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 <assert.h>
  35
  36 #include "subsdec.h"
  37 #include <vlc_plugin.h>
  38
  39 /*****************************************************************************
  40  * Local prototypes
  41  *****************************************************************************/
  42 static int  OpenDecoder   ( vlc_object_t * );
  43 static void CloseDecoder  ( vlc_object_t * );
  44
  45 static subpicture_t   *DecodeBlock   ( decoder_t *, block_t ** );
  46 static subpicture_t   *ParseText     ( decoder_t *, block_t * );
  47 static char           *StripTags      ( char * );
  48 static char           *CreateHtmlSubtitle( int *pi_align, char * );
  49
  50
  51 /*****************************************************************************
  52  * Module descriptor.
  53  *****************************************************************************/
  54 static const char *const ppsz_encodings[] = {
  55     "",
  56     "UTF-8",
  57     "UTF-16",
  58     "UTF-16BE",
  59     "UTF-16LE",
  60     "GB18030",
  61     "ISO-8859-15",
  62     "Windows-1252",
  63     "ISO-8859-2",
  64     "Windows-1250",
  65     "ISO-8859-3",
  66     "ISO-8859-10",
  67     "Windows-1251",
  68     "KOI8-R",
  69     "KOI8-U",
  70     "ISO-8859-6",
  71     "Windows-1256",
  72     "ISO-8859-7",
  73     "Windows-1253",
  74     "ISO-8859-8",
  75     "Windows-1255",
  76     "ISO-8859-9",
  77     "Windows-1254",
  78     "ISO-8859-11",
  79     "Windows-874",
  80     "ISO-8859-13",
  81     "Windows-1257",
  82     "ISO-8859-14",
  83     "ISO-8859-16",
  84     "ISO-2022-CN-EXT",
  85     "EUC-CN",
  86     "ISO-2022-JP-2",
  87     "EUC-JP",
  88     "Shift_JIS",
  89     "CP949",
  90     "ISO-2022-KR",
  91     "Big5",
  92     "ISO-2022-TW",
  93     "Big5-HKSCS",
  94     "VISCII",
  95     "Windows-1258",
  96 };
  97
  98 static const char *const ppsz_encoding_names[] = {
  99     N_("Auto"),
 100     N_("Universal (UTF-8)"),
 101     N_("Universal (UTF-16)"),
 102     N_("Universal (big endian UTF-16)"),
 103     N_("Universal (little endian UTF-16)"),
 104     N_("Universal, Chinese (GB18030)"),
 105
 106   /* ISO 8859 and the likes */
 107     /* 1 */
 108     N_("Western European (Latin-9)"), /* mostly superset of Latin-1 */
 109     N_("Western European (Windows-1252)"),
 110     /* 2 */
 111     N_("Eastern European (Latin-2)"),
 112     N_("Eastern European (Windows-1250)"),
 113     /* 3 */
 114     N_("Esperanto (Latin-3)"),
 115     /* 4 */
 116     N_("Nordic (Latin-6)"), /* Latin 6 supersedes Latin 4 */
 117     /* 5 */
 118     N_("Cyrillic (Windows-1251)"), /* ISO 8859-5 is not practically used */
 119     N_("Russian (KOI8-R)"),
 120     N_("Ukrainian (KOI8-U)"),
 121     /* 6 */
 122     N_("Arabic (ISO 8859-6)"),
 123     N_("Arabic (Windows-1256)"),
 124     /* 7 */
 125     N_("Greek (ISO 8859-7)"),
 126     N_("Greek (Windows-1253)"),
 127     /* 8 */
 128     N_("Hebrew (ISO 8859-8)"),
 129     N_("Hebrew (Windows-1255)"),
 130     /* 9 */
 131     N_("Turkish (ISO 8859-9)"),
 132     N_("Turkish (Windows-1254)"),
 133     /* 10 -> 4 */
 134     /* 11 */
 135     N_("Thai (TIS 620-2533/ISO 8859-11)"),
 136     N_("Thai (Windows-874)"),
 137     /* 13 */
 138     N_("Baltic (Latin-7)"),
 139     N_("Baltic (Windows-1257)"),
 140     /* 12 -> /dev/null */
 141     /* 14 */
 142     N_("Celtic (Latin-8)"),
 143     /* 15 -> 1 */
 144     /* 16 */
 145     N_("South-Eastern European (Latin-10)"),
 146   /* CJK families */
 147     N_("Simplified Chinese (ISO-2022-CN-EXT)"),
 148     N_("Simplified Chinese Unix (EUC-CN)"),
 149     N_("Japanese (7-bits JIS/ISO-2022-JP-2)"),
 150     N_("Japanese Unix (EUC-JP)"),
 151     N_("Japanese (Shift JIS)"),
 152     N_("Korean (EUC-KR/CP949)"),
 153     N_("Korean (ISO-2022-KR)"),
 154     N_("Traditional Chinese (Big5)"),
 155     N_("Traditional Chinese Unix (EUC-TW)"),
 156     N_("Hong-Kong Supplementary (HKSCS)"),
 157   /* Other */
 158     N_("Vietnamese (VISCII)"),
 159     N_("Vietnamese (Windows-1258)"),
 160 };
 161 /*
 162 SSA supports charset selection.
 163 The following known charsets are used:
 164
 165 0 = Ansi - Western European
 166 1 = default
 167 2 = symbol
 168 3 = invalid
 169 77 = Mac
 170 128 = Japanese (Shift JIS)
 171 129 = Hangul
 172 130 = Johab
 173 134 = GB2312 Simplified Chinese
 174 136 = Big5 Traditional Chinese
 175 161 = Greek
 176 162 = Turkish
 177 163 = Vietnamese
 178 177 = Hebrew
 179 178 = Arabic
 180 186 = Baltic
 181 204 = Russian (Cyrillic)
 182 222 = Thai
 183 238 = Eastern European
 184 254 = PC 437
 185 */
 186
 187 static const int  pi_justification[] = { 0, 1, 2 };
 188 static const char *const ppsz_justification_text[] = {
 189     N_("Center"),N_("Left"),N_("Right")};
 190
 191 #define ENCODING_TEXT N_("Subtitles text encoding")
 192 #define ENCODING_LONGTEXT N_("Set the encoding used in text subtitles")
 193 #define ALIGN_TEXT N_("Subtitles justification")
 194 #define ALIGN_LONGTEXT N_("Set the justification of subtitles")
 195 #define AUTODETECT_UTF8_TEXT N_("UTF-8 subtitles autodetection")
 196 #define AUTODETECT_UTF8_LONGTEXT N_("This enables automatic detection of " \
 197             "UTF-8 encoding within subtitles files.")
 198 #define FORMAT_TEXT N_("Formatted Subtitles")
 199 #define FORMAT_LONGTEXT N_("Some subtitle formats allow for text formatting. " \
 200  "VLC partly implements this, but you can choose to disable all formatting.")
 201
 202
 203 vlc_module_begin ()
 204     set_shortname( N_("Subtitles"))
 205     set_description( N_("Text subtitles decoder") )
 206     set_capability( "decoder", 50 )
 207     set_callbacks( OpenDecoder, CloseDecoder )
 208     set_category( CAT_INPUT )
 209     set_subcategory( SUBCAT_INPUT_SCODEC )
 210
 211     add_integer( "subsdec-align", 0, NULL, ALIGN_TEXT, ALIGN_LONGTEXT,
 212                  false )
 213         change_integer_list( pi_justification, ppsz_justification_text, NULL )
 214     add_string( "subsdec-encoding", "", NULL,
 215                 ENCODING_TEXT, ENCODING_LONGTEXT, false )
 216         change_string_list( ppsz_encodings, ppsz_encoding_names, 0 )
 217     add_bool( "subsdec-autodetect-utf8", true, NULL,
 218               AUTODETECT_UTF8_TEXT, AUTODETECT_UTF8_LONGTEXT, false )
 219     add_bool( "subsdec-formatted", true, NULL, FORMAT_TEXT, FORMAT_LONGTEXT,
 220                  false )
 221 vlc_module_end ()
 222
 223 /*****************************************************************************
 224  * OpenDecoder: probe the decoder and return score
 225  *****************************************************************************
 226  * Tries to launch a decoder and return score so that the interface is able
 227  * to chose.
 228  *****************************************************************************/
 229 static int OpenDecoder( vlc_object_t *p_this )
 230 {
 231     decoder_t     *p_dec = (decoder_t*)p_this;
 232     decoder_sys_t *p_sys;
 233
 234     switch( p_dec->fmt_in.i_codec )
 235     {
 236         case VLC_CODEC_SUBT:
 237         case VLC_CODEC_SSA:
 238         case VLC_CODEC_ITU_T140:
 239             break;
 240         default:
 241             return VLC_EGENERIC;
 242     }
 243
 244     p_dec->pf_decode_sub = DecodeBlock;
 245     p_dec->fmt_out.i_cat = SPU_ES;
 246     p_dec->fmt_out.i_codec = 0;
 247
 248     /* Allocate the memory needed to store the decoder's structure */
 249     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 250     if( p_sys == NULL )
 251         return VLC_ENOMEM;
 252
 253     /* init of p_sys */
 254     p_sys->i_align = 0;
 255     p_sys->iconv_handle = (vlc_iconv_t)-1;
 256     p_sys->b_autodetect_utf8 = false;
 257     p_sys->b_ass = false;
 258     p_sys->i_original_height = -1;
 259     p_sys->i_original_width = -1;
 260     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 261     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 262
 263     char *psz_charset = NULL;
 264
 265     /* First try demux-specified encoding */
 266     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 267         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 268     else
 269     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 270     {
 271         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 272         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 273                  p_dec->fmt_in.subs.psz_encoding ?
 274                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 275     }
 276
 277     /* Second, try configured encoding */
 278     if (psz_charset == NULL)
 279     {
 280         psz_charset = var_CreateGetNonEmptyString (p_dec, "subsdec-encoding");
 281         msg_Dbg (p_dec, "trying configured character encoding: %s",
 282                  psz_charset ? psz_charset : "not specified");
 283     }
 284
 285     /* Third, try "local" encoding with optional UTF-8 autodetection */
 286     if (psz_charset == NULL)
 287     {
 288         psz_charset = strdup (GetFallbackEncoding ());
 289         msg_Dbg (p_dec, "trying default character encoding: %s",
 290                  psz_charset ? psz_charset : "not specified");
 291
 292         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 293         {
 294             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 295             p_sys->b_autodetect_utf8 = true;
 296         }
 297     }
 298
 299     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 300     if (psz_charset == NULL)
 301     {
 302         psz_charset = strdup ("UTF-8");
 303         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 304     }
 305
 306     if ((psz_charset != NULL)
 307      && strcasecmp (psz_charset, "UTF-8")
 308      && strcasecmp (psz_charset, "utf8"))
 309     {
 310         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 311         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 312             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 313     }
 314     free (psz_charset);
 315
 316     p_sys->i_align = var_CreateGetInteger( p_dec, "subsdec-align" );
 317
 318     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 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             assert( psz_new_subtitle );
 464
 465             size_t ret = vlc_iconv( p_sys->iconv_handle,
 466                                     &psz_convert_buffer_in, &inbytes_left,
 467                                     &psz_convert_buffer_out, &outbytes_left );
 468
 469             *psz_convert_buffer_out++ = '\0';
 470             free( psz_subtitle );
 471
 472             if( ( ret == (size_t)(-1) ) || inbytes_left )
 473             {
 474                 free( psz_new_subtitle );
 475                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 476                         "Try manually setting a character-encoding "
 477                                 "before you open the file." );
 478                 return NULL;
 479             }
 480
 481             psz_subtitle = realloc( psz_new_subtitle,
 482                                     psz_convert_buffer_out - psz_new_subtitle );
 483             if( !psz_subtitle )
 484                 psz_subtitle = psz_new_subtitle;
 485         }
 486     }
 487
 488     /* Create the subpicture unit */
 489     p_spu = decoder_NewSubpicture( p_dec );
 490     if( !p_spu )
 491     {
 492         msg_Warn( p_dec, "can't get spu buffer" );
 493         free( psz_subtitle );
 494         return NULL;
 495     }
 496
 497     /* Create a new subpicture region */
 498     memset( &fmt, 0, sizeof(video_format_t) );
 499     fmt.i_chroma = VLC_CODEC_TEXT;
 500     fmt.i_aspect = 0;
 501     fmt.i_width = fmt.i_height = 0;
 502     fmt.i_x_offset = fmt.i_y_offset = 0;
 503     p_spu->p_region = subpicture_region_New( &fmt );
 504     if( !p_spu->p_region )
 505     {
 506         msg_Err( p_dec, "cannot allocate SPU region" );
 507         free( psz_subtitle );
 508         decoder_DeleteSubpicture( p_dec, p_spu );
 509         return NULL;
 510     }
 511
 512     /* Decode and format the subpicture unit */
 513     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 514     {
 515         /* Normal text subs, easy markup */
 516         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 517         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 518         p_spu->p_region->i_y = 10;
 519
 520         /* Remove formatting from string */
 521
 522         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 523         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 524         {
 525             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 526         }
 527
 528         p_spu->i_start = p_block->i_pts;
 529         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 530         p_spu->b_ephemer = (p_block->i_length == 0);
 531         p_spu->b_absolute = false;
 532     }
 533     else
 534     {
 535         /* Decode SSA/USF strings */
 536         ParseSSAString( p_dec, psz_subtitle, p_spu );
 537
 538         p_spu->i_start = p_block->i_pts;
 539         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 540         p_spu->b_ephemer = (p_block->i_length == 0);
 541         p_spu->b_absolute = false;
 542         p_spu->i_original_picture_width = p_sys->i_original_width;
 543         p_spu->i_original_picture_height = p_sys->i_original_height;
 544     }
 545     free( psz_subtitle );
 546
 547     return p_spu;
 548 }
 549
 550 char* GotoNextLine( char *psz_text )
 551 {
 552     char *p_newline = psz_text;
 553
 554     while( p_newline[0] != '\0' )
 555     {
 556         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 557         {
 558             p_newline++;
 559             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 560                 p_newline++;
 561             break;
 562         }
 563         else p_newline++;
 564     }
 565     return p_newline;
 566 }
 567
 568 /* Function now handles tags with attribute values, and tries
 569  * to deal with &' commands too. It no longer modifies the string
 570  * in place, so that the original text can be reused
 571  */
 572 static char *StripTags( char *psz_subtitle )
 573 {
 574     char *psz_text_start;
 575     char *psz_text;
 576
 577     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 578     if( !psz_text_start )
 579         return NULL;
 580
 581     while( *psz_subtitle )
 582     {
 583         if( *psz_subtitle == '<' )
 584         {
 585             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 586                 *psz_text++ = '\n';
 587
 588             psz_subtitle += strcspn( psz_subtitle, ">" );
 589         }
 590         else if( *psz_subtitle == '&' )
 591         {
 592             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 593             {
 594                 *psz_text++ = '<';
 595                 psz_subtitle += strcspn( psz_subtitle, ";" );
 596             }
 597             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 598             {
 599                 *psz_text++ = '>';
 600                 psz_subtitle += strcspn( psz_subtitle, ";" );
 601             }
 602             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 603             {
 604                 *psz_text++ = '&';
 605                 psz_subtitle += strcspn( psz_subtitle, ";" );
 606             }
 607             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 608             {
 609                 *psz_text++ = '\"';
 610                 psz_subtitle += strcspn( psz_subtitle, ";" );
 611             }
 612             else
 613             {
 614                 /* Assume it is just a normal ampersand */
 615                 *psz_text++ = '&';
 616             }
 617         }
 618         else
 619         {
 620             *psz_text++ = *psz_subtitle;
 621         }
 622
 623         psz_subtitle++;
 624     }
 625     *psz_text = '\0';
 626     char *psz = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 627     if( psz ) psz_text_start = psz;
 628
 629     return psz_text_start;
 630 }
 631
 632 /* Try to respect any style tags present in the subtitle string. The main
 633  * problem here is a lack of adequate specs for the subtitle formats.
 634  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 635  * SAMI has a detailed spec, but extensive rework is needed in the demux
 636  * code to prevent all this style information being excised, as it presently
 637  * does.
 638  * That leaves the others - none of which were (I guess) originally intended
 639  * to be carrying style information. Over time people have used them that way.
 640  * In the absence of specifications from which to work, the tags supported
 641  * have been restricted to the simple set permitted by the USF DTD, ie. :
 642  *  Basic: <br>, <i>, <b>, <u>
 643  *  Extended: <font>
 644  *    Attributes: face
 645  *                family
 646  *                size
 647  *                color
 648  *                outline-color
 649  *                shadow-color
 650  *                outline-level
 651  *                shadow-level
 652  *                back-color
 653  *                alpha
 654  * There is also the further restriction that the subtitle be well-formed
 655  * as an XML entity, ie. the HTML sentence:
 656  *        <b><i>Bold and Italics</b></i>
 657  * doesn't qualify because the tags aren't nested one inside the other.
 658  * <text> tags are automatically added to the output to ensure
 659  * well-formedness.
 660  * If the text doesn't qualify for any reason, a NULL string is
 661  * returned, and the rendering engine will fall back to the
 662  * plain text version of the subtitle.
 663  */
 664 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 665 {
 666     const int i_len = strlen(psz_text);
 667
 668     strncpy( *ppsz_html, psz_text, i_max );
 669     *ppsz_html += __MIN(i_max,i_len);
 670 }
 671
 672 static void HtmlPut( char **ppsz_html, const char *psz_text )
 673 {
 674     strcpy( *ppsz_html, psz_text );
 675     *ppsz_html += strlen(psz_text);
 676 }
 677 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 678 {
 679     HtmlPut( ppsz_html, psz_text );
 680     *ppsz_subtitle += strlen(psz_text);
 681 }
 682
 683 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 684 {
 685     /* */
 686     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 687     if( !psz_tag )
 688         return NULL;
 689     psz_tag[ 0 ] = '\0';
 690
 691     /* */
 692     //Oo + 100 ???
 693     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 694     char   *psz_html_start = malloc( i_buf_size );
 695     char   *psz_html = psz_html_start;
 696     if( psz_html_start == NULL )
 697     {
 698         free( psz_tag );
 699         return NULL;
 700     }
 701     psz_html[0] = '\0';
 702
 703     bool b_has_align = false;
 704
 705     HtmlPut( &psz_html, "<text>" );
 706
 707     /* */
 708     while( *psz_subtitle )
 709     {
 710         if( *psz_subtitle == '\n' )
 711         {
 712             HtmlPut( &psz_html, "<br/>" );
 713             psz_subtitle++;
 714         }
 715         else if( *psz_subtitle == '<' )
 716         {
 717             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 718             {
 719                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 720             }
 721             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 722             {
 723                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 724                 strcat( psz_tag, "b" );
 725             }
 726             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 727             {
 728                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 729                 strcat( psz_tag, "i" );
 730             }
 731             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 732             {
 733                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 734                 strcat( psz_tag, "u" );
 735             }
 736             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 737             {
 738                 const char *psz_attribs[] = { "face=", "family=", "size=",
 739                         "color=", "outline-color=", "shadow-color=",
 740                         "outline-level=", "shadow-level=", "back-color=",
 741                         "alpha=", NULL };
 742
 743                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 744                 strcat( psz_tag, "f" );
 745
 746                 while( *psz_subtitle != '>' )
 747                 {
 748                     int  k;
 749
 750                     for( k=0; psz_attribs[ k ]; k++ )
 751                     {
 752                         int i_len = strlen( psz_attribs[ k ] );
 753
 754                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 755                         {
 756                             /* */
 757                             HtmlPut( &psz_html, psz_attribs[k] );
 758                             psz_subtitle += i_len;
 759
 760                             /* */
 761                             if( *psz_subtitle == '"' )
 762                             {
 763                                 psz_subtitle++;
 764                                 i_len = strcspn( psz_subtitle, "\"" );
 765                             }
 766                             else
 767                             {
 768                                 i_len = strcspn( psz_subtitle, " \t>" );
 769                             }
 770                             HtmlPut( &psz_html, "\"" );
 771                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 772                                 HtmlPut( &psz_html, "#" );
 773                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 774                             HtmlPut( &psz_html, "\"" );
 775
 776                             psz_subtitle += i_len;
 777                             if( *psz_subtitle == '\"' )
 778                                 psz_subtitle++;
 779                             break;
 780                         }
 781                     }
 782                     if( psz_attribs[ k ] == NULL )
 783                     {
 784                         /* Jump over unrecognised tag */
 785                         int i_len = strcspn( psz_subtitle, "\"" ) + 1;
 786
 787                         i_len += strcspn( psz_subtitle + i_len, "\"" ) + 1;
 788                         psz_subtitle += i_len;
 789                     }
 790                     while (*psz_subtitle == ' ')
 791                         *psz_html++ = *psz_subtitle++;
 792                 }
 793                 *psz_html++ = *psz_subtitle++;
 794             }
 795             else if( !strncmp( psz_subtitle, "</", 2 ))
 796             {
 797                 bool   b_match     = false;
 798                 bool   b_ignore    = false;
 799                 int    i_len       = strlen( psz_tag ) - 1;
 800                 char  *psz_lastTag = NULL;
 801
 802                 if( i_len >= 0 )
 803                 {
 804                     psz_lastTag = psz_tag + i_len;
 805                     i_len = 0;
 806
 807                     switch( *psz_lastTag )
 808                     {
 809                     case 'b':
 810                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 811                         i_len   = 4;
 812                         break;
 813                     case 'i':
 814                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 815                         i_len   = 4;
 816                         break;
 817                     case 'u':
 818                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 819                         i_len   = 4;
 820                         break;
 821                     case 'f':
 822                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 823                         i_len   = 7;
 824                         break;
 825                     case 'I':
 826                         i_len = strcspn( psz_subtitle, ">" );
 827                         b_match = psz_subtitle[i_len] == '>';
 828                         b_ignore = true;
 829                         if( b_match )
 830                             i_len++;
 831                         break;
 832                     }
 833                 }
 834                 if( !b_match )
 835                 {
 836                     /* Not well formed -- kill everything */
 837                     free( psz_html_start );
 838                     psz_html_start = NULL;
 839                     break;
 840                 }
 841                 *psz_lastTag = '\0';
 842                 if( !b_ignore )
 843                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 844
 845                 psz_subtitle += i_len;
 846             }
 847             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 848                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 849             {
 850                 /* We have a single < */
 851                 HtmlPut( &psz_html, "&lt;" );
 852                 psz_subtitle++;
 853             }
 854             else
 855             {
 856                 /* We have an unknown tag or a single < */
 857
 858                 /* Search for the next tag or end of tag or end of string */
 859                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 860                 char *psz_closing = strstr( psz_subtitle, "/>" );
 861
 862                 if( psz_closing && psz_closing < psz_stop )
 863                 {
 864                     /* We have a self closed tag, remove it */
 865                     psz_subtitle = &psz_closing[2];
 866                 }
 867                 else if( *psz_stop == '>' )
 868                 {
 869                     char psz_match[256];
 870
 871                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 872                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 873
 874                     if( strstr( psz_subtitle, psz_match ) )
 875                     {
 876                         /* We have the closing tag, ignore it TODO */
 877                         psz_subtitle = &psz_stop[1];
 878                         strcat( psz_tag, "I" );
 879                     }
 880                     else
 881                     {
 882                         int i_len = psz_stop + 1 - psz_subtitle;
 883
 884                         /* Copy the whole data */
 885                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 886                         {
 887                             if( *psz_subtitle == '<' )
 888                                 HtmlPut( &psz_html, "&lt;" );
 889                             else if( *psz_subtitle == '>' )
 890                                 HtmlPut( &psz_html, "&gt;" );
 891                             else
 892                                 *psz_html++ = *psz_subtitle;
 893                         }
 894                     }
 895                 }
 896                 else
 897                 {
 898                     /* We have a single < */
 899                     HtmlPut( &psz_html, "&lt;" );
 900                     psz_subtitle++;
 901                 }
 902             }
 903         }
 904         else if( *psz_subtitle == '&' )
 905         {
 906             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 907             {
 908                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 909             }
 910             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 911             {
 912                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 913             }
 914             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 915             {
 916                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 917             }
 918             else
 919             {
 920                 HtmlPut( &psz_html, "&amp;" );
 921                 psz_subtitle++;
 922             }
 923         }
 924         else if( *psz_subtitle == '>' )
 925         {
 926             HtmlPut( &psz_html, "&gt;" );
 927             psz_subtitle++;
 928         }
 929         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 930                  strchr( psz_subtitle, '}' ) )
 931         {
 932             /* Check for forced alignment */
 933             if( !b_has_align &&
 934                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 935             {
 936                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 937                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 938                 const int i_id = psz_subtitle[4] - '1';
 939
 940                 b_has_align = true;
 941                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 942             }
 943             /* TODO fr -> rotation */
 944
 945             /* Hide {\stupidity} */
 946             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 947         }
 948         else
 949         {
 950             *psz_html = *psz_subtitle;
 951             if( psz_html > psz_html_start )
 952             {
 953                 /* Check for double whitespace */
 954                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
 955                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
 956                 {
 957                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
 958                     psz_html--;
 959                 }
 960             }
 961             psz_html++;
 962             psz_subtitle++;
 963         }
 964
 965         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
 966         {
 967             const int i_len = psz_html - psz_html_start;
 968
 969             i_buf_size += 200;
 970             char *psz_new = realloc( psz_html_start, i_buf_size );
 971             if( !psz_new )
 972                 break;
 973             psz_html_start = psz_new;
 974             psz_html = &psz_new[i_len];
 975         }
 976     }
 977     if( psz_html_start )
 978     {
 979         static const char *psz_text_close = "</text>";
 980         static const char *psz_tag_long = "/font>";
 981
 982         /* Realloc for closing tags and shrink memory */
 983         const size_t i_length = (size_t)( psz_html - psz_html_start );
 984
 985         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
 986         char *psz_new = realloc( psz_html_start, i_size );
 987         if( psz_new )
 988         {
 989             psz_html_start = psz_new;
 990             psz_html = &psz_new[i_length];
 991
 992             /* Close not well formed subtitle */
 993             while( *psz_tag )
 994             {
 995                 /* */
 996                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
 997                 switch( *psz_last )
 998                 {
 999                 case 'b':
1000                     HtmlPut( &psz_html, "</b>" );
1001                     break;
1002                 case 'i':
1003                     HtmlPut( &psz_html, "</i>" );
1004                     break;
1005                 case 'u':
1006                     HtmlPut( &psz_html, "</u>" );
1007                     break;
1008                 case 'f':
1009                     HtmlPut( &psz_html, "/font>" );
1010                     break;
1011                 case 'I':
1012                     break;
1013                 }
1014
1015                 *psz_last = '\0';
1016             }
1017             HtmlPut( &psz_html, psz_text_close );
1018         }
1019     }
1020     free( psz_tag );
1021
1022     return psz_html_start;
1023 }
1024