool, GP<DjVuPort> xport=0, DjVuFileCache * const xcache=0);

   /** Create a version of DjVuDocument which has begun initializing. */

   static GP<DjVuDocument> create(

     const GP<ByteStream> &bs, GP<DjVuPort> xport=0,

     DjVuFileCache * const xcache=0);

      /** Call this function when you don't need the #DjVuDocument# any more.

	  In a multi-threaded environment it will stop initialization

	  thread, if it is currently running. {\bf You will not be able

	  to start the initialization again. Thus, after calling this

          function the document should not be used any more}. */

   void		stop_init(void);

      /** Initializes the document.

	  Contrary to \Ref{start_init}(), which just starts the initialization

	  thread in a multi-threaded environment, this function does not

	  return until the initialization completes (either successfully or

	  not). Basically, it calls \Ref{start_init}() and then

	  \Ref{wait_for_complete_init}().

	  */

   void		init(const GURL & url, GP<DjVuPort> port=0,

		     DjVuFileCache * cache=0);

      /** Returns #TRUE# if the initialization thread finished (does not

	  matter successfully or not). As soon as it happens, the document

	  becomes completely initialized and its every function should work

	  properly. Please refer to the description of \Ref{init}() function

	  and of the #DjVuDocument# class to learn about the initializing

	  stages.

	  To wait for the initialization to complete use

	  \Ref{wait_for_complete_init}() function.

	  To query the initialization stage use \Ref{get_flags}() function.

	  To learn whether initialization was successful or not,

	  use \Ref{is_init_ok}() and \Ref{is_init_failed}().

	  {\bf Note:} In a single threaded application the initialization

	  completes before the \Ref{init}() function returns. */

   bool		is_init_complete(void) const;

      /** Returns #TRUE# is the initialization thread finished successfully.

	  See \Ref{is_init_complete}() and \Ref{wait_for_complete_init}()

	  for more details. */

   bool		is_init_ok(void) const;

      /** Forces compression with the next save_as function. */

   void		set_needs_compression(void);

      /** Returns #TRUE# if there are uncompressed pages in this document. */

   bool		needs_compression(void) const;

      /** Returns #TRUE# if this file must be renamed before saving. */

   bool		needs_rename(void) const;

      /** Returns #TRUE# if this file must be renamed before saving. */

   bool		can_compress(void) const;

      /** Returns #TRUE# is the initialization thread failed.

	  See \Ref{is_init_complete}() and \Ref{wait_for_complete_init}()

	  for more details. */

   bool		is_init_failed(void) const;

      /** If the document has already learnt its type, the function will

	  returns it: #DjVuDocument::OLD_BUNDLED# or

	  #DjVuDocument::OLD_INDEXED# or #DjVuDocument::SINGLE_PAGE# or

	  #DjVuDocument:BUNDLED# or #DjVuDocument::INDIRECT#. The first

	  two formats are obsolete. Otherwise (if the type is unknown yet),

	  #UNKNOWN_TYPE# will be returned.

	  {\bf Note:} To check the stage of the document initialization

	  use \Ref{get_flags}() or \Ref{is_init_complete}() functions. To

	  wait for the initialization to complete use \Ref{wait_for_complete_init}().

	  For single threaded applications the initialization completes

	  before the \Ref{init}() function returns. */

   int		get_doc_type(void) const;

      /** Returns the document flags. The flags describe the degree in which

	  the #DjVuDocument# object is initialized. Every time the flags

	  are changed, a #notify_doc_flags_changed()# notification is

	  distributed using the \Ref{DjVuPortcaster} communication

	  mechanism.

	  {\bf Note:} To wait for the initialization to complete use

	  \Ref{wait_for_complete_init}(). For single threaded applications

	  the initialization completes before the \Ref{init}() function

	  returns. */

   long		get_doc_flags(void) const;

      /** Returns #TRUE# if the document is in bundled format (either in

	  #DjVuDocument::OLD_BUNDLED# or #DjVuDocument::BUNDLED# formats). */

   bool		is_bundled(void) const;

      /// Returns the URL passed to the \Ref{init}() function

   GURL		get_init_url(void) const;

      /// Returns a listing of id's used by this document.

   GList<GUTF8String> get_id_list(void);

      /// Fill the id's into a GMap.

   void map_ids( GMap<GUTF8String,void *> &map);

      /** Returns data corresponding to the URL passed to the \Ref{init}()

	  function.

	  {\bf Note:} The pointer returned is guaranteed to be non-#ZERO#

	  only after the #DjVuDocument# learns its type (passes through

	  the first stage of initialization process). Please refer to

	  \Ref{init}() for details. */

   GP<DataPool>	get_init_data_pool(void) const;

      /** @name Accessing pages */

      //@{

      /** Returns the number of pages in the document. If there is still

	  insufficient information about the document structure (initialization

	  has not finished yet), #1# will be returned. Please refer to

          \Ref{init}() for details. */

   int		get_pages_num(void) const;

      /** Translates the page number to the full URL of the page. This URL

	  is "artificial" for the {\em bundled} formats and is obtained

	  by appending the page name to the document's URL honoring possible

	  #;# and #?# in it. Negative page number has a special meaning for

	  #OLD_INDEXED# documents: it points to the URL, which the

	  #DjVuDocument# has been initialized with. For other formats this

	  is the same as page #0#.

	  The function tries it best to map the page number to the URL.

	  Although, if the document structure has not been fully discovered

	  yet, an empty URL will be returned. Use \Ref{wait_for_complete_init}()

	  to wait until the document initialization completes. Refer to

	  \Ref{init}() for details.

	  Depending on the document format, the function assumes, that there

	  is enough information to complete the request when:

	  \begin{itemize}

	     \item #OLD_INDEXED#: If #page_num<0#, #DOC_TYPE_KNOWN# flag must

		   be set. Otherwise #DOC_NDIR_KNOWN# must be set.

	     \item #OLD_BUNDLED#: If #page_num=0#, #DOC_DIR_KNOWN# flag must

		   be set. Otherwise #DOC_NDIR_KNOWN# flag must be set.

	     \item #INDIRECT# and #BUNDLED#: #DOC_DIR_KNOWN# flag must be set.

	  \end{itemize} */

   GURL		page_to_url(int page_num) const;

   /// Tranlate the page number to id...

   GUTF8String page_to_id(int page_num) const

   { return url_to_id(page_to_url(page_num)); }

      /** Translates the page URL back to page number. Returns #-1# if the

	  page is not in the document or the document's structure

          has not been learnt yet.

	  Depending on the document format, the function starts working

	  properly as soon as:

	  \begin{itemize}

	     \item #OLD_INDEXED# and #OLD_BUNDLED# and #SINGLE_PAGE#:

	           #DOC_NDIR_KNOWN# is set

	     \item #INDIRECT# and #BUNDLED#: #DOC_DIR_KNOWN# is set.

	  \end{itemize} */

   int		url_to_page(const GURL & url) const;

   /// Map the specified url to it's id.

   GUTF8String  url_to_id(const GURL &url) const

   { return url.fname(); }

      /** Translates the textual ID to the complete URL if possible.

	  Depending on the document format the translation is done in the

	  following way:

	  \begin{itemize}

	     \item For #BUNDLED# and #INDIRECT# documents the function

		   scans the \Ref{DjVmDir} (the document directory) and

		   matches the ID against:

		   \begin{enumerate}

		      \item File ID from the \Ref{DjVmDir}

		      \item File name from the \Ref{DjVmDir}

		      \item File title from the \Ref{DjVmDir}

		   \end{enumerate}

		   Then for #BUNDLED# document the URL is obtained by

		   appending the #name# of the found file to the document's

		   URL.

		   For #INDIRECT# documents the URL is obtained by

		   appending the #name# of the found file to the URL of

		   the directory containing the document.

	     \item For #OLD_BUNDLED# documents the function compares the ID

		   with internal name of every file inside the bundle and

		   composes an artificial URL by appending the file name to

		   the document's URL.

	     \item For #OLD_INDEXED# or #SINGLE_PAGE# documents the function

		   composes the URL by appending the ID to the URL of the

		   directory containing the document.

	  \end{itemize}

	  If information obtained by the initialization thread is not

	  sufficient yet, the #id_to_url()# may return an empty URL.

	  Depending on the document type, the information is sufficient when

	  \begin{itemize}

	     \item #BUNDLED# and #INDIRECT#: #DOC_DIR_KNOWN# flag is set.

	     \item #OLD_BUNDLED# and #OLD_INDEXED# and #SINGLE_PAGE#:

	           #DOC_TYPE_KNOWN# flag is set.

	  \end{itemize} */

   GURL		id_to_url(const GUTF8String &id) const;

   /// Find out which page this id is...

   int		id_to_page(const GUTF8String &id) const

   {  return url_to_page(id_to_url(id)); }

      /** Returns \Ref{GP} pointer to \Ref{DjVuImage} corresponding to page

          #page_num#. If caching is enabled, and there is a {\em fully decoded}

	  \Ref{DjVuFile} in the cache, the image will be reused and will

	  be returned fully decoded. Otherwise, if multi-threaded behavior

	  is allowed, and #sync# is set to #FALSE#, the decoding will be

	  started in a separate thread, which enables to do progressive

	  redisplay. Thus, in this case the image returned may be partially

	  decoded.

	  Negative #page_num# has a special meaning for the {\em old indexed}

	  multipage documents: the #DjVuDocument# will start decoding of the

	  URL with which it has been initialized. For other formats page

	  #-1# is the same as page #0#.

	  #DjVuDocument# can also connect the created page to the specified

	  #port# {\em before starting decoding}. This option will allow

	  the future owner of \Ref{DjVuImage} to receive all messages and

	  requests generated during its decoding.

	  If this function is called before the document's structure becomes

	  known (the initialization process completes), the \Ref{DjVuFile},

	  which the returned image will be attached to, will be assigned a

	  temporary artificial URL, which will be corrected as soon as enough

	  information becomes available. The trick prevents the main thread

	  from blocking and in some cases helps to start decoding earlier.

	  The URL is corrected and decoding will start as soon as

	  #DjVuDocument# passes some given stages of initialization and

	  \Ref{page_to_url}(), \Ref{id_to_url}() functions start working

	  properly. Please look through their description for details.

	  {\bf Note:} To wait for the initialization to complete use

	  \Ref{wait_for_complete_init}(). For single threaded applications

	  the initialization completes before the \Ref{init}() function

	  returns.

	  @param page_num Number of the page to be decoded

	  @param sync When set to #TRUE# the function will not return

	  	      until the page is completely decoded. Otherwise,

		      in a multi-threaded program, this function will

		      start decoding in a new thread and will return

		      a partially decoded image. Refer to

		      \Ref{DjVuImage::wait_for_complete_decode}() and

		      \Ref{DjVuFile::is_decode_ok}().

	  @param port A pointer to \Ref{DjVuPort}, that the created image

	  	      will be connected to. */

   GP<DjVuImage> get_page(int page_num, bool sync=true, DjVuPort * port=0) const;

   GP<DjVuImage> get_page(int page_num, bool sync=true, DjVuPort * port=0)

   { return const_cast<const DjVuDocument *>(this)->get_page(page_num,sync,port); }

      /** Returns \Ref{GP} pointer to \Ref{DjVuImage} corresponding to the

	  specified ID. This function behaves exactly as the #get_page()#

	  function above. The only thing worth mentioning here is how the #ID#

	  parameter is treated.

	  First of all the function checks, if the ID contains a number.

	  If so, it just calls the #get_page()# function above. If ID is

	  #ZERO# or just empty, page number #-1# is assumed. Otherwise

	  the ID is translated to the URL using \Ref{id_to_url}(). */

   GP<DjVuImage> get_page(const GUTF8String &id, bool sync=true, DjVuPort * port=0);

      /** Returns \Ref{DjVuFile} corresponding to the specified page.

	  Normally it translates the page number to the URL using

	  \Ref{page_to_url}() and then creates \Ref{DjVuFile} initializing

	  it with data from the URL.

	  The behavior becomes different, though in the case when the

	  document structure is unknown at the moment this function is called.

	  In this situations it invents a temporary URL, creates a

	  \Ref{DjVuFile}, initializes it with this URL and returns

	  immediately. The caller may start decoding the file right away

	  (if necessary). The decoding will block but will automatically

	  continue as soon as enough information is collected about the

	  document. This trick should be quite transparent to the user and

	  helps to prevent the main thread from blocking. The decoding will

	  unblock and this function will stop using this "trick" as soon

	  as #DjVuDocument# passes some given stages of initialization and

	  \Ref{page_to_url}(), \Ref{id_to_url}() functions start working

	  properly.

	  If #dont_create# is #FALSE# the function will return the file

	  only if it already exists.

	  {\bf Note:} To wait for the initialization to complete use

	  \Ref{wait_for_complete_init}(). For single threaded applications

	  the initialization completes before the \Ref{init}() function

	  returns. */

   GP<DjVuFile>	get_djvu_file(int page_num, bool dont_create=false) const;

   GP<DjVuFile> get_djvu_file(int page_num, bool dont_create=false)

   { return const_cast<const DjVuDocument *>(this)->get_djvu_file(page_num,dont_create); }

      /** Returns \Ref{DjVuFile} corresponding to the specified ID.

          This function behaves exactly as the #get_djvu_file()# function

	  above. The only thing worth mentioning here is how the #ID#

	  parameter is treated.

          First off, \Ref{id_to_url}() is called.  If not successfull,

	  the function checks, if the ID contains a number.

	  If so, it just calls the #get_djvu_file()# function above. If ID is

	  #ZERO# or just empty, page number #-1# is assumed.

	  If #dont_create# is #FALSE# the function will return the file

	  only if it already exists. */

   GP<DjVuFile>	get_djvu_file(const GUTF8String &id, bool dont_create=false);

   GP<DjVuFile>	get_djvu_file(const GURL &url, bool dont_create=false);

      /** Returns a \Ref{DataPool} containing one chunk #TH44# with

	  the encoded thumbnail for the specified page. The function

	  first looks for thumbnails enclosed into the document and if

	  it fails to find one, it decodes the required page and creates

	  the thumbnail on the fly (unless #dont_decode# is true).

	  {\bf Note:} It may happen that the returned \Ref{DataPool} will

	  not contain all the data you need. In this case you will need

	  to install a trigger into the \Ref{DataPool} to learn when the

	  data actually arrives. */

   virtual GP<DataPool> get_thumbnail(int page_num, bool dont_decode);

      /* Will return gamma correction, which was used when creating

	 thumbnail images. If you need other gamma correction, you will

	 need to correct the thumbnails again. */

   float	get_thumbnails_gamma(void) const;

      //@}

      /** Waits until the document initialization process finishes.

	  It can finish either successfully or not. Use \Ref{is_init_ok}()

	  and \Ref{is_init_failed}() to learn the result code.

	  As described in \Ref{start_init}(), for multi-threaded applications the

	  initialization is carried out in parallel with the main thread.

	  This function blocks the calling thread until the initializing

	  thread reads enough data, receives information about the document

	  format and exits.  This function returns #true# if the

	  initialization is successful. You can use \Ref{get_flags}() or

	  \Ref{is_init_complete}() to check more precisely the degree of

	  initialization. Use \Ref{stop_init}() to interrupt initialization. */

   bool    	   wait_for_complete_init(void);

          /** Wait until we known the number of pages and return. */

   int wait_get_pages_num(void) const;

      /// Returns cache being used.

   DjVuFileCache * get_cache(void) const;

      /** @name Saving document to disk */

      //@{

      /** Returns pointer to the \Ref{DjVmDoc} class, which can save the

	  document contents on the hard disk in one of the two new formats:

	  {\em bundled} and {\em indirect}. You may also want to look

	  at \Ref{write}() and \Ref{expand}() if you are interested in

	  how to save the document.

	  {\bf Plugin Warning}. This function will read contents of the whole

	  document. Thus, if you call it from the main thread (the thread,

	  which transfers data from Netscape), the plugin will block. */

   GP<DjVmDoc>		get_djvm_doc(void);

      /** Saves the document in the {\em new bundled} format. All the data

	  is "bundled" into one file and this file is written into the

	  passed stream.

	  If #force_djvm# is #TRUE# then even one page documents will be

	  saved in the #DJVM BUNDLED# format (inside a #FORM:DJVM#);

	  {\bf Plugin Warning}. This function will read contents of the whole

	  document. Thus, if you call it from the main thread (the thread,

	  which transfers data from Netscape), the plugin will block. */

   virtual void write(const GP<ByteStream> &str, bool force_djvm=false);

     /** Always save as bundled, renaming any files conflicting with the

         the names in the supplied GMap. */

   virtual void write(const GP<ByteStream> &str,

     const GMap<GUTF8String,void *> &reserved);

      /** Saves the document in the {\em new indirect} format when every

	  page and component are stored in separate files. This format

	  is ideal for web publishing because it allows direct access to

	  any page and component. In addition to it, a top-level file

	  containing the list of all components will be created. To view

	  the document later in the plugin or in the viewer one should

	  load the top-level file.

	  {\bf Plugin Warning}. This function will read contents of the whole

	  document. Thus, if you call it from the main thread (the thread,

	  which transfers data from Netscape), the plugin will block.

	  @param codebase - Name of the directory which the document should

	         be expanded into.

	  @param idx_name - Name of the top-level file containing the document

	         directory (basically, list of all files composing the document).

      */

   void			expand(const GURL &codebase, const GUTF8String &idx_name);

      /** This function can be used instead of \Ref{write}() and \Ref{expand}().

	  It allows to save the document either in the new #BUNDLED# format

	  or in the new #INDIRECT# format depending on the value of parameter

	  #bundled#.

	  Depending on the document's type, the meaning of #where# is:

	  \begin{itemize}

	     \item For #BUNDLED# documents this is the name of the file

	     \item For #INDIRECT# documents this is the name of top-level

	           index file. All document files will be saved into the

		   save directory where the index file will resize. */

   virtual void		save_as(const GURL &where, bool bundled=0);

      //@}

      /** Returns pointer to the internal directory of the document, if it

	  is in one of the new formats: #BUNDLED# or #INDIRECT#.

	  Otherwise (if the format of the input document is obsolete),

	  #ZERO# is returned.

	  #ZERO# will also be returned if the initializing thread has not

	  learnt enough information about the document (#DOC_DIR_KNOWN# has

	  not been set yet). Check \Ref{is_init_complete}() and \Ref{init}()

          for details. */

   GP<DjVmDir>		get_djvm_dir(void) const;

      /** Returns pointer to the document bookmarks.

          This applies to #BUNDLED# and #INDIRECT# documents.

	  #ZERO# will also be returned if the initializing thread has not

	  learnt enough information about the document (#DOC_DIR_KNOWN# has

	  not been set yet). Check \Ref{is_init_complete}() and \Ref{init}()

          for details. */

   GP<DjVmNav>		get_djvm_nav(void) const;

      /** Returns pointer to the internal directory of the document, if it

	  is in obsolete #OLD_BUNDLED# format.

	  #ZERO# will also be returned if the initializing thread has not

	  learnt enough information about the document (#DOC_DIR_KNOWN# has

	  not been set yet). Check \Ref{is_init_complete}() and \Ref{init}()

          for details. */

   GP<DjVmDir0>		get_djvm_dir0(void) const;

      /** Returns pointer to {\em navigation directory} of the document.

	  The navigation directory is a DjVu file containing only one

	  chunk #NDIR# inside a #FORM:DJVI# with the list of all

	  document pages. */

   GP<DjVuNavDir>	get_nav_dir(void) const;

   /// Create a complete DjVuXML file.

   void writeDjVuXML(const GP<ByteStream> &gstr_out,

                     int flags, int page = -1) const;

      /// Returns TRUE if #class_name# is #"DjVuDocument"# or #"DjVuPort"#

   virtual bool		inherits(const GUTF8String &class_name) const;

      /// Converts the specified id to a URL.

   virtual GURL		id_to_url(const DjVuPort * source, const GUTF8String &id;;

   virtual GP<DjVuFile>	id_to_file(const DjVuPort * source, const GUTF8String &id;;

   virtual GP<DataPool>	request_data(const DjVuPort * source, const GURL & url);

   virtual void		notify_file_flags_changed(const DjVuFile * source,

 			long set_mask, long clr_mask);

   virtual GList<GURL>	get_url_names(void);

   virtual void 	set_recover_errors(ErrorRecoveryAction=ABORT);

   virtual void 	set_verbose_eof(bool=true);

   static void set_compress_codec(

     void (*codec)(GP<ByteStream> &, const GURL &where, bool bundled));

   static void set_import_codec(

     void (*codec)(GP<DataPool> &,const GURL &url,bool &, bool &);;

protected:

   static void (*djvu_import_codec) (

     GP<DataPool> &pool, const GURL &url,bool &needs_compression, bool &needs_rename );

   static void (*djvu_compress_codec) (

     GP<ByteStream> &bs, const GURL &where, bool bundled);

   virtual GP<DjVuFile>	url_to_file(const GURL & url, bool dont_create=false) const;

   GURL			init_url;

   GP<DataPool>		init_data_pool;

   GP<DjVmDir>		djvm_dir;	// New-style DjVm directory

   GP<DjVmNav>          djvm_nav;

   int	doc_type;

   bool needs_compression_flag;

   bool can_compress_flag;

   bool needs_rename_flag;

   bool			has_url_names;

   GCriticalSection	url_names_lock;

   GList<GURL>	url_names;

   ErrorRecoveryAction	recover_errors;

   bool			verbose_eof;

public:

   class UnnamedFile; // This really should be protected ...

   class ThumbReq; // This really should be protected ...

protected:

   bool                 init_started;

   GSafeFlags		flags;

   GSafeFlags		init_thread_flags;

   DjVuFileCache	* cache;

   GP<DjVuSimplePort>	simple_port;

   GP<DjVmDir0>		djvm_dir0;	// Old-style DjVm directory

   GP<DjVuNavDir>	ndir;		// Old-style navigation directory

   GUTF8String		first_page_name;// For OLD_BUNDLED docs only

      // The following is used in init() and destructor to query NDIR

      // DO NOT USE FOR ANYTHING ELSE. THE FILE IS ZEROED IMMEDIATELY

      // AFTER IT'S NO LONGER NEEDED. If you don't zero it, ~DjVuDocument()

      // will kill it, which is a BAD thing if the file's already in cache.

   GP<DjVuFile>		ndir_file;

   GPList<UnnamedFile>	ufiles_list;

   GCriticalSection	ufiles_lock;

   GPList<ThumbReq>	threqs_list;

   GCriticalSection	threqs_lock;

   GP<DjVuDocument>	init_life_saver;

   static const float	thumb_gamma;

      // Reads document contents in another thread trying to determine

      // its type and structure

   GThread		init_thr;

   static void		static_init_thread(void *);

   void			init_thread(void);

   void                 check() const;

   void			process_threqs(void);

   GP<ThumbReq>		add_thumb_req(const GP<ThumbReq> & thumb_req);

   void			add_to_cache(const GP<DjVuFile> & f);

   void			check_unnamed_files(void);

   GUTF8String		get_int_prefix(void) const;

   void			set_file_aliases(const DjVuFile * file);