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-1256",
  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     "ISO-2022-KR",
  88     "EUC-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-1256)"),
 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 (ISO-2022-KR)"),
 151     N_("Korean Unix (EUC-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     vlc_value_t    val;
 232
 233     switch( p_dec->fmt_in.i_codec )
 234     {
 235         case VLC_CODEC_SUBT:
 236         case VLC_CODEC_SSA:
 237         case VLC_CODEC_ITU_T140:
 238             break;
 239         default:
 240             return VLC_EGENERIC;
 241     }
 242
 243     p_dec->pf_decode_sub = DecodeBlock;
 244     p_dec->fmt_out.i_cat = SPU_ES;
 245     p_dec->fmt_out.i_codec = 0;
 246
 247     /* Allocate the memory needed to store the decoder's structure */
 248     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 249     if( p_sys == NULL )
 250         return VLC_ENOMEM;
 251
 252     /* init of p_sys */
 253     p_sys->i_align = 0;
 254     p_sys->iconv_handle = (vlc_iconv_t)-1;
 255     p_sys->b_autodetect_utf8 = false;
 256     p_sys->b_ass = false;
 257     p_sys->i_original_height = -1;
 258     p_sys->i_original_width = -1;
 259     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 260     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 261
 262     char *psz_charset = NULL;
 263
 264     /* First try demux-specified encoding */
 265     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 266         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 267     else
 268     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 269     {
 270         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 271         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 272                  p_dec->fmt_in.subs.psz_encoding ?
 273                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 274     }
 275
 276     /* Second, try configured encoding */
 277     if (psz_charset == NULL)
 278     {
 279         psz_charset = var_CreateGetNonEmptyString (p_dec, "subsdec-encoding");
 280         msg_Dbg (p_dec, "trying configured character encoding: %s",
 281                  psz_charset ? psz_charset : "not specified");
 282     }
 283
 284     /* Third, try "local" encoding with optional UTF-8 autodetection */
 285     if (psz_charset == NULL)
 286     {
 287         psz_charset = strdup (GetFallbackEncoding ());
 288         msg_Dbg (p_dec, "trying default character encoding: %s",
 289                  psz_charset ? psz_charset : "not specified");
 290
 291         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 292         {
 293             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 294             p_sys->b_autodetect_utf8 = true;
 295         }
 296     }
 297
 298     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 299     if (psz_charset == NULL)
 300     {
 301         psz_charset = strdup ("UTF-8");
 302         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 303     }
 304
 305     if ((psz_charset != NULL)
 306      && strcasecmp (psz_charset, "UTF-8")
 307      && strcasecmp (psz_charset, "utf8"))
 308     {
 309         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 310         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 311             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 312     }
 313     free (psz_charset);
 314
 315     var_Create( p_dec, "subsdec-align", VLC_VAR_INTEGER | VLC_VAR_DOINHERIT );
 316     var_Get( p_dec, "subsdec-align", &val );
 317     p_sys->i_align = val.i_int;
 318
 319     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 320      && var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 321     {
 322         if( p_dec->fmt_in.i_extra > 0 )
 323             ParseSSAHeader( p_dec );
 324     }
 325
 326     return VLC_SUCCESS;
 327 }
 328
 329 /****************************************************************************
 330  * DecodeBlock: the whole thing
 331  ****************************************************************************
 332  * This function must be fed with complete subtitles units.
 333  ****************************************************************************/
 334 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 335 {
 336     subpicture_t *p_spu;
 337     block_t *p_block;
 338
 339     if( !pp_block || *pp_block == NULL )
 340         return NULL;
 341
 342     p_block = *pp_block;
 343     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 344     {
 345         block_Release( p_block );
 346         return NULL;
 347     }
 348
 349     p_spu = ParseText( p_dec, p_block );
 350
 351     block_Release( p_block );
 352     *pp_block = NULL;
 353
 354     return p_spu;
 355 }
 356
 357 /*****************************************************************************
 358  * CloseDecoder: clean up the decoder
 359  *****************************************************************************/
 360 static void CloseDecoder( vlc_object_t *p_this )
 361 {
 362     decoder_t *p_dec = (decoder_t *)p_this;
 363     decoder_sys_t *p_sys = p_dec->p_sys;
 364
 365     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 366         vlc_iconv_close( p_sys->iconv_handle );
 367
 368     if( p_sys->pp_ssa_styles )
 369     {
 370         int i;
 371         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 372         {
 373             if( !p_sys->pp_ssa_styles[i] )
 374                 continue;
 375
 376             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 377             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 378             free( p_sys->pp_ssa_styles[i] );
 379         }
 380         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 381     }
 382     if( p_sys->pp_images )
 383     {
 384         int i;
 385         for( i = 0; i < p_sys->i_images; i++ )
 386         {
 387             if( !p_sys->pp_images[i] )
 388                 continue;
 389
 390             if( p_sys->pp_images[i]->p_pic )
 391                 picture_Release( p_sys->pp_images[i]->p_pic );
 392             free( p_sys->pp_images[i]->psz_filename );
 393
 394             free( p_sys->pp_images[i] );
 395         }
 396         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 397     }
 398
 399     free( p_sys );
 400 }
 401
 402 /*****************************************************************************
 403  * ParseText: parse an text subtitle packet and send it to the video output
 404  *****************************************************************************/
 405 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 406 {
 407     decoder_sys_t *p_sys = p_dec->p_sys;
 408     subpicture_t *p_spu = NULL;
 409     char *psz_subtitle = NULL;
 410     video_format_t fmt;
 411
 412     /* We cannot display a subpicture with no date */
 413     if( p_block->i_pts == 0 )
 414     {
 415         msg_Warn( p_dec, "subtitle without a date" );
 416         return NULL;
 417     }
 418
 419     /* Check validity of packet data */
 420     /* An "empty" line containing only \0 can be used to force
 421        and ephemer picture from the screen */
 422     if( p_block->i_buffer < 1 )
 423     {
 424         msg_Warn( p_dec, "no subtitle data" );
 425         return NULL;
 426     }
 427
 428     /* Should be resiliant against bad subtitles */
 429     psz_subtitle = strndup( (const char *)p_block->p_buffer,
 430                             p_block->i_buffer );
 431     if( psz_subtitle == NULL )
 432         return NULL;
 433
 434     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 435     {
 436         if (EnsureUTF8( psz_subtitle ) == NULL)
 437         {
 438             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 439                      "Try manually setting a character-encoding "
 440                      "before you open the file." );
 441         }
 442     }
 443     else
 444     {
 445
 446         if( p_sys->b_autodetect_utf8 )
 447         {
 448             if( IsUTF8( psz_subtitle ) == NULL )
 449             {
 450                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 451                          "disabling UTF-8 subtitles autodetection" );
 452                 p_sys->b_autodetect_utf8 = false;
 453             }
 454         }
 455
 456         if( !p_sys->b_autodetect_utf8 )
 457         {
 458             size_t inbytes_left = strlen( psz_subtitle );
 459             size_t outbytes_left = 6 * inbytes_left;
 460             char *psz_new_subtitle = malloc( outbytes_left + 1 );
 461             char *psz_convert_buffer_out = psz_new_subtitle;
 462             const char *psz_convert_buffer_in = psz_subtitle;
 463
 464             size_t ret = vlc_iconv( p_sys->iconv_handle,
 465                                     &psz_convert_buffer_in, &inbytes_left,
 466                                     &psz_convert_buffer_out, &outbytes_left );
 467
 468             *psz_convert_buffer_out++ = '\0';
 469             free( psz_subtitle );
 470
 471             if( ( ret == (size_t)(-1) ) || inbytes_left )
 472             {
 473                 free( psz_new_subtitle );
 474                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 475                         "Try manually setting a character-encoding "
 476                                 "before you open the file." );
 477                 return NULL;
 478             }
 479
 480             psz_subtitle = realloc( psz_new_subtitle,
 481                                     psz_convert_buffer_out - psz_new_subtitle );
 482         }
 483     }
 484
 485     /* Create the subpicture unit */
 486     p_spu = decoder_NewSubpicture( p_dec );
 487     if( !p_spu )
 488     {
 489         msg_Warn( p_dec, "can't get spu buffer" );
 490         free( psz_subtitle );
 491         return NULL;
 492     }
 493
 494     /* Create a new subpicture region */
 495     memset( &fmt, 0, sizeof(video_format_t) );
 496     fmt.i_chroma = VLC_CODEC_TEXT;
 497     fmt.i_aspect = 0;
 498     fmt.i_width = fmt.i_height = 0;
 499     fmt.i_x_offset = fmt.i_y_offset = 0;
 500     p_spu->p_region = subpicture_region_New( &fmt );
 501     if( !p_spu->p_region )
 502     {
 503         msg_Err( p_dec, "cannot allocate SPU region" );
 504         free( psz_subtitle );
 505         decoder_DeleteSubpicture( p_dec, p_spu );
 506         return NULL;
 507     }
 508
 509     /* Decode and format the subpicture unit */
 510     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 511     {
 512         /* Normal text subs, easy markup */
 513         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 514         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 515         p_spu->p_region->i_y = 10;
 516
 517         /* Remove formatting from string */
 518
 519         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 520         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 521         {
 522             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 523         }
 524
 525         p_spu->i_start = p_block->i_pts;
 526         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 527         p_spu->b_ephemer = (p_block->i_length == 0);
 528         p_spu->b_absolute = false;
 529     }
 530     else
 531     {
 532         /* Decode SSA/USF strings */
 533         ParseSSAString( p_dec, psz_subtitle, p_spu );
 534
 535         p_spu->i_start = p_block->i_pts;
 536         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 537         p_spu->b_ephemer = (p_block->i_length == 0);
 538         p_spu->b_absolute = false;
 539         p_spu->i_original_picture_width = p_sys->i_original_width;
 540         p_spu->i_original_picture_height = p_sys->i_original_height;
 541     }
 542     free( psz_subtitle );
 543
 544     return p_spu;
 545 }
 546
 547 char* GotoNextLine( char *psz_text )
 548 {
 549     char *p_newline = psz_text;
 550
 551     while( p_newline[0] != '\0' )
 552     {
 553         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 554         {
 555             p_newline++;
 556             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 557                 p_newline++;
 558             break;
 559         }
 560         else p_newline++;
 561     }
 562     return p_newline;
 563 }
 564
 565 /* Function now handles tags with attribute values, and tries
 566  * to deal with &' commands too. It no longer modifies the string
 567  * in place, so that the original text can be reused
 568  */
 569 static char *StripTags( char *psz_subtitle )
 570 {
 571     char *psz_text_start;
 572     char *psz_text;
 573
 574     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 575     if( !psz_text_start )
 576         return NULL;
 577
 578     while( *psz_subtitle )
 579     {
 580         if( *psz_subtitle == '<' )
 581         {
 582             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 583                 *psz_text++ = '\n';
 584
 585             psz_subtitle += strcspn( psz_subtitle, ">" );
 586         }
 587         else if( *psz_subtitle == '&' )
 588         {
 589             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 590             {
 591                 *psz_text++ = '<';
 592                 psz_subtitle += strcspn( psz_subtitle, ";" );
 593             }
 594             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 595             {
 596                 *psz_text++ = '>';
 597                 psz_subtitle += strcspn( psz_subtitle, ";" );
 598             }
 599             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 600             {
 601                 *psz_text++ = '&';
 602                 psz_subtitle += strcspn( psz_subtitle, ";" );
 603             }
 604             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 605             {
 606                 *psz_text++ = '\"';
 607                 psz_subtitle += strcspn( psz_subtitle, ";" );
 608             }
 609             else
 610             {
 611                 /* Assume it is just a normal ampersand */
 612                 *psz_text++ = '&';
 613             }
 614         }
 615         else
 616         {
 617             *psz_text++ = *psz_subtitle;
 618         }
 619
 620         psz_subtitle++;
 621     }
 622     *psz_text = '\0';
 623     psz_text_start = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 624
 625     return psz_text_start;
 626 }
 627
 628 /* Try to respect any style tags present in the subtitle string. The main
 629  * problem here is a lack of adequate specs for the subtitle formats.
 630  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 631  * SAMI has a detailed spec, but extensive rework is needed in the demux
 632  * code to prevent all this style information being excised, as it presently
 633  * does.
 634  * That leaves the others - none of which were (I guess) originally intended
 635  * to be carrying style information. Over time people have used them that way.
 636  * In the absence of specifications from which to work, the tags supported
 637  * have been restricted to the simple set permitted by the USF DTD, ie. :
 638  *  Basic: <br>, <i>, <b>, <u>
 639  *  Extended: <font>
 640  *    Attributes: face
 641  *                family
 642  *                size
 643  *                color
 644  *                outline-color
 645  *                shadow-color
 646  *                outline-level
 647  *                shadow-level
 648  *                back-color
 649  *                alpha
 650  * There is also the further restriction that the subtitle be well-formed
 651  * as an XML entity, ie. the HTML sentence:
 652  *        <b><i>Bold and Italics</b></i>
 653  * doesn't qualify because the tags aren't nested one inside the other.
 654  * <text> tags are automatically added to the output to ensure
 655  * well-formedness.
 656  * If the text doesn't qualify for any reason, a NULL string is
 657  * returned, and the rendering engine will fall back to the
 658  * plain text version of the subtitle.
 659  */
 660 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 661 {
 662     const int i_len = strlen(psz_text);
 663
 664     strncpy( *ppsz_html, psz_text, i_max );
 665     *ppsz_html += __MIN(i_max,i_len);
 666 }
 667
 668 static void HtmlPut( char **ppsz_html, const char *psz_text )
 669 {
 670     strcpy( *ppsz_html, psz_text );
 671     *ppsz_html += strlen(psz_text);
 672 }
 673 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 674 {
 675     HtmlPut( ppsz_html, psz_text );
 676     *ppsz_subtitle += strlen(psz_text);
 677 }
 678
 679 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 680 {
 681     /* */
 682     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 683     if( !psz_tag )
 684         return NULL;
 685     psz_tag[ 0 ] = '\0';
 686
 687     /* */
 688     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 689     char   *psz_html_start = malloc( i_buf_size );
 690     char   *psz_html = psz_html_start;
 691     if( psz_html_start == NULL )
 692     {
 693         free( psz_tag );
 694         return NULL;
 695     }
 696     psz_html[0] = '\0';
 697
 698     bool b_has_align = false;
 699
 700     HtmlPut( &psz_html, "<text>" );
 701
 702     /* */
 703     while( *psz_subtitle )
 704     {
 705         if( *psz_subtitle == '\n' )
 706         {
 707             HtmlPut( &psz_html, "<br/>" );
 708             psz_subtitle++;
 709         }
 710         else if( *psz_subtitle == '<' )
 711         {
 712             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 713             {
 714                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 715             }
 716             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 717             {
 718                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 719                 strcat( psz_tag, "b" );
 720             }
 721             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 722             {
 723                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 724                 strcat( psz_tag, "i" );
 725             }
 726             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 727             {
 728                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 729                 strcat( psz_tag, "u" );
 730             }
 731             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 732             {
 733                 const char *psz_attribs[] = { "face=", "family=", "size=",
 734                         "color=", "outline-color=", "shadow-color=",
 735                         "outline-level=", "shadow-level=", "back-color=",
 736                         "alpha=", NULL };
 737
 738                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 739                 strcat( psz_tag, "f" );
 740
 741                 while( *psz_subtitle != '>' )
 742                 {
 743                     int  k;
 744
 745                     for( k=0; psz_attribs[ k ]; k++ )
 746                     {
 747                         int i_len = strlen( psz_attribs[ k ] );
 748
 749                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 750                         {
 751                             /* */
 752                             HtmlPut( &psz_html, psz_attribs[k] );
 753                             psz_subtitle += i_len;
 754
 755                             /* */
 756                             if( *psz_subtitle == '"' )
 757                             {
 758                                 psz_subtitle++;
 759                                 i_len = strcspn( psz_subtitle, "\"" );
 760                             }
 761                             else
 762                             {
 763                                 i_len = strcspn( psz_subtitle, " \t>" );
 764                             }
 765                             HtmlPut( &psz_html, "\"" );
 766                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 767                                 HtmlPut( &psz_html, "#" );
 768                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 769                             HtmlPut( &psz_html, "\"" );
 770
 771                             psz_subtitle += i_len;
 772                             if( *psz_subtitle == '\"' )
 773                                 psz_subtitle++;
 774                             break;
 775                         }
 776                     }
 777                     if( psz_attribs[ k ] == NULL )
 778                     {
 779                         /* Jump over unrecognised tag */
 780                         int i_len = strcspn( psz_subtitle, "\"" ) + 1;
 781
 782                         i_len += strcspn( psz_subtitle + i_len, "\"" ) + 1;
 783                         psz_subtitle += i_len;
 784                     }
 785                     while (*psz_subtitle == ' ')
 786                         *psz_html++ = *psz_subtitle++;
 787                 }
 788                 *psz_html++ = *psz_subtitle++;
 789             }
 790             else if( !strncmp( psz_subtitle, "</", 2 ))
 791             {
 792                 bool   b_match     = false;
 793                 bool   b_ignore    = false;
 794                 int    i_len       = strlen( psz_tag ) - 1;
 795                 char  *psz_lastTag = NULL;
 796
 797                 if( i_len >= 0 )
 798                 {
 799                     psz_lastTag = psz_tag + i_len;
 800                     i_len = 0;
 801
 802                     switch( *psz_lastTag )
 803                     {
 804                     case 'b':
 805                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 806                         i_len   = 4;
 807                         break;
 808                     case 'i':
 809                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 810                         i_len   = 4;
 811                         break;
 812                     case 'u':
 813                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 814                         i_len   = 4;
 815                         break;
 816                     case 'f':
 817                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 818                         i_len   = 7;
 819                         break;
 820                     case 'I':
 821                         i_len = strcspn( psz_subtitle, ">" );
 822                         b_match = psz_subtitle[i_len] == '>';
 823                         b_ignore = true;
 824                         if( b_match )
 825                             i_len++;
 826                         break;
 827                     }
 828                 }
 829                 if( !b_match )
 830                 {
 831                     /* Not well formed -- kill everything */
 832                     free( psz_html_start );
 833                     psz_html_start = NULL;
 834                     break;
 835                 }
 836                 *psz_lastTag = '\0';
 837                 if( !b_ignore )
 838                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 839
 840                 psz_subtitle += i_len;
 841             }
 842             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 843                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 844             {
 845                 /* We have a single < */
 846                 HtmlPut( &psz_html, "&lt;" );
 847                 psz_subtitle++;
 848             }
 849             else
 850             {
 851                 /* We have an unknown tag or a single < */
 852
 853                 /* Search for the next tag or end of tag or end of string */
 854                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 855                 char *psz_closing = strstr( psz_subtitle, "/>" );
 856
 857                 if( psz_closing && psz_closing < psz_stop )
 858                 {
 859                     /* We have a self closed tag, remove it */
 860                     psz_subtitle = &psz_closing[2];
 861                 }
 862                 else if( *psz_stop == '>' )
 863                 {
 864                     char psz_match[256];
 865
 866                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 867                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 868
 869                     if( strstr( psz_subtitle, psz_match ) )
 870                     {
 871                         /* We have the closing tag, ignore it TODO */
 872                         psz_subtitle = &psz_stop[1];
 873                         strcat( psz_tag, "I" );
 874                     }
 875                     else
 876                     {
 877                         int i_len = psz_stop + 1 - psz_subtitle;
 878
 879                         /* Copy the whole data */
 880                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 881                         {
 882                             if( *psz_subtitle == '<' )
 883                                 HtmlPut( &psz_html, "&lt;" );
 884                             else if( *psz_subtitle == '>' )
 885                                 HtmlPut( &psz_html, "&gt;" );
 886                             else
 887                                 *psz_html++ = *psz_subtitle;
 888                         }
 889                     }
 890                 }
 891                 else
 892                 {
 893                     /* We have a single < */
 894                     HtmlPut( &psz_html, "&lt;" );
 895                     psz_subtitle++;
 896                 }
 897             }
 898         }
 899         else if( *psz_subtitle == '&' )
 900         {
 901             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 902             {
 903                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 904             }
 905             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 906             {
 907                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 908             }
 909             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 910             {
 911                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 912             }
 913             else
 914             {
 915                 HtmlPut( &psz_html, "&amp;" );
 916                 psz_subtitle++;
 917             }
 918         }
 919         else if( *psz_subtitle == '>' )
 920         {
 921             HtmlPut( &psz_html, "&gt;" );
 922             psz_subtitle++;
 923         }
 924         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 925                  strchr( psz_subtitle, '}' ) )
 926         {
 927             /* Check for forced alignment */
 928             if( !b_has_align &&
 929                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 930             {
 931                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 932                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 933                 const int i_id = psz_subtitle[4] - '1';
 934
 935                 b_has_align = true;
 936                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 937             }
 938             /* TODO fr -> rotation */
 939
 940             /* Hide {\stupidity} */
 941             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 942         }
 943         else
 944         {
 945             *psz_html = *psz_subtitle;
 946             if( psz_html > psz_html_start )
 947             {
 948                 /* Check for double whitespace */
 949                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
 950                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
 951                 {
 952                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
 953                     psz_html--;
 954                 }
 955             }
 956             psz_html++;
 957             psz_subtitle++;
 958         }
 959
 960         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
 961         {
 962             const int i_len = psz_html - psz_html_start;
 963
 964             i_buf_size += 200;
 965             char *psz_new = realloc( psz_html_start, i_buf_size );
 966             if( !psz_new )
 967                 break;
 968             psz_html_start = psz_new;
 969             psz_html = &psz_new[i_len];
 970         }
 971     }
 972     if( psz_html_start )
 973     {
 974         static const char *psz_text_close = "</text>";
 975         static const char *psz_tag_long = "/font>";
 976
 977         /* Realloc for closing tags and shrink memory */
 978         const size_t i_length = (size_t)( psz_html - psz_html_start );
 979
 980         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
 981         char *psz_new = realloc( psz_html_start, i_size );
 982         if( psz_new )
 983         {
 984             psz_html_start = psz_new;
 985             psz_html = &psz_new[i_length];
 986
 987             /* Close not well formed subtitle */
 988             while( *psz_tag )
 989             {
 990                 /* */
 991                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
 992                 switch( *psz_last )
 993                 {
 994                 case 'b':
 995                     HtmlPut( &psz_html, "</b>" );
 996                     break;
 997                 case 'i':
 998                     HtmlPut( &psz_html, "</i>" );
 999                     break;
1000                 case 'u':
1001                     HtmlPut( &psz_html, "</u>" );
1002                     break;
1003                 case 'f':
1004                     HtmlPut( &psz_html, "/font>" );
1005                     break;
1006                 case 'I':
1007                     break;
1008                 }
1009
1010                 *psz_last = '\0';
1011             }
1012             HtmlPut( &psz_html, psz_text_close );
1013         }
1014     }
1015     free( psz_tag );
1016
1017     return psz_html_start;
1018 }
1019