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