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     /* xgettext:
  98       The character encoding name in parenthesis corresponds to that used for
  99       the GetACP translation. "Windows-1252" applies to Western European
 100       languages using the Latin alphabet. */
 101     N_("Default (Windows-1252)"),
 102     N_("Universal (UTF-8)"),
 103     N_("Universal (UTF-16)"),
 104     N_("Universal (big endian UTF-16)"),
 105     N_("Universal (little endian UTF-16)"),
 106     N_("Universal, Chinese (GB18030)"),
 107
 108   /* ISO 8859 and the likes */
 109     /* 1 */
 110     N_("Western European (Latin-9)"), /* mostly superset of Latin-1 */
 111     N_("Western European (Windows-1252)"),
 112     /* 2 */
 113     N_("Eastern European (Latin-2)"),
 114     N_("Eastern European (Windows-1250)"),
 115     /* 3 */
 116     N_("Esperanto (Latin-3)"),
 117     /* 4 */
 118     N_("Nordic (Latin-6)"), /* Latin 6 supersedes Latin 4 */
 119     /* 5 */
 120     N_("Cyrillic (Windows-1251)"), /* ISO 8859-5 is not practically used */
 121     N_("Russian (KOI8-R)"),
 122     N_("Ukrainian (KOI8-U)"),
 123     /* 6 */
 124     N_("Arabic (ISO 8859-6)"),
 125     N_("Arabic (Windows-1256)"),
 126     /* 7 */
 127     N_("Greek (ISO 8859-7)"),
 128     N_("Greek (Windows-1253)"),
 129     /* 8 */
 130     N_("Hebrew (ISO 8859-8)"),
 131     N_("Hebrew (Windows-1255)"),
 132     /* 9 */
 133     N_("Turkish (ISO 8859-9)"),
 134     N_("Turkish (Windows-1254)"),
 135     /* 10 -> 4 */
 136     /* 11 */
 137     N_("Thai (TIS 620-2533/ISO 8859-11)"),
 138     N_("Thai (Windows-874)"),
 139     /* 13 */
 140     N_("Baltic (Latin-7)"),
 141     N_("Baltic (Windows-1257)"),
 142     /* 12 -> /dev/null */
 143     /* 14 */
 144     N_("Celtic (Latin-8)"),
 145     /* 15 -> 1 */
 146     /* 16 */
 147     N_("South-Eastern European (Latin-10)"),
 148   /* CJK families */
 149     N_("Simplified Chinese (ISO-2022-CN-EXT)"),
 150     N_("Simplified Chinese Unix (EUC-CN)"),
 151     N_("Japanese (7-bits JIS/ISO-2022-JP-2)"),
 152     N_("Japanese Unix (EUC-JP)"),
 153     N_("Japanese (Shift JIS)"),
 154     N_("Korean (EUC-KR/CP949)"),
 155     N_("Korean (ISO-2022-KR)"),
 156     N_("Traditional Chinese (Big5)"),
 157     N_("Traditional Chinese Unix (EUC-TW)"),
 158     N_("Hong-Kong Supplementary (HKSCS)"),
 159   /* Other */
 160     N_("Vietnamese (VISCII)"),
 161     N_("Vietnamese (Windows-1258)"),
 162 };
 163 /*
 164 SSA supports charset selection.
 165 The following known charsets are used:
 166
 167 0 = Ansi - Western European
 168 1 = default
 169 2 = symbol
 170 3 = invalid
 171 77 = Mac
 172 128 = Japanese (Shift JIS)
 173 129 = Hangul
 174 130 = Johab
 175 134 = GB2312 Simplified Chinese
 176 136 = Big5 Traditional Chinese
 177 161 = Greek
 178 162 = Turkish
 179 163 = Vietnamese
 180 177 = Hebrew
 181 178 = Arabic
 182 186 = Baltic
 183 204 = Russian (Cyrillic)
 184 222 = Thai
 185 238 = Eastern European
 186 254 = PC 437
 187 */
 188
 189 static const int  pi_justification[] = { 0, 1, 2 };
 190 static const char *const ppsz_justification_text[] = {
 191     N_("Center"),N_("Left"),N_("Right")};
 192
 193 #define ENCODING_TEXT N_("Subtitles text encoding")
 194 #define ENCODING_LONGTEXT N_("Set the encoding used in text subtitles")
 195 #define ALIGN_TEXT N_("Subtitles justification")
 196 #define ALIGN_LONGTEXT N_("Set the justification of subtitles")
 197 #define AUTODETECT_UTF8_TEXT N_("UTF-8 subtitles autodetection")
 198 #define AUTODETECT_UTF8_LONGTEXT N_("This enables automatic detection of " \
 199             "UTF-8 encoding within subtitles files.")
 200 #define FORMAT_TEXT N_("Formatted Subtitles")
 201 #define FORMAT_LONGTEXT N_("Some subtitle formats allow for text formatting. " \
 202  "VLC partly implements this, but you can choose to disable all formatting.")
 203
 204
 205 vlc_module_begin ()
 206     set_shortname( N_("Subtitles"))
 207     set_description( N_("Text subtitles decoder") )
 208     set_capability( "decoder", 50 )
 209     set_callbacks( OpenDecoder, CloseDecoder )
 210     set_category( CAT_INPUT )
 211     set_subcategory( SUBCAT_INPUT_SCODEC )
 212
 213     add_integer( "subsdec-align", 0, NULL, ALIGN_TEXT, ALIGN_LONGTEXT,
 214                  false )
 215         change_integer_list( pi_justification, ppsz_justification_text, NULL )
 216     add_string( "subsdec-encoding", "", NULL,
 217                 ENCODING_TEXT, ENCODING_LONGTEXT, false )
 218         change_string_list( ppsz_encodings, ppsz_encoding_names, 0 )
 219     add_bool( "subsdec-autodetect-utf8", true, NULL,
 220               AUTODETECT_UTF8_TEXT, AUTODETECT_UTF8_LONGTEXT, false )
 221     add_bool( "subsdec-formatted", true, NULL, FORMAT_TEXT, FORMAT_LONGTEXT,
 222                  false )
 223 vlc_module_end ()
 224
 225 /*****************************************************************************
 226  * OpenDecoder: probe the decoder and return score
 227  *****************************************************************************
 228  * Tries to launch a decoder and return score so that the interface is able
 229  * to chose.
 230  *****************************************************************************/
 231 static int OpenDecoder( vlc_object_t *p_this )
 232 {
 233     decoder_t     *p_dec = (decoder_t*)p_this;
 234     decoder_sys_t *p_sys;
 235
 236     switch( p_dec->fmt_in.i_codec )
 237     {
 238         case VLC_CODEC_SUBT:
 239         case VLC_CODEC_SSA:
 240         case VLC_CODEC_ITU_T140:
 241             break;
 242         default:
 243             return VLC_EGENERIC;
 244     }
 245
 246     p_dec->pf_decode_sub = DecodeBlock;
 247     p_dec->fmt_out.i_cat = SPU_ES;
 248     p_dec->fmt_out.i_codec = 0;
 249
 250     /* Allocate the memory needed to store the decoder's structure */
 251     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 252     if( p_sys == NULL )
 253         return VLC_ENOMEM;
 254
 255     /* init of p_sys */
 256     p_sys->i_align = 0;
 257     p_sys->iconv_handle = (vlc_iconv_t)-1;
 258     p_sys->b_autodetect_utf8 = false;
 259     p_sys->b_ass = false;
 260     p_sys->i_original_height = -1;
 261     p_sys->i_original_width = -1;
 262     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 263     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 264
 265     char *psz_charset = NULL;
 266
 267     /* First try demux-specified encoding */
 268     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 269         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 270     else
 271     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 272     {
 273         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 274         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 275                  p_dec->fmt_in.subs.psz_encoding ?
 276                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 277     }
 278
 279     /* Second, try configured encoding */
 280     if (psz_charset == NULL)
 281     {
 282         psz_charset = var_CreateGetNonEmptyString (p_dec, "subsdec-encoding");
 283         msg_Dbg (p_dec, "trying configured character encoding: %s",
 284                  psz_charset ? psz_charset : "not specified");
 285     }
 286
 287     /* Third, try "local" encoding with optional UTF-8 autodetection */
 288     if (psz_charset == NULL)
 289     {
 290         /* xgettext:
 291            The Windows ANSI code page most commonly used for this language.
 292            VLC uses this as a guess of the subtitle files character set
 293            (if UTF-8 and UTF-16 autodetection fails).
 294            Western European languages normally use "CP1252", which is a
 295            Microsoft-variant of ISO 8859-1. That suits the Latin alphabet.
 296            Other scripts use other code pages.
 297
 298            This MUST be a valid iconv character set. If unsure, please refer
 299            the VideoLAN translators mailing list. */
 300         const char *acp = vlc_pgettext("GetACP", "CP1252");
 301
 302         psz_charset = strdup (acp);
 303         msg_Dbg (p_dec, "trying default character encoding: %s",
 304                  psz_charset ? psz_charset : "not specified");
 305
 306         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 307         {
 308             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 309             p_sys->b_autodetect_utf8 = true;
 310         }
 311     }
 312
 313     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 314     if (psz_charset == NULL)
 315     {
 316         psz_charset = strdup ("UTF-8");
 317         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 318     }
 319
 320     if ((psz_charset != NULL)
 321      && strcasecmp (psz_charset, "UTF-8")
 322      && strcasecmp (psz_charset, "utf8"))
 323     {
 324         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 325         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 326             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 327     }
 328     free (psz_charset);
 329
 330     p_sys->i_align = var_CreateGetInteger( p_dec, "subsdec-align" );
 331
 332     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 333      && var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 334     {
 335         if( p_dec->fmt_in.i_extra > 0 )
 336             ParseSSAHeader( p_dec );
 337     }
 338
 339     return VLC_SUCCESS;
 340 }
 341
 342 /****************************************************************************
 343  * DecodeBlock: the whole thing
 344  ****************************************************************************
 345  * This function must be fed with complete subtitles units.
 346  ****************************************************************************/
 347 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 348 {
 349     subpicture_t *p_spu;
 350     block_t *p_block;
 351
 352     if( !pp_block || *pp_block == NULL )
 353         return NULL;
 354
 355     p_block = *pp_block;
 356     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 357     {
 358         block_Release( p_block );
 359         return NULL;
 360     }
 361
 362     p_spu = ParseText( p_dec, p_block );
 363
 364     block_Release( p_block );
 365     *pp_block = NULL;
 366
 367     return p_spu;
 368 }
 369
 370 /*****************************************************************************
 371  * CloseDecoder: clean up the decoder
 372  *****************************************************************************/
 373 static void CloseDecoder( vlc_object_t *p_this )
 374 {
 375     decoder_t *p_dec = (decoder_t *)p_this;
 376     decoder_sys_t *p_sys = p_dec->p_sys;
 377
 378     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 379         vlc_iconv_close( p_sys->iconv_handle );
 380
 381     if( p_sys->pp_ssa_styles )
 382     {
 383         int i;
 384         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 385         {
 386             if( !p_sys->pp_ssa_styles[i] )
 387                 continue;
 388
 389             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 390             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 391             free( p_sys->pp_ssa_styles[i] );
 392         }
 393         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 394     }
 395     if( p_sys->pp_images )
 396     {
 397         int i;
 398         for( i = 0; i < p_sys->i_images; i++ )
 399         {
 400             if( !p_sys->pp_images[i] )
 401                 continue;
 402
 403             if( p_sys->pp_images[i]->p_pic )
 404                 picture_Release( p_sys->pp_images[i]->p_pic );
 405             free( p_sys->pp_images[i]->psz_filename );
 406
 407             free( p_sys->pp_images[i] );
 408         }
 409         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 410     }
 411
 412     free( p_sys );
 413 }
 414
 415 /*****************************************************************************
 416  * ParseText: parse an text subtitle packet and send it to the video output
 417  *****************************************************************************/
 418 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 419 {
 420     decoder_sys_t *p_sys = p_dec->p_sys;
 421     subpicture_t *p_spu = NULL;
 422     char *psz_subtitle = NULL;
 423     video_format_t fmt;
 424
 425     /* We cannot display a subpicture with no date */
 426     if( p_block->i_pts <= VLC_TS_INVALID )
 427     {
 428         msg_Warn( p_dec, "subtitle without a date" );
 429         return NULL;
 430     }
 431
 432     /* Check validity of packet data */
 433     /* An "empty" line containing only \0 can be used to force
 434        and ephemer picture from the screen */
 435     if( p_block->i_buffer < 1 )
 436     {
 437         msg_Warn( p_dec, "no subtitle data" );
 438         return NULL;
 439     }
 440
 441     /* Should be resiliant against bad subtitles */
 442     psz_subtitle = malloc( p_block->i_buffer + 1 );
 443     if( psz_subtitle == NULL )
 444         return NULL;
 445     memcpy( psz_subtitle, p_block->p_buffer, p_block->i_buffer );
 446     psz_subtitle[p_block->i_buffer] = '\0';
 447
 448     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 449     {
 450         if (EnsureUTF8( psz_subtitle ) == NULL)
 451         {
 452             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 453                      "Try manually setting a character-encoding "
 454                      "before you open the file." );
 455         }
 456     }
 457     else
 458     {
 459
 460         if( p_sys->b_autodetect_utf8 )
 461         {
 462             if( IsUTF8( psz_subtitle ) == NULL )
 463             {
 464                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 465                          "disabling UTF-8 subtitles autodetection" );
 466                 p_sys->b_autodetect_utf8 = false;
 467             }
 468         }
 469
 470         if( !p_sys->b_autodetect_utf8 )
 471         {
 472             size_t inbytes_left = strlen( psz_subtitle );
 473             size_t outbytes_left = 6 * inbytes_left;
 474             char *psz_new_subtitle = xmalloc( outbytes_left + 1 );
 475             char *psz_convert_buffer_out = psz_new_subtitle;
 476             const char *psz_convert_buffer_in = psz_subtitle;
 477
 478             size_t ret = vlc_iconv( p_sys->iconv_handle,
 479                                     &psz_convert_buffer_in, &inbytes_left,
 480                                     &psz_convert_buffer_out, &outbytes_left );
 481
 482             *psz_convert_buffer_out++ = '\0';
 483             free( psz_subtitle );
 484
 485             if( ( ret == (size_t)(-1) ) || inbytes_left )
 486             {
 487                 free( psz_new_subtitle );
 488                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 489                         "Try manually setting a character-encoding "
 490                                 "before you open the file." );
 491                 return NULL;
 492             }
 493
 494             psz_subtitle = realloc( psz_new_subtitle,
 495                                     psz_convert_buffer_out - psz_new_subtitle );
 496             if( !psz_subtitle )
 497                 psz_subtitle = psz_new_subtitle;
 498         }
 499     }
 500
 501     /* Create the subpicture unit */
 502     p_spu = decoder_NewSubpicture( p_dec, NULL );
 503     if( !p_spu )
 504     {
 505         msg_Warn( p_dec, "can't get spu buffer" );
 506         free( psz_subtitle );
 507         return NULL;
 508     }
 509
 510     /* Create a new subpicture region */
 511     memset( &fmt, 0, sizeof(video_format_t) );
 512     fmt.i_chroma = VLC_CODEC_TEXT;
 513     fmt.i_width = fmt.i_height = 0;
 514     fmt.i_x_offset = fmt.i_y_offset = 0;
 515     p_spu->p_region = subpicture_region_New( &fmt );
 516     if( !p_spu->p_region )
 517     {
 518         msg_Err( p_dec, "cannot allocate SPU region" );
 519         free( psz_subtitle );
 520         decoder_DeleteSubpicture( p_dec, p_spu );
 521         return NULL;
 522     }
 523
 524     /* Decode and format the subpicture unit */
 525     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 526     {
 527         /* Normal text subs, easy markup */
 528         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 529         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 530         p_spu->p_region->i_y = 10;
 531
 532         /* Remove formatting from string */
 533
 534         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 535         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 536         {
 537             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 538         }
 539
 540         p_spu->i_start = p_block->i_pts;
 541         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 542         p_spu->b_ephemer = (p_block->i_length == 0);
 543         p_spu->b_absolute = false;
 544     }
 545     else
 546     {
 547         /* Decode SSA/USF strings */
 548         ParseSSAString( p_dec, psz_subtitle, p_spu );
 549
 550         p_spu->i_start = p_block->i_pts;
 551         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 552         p_spu->b_ephemer = (p_block->i_length == 0);
 553         p_spu->b_absolute = false;
 554         p_spu->i_original_picture_width = p_sys->i_original_width;
 555         p_spu->i_original_picture_height = p_sys->i_original_height;
 556     }
 557     free( psz_subtitle );
 558
 559     return p_spu;
 560 }
 561
 562 char* GotoNextLine( char *psz_text )
 563 {
 564     char *p_newline = psz_text;
 565
 566     while( p_newline[0] != '\0' )
 567     {
 568         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 569         {
 570             p_newline++;
 571             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 572                 p_newline++;
 573             break;
 574         }
 575         else p_newline++;
 576     }
 577     return p_newline;
 578 }
 579
 580 /* Function now handles tags with attribute values, and tries
 581  * to deal with &' commands too. It no longer modifies the string
 582  * in place, so that the original text can be reused
 583  */
 584 static char *StripTags( char *psz_subtitle )
 585 {
 586     char *psz_text_start;
 587     char *psz_text;
 588
 589     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 590     if( !psz_text_start )
 591         return NULL;
 592
 593     while( *psz_subtitle )
 594     {
 595         if( *psz_subtitle == '<' )
 596         {
 597             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 598                 *psz_text++ = '\n';
 599
 600             psz_subtitle += strcspn( psz_subtitle, ">" );
 601         }
 602         else if( *psz_subtitle == '&' )
 603         {
 604             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 605             {
 606                 *psz_text++ = '<';
 607                 psz_subtitle += strcspn( psz_subtitle, ";" );
 608             }
 609             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 610             {
 611                 *psz_text++ = '>';
 612                 psz_subtitle += strcspn( psz_subtitle, ";" );
 613             }
 614             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 615             {
 616                 *psz_text++ = '&';
 617                 psz_subtitle += strcspn( psz_subtitle, ";" );
 618             }
 619             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 620             {
 621                 *psz_text++ = '\"';
 622                 psz_subtitle += strcspn( psz_subtitle, ";" );
 623             }
 624             else
 625             {
 626                 /* Assume it is just a normal ampersand */
 627                 *psz_text++ = '&';
 628             }
 629         }
 630         else
 631         {
 632             *psz_text++ = *psz_subtitle;
 633         }
 634
 635         psz_subtitle++;
 636     }
 637     *psz_text = '\0';
 638     char *psz = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 639     if( psz ) psz_text_start = psz;
 640
 641     return psz_text_start;
 642 }
 643
 644 /* Try to respect any style tags present in the subtitle string. The main
 645  * problem here is a lack of adequate specs for the subtitle formats.
 646  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 647  * SAMI has a detailed spec, but extensive rework is needed in the demux
 648  * code to prevent all this style information being excised, as it presently
 649  * does.
 650  * That leaves the others - none of which were (I guess) originally intended
 651  * to be carrying style information. Over time people have used them that way.
 652  * In the absence of specifications from which to work, the tags supported
 653  * have been restricted to the simple set permitted by the USF DTD, ie. :
 654  *  Basic: <br>, <i>, <b>, <u>, <s>
 655  *  Extended: <font>
 656  *    Attributes: face
 657  *                family
 658  *                size
 659  *                color
 660  *                outline-color
 661  *                shadow-color
 662  *                outline-level
 663  *                shadow-level
 664  *                back-color
 665  *                alpha
 666  * There is also the further restriction that the subtitle be well-formed
 667  * as an XML entity, ie. the HTML sentence:
 668  *        <b><i>Bold and Italics</b></i>
 669  * doesn't qualify because the tags aren't nested one inside the other.
 670  * <text> tags are automatically added to the output to ensure
 671  * well-formedness.
 672  * If the text doesn't qualify for any reason, a NULL string is
 673  * returned, and the rendering engine will fall back to the
 674  * plain text version of the subtitle.
 675  */
 676 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 677 {
 678     const int i_len = strlen(psz_text);
 679
 680     strncpy( *ppsz_html, psz_text, i_max );
 681     *ppsz_html += __MIN(i_max,i_len);
 682 }
 683
 684 static void HtmlPut( char **ppsz_html, const char *psz_text )
 685 {
 686     strcpy( *ppsz_html, psz_text );
 687     *ppsz_html += strlen(psz_text);
 688 }
 689 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 690 {
 691     HtmlPut( ppsz_html, psz_text );
 692     *ppsz_subtitle += strlen(psz_text);
 693 }
 694
 695 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 696 {
 697     /* */
 698     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 699     if( !psz_tag )
 700         return NULL;
 701     psz_tag[ 0 ] = '\0';
 702
 703     /* */
 704     //Oo + 100 ???
 705     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 706     char   *psz_html_start = malloc( i_buf_size );
 707     char   *psz_html = psz_html_start;
 708     if( psz_html_start == NULL )
 709     {
 710         free( psz_tag );
 711         return NULL;
 712     }
 713     psz_html[0] = '\0';
 714
 715     bool b_has_align = false;
 716
 717     HtmlPut( &psz_html, "<text>" );
 718
 719     /* */
 720     while( *psz_subtitle )
 721     {
 722         if( *psz_subtitle == '\n' )
 723         {
 724             HtmlPut( &psz_html, "<br/>" );
 725             psz_subtitle++;
 726         }
 727         else if( *psz_subtitle == '<' )
 728         {
 729             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 730             {
 731                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 732             }
 733             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 734             {
 735                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 736                 strcat( psz_tag, "b" );
 737             }
 738             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 739             {
 740                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 741                 strcat( psz_tag, "i" );
 742             }
 743             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 744             {
 745                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 746                 strcat( psz_tag, "u" );
 747             }
 748             else if( !strncasecmp( psz_subtitle, "<s>", 3 ) )
 749             {
 750                 HtmlCopy( &psz_html, &psz_subtitle, "<s>" );
 751                 strcat( psz_tag, "s" );
 752             }
 753             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 754             {
 755                 const char *psz_attribs[] = { "face=", "family=", "size=",
 756                         "color=", "outline-color=", "shadow-color=",
 757                         "outline-level=", "shadow-level=", "back-color=",
 758                         "alpha=", NULL };
 759
 760                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 761                 strcat( psz_tag, "f" );
 762
 763                 while( *psz_subtitle != '>' )
 764                 {
 765                     int  k;
 766
 767                     for( k=0; psz_attribs[ k ]; k++ )
 768                     {
 769                         int i_len = strlen( psz_attribs[ k ] );
 770
 771                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 772                         {
 773                             /* */
 774                             HtmlPut( &psz_html, psz_attribs[k] );
 775                             psz_subtitle += i_len;
 776
 777                             /* */
 778                             if( *psz_subtitle == '"' )
 779                             {
 780                                 psz_subtitle++;
 781                                 i_len = strcspn( psz_subtitle, "\"" );
 782                             }
 783                             else
 784                             {
 785                                 i_len = strcspn( psz_subtitle, " \t>" );
 786                             }
 787                             HtmlPut( &psz_html, "\"" );
 788                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 789                                 HtmlPut( &psz_html, "#" );
 790                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 791                             HtmlPut( &psz_html, "\"" );
 792
 793                             psz_subtitle += i_len;
 794                             if( *psz_subtitle == '\"' )
 795                                 psz_subtitle++;
 796                             break;
 797                         }
 798                     }
 799                     if( psz_attribs[ k ] == NULL )
 800                     {
 801                         /* Jump over unrecognised tag */
 802                         int i_len = strcspn( psz_subtitle, "\"" );
 803                         if( psz_subtitle[i_len] == '\"' )
 804                         {
 805                             i_len += 1 + strcspn( &psz_subtitle[i_len + 1], "\"" );
 806                             if( psz_subtitle[i_len] == '\"' )
 807                                 i_len++;
 808                         }
 809                         psz_subtitle += i_len;
 810                     }
 811                     while (*psz_subtitle == ' ')
 812                         *psz_html++ = *psz_subtitle++;
 813                 }
 814                 *psz_html++ = *psz_subtitle++;
 815             }
 816             else if( !strncmp( psz_subtitle, "</", 2 ))
 817             {
 818                 bool   b_match     = false;
 819                 bool   b_ignore    = false;
 820                 int    i_len       = strlen( psz_tag ) - 1;
 821                 char  *psz_lastTag = NULL;
 822
 823                 if( i_len >= 0 )
 824                 {
 825                     psz_lastTag = psz_tag + i_len;
 826                     i_len = 0;
 827
 828                     switch( *psz_lastTag )
 829                     {
 830                     case 'b':
 831                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 832                         i_len   = 4;
 833                         break;
 834                     case 'i':
 835                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 836                         i_len   = 4;
 837                         break;
 838                     case 'u':
 839                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 840                         i_len   = 4;
 841                         break;
 842                     case 's':
 843                         b_match = !strncasecmp( psz_subtitle, "</s>", 4 );
 844                         i_len   = 4;
 845                         break;
 846                     case 'f':
 847                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 848                         i_len   = 7;
 849                         break;
 850                     case 'I':
 851                         i_len = strcspn( psz_subtitle, ">" );
 852                         b_match = psz_subtitle[i_len] == '>';
 853                         b_ignore = true;
 854                         if( b_match )
 855                             i_len++;
 856                         break;
 857                     }
 858                 }
 859                 if( !b_match )
 860                 {
 861                     /* Not well formed -- kill everything */
 862                     free( psz_html_start );
 863                     psz_html_start = NULL;
 864                     break;
 865                 }
 866                 *psz_lastTag = '\0';
 867                 if( !b_ignore )
 868                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 869
 870                 psz_subtitle += i_len;
 871             }
 872             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 873                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 874             {
 875                 /* We have a single < */
 876                 HtmlPut( &psz_html, "&lt;" );
 877                 psz_subtitle++;
 878             }
 879             else
 880             {
 881                 /* We have an unknown tag or a single < */
 882
 883                 /* Search for the next tag or end of tag or end of string */
 884                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 885                 char *psz_closing = strstr( psz_subtitle, "/>" );
 886
 887                 if( psz_closing && psz_closing < psz_stop )
 888                 {
 889                     /* We have a self closed tag, remove it */
 890                     psz_subtitle = &psz_closing[2];
 891                 }
 892                 else if( *psz_stop == '>' )
 893                 {
 894                     char psz_match[256];
 895
 896                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 897                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 898
 899                     if( strstr( psz_subtitle, psz_match ) )
 900                     {
 901                         /* We have the closing tag, ignore it TODO */
 902                         psz_subtitle = &psz_stop[1];
 903                         strcat( psz_tag, "I" );
 904                     }
 905                     else
 906                     {
 907                         int i_len = psz_stop + 1 - psz_subtitle;
 908
 909                         /* Copy the whole data */
 910                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 911                         {
 912                             if( *psz_subtitle == '<' )
 913                                 HtmlPut( &psz_html, "&lt;" );
 914                             else if( *psz_subtitle == '>' )
 915                                 HtmlPut( &psz_html, "&gt;" );
 916                             else
 917                                 *psz_html++ = *psz_subtitle;
 918                         }
 919                     }
 920                 }
 921                 else
 922                 {
 923                     /* We have a single < */
 924                     HtmlPut( &psz_html, "&lt;" );
 925                     psz_subtitle++;
 926                 }
 927             }
 928         }
 929         else if( *psz_subtitle == '&' )
 930         {
 931             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 932             {
 933                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 934             }
 935             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 936             {
 937                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 938             }
 939             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 940             {
 941                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 942             }
 943             else
 944             {
 945                 HtmlPut( &psz_html, "&amp;" );
 946                 psz_subtitle++;
 947             }
 948         }
 949         else if( *psz_subtitle == '>' )
 950         {
 951             HtmlPut( &psz_html, "&gt;" );
 952             psz_subtitle++;
 953         }
 954         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 955                  strchr( psz_subtitle, '}' ) )
 956         {
 957             /* Check for forced alignment */
 958             if( !b_has_align &&
 959                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 960             {
 961                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 962                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 963                 const int i_id = psz_subtitle[4] - '1';
 964
 965                 b_has_align = true;
 966                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 967             }
 968             /* TODO fr -> rotation */
 969
 970             /* Hide {\stupidity} */
 971             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 972         }
 973         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == 'Y'
 974                 && psz_subtitle[2] == ':' && strchr( psz_subtitle, '}' ) )
 975         {
 976             /* Hide {Y:stupidity} */
 977             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 978         }
 979         else if( psz_subtitle[0] == '\\' && psz_subtitle[1] )
 980         {
 981             if( psz_subtitle[1] == 'N' || psz_subtitle[1] == 'n' )
 982             {
 983                 HtmlPut( &psz_html, "<br/>" );
 984                 psz_subtitle += 2;
 985             }
 986             else if( psz_subtitle[1] == 'h' )
 987             {
 988                 /* Non breakable space */
 989                 HtmlPut( &psz_html, NO_BREAKING_SPACE );
 990                 psz_subtitle += 2;
 991             }
 992             else
 993             {
 994                 HtmlPut( &psz_html, "\\" );
 995                 psz_subtitle++;
 996             }
 997         }
 998         else
 999         {
1000             *psz_html = *psz_subtitle;
1001             if( psz_html > psz_html_start )
1002             {
1003                 /* Check for double whitespace */
1004                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
1005                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
1006                 {
1007                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
1008                     psz_html--;
1009                 }
1010             }
1011             psz_html++;
1012             psz_subtitle++;
1013         }
1014
1015         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
1016         {
1017             const int i_len = psz_html - psz_html_start;
1018
1019             i_buf_size += 200;
1020             char *psz_new = realloc( psz_html_start, i_buf_size );
1021             if( !psz_new )
1022                 break;
1023             psz_html_start = psz_new;
1024             psz_html = &psz_new[i_len];
1025         }
1026     }
1027     if( psz_html_start )
1028     {
1029         static const char *psz_text_close = "</text>";
1030         static const char *psz_tag_long = "/font>";
1031
1032         /* Realloc for closing tags and shrink memory */
1033         const size_t i_length = (size_t)( psz_html - psz_html_start );
1034
1035         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
1036         char *psz_new = realloc( psz_html_start, i_size );
1037         if( psz_new )
1038         {
1039             psz_html_start = psz_new;
1040             psz_html = &psz_new[i_length];
1041
1042             /* Close not well formed subtitle */
1043             while( *psz_tag )
1044             {
1045                 /* */
1046                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
1047                 switch( *psz_last )
1048                 {
1049                 case 'b':
1050                     HtmlPut( &psz_html, "</b>" );
1051                     break;
1052                 case 'i':
1053                     HtmlPut( &psz_html, "</i>" );
1054                     break;
1055                 case 'u':
1056                     HtmlPut( &psz_html, "</u>" );
1057                     break;
1058                 case 's':
1059                     HtmlPut( &psz_html, "</s>" );
1060                     break;
1061                 case 'f':
1062                     HtmlPut( &psz_html, "/font>" );
1063                     break;
1064                 case 'I':
1065                     break;
1066                 }
1067
1068                 *psz_last = '\0';
1069             }
1070             HtmlPut( &psz_html, psz_text_close );
1071         }
1072     }
1073     free( psz_tag );
1074
1075     return psz_html_start;
1076 }
1077