git.sesse.net Git - vlc/blob - modules/codec/subtitles/subsdec.c

   1 /*****************************************************************************
   2  * subsdec.c : text subtitles decoder
   3  *****************************************************************************
   4  * Copyright (C) 2000-2006 the VideoLAN team
   5  * $Id$
   6  *
   7  * Authors: Gildas Bazin <gbazin@videolan.org>
   8  *          Samuel Hocevar <sam@zoy.org>
   9  *          Derk-Jan Hartman <hartman at videolan dot org>
  10  *          Bernie Purcell <bitmap@videolan.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License as published by
  14  * the Free Software Foundation; either version 2 of the License, or
  15  * (at your option) any later version.
  16  *
  17  * This program is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  * GNU General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU General Public License
  23  * along with this program; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  25  *****************************************************************************/
  26
  27 /*****************************************************************************
  28  * Preamble
  29  *****************************************************************************/
  30 #ifdef HAVE_CONFIG_H
  31 # include "config.h"
  32 #endif
  33
  34 #include "subsdec.h"
  35 #include <vlc_plugin.h>
  36
  37 /*****************************************************************************
  38  * Local prototypes
  39  *****************************************************************************/
  40 static int  OpenDecoder   ( vlc_object_t * );
  41 static void CloseDecoder  ( vlc_object_t * );
  42
  43 static subpicture_t   *DecodeBlock   ( decoder_t *, block_t ** );
  44 static subpicture_t   *ParseText     ( decoder_t *, block_t * );
  45 static char           *StripTags      ( char * );
  46 static char           *CreateHtmlSubtitle( int *pi_align, char * );
  47
  48
  49 /*****************************************************************************
  50  * Module descriptor.
  51  *****************************************************************************/
  52 static const char *const ppsz_encodings[] = {
  53     "",
  54     "UTF-8",
  55     "UTF-16",
  56     "UTF-16BE",
  57     "UTF-16LE",
  58     "GB18030",
  59     "ISO-8859-15",
  60     "Windows-1252",
  61     "ISO-8859-2",
  62     "Windows-1250",
  63     "ISO-8859-3",
  64     "ISO-8859-10",
  65     "Windows-1251",
  66     "KOI8-R",
  67     "KOI8-U",
  68     "ISO-8859-6",
  69     "Windows-1256",
  70     "ISO-8859-7",
  71     "Windows-1253",
  72     "ISO-8859-8",
  73     "Windows-1255",
  74     "ISO-8859-9",
  75     "Windows-1254",
  76     "ISO-8859-11",
  77     "Windows-874",
  78     "ISO-8859-13",
  79     "Windows-1257",
  80     "ISO-8859-14",
  81     "ISO-8859-16",
  82     "ISO-2022-CN-EXT",
  83     "EUC-CN",
  84     "ISO-2022-JP-2",
  85     "EUC-JP",
  86     "Shift_JIS",
  87     "CP949",
  88     "ISO-2022-KR",
  89     "Big5",
  90     "ISO-2022-TW",
  91     "Big5-HKSCS",
  92     "VISCII",
  93     "Windows-1258",
  94 };
  95
  96 static const char *const ppsz_encoding_names[] = {
  97     N_("Auto"),
  98     N_("Universal (UTF-8)"),
  99     N_("Universal (UTF-16)"),
 100     N_("Universal (big endian UTF-16)"),
 101     N_("Universal (little endian UTF-16)"),
 102     N_("Universal, Chinese (GB18030)"),
 103
 104   /* ISO 8859 and the likes */
 105     /* 1 */
 106     N_("Western European (Latin-9)"), /* mostly superset of Latin-1 */
 107     N_("Western European (Windows-1252)"),
 108     /* 2 */
 109     N_("Eastern European (Latin-2)"),
 110     N_("Eastern European (Windows-1250)"),
 111     /* 3 */
 112     N_("Esperanto (Latin-3)"),
 113     /* 4 */
 114     N_("Nordic (Latin-6)"), /* Latin 6 supersedes Latin 4 */
 115     /* 5 */
 116     N_("Cyrillic (Windows-1251)"), /* ISO 8859-5 is not practically used */
 117     N_("Russian (KOI8-R)"),
 118     N_("Ukrainian (KOI8-U)"),
 119     /* 6 */
 120     N_("Arabic (ISO 8859-6)"),
 121     N_("Arabic (Windows-1256)"),
 122     /* 7 */
 123     N_("Greek (ISO 8859-7)"),
 124     N_("Greek (Windows-1253)"),
 125     /* 8 */
 126     N_("Hebrew (ISO 8859-8)"),
 127     N_("Hebrew (Windows-1255)"),
 128     /* 9 */
 129     N_("Turkish (ISO 8859-9)"),
 130     N_("Turkish (Windows-1254)"),
 131     /* 10 -> 4 */
 132     /* 11 */
 133     N_("Thai (TIS 620-2533/ISO 8859-11)"),
 134     N_("Thai (Windows-874)"),
 135     /* 13 */
 136     N_("Baltic (Latin-7)"),
 137     N_("Baltic (Windows-1257)"),
 138     /* 12 -> /dev/null */
 139     /* 14 */
 140     N_("Celtic (Latin-8)"),
 141     /* 15 -> 1 */
 142     /* 16 */
 143     N_("South-Eastern European (Latin-10)"),
 144   /* CJK families */
 145     N_("Simplified Chinese (ISO-2022-CN-EXT)"),
 146     N_("Simplified Chinese Unix (EUC-CN)"),
 147     N_("Japanese (7-bits JIS/ISO-2022-JP-2)"),
 148     N_("Japanese Unix (EUC-JP)"),
 149     N_("Japanese (Shift JIS)"),
 150     N_("Korean (EUC-KR/CP949)"),
 151     N_("Korean (ISO-2022-KR)"),
 152     N_("Traditional Chinese (Big5)"),
 153     N_("Traditional Chinese Unix (EUC-TW)"),
 154     N_("Hong-Kong Supplementary (HKSCS)"),
 155   /* Other */
 156     N_("Vietnamese (VISCII)"),
 157     N_("Vietnamese (Windows-1258)"),
 158 };
 159 /*
 160 SSA supports charset selection.
 161 The following known charsets are used:
 162
 163 0 = Ansi - Western European
 164 1 = default
 165 2 = symbol
 166 3 = invalid
 167 77 = Mac
 168 128 = Japanese (Shift JIS)
 169 129 = Hangul
 170 130 = Johab
 171 134 = GB2312 Simplified Chinese
 172 136 = Big5 Traditional Chinese
 173 161 = Greek
 174 162 = Turkish
 175 163 = Vietnamese
 176 177 = Hebrew
 177 178 = Arabic
 178 186 = Baltic
 179 204 = Russian (Cyrillic)
 180 222 = Thai
 181 238 = Eastern European
 182 254 = PC 437
 183 */
 184
 185 static const int  pi_justification[] = { 0, 1, 2 };
 186 static const char *const ppsz_justification_text[] = {
 187     N_("Center"),N_("Left"),N_("Right")};
 188
 189 #define ENCODING_TEXT N_("Subtitles text encoding")
 190 #define ENCODING_LONGTEXT N_("Set the encoding used in text subtitles")
 191 #define ALIGN_TEXT N_("Subtitles justification")
 192 #define ALIGN_LONGTEXT N_("Set the justification of subtitles")
 193 #define AUTODETECT_UTF8_TEXT N_("UTF-8 subtitles autodetection")
 194 #define AUTODETECT_UTF8_LONGTEXT N_("This enables automatic detection of " \
 195             "UTF-8 encoding within subtitles files.")
 196 #define FORMAT_TEXT N_("Formatted Subtitles")
 197 #define FORMAT_LONGTEXT N_("Some subtitle formats allow for text formatting. " \
 198  "VLC partly implements this, but you can choose to disable all formatting.")
 199
 200
 201 vlc_module_begin ()
 202     set_shortname( N_("Subtitles"))
 203     set_description( N_("Text subtitles decoder") )
 204     set_capability( "decoder", 50 )
 205     set_callbacks( OpenDecoder, CloseDecoder )
 206     set_category( CAT_INPUT )
 207     set_subcategory( SUBCAT_INPUT_SCODEC )
 208
 209     add_integer( "subsdec-align", 0, NULL, ALIGN_TEXT, ALIGN_LONGTEXT,
 210                  false )
 211         change_integer_list( pi_justification, ppsz_justification_text, NULL )
 212     add_string( "subsdec-encoding", "", NULL,
 213                 ENCODING_TEXT, ENCODING_LONGTEXT, false )
 214         change_string_list( ppsz_encodings, ppsz_encoding_names, 0 )
 215     add_bool( "subsdec-autodetect-utf8", true, NULL,
 216               AUTODETECT_UTF8_TEXT, AUTODETECT_UTF8_LONGTEXT, false )
 217     add_bool( "subsdec-formatted", true, NULL, FORMAT_TEXT, FORMAT_LONGTEXT,
 218                  false )
 219 vlc_module_end ()
 220
 221 /*****************************************************************************
 222  * OpenDecoder: probe the decoder and return score
 223  *****************************************************************************
 224  * Tries to launch a decoder and return score so that the interface is able
 225  * to chose.
 226  *****************************************************************************/
 227 static int OpenDecoder( vlc_object_t *p_this )
 228 {
 229     decoder_t     *p_dec = (decoder_t*)p_this;
 230     decoder_sys_t *p_sys;
 231
 232     switch( p_dec->fmt_in.i_codec )
 233     {
 234         case VLC_CODEC_SUBT:
 235         case VLC_CODEC_SSA:
 236         case VLC_CODEC_ITU_T140:
 237             break;
 238         default:
 239             return VLC_EGENERIC;
 240     }
 241
 242     p_dec->pf_decode_sub = DecodeBlock;
 243     p_dec->fmt_out.i_cat = SPU_ES;
 244     p_dec->fmt_out.i_codec = 0;
 245
 246     /* Allocate the memory needed to store the decoder's structure */
 247     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 248     if( p_sys == NULL )
 249         return VLC_ENOMEM;
 250
 251     /* init of p_sys */
 252     p_sys->i_align = 0;
 253     p_sys->iconv_handle = (vlc_iconv_t)-1;
 254     p_sys->b_autodetect_utf8 = false;
 255     p_sys->b_ass = false;
 256     p_sys->i_original_height = -1;
 257     p_sys->i_original_width = -1;
 258     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 259     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 260
 261     char *psz_charset = NULL;
 262
 263     /* First try demux-specified encoding */
 264     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 265         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 266     else
 267     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 268     {
 269         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 270         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 271                  p_dec->fmt_in.subs.psz_encoding ?
 272                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 273     }
 274
 275     /* Second, try configured encoding */
 276     if (psz_charset == NULL)
 277     {
 278         psz_charset = var_CreateGetNonEmptyString (p_dec, "subsdec-encoding");
 279         msg_Dbg (p_dec, "trying configured character encoding: %s",
 280                  psz_charset ? psz_charset : "not specified");
 281     }
 282
 283     /* Third, try "local" encoding with optional UTF-8 autodetection */
 284     if (psz_charset == NULL)
 285     {
 286         /* xgettext:
 287            The Windows ANSI code page most commonly used for this language.
 288            VLC uses this as a guess of the subtitle files character set
 289            (if UTF-8 and UTF-16 autodetection fails).
 290            Western European languages normally use "CP1252", which is a
 291            Microsoft-variant of ISO 8859-1. That suits the Latin alphabet.
 292            Other scripts use other code pages.
 293
 294            This MUST be a valid iconv character set. If unsure, please refer
 295            the VideoLAN translators mailing list. */
 296         const char *acp = vlc_pgettext("GetACP", "CP1252");
 297
 298         psz_charset = strdup (acp);
 299         msg_Dbg (p_dec, "trying default character encoding: %s",
 300                  psz_charset ? psz_charset : "not specified");
 301
 302         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 303         {
 304             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 305             p_sys->b_autodetect_utf8 = true;
 306         }
 307     }
 308
 309     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 310     if (psz_charset == NULL)
 311     {
 312         psz_charset = strdup ("UTF-8");
 313         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 314     }
 315
 316     if ((psz_charset != NULL)
 317      && strcasecmp (psz_charset, "UTF-8")
 318      && strcasecmp (psz_charset, "utf8"))
 319     {
 320         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 321         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 322             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 323     }
 324     free (psz_charset);
 325
 326     p_sys->i_align = var_CreateGetInteger( p_dec, "subsdec-align" );
 327
 328     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 329      && var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 330     {
 331         if( p_dec->fmt_in.i_extra > 0 )
 332             ParseSSAHeader( p_dec );
 333     }
 334
 335     return VLC_SUCCESS;
 336 }
 337
 338 /****************************************************************************
 339  * DecodeBlock: the whole thing
 340  ****************************************************************************
 341  * This function must be fed with complete subtitles units.
 342  ****************************************************************************/
 343 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 344 {
 345     subpicture_t *p_spu;
 346     block_t *p_block;
 347
 348     if( !pp_block || *pp_block == NULL )
 349         return NULL;
 350
 351     p_block = *pp_block;
 352     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 353     {
 354         block_Release( p_block );
 355         return NULL;
 356     }
 357
 358     p_spu = ParseText( p_dec, p_block );
 359
 360     block_Release( p_block );
 361     *pp_block = NULL;
 362
 363     return p_spu;
 364 }
 365
 366 /*****************************************************************************
 367  * CloseDecoder: clean up the decoder
 368  *****************************************************************************/
 369 static void CloseDecoder( vlc_object_t *p_this )
 370 {
 371     decoder_t *p_dec = (decoder_t *)p_this;
 372     decoder_sys_t *p_sys = p_dec->p_sys;
 373
 374     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 375         vlc_iconv_close( p_sys->iconv_handle );
 376
 377     if( p_sys->pp_ssa_styles )
 378     {
 379         int i;
 380         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 381         {
 382             if( !p_sys->pp_ssa_styles[i] )
 383                 continue;
 384
 385             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 386             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 387             free( p_sys->pp_ssa_styles[i] );
 388         }
 389         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 390     }
 391     if( p_sys->pp_images )
 392     {
 393         int i;
 394         for( i = 0; i < p_sys->i_images; i++ )
 395         {
 396             if( !p_sys->pp_images[i] )
 397                 continue;
 398
 399             if( p_sys->pp_images[i]->p_pic )
 400                 picture_Release( p_sys->pp_images[i]->p_pic );
 401             free( p_sys->pp_images[i]->psz_filename );
 402
 403             free( p_sys->pp_images[i] );
 404         }
 405         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 406     }
 407
 408     free( p_sys );
 409 }
 410
 411 /*****************************************************************************
 412  * ParseText: parse an text subtitle packet and send it to the video output
 413  *****************************************************************************/
 414 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 415 {
 416     decoder_sys_t *p_sys = p_dec->p_sys;
 417     subpicture_t *p_spu = NULL;
 418     char *psz_subtitle = NULL;
 419     video_format_t fmt;
 420
 421     /* We cannot display a subpicture with no date */
 422     if( p_block->i_pts <= VLC_TS_INVALID )
 423     {
 424         msg_Warn( p_dec, "subtitle without a date" );
 425         return NULL;
 426     }
 427
 428     /* Check validity of packet data */
 429     /* An "empty" line containing only \0 can be used to force
 430        and ephemer picture from the screen */
 431     if( p_block->i_buffer < 1 )
 432     {
 433         msg_Warn( p_dec, "no subtitle data" );
 434         return NULL;
 435     }
 436
 437     /* Should be resiliant against bad subtitles */
 438     psz_subtitle = malloc( p_block->i_buffer + 1 );
 439     if( psz_subtitle == NULL )
 440         return NULL;
 441     memcpy( psz_subtitle, p_block->p_buffer, p_block->i_buffer );
 442     psz_subtitle[p_block->i_buffer] = '\0';
 443
 444     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 445     {
 446         if (EnsureUTF8( psz_subtitle ) == NULL)
 447         {
 448             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 449                      "Try manually setting a character-encoding "
 450                      "before you open the file." );
 451         }
 452     }
 453     else
 454     {
 455
 456         if( p_sys->b_autodetect_utf8 )
 457         {
 458             if( IsUTF8( psz_subtitle ) == NULL )
 459             {
 460                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 461                          "disabling UTF-8 subtitles autodetection" );
 462                 p_sys->b_autodetect_utf8 = false;
 463             }
 464         }
 465
 466         if( !p_sys->b_autodetect_utf8 )
 467         {
 468             size_t inbytes_left = strlen( psz_subtitle );
 469             size_t outbytes_left = 6 * inbytes_left;
 470             char *psz_new_subtitle = xmalloc( outbytes_left + 1 );
 471             char *psz_convert_buffer_out = psz_new_subtitle;
 472             const char *psz_convert_buffer_in = psz_subtitle;
 473
 474             size_t ret = vlc_iconv( p_sys->iconv_handle,
 475                                     &psz_convert_buffer_in, &inbytes_left,
 476                                     &psz_convert_buffer_out, &outbytes_left );
 477
 478             *psz_convert_buffer_out++ = '\0';
 479             free( psz_subtitle );
 480
 481             if( ( ret == (size_t)(-1) ) || inbytes_left )
 482             {
 483                 free( psz_new_subtitle );
 484                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 485                         "Try manually setting a character-encoding "
 486                                 "before you open the file." );
 487                 return NULL;
 488             }
 489
 490             psz_subtitle = realloc( psz_new_subtitle,
 491                                     psz_convert_buffer_out - psz_new_subtitle );
 492             if( !psz_subtitle )
 493                 psz_subtitle = psz_new_subtitle;
 494         }
 495     }
 496
 497     /* Create the subpicture unit */
 498     p_spu = decoder_NewSubpicture( p_dec, NULL );
 499     if( !p_spu )
 500     {
 501         msg_Warn( p_dec, "can't get spu buffer" );
 502         free( psz_subtitle );
 503         return NULL;
 504     }
 505
 506     /* Create a new subpicture region */
 507     memset( &fmt, 0, sizeof(video_format_t) );
 508     fmt.i_chroma = VLC_CODEC_TEXT;
 509     fmt.i_width = fmt.i_height = 0;
 510     fmt.i_x_offset = fmt.i_y_offset = 0;
 511     p_spu->p_region = subpicture_region_New( &fmt );
 512     if( !p_spu->p_region )
 513     {
 514         msg_Err( p_dec, "cannot allocate SPU region" );
 515         free( psz_subtitle );
 516         decoder_DeleteSubpicture( p_dec, p_spu );
 517         return NULL;
 518     }
 519
 520     /* Decode and format the subpicture unit */
 521     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 522     {
 523         /* Normal text subs, easy markup */
 524         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 525         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 526         p_spu->p_region->i_y = 10;
 527
 528         /* Remove formatting from string */
 529
 530         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 531         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 532         {
 533             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 534         }
 535
 536         p_spu->i_start = p_block->i_pts;
 537         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 538         p_spu->b_ephemer = (p_block->i_length == 0);
 539         p_spu->b_absolute = false;
 540     }
 541     else
 542     {
 543         /* Decode SSA/USF strings */
 544         ParseSSAString( p_dec, psz_subtitle, p_spu );
 545
 546         p_spu->i_start = p_block->i_pts;
 547         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 548         p_spu->b_ephemer = (p_block->i_length == 0);
 549         p_spu->b_absolute = false;
 550         p_spu->i_original_picture_width = p_sys->i_original_width;
 551         p_spu->i_original_picture_height = p_sys->i_original_height;
 552     }
 553     free( psz_subtitle );
 554
 555     return p_spu;
 556 }
 557
 558 char* GotoNextLine( char *psz_text )
 559 {
 560     char *p_newline = psz_text;
 561
 562     while( p_newline[0] != '\0' )
 563     {
 564         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 565         {
 566             p_newline++;
 567             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 568                 p_newline++;
 569             break;
 570         }
 571         else p_newline++;
 572     }
 573     return p_newline;
 574 }
 575
 576 /* Function now handles tags with attribute values, and tries
 577  * to deal with &' commands too. It no longer modifies the string
 578  * in place, so that the original text can be reused
 579  */
 580 static char *StripTags( char *psz_subtitle )
 581 {
 582     char *psz_text_start;
 583     char *psz_text;
 584
 585     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 586     if( !psz_text_start )
 587         return NULL;
 588
 589     while( *psz_subtitle )
 590     {
 591         if( *psz_subtitle == '<' )
 592         {
 593             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 594                 *psz_text++ = '\n';
 595
 596             psz_subtitle += strcspn( psz_subtitle, ">" );
 597         }
 598         else if( *psz_subtitle == '&' )
 599         {
 600             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 601             {
 602                 *psz_text++ = '<';
 603                 psz_subtitle += strcspn( psz_subtitle, ";" );
 604             }
 605             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 606             {
 607                 *psz_text++ = '>';
 608                 psz_subtitle += strcspn( psz_subtitle, ";" );
 609             }
 610             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 611             {
 612                 *psz_text++ = '&';
 613                 psz_subtitle += strcspn( psz_subtitle, ";" );
 614             }
 615             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 616             {
 617                 *psz_text++ = '\"';
 618                 psz_subtitle += strcspn( psz_subtitle, ";" );
 619             }
 620             else
 621             {
 622                 /* Assume it is just a normal ampersand */
 623                 *psz_text++ = '&';
 624             }
 625         }
 626         else
 627         {
 628             *psz_text++ = *psz_subtitle;
 629         }
 630
 631         psz_subtitle++;
 632     }
 633     *psz_text = '\0';
 634     char *psz = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 635     if( psz ) psz_text_start = psz;
 636
 637     return psz_text_start;
 638 }
 639
 640 /* Try to respect any style tags present in the subtitle string. The main
 641  * problem here is a lack of adequate specs for the subtitle formats.
 642  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 643  * SAMI has a detailed spec, but extensive rework is needed in the demux
 644  * code to prevent all this style information being excised, as it presently
 645  * does.
 646  * That leaves the others - none of which were (I guess) originally intended
 647  * to be carrying style information. Over time people have used them that way.
 648  * In the absence of specifications from which to work, the tags supported
 649  * have been restricted to the simple set permitted by the USF DTD, ie. :
 650  *  Basic: <br>, <i>, <b>, <u>, <s>
 651  *  Extended: <font>
 652  *    Attributes: face
 653  *                family
 654  *                size
 655  *                color
 656  *                outline-color
 657  *                shadow-color
 658  *                outline-level
 659  *                shadow-level
 660  *                back-color
 661  *                alpha
 662  * There is also the further restriction that the subtitle be well-formed
 663  * as an XML entity, ie. the HTML sentence:
 664  *        <b><i>Bold and Italics</b></i>
 665  * doesn't qualify because the tags aren't nested one inside the other.
 666  * <text> tags are automatically added to the output to ensure
 667  * well-formedness.
 668  * If the text doesn't qualify for any reason, a NULL string is
 669  * returned, and the rendering engine will fall back to the
 670  * plain text version of the subtitle.
 671  */
 672 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 673 {
 674     const int i_len = strlen(psz_text);
 675
 676     strncpy( *ppsz_html, psz_text, i_max );
 677     *ppsz_html += __MIN(i_max,i_len);
 678 }
 679
 680 static void HtmlPut( char **ppsz_html, const char *psz_text )
 681 {
 682     strcpy( *ppsz_html, psz_text );
 683     *ppsz_html += strlen(psz_text);
 684 }
 685 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 686 {
 687     HtmlPut( ppsz_html, psz_text );
 688     *ppsz_subtitle += strlen(psz_text);
 689 }
 690
 691 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 692 {
 693     /* */
 694     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 695     if( !psz_tag )
 696         return NULL;
 697     psz_tag[ 0 ] = '\0';
 698
 699     /* */
 700     //Oo + 100 ???
 701     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 702     char   *psz_html_start = malloc( i_buf_size );
 703     char   *psz_html = psz_html_start;
 704     if( psz_html_start == NULL )
 705     {
 706         free( psz_tag );
 707         return NULL;
 708     }
 709     psz_html[0] = '\0';
 710
 711     bool b_has_align = false;
 712
 713     HtmlPut( &psz_html, "<text>" );
 714
 715     /* */
 716     while( *psz_subtitle )
 717     {
 718         if( *psz_subtitle == '\n' )
 719         {
 720             HtmlPut( &psz_html, "<br/>" );
 721             psz_subtitle++;
 722         }
 723         else if( *psz_subtitle == '<' )
 724         {
 725             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 726             {
 727                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 728             }
 729             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 730             {
 731                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 732                 strcat( psz_tag, "b" );
 733             }
 734             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 735             {
 736                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 737                 strcat( psz_tag, "i" );
 738             }
 739             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 740             {
 741                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 742                 strcat( psz_tag, "u" );
 743             }
 744             else if( !strncasecmp( psz_subtitle, "<s>", 3 ) )
 745             {
 746                 HtmlCopy( &psz_html, &psz_subtitle, "<s>" );
 747                 strcat( psz_tag, "s" );
 748             }
 749             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 750             {
 751                 const char *psz_attribs[] = { "face=", "family=", "size=",
 752                         "color=", "outline-color=", "shadow-color=",
 753                         "outline-level=", "shadow-level=", "back-color=",
 754                         "alpha=", NULL };
 755
 756                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 757                 strcat( psz_tag, "f" );
 758
 759                 while( *psz_subtitle != '>' )
 760                 {
 761                     int  k;
 762
 763                     for( k=0; psz_attribs[ k ]; k++ )
 764                     {
 765                         int i_len = strlen( psz_attribs[ k ] );
 766
 767                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 768                         {
 769                             /* */
 770                             HtmlPut( &psz_html, psz_attribs[k] );
 771                             psz_subtitle += i_len;
 772
 773                             /* */
 774                             if( *psz_subtitle == '"' )
 775                             {
 776                                 psz_subtitle++;
 777                                 i_len = strcspn( psz_subtitle, "\"" );
 778                             }
 779                             else
 780                             {
 781                                 i_len = strcspn( psz_subtitle, " \t>" );
 782                             }
 783                             HtmlPut( &psz_html, "\"" );
 784                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 785                                 HtmlPut( &psz_html, "#" );
 786                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 787                             HtmlPut( &psz_html, "\"" );
 788
 789                             psz_subtitle += i_len;
 790                             if( *psz_subtitle == '\"' )
 791                                 psz_subtitle++;
 792                             break;
 793                         }
 794                     }
 795                     if( psz_attribs[ k ] == NULL )
 796                     {
 797                         /* Jump over unrecognised tag */
 798                         int i_len = strcspn( psz_subtitle, "\"" );
 799                         if( psz_subtitle[i_len] == '\"' )
 800                         {
 801                             i_len += 1 + strcspn( &psz_subtitle[i_len + 1], "\"" );
 802                             if( psz_subtitle[i_len] == '\"' )
 803                                 i_len++;
 804                         }
 805                         psz_subtitle += i_len;
 806                     }
 807                     while (*psz_subtitle == ' ')
 808                         *psz_html++ = *psz_subtitle++;
 809                 }
 810                 *psz_html++ = *psz_subtitle++;
 811             }
 812             else if( !strncmp( psz_subtitle, "</", 2 ))
 813             {
 814                 bool   b_match     = false;
 815                 bool   b_ignore    = false;
 816                 int    i_len       = strlen( psz_tag ) - 1;
 817                 char  *psz_lastTag = NULL;
 818
 819                 if( i_len >= 0 )
 820                 {
 821                     psz_lastTag = psz_tag + i_len;
 822                     i_len = 0;
 823
 824                     switch( *psz_lastTag )
 825                     {
 826                     case 'b':
 827                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 828                         i_len   = 4;
 829                         break;
 830                     case 'i':
 831                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 832                         i_len   = 4;
 833                         break;
 834                     case 'u':
 835                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 836                         i_len   = 4;
 837                         break;
 838                     case 's':
 839                         b_match = !strncasecmp( psz_subtitle, "</s>", 4 );
 840                         i_len   = 4;
 841                         break;
 842                     case 'f':
 843                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 844                         i_len   = 7;
 845                         break;
 846                     case 'I':
 847                         i_len = strcspn( psz_subtitle, ">" );
 848                         b_match = psz_subtitle[i_len] == '>';
 849                         b_ignore = true;
 850                         if( b_match )
 851                             i_len++;
 852                         break;
 853                     }
 854                 }
 855                 if( !b_match )
 856                 {
 857                     /* Not well formed -- kill everything */
 858                     free( psz_html_start );
 859                     psz_html_start = NULL;
 860                     break;
 861                 }
 862                 *psz_lastTag = '\0';
 863                 if( !b_ignore )
 864                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 865
 866                 psz_subtitle += i_len;
 867             }
 868             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 869                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 870             {
 871                 /* We have a single < */
 872                 HtmlPut( &psz_html, "&lt;" );
 873                 psz_subtitle++;
 874             }
 875             else
 876             {
 877                 /* We have an unknown tag or a single < */
 878
 879                 /* Search for the next tag or end of tag or end of string */
 880                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 881                 char *psz_closing = strstr( psz_subtitle, "/>" );
 882
 883                 if( psz_closing && psz_closing < psz_stop )
 884                 {
 885                     /* We have a self closed tag, remove it */
 886                     psz_subtitle = &psz_closing[2];
 887                 }
 888                 else if( *psz_stop == '>' )
 889                 {
 890                     char psz_match[256];
 891
 892                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 893                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 894
 895                     if( strstr( psz_subtitle, psz_match ) )
 896                     {
 897                         /* We have the closing tag, ignore it TODO */
 898                         psz_subtitle = &psz_stop[1];
 899                         strcat( psz_tag, "I" );
 900                     }
 901                     else
 902                     {
 903                         int i_len = psz_stop + 1 - psz_subtitle;
 904
 905                         /* Copy the whole data */
 906                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 907                         {
 908                             if( *psz_subtitle == '<' )
 909                                 HtmlPut( &psz_html, "&lt;" );
 910                             else if( *psz_subtitle == '>' )
 911                                 HtmlPut( &psz_html, "&gt;" );
 912                             else
 913                                 *psz_html++ = *psz_subtitle;
 914                         }
 915                     }
 916                 }
 917                 else
 918                 {
 919                     /* We have a single < */
 920                     HtmlPut( &psz_html, "&lt;" );
 921                     psz_subtitle++;
 922                 }
 923             }
 924         }
 925         else if( *psz_subtitle == '&' )
 926         {
 927             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 928             {
 929                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 930             }
 931             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 932             {
 933                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 934             }
 935             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 936             {
 937                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 938             }
 939             else
 940             {
 941                 HtmlPut( &psz_html, "&amp;" );
 942                 psz_subtitle++;
 943             }
 944         }
 945         else if( *psz_subtitle == '>' )
 946         {
 947             HtmlPut( &psz_html, "&gt;" );
 948             psz_subtitle++;
 949         }
 950         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 951                  strchr( psz_subtitle, '}' ) )
 952         {
 953             /* Check for forced alignment */
 954             if( !b_has_align &&
 955                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 956             {
 957                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 958                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 959                 const int i_id = psz_subtitle[4] - '1';
 960
 961                 b_has_align = true;
 962                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 963             }
 964             /* TODO fr -> rotation */
 965
 966             /* Hide {\stupidity} */
 967             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 968         }
 969         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == 'Y'
 970                 && psz_subtitle[2] == ':' && strchr( psz_subtitle, '}' ) )
 971         {
 972             /* Hide {Y:stupidity} */
 973             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 974         }
 975         else if( psz_subtitle[0] == '\\' && psz_subtitle[1] )
 976         {
 977             if( psz_subtitle[1] == 'N' || psz_subtitle[1] == 'n' )
 978             {
 979                 HtmlPut( &psz_html, "<br/>" );
 980                 psz_subtitle += 2;
 981             }
 982             else if( psz_subtitle[1] == 'h' )
 983             {
 984                 /* Non breakable space */
 985                 HtmlPut( &psz_html, NO_BREAKING_SPACE );
 986                 psz_subtitle += 2;
 987             }
 988             else
 989             {
 990                 HtmlPut( &psz_html, "\\" );
 991                 psz_subtitle++;
 992             }
 993         }
 994         else
 995         {
 996             *psz_html = *psz_subtitle;
 997             if( psz_html > psz_html_start )
 998             {
 999                 /* Check for double whitespace */
1000                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
1001                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
1002                 {
1003                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
1004                     psz_html--;
1005                 }
1006             }
1007             psz_html++;
1008             psz_subtitle++;
1009         }
1010
1011         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
1012         {
1013             const int i_len = psz_html - psz_html_start;
1014
1015             i_buf_size += 200;
1016             char *psz_new = realloc( psz_html_start, i_buf_size );
1017             if( !psz_new )
1018                 break;
1019             psz_html_start = psz_new;
1020             psz_html = &psz_new[i_len];
1021         }
1022     }
1023     if( psz_html_start )
1024     {
1025         static const char *psz_text_close = "</text>";
1026         static const char *psz_tag_long = "/font>";
1027
1028         /* Realloc for closing tags and shrink memory */
1029         const size_t i_length = (size_t)( psz_html - psz_html_start );
1030
1031         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
1032         char *psz_new = realloc( psz_html_start, i_size );
1033         if( psz_new )
1034         {
1035             psz_html_start = psz_new;
1036             psz_html = &psz_new[i_length];
1037
1038             /* Close not well formed subtitle */
1039             while( *psz_tag )
1040             {
1041                 /* */
1042                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
1043                 switch( *psz_last )
1044                 {
1045                 case 'b':
1046                     HtmlPut( &psz_html, "</b>" );
1047                     break;
1048                 case 'i':
1049                     HtmlPut( &psz_html, "</i>" );
1050                     break;
1051                 case 'u':
1052                     HtmlPut( &psz_html, "</u>" );
1053                     break;
1054                 case 's':
1055                     HtmlPut( &psz_html, "</s>" );
1056                     break;
1057                 case 'f':
1058                     HtmlPut( &psz_html, "/font>" );
1059                     break;
1060                 case 'I':
1061                     break;
1062                 }
1063
1064                 *psz_last = '\0';
1065             }
1066             HtmlPut( &psz_html, psz_text_close );
1067         }
1068     }
1069     free( psz_tag );
1070
1071     return psz_html_start;
1072 }
1073