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         psz_charset = strdup (GetFallbackEncoding ());
 287         msg_Dbg (p_dec, "trying default character encoding: %s",
 288                  psz_charset ? psz_charset : "not specified");
 289
 290         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 291         {
 292             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 293             p_sys->b_autodetect_utf8 = true;
 294         }
 295     }
 296
 297     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 298     if (psz_charset == NULL)
 299     {
 300         psz_charset = strdup ("UTF-8");
 301         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 302     }
 303
 304     if ((psz_charset != NULL)
 305      && strcasecmp (psz_charset, "UTF-8")
 306      && strcasecmp (psz_charset, "utf8"))
 307     {
 308         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 309         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 310             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 311     }
 312     free (psz_charset);
 313
 314     p_sys->i_align = var_CreateGetInteger( p_dec, "subsdec-align" );
 315
 316     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 317      && var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 318     {
 319         if( p_dec->fmt_in.i_extra > 0 )
 320             ParseSSAHeader( p_dec );
 321     }
 322
 323     return VLC_SUCCESS;
 324 }
 325
 326 /****************************************************************************
 327  * DecodeBlock: the whole thing
 328  ****************************************************************************
 329  * This function must be fed with complete subtitles units.
 330  ****************************************************************************/
 331 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 332 {
 333     subpicture_t *p_spu;
 334     block_t *p_block;
 335
 336     if( !pp_block || *pp_block == NULL )
 337         return NULL;
 338
 339     p_block = *pp_block;
 340     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 341     {
 342         block_Release( p_block );
 343         return NULL;
 344     }
 345
 346     p_spu = ParseText( p_dec, p_block );
 347
 348     block_Release( p_block );
 349     *pp_block = NULL;
 350
 351     return p_spu;
 352 }
 353
 354 /*****************************************************************************
 355  * CloseDecoder: clean up the decoder
 356  *****************************************************************************/
 357 static void CloseDecoder( vlc_object_t *p_this )
 358 {
 359     decoder_t *p_dec = (decoder_t *)p_this;
 360     decoder_sys_t *p_sys = p_dec->p_sys;
 361
 362     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 363         vlc_iconv_close( p_sys->iconv_handle );
 364
 365     if( p_sys->pp_ssa_styles )
 366     {
 367         int i;
 368         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 369         {
 370             if( !p_sys->pp_ssa_styles[i] )
 371                 continue;
 372
 373             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 374             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 375             free( p_sys->pp_ssa_styles[i] );
 376         }
 377         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 378     }
 379     if( p_sys->pp_images )
 380     {
 381         int i;
 382         for( i = 0; i < p_sys->i_images; i++ )
 383         {
 384             if( !p_sys->pp_images[i] )
 385                 continue;
 386
 387             if( p_sys->pp_images[i]->p_pic )
 388                 picture_Release( p_sys->pp_images[i]->p_pic );
 389             free( p_sys->pp_images[i]->psz_filename );
 390
 391             free( p_sys->pp_images[i] );
 392         }
 393         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 394     }
 395
 396     free( p_sys );
 397 }
 398
 399 /*****************************************************************************
 400  * ParseText: parse an text subtitle packet and send it to the video output
 401  *****************************************************************************/
 402 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 403 {
 404     decoder_sys_t *p_sys = p_dec->p_sys;
 405     subpicture_t *p_spu = NULL;
 406     char *psz_subtitle = NULL;
 407     video_format_t fmt;
 408
 409     /* We cannot display a subpicture with no date */
 410     if( p_block->i_pts == 0 )
 411     {
 412         msg_Warn( p_dec, "subtitle without a date" );
 413         return NULL;
 414     }
 415
 416     /* Check validity of packet data */
 417     /* An "empty" line containing only \0 can be used to force
 418        and ephemer picture from the screen */
 419     if( p_block->i_buffer < 1 )
 420     {
 421         msg_Warn( p_dec, "no subtitle data" );
 422         return NULL;
 423     }
 424
 425     /* Should be resiliant against bad subtitles */
 426     psz_subtitle = strndup( (const char *)p_block->p_buffer,
 427                             p_block->i_buffer );
 428     if( psz_subtitle == NULL )
 429         return NULL;
 430
 431     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 432     {
 433         if (EnsureUTF8( psz_subtitle ) == NULL)
 434         {
 435             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 436                      "Try manually setting a character-encoding "
 437                      "before you open the file." );
 438         }
 439     }
 440     else
 441     {
 442
 443         if( p_sys->b_autodetect_utf8 )
 444         {
 445             if( IsUTF8( psz_subtitle ) == NULL )
 446             {
 447                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 448                          "disabling UTF-8 subtitles autodetection" );
 449                 p_sys->b_autodetect_utf8 = false;
 450             }
 451         }
 452
 453         if( !p_sys->b_autodetect_utf8 )
 454         {
 455             size_t inbytes_left = strlen( psz_subtitle );
 456             size_t outbytes_left = 6 * inbytes_left;
 457             char *psz_new_subtitle = malloc( outbytes_left + 1 );
 458             char *psz_convert_buffer_out = psz_new_subtitle;
 459             const char *psz_convert_buffer_in = psz_subtitle;
 460
 461             size_t ret = vlc_iconv( p_sys->iconv_handle,
 462                                     &psz_convert_buffer_in, &inbytes_left,
 463                                     &psz_convert_buffer_out, &outbytes_left );
 464
 465             *psz_convert_buffer_out++ = '\0';
 466             free( psz_subtitle );
 467
 468             if( ( ret == (size_t)(-1) ) || inbytes_left )
 469             {
 470                 free( psz_new_subtitle );
 471                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 472                         "Try manually setting a character-encoding "
 473                                 "before you open the file." );
 474                 return NULL;
 475             }
 476
 477             psz_subtitle = realloc( psz_new_subtitle,
 478                                     psz_convert_buffer_out - psz_new_subtitle );
 479         }
 480     }
 481
 482     /* Create the subpicture unit */
 483     p_spu = decoder_NewSubpicture( p_dec );
 484     if( !p_spu )
 485     {
 486         msg_Warn( p_dec, "can't get spu buffer" );
 487         free( psz_subtitle );
 488         return NULL;
 489     }
 490
 491     /* Create a new subpicture region */
 492     memset( &fmt, 0, sizeof(video_format_t) );
 493     fmt.i_chroma = VLC_CODEC_TEXT;
 494     fmt.i_aspect = 0;
 495     fmt.i_width = fmt.i_height = 0;
 496     fmt.i_x_offset = fmt.i_y_offset = 0;
 497     p_spu->p_region = subpicture_region_New( &fmt );
 498     if( !p_spu->p_region )
 499     {
 500         msg_Err( p_dec, "cannot allocate SPU region" );
 501         free( psz_subtitle );
 502         decoder_DeleteSubpicture( p_dec, p_spu );
 503         return NULL;
 504     }
 505
 506     /* Decode and format the subpicture unit */
 507     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 508     {
 509         /* Normal text subs, easy markup */
 510         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 511         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 512         p_spu->p_region->i_y = 10;
 513
 514         /* Remove formatting from string */
 515
 516         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 517         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 518         {
 519             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 520         }
 521
 522         p_spu->i_start = p_block->i_pts;
 523         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 524         p_spu->b_ephemer = (p_block->i_length == 0);
 525         p_spu->b_absolute = false;
 526     }
 527     else
 528     {
 529         /* Decode SSA/USF strings */
 530         ParseSSAString( p_dec, psz_subtitle, p_spu );
 531
 532         p_spu->i_start = p_block->i_pts;
 533         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 534         p_spu->b_ephemer = (p_block->i_length == 0);
 535         p_spu->b_absolute = false;
 536         p_spu->i_original_picture_width = p_sys->i_original_width;
 537         p_spu->i_original_picture_height = p_sys->i_original_height;
 538     }
 539     free( psz_subtitle );
 540
 541     return p_spu;
 542 }
 543
 544 char* GotoNextLine( char *psz_text )
 545 {
 546     char *p_newline = psz_text;
 547
 548     while( p_newline[0] != '\0' )
 549     {
 550         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 551         {
 552             p_newline++;
 553             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 554                 p_newline++;
 555             break;
 556         }
 557         else p_newline++;
 558     }
 559     return p_newline;
 560 }
 561
 562 /* Function now handles tags with attribute values, and tries
 563  * to deal with &' commands too. It no longer modifies the string
 564  * in place, so that the original text can be reused
 565  */
 566 static char *StripTags( char *psz_subtitle )
 567 {
 568     char *psz_text_start;
 569     char *psz_text;
 570
 571     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 572     if( !psz_text_start )
 573         return NULL;
 574
 575     while( *psz_subtitle )
 576     {
 577         if( *psz_subtitle == '<' )
 578         {
 579             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 580                 *psz_text++ = '\n';
 581
 582             psz_subtitle += strcspn( psz_subtitle, ">" );
 583         }
 584         else if( *psz_subtitle == '&' )
 585         {
 586             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 587             {
 588                 *psz_text++ = '<';
 589                 psz_subtitle += strcspn( psz_subtitle, ";" );
 590             }
 591             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 592             {
 593                 *psz_text++ = '>';
 594                 psz_subtitle += strcspn( psz_subtitle, ";" );
 595             }
 596             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 597             {
 598                 *psz_text++ = '&';
 599                 psz_subtitle += strcspn( psz_subtitle, ";" );
 600             }
 601             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 602             {
 603                 *psz_text++ = '\"';
 604                 psz_subtitle += strcspn( psz_subtitle, ";" );
 605             }
 606             else
 607             {
 608                 /* Assume it is just a normal ampersand */
 609                 *psz_text++ = '&';
 610             }
 611         }
 612         else
 613         {
 614             *psz_text++ = *psz_subtitle;
 615         }
 616
 617         psz_subtitle++;
 618     }
 619     *psz_text = '\0';
 620     psz_text_start = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 621
 622     return psz_text_start;
 623 }
 624
 625 /* Try to respect any style tags present in the subtitle string. The main
 626  * problem here is a lack of adequate specs for the subtitle formats.
 627  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 628  * SAMI has a detailed spec, but extensive rework is needed in the demux
 629  * code to prevent all this style information being excised, as it presently
 630  * does.
 631  * That leaves the others - none of which were (I guess) originally intended
 632  * to be carrying style information. Over time people have used them that way.
 633  * In the absence of specifications from which to work, the tags supported
 634  * have been restricted to the simple set permitted by the USF DTD, ie. :
 635  *  Basic: <br>, <i>, <b>, <u>
 636  *  Extended: <font>
 637  *    Attributes: face
 638  *                family
 639  *                size
 640  *                color
 641  *                outline-color
 642  *                shadow-color
 643  *                outline-level
 644  *                shadow-level
 645  *                back-color
 646  *                alpha
 647  * There is also the further restriction that the subtitle be well-formed
 648  * as an XML entity, ie. the HTML sentence:
 649  *        <b><i>Bold and Italics</b></i>
 650  * doesn't qualify because the tags aren't nested one inside the other.
 651  * <text> tags are automatically added to the output to ensure
 652  * well-formedness.
 653  * If the text doesn't qualify for any reason, a NULL string is
 654  * returned, and the rendering engine will fall back to the
 655  * plain text version of the subtitle.
 656  */
 657 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 658 {
 659     const int i_len = strlen(psz_text);
 660
 661     strncpy( *ppsz_html, psz_text, i_max );
 662     *ppsz_html += __MIN(i_max,i_len);
 663 }
 664
 665 static void HtmlPut( char **ppsz_html, const char *psz_text )
 666 {
 667     strcpy( *ppsz_html, psz_text );
 668     *ppsz_html += strlen(psz_text);
 669 }
 670 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 671 {
 672     HtmlPut( ppsz_html, psz_text );
 673     *ppsz_subtitle += strlen(psz_text);
 674 }
 675
 676 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 677 {
 678     /* */
 679     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 680     if( !psz_tag )
 681         return NULL;
 682     psz_tag[ 0 ] = '\0';
 683
 684     /* */
 685     //Oo + 100 ???
 686     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 687     char   *psz_html_start = malloc( i_buf_size );
 688     char   *psz_html = psz_html_start;
 689     if( psz_html_start == NULL )
 690     {
 691         free( psz_tag );
 692         return NULL;
 693     }
 694     psz_html[0] = '\0';
 695
 696     bool b_has_align = false;
 697
 698     HtmlPut( &psz_html, "<text>" );
 699
 700     /* */
 701     while( *psz_subtitle )
 702     {
 703         if( *psz_subtitle == '\n' )
 704         {
 705             HtmlPut( &psz_html, "<br/>" );
 706             psz_subtitle++;
 707         }
 708         else if( *psz_subtitle == '<' )
 709         {
 710             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 711             {
 712                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 713             }
 714             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 715             {
 716                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 717                 strcat( psz_tag, "b" );
 718             }
 719             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 720             {
 721                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 722                 strcat( psz_tag, "i" );
 723             }
 724             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 725             {
 726                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 727                 strcat( psz_tag, "u" );
 728             }
 729             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 730             {
 731                 const char *psz_attribs[] = { "face=", "family=", "size=",
 732                         "color=", "outline-color=", "shadow-color=",
 733                         "outline-level=", "shadow-level=", "back-color=",
 734                         "alpha=", NULL };
 735
 736                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 737                 strcat( psz_tag, "f" );
 738
 739                 while( *psz_subtitle != '>' )
 740                 {
 741                     int  k;
 742
 743                     for( k=0; psz_attribs[ k ]; k++ )
 744                     {
 745                         int i_len = strlen( psz_attribs[ k ] );
 746
 747                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 748                         {
 749                             /* */
 750                             HtmlPut( &psz_html, psz_attribs[k] );
 751                             psz_subtitle += i_len;
 752
 753                             /* */
 754                             if( *psz_subtitle == '"' )
 755                             {
 756                                 psz_subtitle++;
 757                                 i_len = strcspn( psz_subtitle, "\"" );
 758                             }
 759                             else
 760                             {
 761                                 i_len = strcspn( psz_subtitle, " \t>" );
 762                             }
 763                             HtmlPut( &psz_html, "\"" );
 764                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 765                                 HtmlPut( &psz_html, "#" );
 766                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 767                             HtmlPut( &psz_html, "\"" );
 768
 769                             psz_subtitle += i_len;
 770                             if( *psz_subtitle == '\"' )
 771                                 psz_subtitle++;
 772                             break;
 773                         }
 774                     }
 775                     if( psz_attribs[ k ] == NULL )
 776                     {
 777                         /* Jump over unrecognised tag */
 778                         int i_len = strcspn( psz_subtitle, "\"" ) + 1;
 779
 780                         i_len += strcspn( psz_subtitle + i_len, "\"" ) + 1;
 781                         psz_subtitle += i_len;
 782                     }
 783                     while (*psz_subtitle == ' ')
 784                         *psz_html++ = *psz_subtitle++;
 785                 }
 786                 *psz_html++ = *psz_subtitle++;
 787             }
 788             else if( !strncmp( psz_subtitle, "</", 2 ))
 789             {
 790                 bool   b_match     = false;
 791                 bool   b_ignore    = false;
 792                 int    i_len       = strlen( psz_tag ) - 1;
 793                 char  *psz_lastTag = NULL;
 794
 795                 if( i_len >= 0 )
 796                 {
 797                     psz_lastTag = psz_tag + i_len;
 798                     i_len = 0;
 799
 800                     switch( *psz_lastTag )
 801                     {
 802                     case 'b':
 803                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 804                         i_len   = 4;
 805                         break;
 806                     case 'i':
 807                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 808                         i_len   = 4;
 809                         break;
 810                     case 'u':
 811                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 812                         i_len   = 4;
 813                         break;
 814                     case 'f':
 815                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 816                         i_len   = 7;
 817                         break;
 818                     case 'I':
 819                         i_len = strcspn( psz_subtitle, ">" );
 820                         b_match = psz_subtitle[i_len] == '>';
 821                         b_ignore = true;
 822                         if( b_match )
 823                             i_len++;
 824                         break;
 825                     }
 826                 }
 827                 if( !b_match )
 828                 {
 829                     /* Not well formed -- kill everything */
 830                     free( psz_html_start );
 831                     psz_html_start = NULL;
 832                     break;
 833                 }
 834                 *psz_lastTag = '\0';
 835                 if( !b_ignore )
 836                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 837
 838                 psz_subtitle += i_len;
 839             }
 840             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 841                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 842             {
 843                 /* We have a single < */
 844                 HtmlPut( &psz_html, "&lt;" );
 845                 psz_subtitle++;
 846             }
 847             else
 848             {
 849                 /* We have an unknown tag or a single < */
 850
 851                 /* Search for the next tag or end of tag or end of string */
 852                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 853                 char *psz_closing = strstr( psz_subtitle, "/>" );
 854
 855                 if( psz_closing && psz_closing < psz_stop )
 856                 {
 857                     /* We have a self closed tag, remove it */
 858                     psz_subtitle = &psz_closing[2];
 859                 }
 860                 else if( *psz_stop == '>' )
 861                 {
 862                     char psz_match[256];
 863
 864                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 865                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 866
 867                     if( strstr( psz_subtitle, psz_match ) )
 868                     {
 869                         /* We have the closing tag, ignore it TODO */
 870                         psz_subtitle = &psz_stop[1];
 871                         strcat( psz_tag, "I" );
 872                     }
 873                     else
 874                     {
 875                         int i_len = psz_stop + 1 - psz_subtitle;
 876
 877                         /* Copy the whole data */
 878                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 879                         {
 880                             if( *psz_subtitle == '<' )
 881                                 HtmlPut( &psz_html, "&lt;" );
 882                             else if( *psz_subtitle == '>' )
 883                                 HtmlPut( &psz_html, "&gt;" );
 884                             else
 885                                 *psz_html++ = *psz_subtitle;
 886                         }
 887                     }
 888                 }
 889                 else
 890                 {
 891                     /* We have a single < */
 892                     HtmlPut( &psz_html, "&lt;" );
 893                     psz_subtitle++;
 894                 }
 895             }
 896         }
 897         else if( *psz_subtitle == '&' )
 898         {
 899             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 900             {
 901                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 902             }
 903             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 904             {
 905                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 906             }
 907             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 908             {
 909                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 910             }
 911             else
 912             {
 913                 HtmlPut( &psz_html, "&amp;" );
 914                 psz_subtitle++;
 915             }
 916         }
 917         else if( *psz_subtitle == '>' )
 918         {
 919             HtmlPut( &psz_html, "&gt;" );
 920             psz_subtitle++;
 921         }
 922         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 923                  strchr( psz_subtitle, '}' ) )
 924         {
 925             /* Check for forced alignment */
 926             if( !b_has_align &&
 927                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 928             {
 929                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 930                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 931                 const int i_id = psz_subtitle[4] - '1';
 932
 933                 b_has_align = true;
 934                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 935             }
 936             /* TODO fr -> rotation */
 937
 938             /* Hide {\stupidity} */
 939             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 940         }
 941         else
 942         {
 943             *psz_html = *psz_subtitle;
 944             if( psz_html > psz_html_start )
 945             {
 946                 /* Check for double whitespace */
 947                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
 948                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
 949                 {
 950                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
 951                     psz_html--;
 952                 }
 953             }
 954             psz_html++;
 955             psz_subtitle++;
 956         }
 957
 958         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
 959         {
 960             const int i_len = psz_html - psz_html_start;
 961
 962             i_buf_size += 200;
 963             char *psz_new = realloc( psz_html_start, i_buf_size );
 964             if( !psz_new )
 965                 break;
 966             psz_html_start = psz_new;
 967             psz_html = &psz_new[i_len];
 968         }
 969     }
 970     if( psz_html_start )
 971     {
 972         static const char *psz_text_close = "</text>";
 973         static const char *psz_tag_long = "/font>";
 974
 975         /* Realloc for closing tags and shrink memory */
 976         const size_t i_length = (size_t)( psz_html - psz_html_start );
 977
 978         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
 979         char *psz_new = realloc( psz_html_start, i_size );
 980         if( psz_new )
 981         {
 982             psz_html_start = psz_new;
 983             psz_html = &psz_new[i_length];
 984
 985             /* Close not well formed subtitle */
 986             while( *psz_tag )
 987             {
 988                 /* */
 989                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
 990                 switch( *psz_last )
 991                 {
 992                 case 'b':
 993                     HtmlPut( &psz_html, "</b>" );
 994                     break;
 995                 case 'i':
 996                     HtmlPut( &psz_html, "</i>" );
 997                     break;
 998                 case 'u':
 999                     HtmlPut( &psz_html, "</u>" );
1000                     break;
1001                 case 'f':
1002                     HtmlPut( &psz_html, "/font>" );
1003                     break;
1004                 case 'I':
1005                     break;
1006                 }
1007
1008                 *psz_last = '\0';
1009             }
1010             HtmlPut( &psz_html, psz_text_close );
1011         }
1012     }
1013     free( psz_tag );
1014
1015     return psz_html_start;
1016 }
1017