git.sesse.net Git - vlc/blob - modules/codec/subtitles/subsdec.c

   1 /*****************************************************************************
   2  * subsdec.c : text subtitles decoder
   3  *****************************************************************************
   4  * Copyright (C) 2000-2006 the VideoLAN team
   5  * $Id$
   6  *
   7  * Authors: Gildas Bazin <gbazin@videolan.org>
   8  *          Samuel Hocevar <sam@zoy.org>
   9  *          Derk-Jan Hartman <hartman at videolan dot org>
  10  *          Bernie Purcell <bitmap@videolan.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License as published by
  14  * the Free Software Foundation; either version 2 of the License, or
  15  * (at your option) any later version.
  16  *
  17  * This program is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  * GNU General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU General Public License
  23  * along with this program; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  25  *****************************************************************************/
  26
  27 /*****************************************************************************
  28  * Preamble
  29  *****************************************************************************/
  30 #ifdef HAVE_CONFIG_H
  31 # include "config.h"
  32 #endif
  33
  34 #include "subsdec.h"
  35 #include <vlc_plugin.h>
  36
  37 /*****************************************************************************
  38  * Local prototypes
  39  *****************************************************************************/
  40 static int  OpenDecoder   ( vlc_object_t * );
  41 static void CloseDecoder  ( vlc_object_t * );
  42
  43 static subpicture_t   *DecodeBlock   ( decoder_t *, block_t ** );
  44 static subpicture_t   *ParseText     ( decoder_t *, block_t * );
  45 static char           *StripTags      ( char * );
  46 static char           *CreateHtmlSubtitle( int *pi_align, char * );
  47
  48
  49 /*****************************************************************************
  50  * Module descriptor.
  51  *****************************************************************************/
  52 static const char *const ppsz_encodings[] = {
  53     "",
  54     "UTF-8",
  55     "UTF-16",
  56     "UTF-16BE",
  57     "UTF-16LE",
  58     "GB18030",
  59     "ISO-8859-15",
  60     "Windows-1252",
  61     "ISO-8859-2",
  62     "Windows-1250",
  63     "ISO-8859-3",
  64     "ISO-8859-10",
  65     "Windows-1251",
  66     "KOI8-R",
  67     "KOI8-U",
  68     "ISO-8859-6",
  69     "Windows-1256",
  70     "ISO-8859-7",
  71     "Windows-1253",
  72     "ISO-8859-8",
  73     "Windows-1255",
  74     "ISO-8859-9",
  75     "Windows-1254",
  76     "ISO-8859-11",
  77     "Windows-874",
  78     "ISO-8859-13",
  79     "Windows-1257",
  80     "ISO-8859-14",
  81     "ISO-8859-16",
  82     "ISO-2022-CN-EXT",
  83     "EUC-CN",
  84     "ISO-2022-JP-2",
  85     "EUC-JP",
  86     "Shift_JIS",
  87     "CP949",
  88     "ISO-2022-KR",
  89     "Big5",
  90     "ISO-2022-TW",
  91     "Big5-HKSCS",
  92     "VISCII",
  93     "Windows-1258",
  94 };
  95
  96 static const char *const ppsz_encoding_names[] = {
  97     N_("Auto"),
  98     N_("Universal (UTF-8)"),
  99     N_("Universal (UTF-16)"),
 100     N_("Universal (big endian UTF-16)"),
 101     N_("Universal (little endian UTF-16)"),
 102     N_("Universal, Chinese (GB18030)"),
 103
 104   /* ISO 8859 and the likes */
 105     /* 1 */
 106     N_("Western European (Latin-9)"), /* mostly superset of Latin-1 */
 107     N_("Western European (Windows-1252)"),
 108     /* 2 */
 109     N_("Eastern European (Latin-2)"),
 110     N_("Eastern European (Windows-1250)"),
 111     /* 3 */
 112     N_("Esperanto (Latin-3)"),
 113     /* 4 */
 114     N_("Nordic (Latin-6)"), /* Latin 6 supersedes Latin 4 */
 115     /* 5 */
 116     N_("Cyrillic (Windows-1251)"), /* ISO 8859-5 is not practically used */
 117     N_("Russian (KOI8-R)"),
 118     N_("Ukrainian (KOI8-U)"),
 119     /* 6 */
 120     N_("Arabic (ISO 8859-6)"),
 121     N_("Arabic (Windows-1256)"),
 122     /* 7 */
 123     N_("Greek (ISO 8859-7)"),
 124     N_("Greek (Windows-1253)"),
 125     /* 8 */
 126     N_("Hebrew (ISO 8859-8)"),
 127     N_("Hebrew (Windows-1255)"),
 128     /* 9 */
 129     N_("Turkish (ISO 8859-9)"),
 130     N_("Turkish (Windows-1254)"),
 131     /* 10 -> 4 */
 132     /* 11 */
 133     N_("Thai (TIS 620-2533/ISO 8859-11)"),
 134     N_("Thai (Windows-874)"),
 135     /* 13 */
 136     N_("Baltic (Latin-7)"),
 137     N_("Baltic (Windows-1257)"),
 138     /* 12 -> /dev/null */
 139     /* 14 */
 140     N_("Celtic (Latin-8)"),
 141     /* 15 -> 1 */
 142     /* 16 */
 143     N_("South-Eastern European (Latin-10)"),
 144   /* CJK families */
 145     N_("Simplified Chinese (ISO-2022-CN-EXT)"),
 146     N_("Simplified Chinese Unix (EUC-CN)"),
 147     N_("Japanese (7-bits JIS/ISO-2022-JP-2)"),
 148     N_("Japanese Unix (EUC-JP)"),
 149     N_("Japanese (Shift JIS)"),
 150     N_("Korean (EUC-KR/CP949)"),
 151     N_("Korean (ISO-2022-KR)"),
 152     N_("Traditional Chinese (Big5)"),
 153     N_("Traditional Chinese Unix (EUC-TW)"),
 154     N_("Hong-Kong Supplementary (HKSCS)"),
 155   /* Other */
 156     N_("Vietnamese (VISCII)"),
 157     N_("Vietnamese (Windows-1258)"),
 158 };
 159 /*
 160 SSA supports charset selection.
 161 The following known charsets are used:
 162
 163 0 = Ansi - Western European
 164 1 = default
 165 2 = symbol
 166 3 = invalid
 167 77 = Mac
 168 128 = Japanese (Shift JIS)
 169 129 = Hangul
 170 130 = Johab
 171 134 = GB2312 Simplified Chinese
 172 136 = Big5 Traditional Chinese
 173 161 = Greek
 174 162 = Turkish
 175 163 = Vietnamese
 176 177 = Hebrew
 177 178 = Arabic
 178 186 = Baltic
 179 204 = Russian (Cyrillic)
 180 222 = Thai
 181 238 = Eastern European
 182 254 = PC 437
 183 */
 184
 185 static const int  pi_justification[] = { 0, 1, 2 };
 186 static const char *const ppsz_justification_text[] = {
 187     N_("Center"),N_("Left"),N_("Right")};
 188
 189 #define ENCODING_TEXT N_("Subtitles text encoding")
 190 #define ENCODING_LONGTEXT N_("Set the encoding used in text subtitles")
 191 #define ALIGN_TEXT N_("Subtitles justification")
 192 #define ALIGN_LONGTEXT N_("Set the justification of subtitles")
 193 #define AUTODETECT_UTF8_TEXT N_("UTF-8 subtitles autodetection")
 194 #define AUTODETECT_UTF8_LONGTEXT N_("This enables automatic detection of " \
 195             "UTF-8 encoding within subtitles files.")
 196 #define FORMAT_TEXT N_("Formatted Subtitles")
 197 #define FORMAT_LONGTEXT N_("Some subtitle formats allow for text formatting. " \
 198  "VLC partly implements this, but you can choose to disable all formatting.")
 199
 200
 201 vlc_module_begin ()
 202     set_shortname( N_("Subtitles"))
 203     set_description( N_("Text subtitles decoder") )
 204     set_capability( "decoder", 50 )
 205     set_callbacks( OpenDecoder, CloseDecoder )
 206     set_category( CAT_INPUT )
 207     set_subcategory( SUBCAT_INPUT_SCODEC )
 208
 209     add_integer( "subsdec-align", 0, NULL, ALIGN_TEXT, ALIGN_LONGTEXT,
 210                  false )
 211         change_integer_list( pi_justification, ppsz_justification_text, NULL )
 212     add_string( "subsdec-encoding", "", NULL,
 213                 ENCODING_TEXT, ENCODING_LONGTEXT, false )
 214         change_string_list( ppsz_encodings, ppsz_encoding_names, 0 )
 215     add_bool( "subsdec-autodetect-utf8", true, NULL,
 216               AUTODETECT_UTF8_TEXT, AUTODETECT_UTF8_LONGTEXT, false )
 217     add_bool( "subsdec-formatted", true, NULL, FORMAT_TEXT, FORMAT_LONGTEXT,
 218                  false )
 219 vlc_module_end ()
 220
 221 /*****************************************************************************
 222  * OpenDecoder: probe the decoder and return score
 223  *****************************************************************************
 224  * Tries to launch a decoder and return score so that the interface is able
 225  * to chose.
 226  *****************************************************************************/
 227 static int OpenDecoder( vlc_object_t *p_this )
 228 {
 229     decoder_t     *p_dec = (decoder_t*)p_this;
 230     decoder_sys_t *p_sys;
 231
 232     switch( p_dec->fmt_in.i_codec )
 233     {
 234         case VLC_CODEC_SUBT:
 235         case VLC_CODEC_SSA:
 236         case VLC_CODEC_ITU_T140:
 237             break;
 238         default:
 239             return VLC_EGENERIC;
 240     }
 241
 242     p_dec->pf_decode_sub = DecodeBlock;
 243     p_dec->fmt_out.i_cat = SPU_ES;
 244     p_dec->fmt_out.i_codec = 0;
 245
 246     /* Allocate the memory needed to store the decoder's structure */
 247     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 248     if( p_sys == NULL )
 249         return VLC_ENOMEM;
 250
 251     /* init of p_sys */
 252     p_sys->i_align = 0;
 253     p_sys->iconv_handle = (vlc_iconv_t)-1;
 254     p_sys->b_autodetect_utf8 = false;
 255     p_sys->b_ass = false;
 256     p_sys->i_original_height = -1;
 257     p_sys->i_original_width = -1;
 258     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 259     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 260
 261     char *psz_charset = NULL;
 262
 263     /* First try demux-specified encoding */
 264     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 265         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 266     else
 267     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 268     {
 269         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 270         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 271                  p_dec->fmt_in.subs.psz_encoding ?
 272                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 273     }
 274
 275     /* Second, try configured encoding */
 276     if (psz_charset == NULL)
 277     {
 278         psz_charset = var_CreateGetNonEmptyString (p_dec, "subsdec-encoding");
 279         msg_Dbg (p_dec, "trying configured character encoding: %s",
 280                  psz_charset ? psz_charset : "not specified");
 281     }
 282
 283     /* Third, try "local" encoding with optional UTF-8 autodetection */
 284     if (psz_charset == NULL)
 285     {
 286         /* xgettext:
 287            The Windows ANSI code page most commonly used for this language.
 288            VLC uses this as a guess of the subtitle files character set
 289            (if UTF-8 and UTF-16 autodetection fails).
 290            Western European languages normally use "CP1252", which is a
 291            Microsoft-variant of ISO 8859-1. That suits the Latin alphabet.
 292            Other scripts use other code pages.
 293
 294            This MUST be a valid iconv character set. If unsure, please refer
 295            the VideoLAN translators mailing list. */
 296         const char *acp = vlc_pgettext("GetACP", "CP1252");
 297
 298         psz_charset = strdup (acp);
 299         msg_Dbg (p_dec, "trying default character encoding: %s",
 300                  psz_charset ? psz_charset : "not specified");
 301
 302         if (var_CreateGetBool (p_dec, "subsdec-autodetect-utf8"))
 303         {
 304             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 305             p_sys->b_autodetect_utf8 = true;
 306         }
 307     }
 308
 309     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 310     if (psz_charset == NULL)
 311     {
 312         psz_charset = strdup ("UTF-8");
 313         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 314     }
 315
 316     if ((psz_charset != NULL)
 317      && strcasecmp (psz_charset, "UTF-8")
 318      && strcasecmp (psz_charset, "utf8"))
 319     {
 320         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 321         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 322             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 323     }
 324     free (psz_charset);
 325
 326     p_sys->i_align = var_CreateGetInteger( p_dec, "subsdec-align" );
 327
 328     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 329      && var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 330     {
 331         if( p_dec->fmt_in.i_extra > 0 )
 332             ParseSSAHeader( p_dec );
 333     }
 334
 335     return VLC_SUCCESS;
 336 }
 337
 338 /****************************************************************************
 339  * DecodeBlock: the whole thing
 340  ****************************************************************************
 341  * This function must be fed with complete subtitles units.
 342  ****************************************************************************/
 343 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 344 {
 345     subpicture_t *p_spu;
 346     block_t *p_block;
 347
 348     if( !pp_block || *pp_block == NULL )
 349         return NULL;
 350
 351     p_block = *pp_block;
 352     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 353     {
 354         block_Release( p_block );
 355         return NULL;
 356     }
 357
 358     p_spu = ParseText( p_dec, p_block );
 359
 360     block_Release( p_block );
 361     *pp_block = NULL;
 362
 363     return p_spu;
 364 }
 365
 366 /*****************************************************************************
 367  * CloseDecoder: clean up the decoder
 368  *****************************************************************************/
 369 static void CloseDecoder( vlc_object_t *p_this )
 370 {
 371     decoder_t *p_dec = (decoder_t *)p_this;
 372     decoder_sys_t *p_sys = p_dec->p_sys;
 373
 374     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 375         vlc_iconv_close( p_sys->iconv_handle );
 376
 377     if( p_sys->pp_ssa_styles )
 378     {
 379         int i;
 380         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 381         {
 382             if( !p_sys->pp_ssa_styles[i] )
 383                 continue;
 384
 385             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 386             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 387             free( p_sys->pp_ssa_styles[i] );
 388         }
 389         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 390     }
 391     if( p_sys->pp_images )
 392     {
 393         int i;
 394         for( i = 0; i < p_sys->i_images; i++ )
 395         {
 396             if( !p_sys->pp_images[i] )
 397                 continue;
 398
 399             if( p_sys->pp_images[i]->p_pic )
 400                 picture_Release( p_sys->pp_images[i]->p_pic );
 401             free( p_sys->pp_images[i]->psz_filename );
 402
 403             free( p_sys->pp_images[i] );
 404         }
 405         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 406     }
 407
 408     free( p_sys );
 409 }
 410
 411 /*****************************************************************************
 412  * ParseText: parse an text subtitle packet and send it to the video output
 413  *****************************************************************************/
 414 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 415 {
 416     decoder_sys_t *p_sys = p_dec->p_sys;
 417     subpicture_t *p_spu = NULL;
 418     char *psz_subtitle = NULL;
 419     video_format_t fmt;
 420
 421     /* We cannot display a subpicture with no date */
 422     if( p_block->i_pts <= VLC_TS_INVALID )
 423     {
 424         msg_Warn( p_dec, "subtitle without a date" );
 425         return NULL;
 426     }
 427
 428     /* Check validity of packet data */
 429     /* An "empty" line containing only \0 can be used to force
 430        and ephemer picture from the screen */
 431     if( p_block->i_buffer < 1 )
 432     {
 433         msg_Warn( p_dec, "no subtitle data" );
 434         return NULL;
 435     }
 436
 437     /* Should be resiliant against bad subtitles */
 438     psz_subtitle = strndup( (const char *)p_block->p_buffer,
 439                             p_block->i_buffer );
 440     if( psz_subtitle == NULL )
 441         return NULL;
 442
 443     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 444     {
 445         if (EnsureUTF8( psz_subtitle ) == NULL)
 446         {
 447             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 448                      "Try manually setting a character-encoding "
 449                      "before you open the file." );
 450         }
 451     }
 452     else
 453     {
 454
 455         if( p_sys->b_autodetect_utf8 )
 456         {
 457             if( IsUTF8( psz_subtitle ) == NULL )
 458             {
 459                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 460                          "disabling UTF-8 subtitles autodetection" );
 461                 p_sys->b_autodetect_utf8 = false;
 462             }
 463         }
 464
 465         if( !p_sys->b_autodetect_utf8 )
 466         {
 467             size_t inbytes_left = strlen( psz_subtitle );
 468             size_t outbytes_left = 6 * inbytes_left;
 469             char *psz_new_subtitle = xmalloc( outbytes_left + 1 );
 470             char *psz_convert_buffer_out = psz_new_subtitle;
 471             const char *psz_convert_buffer_in = psz_subtitle;
 472
 473             size_t ret = vlc_iconv( p_sys->iconv_handle,
 474                                     &psz_convert_buffer_in, &inbytes_left,
 475                                     &psz_convert_buffer_out, &outbytes_left );
 476
 477             *psz_convert_buffer_out++ = '\0';
 478             free( psz_subtitle );
 479
 480             if( ( ret == (size_t)(-1) ) || inbytes_left )
 481             {
 482                 free( psz_new_subtitle );
 483                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 484                         "Try manually setting a character-encoding "
 485                                 "before you open the file." );
 486                 return NULL;
 487             }
 488
 489             psz_subtitle = realloc( psz_new_subtitle,
 490                                     psz_convert_buffer_out - psz_new_subtitle );
 491             if( !psz_subtitle )
 492                 psz_subtitle = psz_new_subtitle;
 493         }
 494     }
 495
 496     /* Create the subpicture unit */
 497     p_spu = decoder_NewSubpicture( p_dec );
 498     if( !p_spu )
 499     {
 500         msg_Warn( p_dec, "can't get spu buffer" );
 501         free( psz_subtitle );
 502         return NULL;
 503     }
 504
 505     /* Create a new subpicture region */
 506     memset( &fmt, 0, sizeof(video_format_t) );
 507     fmt.i_chroma = VLC_CODEC_TEXT;
 508     fmt.i_width = fmt.i_height = 0;
 509     fmt.i_x_offset = fmt.i_y_offset = 0;
 510     p_spu->p_region = subpicture_region_New( &fmt );
 511     if( !p_spu->p_region )
 512     {
 513         msg_Err( p_dec, "cannot allocate SPU region" );
 514         free( psz_subtitle );
 515         decoder_DeleteSubpicture( p_dec, p_spu );
 516         return NULL;
 517     }
 518
 519     /* Decode and format the subpicture unit */
 520     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 521     {
 522         /* Normal text subs, easy markup */
 523         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 524         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 525         p_spu->p_region->i_y = 10;
 526
 527         /* Remove formatting from string */
 528
 529         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 530         if( var_CreateGetBool( p_dec, "subsdec-formatted" ) )
 531         {
 532             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 533         }
 534
 535         p_spu->i_start = p_block->i_pts;
 536         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 537         p_spu->b_ephemer = (p_block->i_length == 0);
 538         p_spu->b_absolute = false;
 539     }
 540     else
 541     {
 542         /* Decode SSA/USF strings */
 543         ParseSSAString( p_dec, psz_subtitle, p_spu );
 544
 545         p_spu->i_start = p_block->i_pts;
 546         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 547         p_spu->b_ephemer = (p_block->i_length == 0);
 548         p_spu->b_absolute = false;
 549         p_spu->i_original_picture_width = p_sys->i_original_width;
 550         p_spu->i_original_picture_height = p_sys->i_original_height;
 551     }
 552     free( psz_subtitle );
 553
 554     return p_spu;
 555 }
 556
 557 char* GotoNextLine( char *psz_text )
 558 {
 559     char *p_newline = psz_text;
 560
 561     while( p_newline[0] != '\0' )
 562     {
 563         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 564         {
 565             p_newline++;
 566             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 567                 p_newline++;
 568             break;
 569         }
 570         else p_newline++;
 571     }
 572     return p_newline;
 573 }
 574
 575 /* Function now handles tags with attribute values, and tries
 576  * to deal with &' commands too. It no longer modifies the string
 577  * in place, so that the original text can be reused
 578  */
 579 static char *StripTags( char *psz_subtitle )
 580 {
 581     char *psz_text_start;
 582     char *psz_text;
 583
 584     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 585     if( !psz_text_start )
 586         return NULL;
 587
 588     while( *psz_subtitle )
 589     {
 590         if( *psz_subtitle == '<' )
 591         {
 592             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 593                 *psz_text++ = '\n';
 594
 595             psz_subtitle += strcspn( psz_subtitle, ">" );
 596         }
 597         else if( *psz_subtitle == '&' )
 598         {
 599             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 600             {
 601                 *psz_text++ = '<';
 602                 psz_subtitle += strcspn( psz_subtitle, ";" );
 603             }
 604             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 605             {
 606                 *psz_text++ = '>';
 607                 psz_subtitle += strcspn( psz_subtitle, ";" );
 608             }
 609             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 610             {
 611                 *psz_text++ = '&';
 612                 psz_subtitle += strcspn( psz_subtitle, ";" );
 613             }
 614             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 615             {
 616                 *psz_text++ = '\"';
 617                 psz_subtitle += strcspn( psz_subtitle, ";" );
 618             }
 619             else
 620             {
 621                 /* Assume it is just a normal ampersand */
 622                 *psz_text++ = '&';
 623             }
 624         }
 625         else
 626         {
 627             *psz_text++ = *psz_subtitle;
 628         }
 629
 630         psz_subtitle++;
 631     }
 632     *psz_text = '\0';
 633     char *psz = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 634     if( psz ) psz_text_start = psz;
 635
 636     return psz_text_start;
 637 }
 638
 639 /* Try to respect any style tags present in the subtitle string. The main
 640  * problem here is a lack of adequate specs for the subtitle formats.
 641  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 642  * SAMI has a detailed spec, but extensive rework is needed in the demux
 643  * code to prevent all this style information being excised, as it presently
 644  * does.
 645  * That leaves the others - none of which were (I guess) originally intended
 646  * to be carrying style information. Over time people have used them that way.
 647  * In the absence of specifications from which to work, the tags supported
 648  * have been restricted to the simple set permitted by the USF DTD, ie. :
 649  *  Basic: <br>, <i>, <b>, <u>, <s>
 650  *  Extended: <font>
 651  *    Attributes: face
 652  *                family
 653  *                size
 654  *                color
 655  *                outline-color
 656  *                shadow-color
 657  *                outline-level
 658  *                shadow-level
 659  *                back-color
 660  *                alpha
 661  * There is also the further restriction that the subtitle be well-formed
 662  * as an XML entity, ie. the HTML sentence:
 663  *        <b><i>Bold and Italics</b></i>
 664  * doesn't qualify because the tags aren't nested one inside the other.
 665  * <text> tags are automatically added to the output to ensure
 666  * well-formedness.
 667  * If the text doesn't qualify for any reason, a NULL string is
 668  * returned, and the rendering engine will fall back to the
 669  * plain text version of the subtitle.
 670  */
 671 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 672 {
 673     const int i_len = strlen(psz_text);
 674
 675     strncpy( *ppsz_html, psz_text, i_max );
 676     *ppsz_html += __MIN(i_max,i_len);
 677 }
 678
 679 static void HtmlPut( char **ppsz_html, const char *psz_text )
 680 {
 681     strcpy( *ppsz_html, psz_text );
 682     *ppsz_html += strlen(psz_text);
 683 }
 684 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 685 {
 686     HtmlPut( ppsz_html, psz_text );
 687     *ppsz_subtitle += strlen(psz_text);
 688 }
 689
 690 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 691 {
 692     /* */
 693     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 694     if( !psz_tag )
 695         return NULL;
 696     psz_tag[ 0 ] = '\0';
 697
 698     /* */
 699     //Oo + 100 ???
 700     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 701     char   *psz_html_start = malloc( i_buf_size );
 702     char   *psz_html = psz_html_start;
 703     if( psz_html_start == NULL )
 704     {
 705         free( psz_tag );
 706         return NULL;
 707     }
 708     psz_html[0] = '\0';
 709
 710     bool b_has_align = false;
 711
 712     HtmlPut( &psz_html, "<text>" );
 713
 714     /* */
 715     while( *psz_subtitle )
 716     {
 717         if( *psz_subtitle == '\n' )
 718         {
 719             HtmlPut( &psz_html, "<br/>" );
 720             psz_subtitle++;
 721         }
 722         else if( *psz_subtitle == '<' )
 723         {
 724             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 725             {
 726                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 727             }
 728             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 729             {
 730                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 731                 strcat( psz_tag, "b" );
 732             }
 733             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 734             {
 735                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 736                 strcat( psz_tag, "i" );
 737             }
 738             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 739             {
 740                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 741                 strcat( psz_tag, "u" );
 742             }
 743             else if( !strncasecmp( psz_subtitle, "<s>", 3 ) )
 744             {
 745                 HtmlCopy( &psz_html, &psz_subtitle, "<s>" );
 746                 strcat( psz_tag, "s" );
 747             }
 748             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 749             {
 750                 const char *psz_attribs[] = { "face=", "family=", "size=",
 751                         "color=", "outline-color=", "shadow-color=",
 752                         "outline-level=", "shadow-level=", "back-color=",
 753                         "alpha=", NULL };
 754
 755                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 756                 strcat( psz_tag, "f" );
 757
 758                 while( *psz_subtitle != '>' )
 759                 {
 760                     int  k;
 761
 762                     for( k=0; psz_attribs[ k ]; k++ )
 763                     {
 764                         int i_len = strlen( psz_attribs[ k ] );
 765
 766                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 767                         {
 768                             /* */
 769                             HtmlPut( &psz_html, psz_attribs[k] );
 770                             psz_subtitle += i_len;
 771
 772                             /* */
 773                             if( *psz_subtitle == '"' )
 774                             {
 775                                 psz_subtitle++;
 776                                 i_len = strcspn( psz_subtitle, "\"" );
 777                             }
 778                             else
 779                             {
 780                                 i_len = strcspn( psz_subtitle, " \t>" );
 781                             }
 782                             HtmlPut( &psz_html, "\"" );
 783                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 784                                 HtmlPut( &psz_html, "#" );
 785                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 786                             HtmlPut( &psz_html, "\"" );
 787
 788                             psz_subtitle += i_len;
 789                             if( *psz_subtitle == '\"' )
 790                                 psz_subtitle++;
 791                             break;
 792                         }
 793                     }
 794                     if( psz_attribs[ k ] == NULL )
 795                     {
 796                         /* Jump over unrecognised tag */
 797                         int i_len = strcspn( psz_subtitle, "\"" ) + 1;
 798
 799                         i_len += strcspn( psz_subtitle + i_len, "\"" ) + 1;
 800                         psz_subtitle += i_len;
 801                     }
 802                     while (*psz_subtitle == ' ')
 803                         *psz_html++ = *psz_subtitle++;
 804                 }
 805                 *psz_html++ = *psz_subtitle++;
 806             }
 807             else if( !strncmp( psz_subtitle, "</", 2 ))
 808             {
 809                 bool   b_match     = false;
 810                 bool   b_ignore    = false;
 811                 int    i_len       = strlen( psz_tag ) - 1;
 812                 char  *psz_lastTag = NULL;
 813
 814                 if( i_len >= 0 )
 815                 {
 816                     psz_lastTag = psz_tag + i_len;
 817                     i_len = 0;
 818
 819                     switch( *psz_lastTag )
 820                     {
 821                     case 'b':
 822                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 823                         i_len   = 4;
 824                         break;
 825                     case 'i':
 826                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 827                         i_len   = 4;
 828                         break;
 829                     case 'u':
 830                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 831                         i_len   = 4;
 832                         break;
 833                     case 's':
 834                         b_match = !strncasecmp( psz_subtitle, "</s>", 4 );
 835                         i_len   = 4;
 836                         break;
 837                     case 'f':
 838                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 839                         i_len   = 7;
 840                         break;
 841                     case 'I':
 842                         i_len = strcspn( psz_subtitle, ">" );
 843                         b_match = psz_subtitle[i_len] == '>';
 844                         b_ignore = true;
 845                         if( b_match )
 846                             i_len++;
 847                         break;
 848                     }
 849                 }
 850                 if( !b_match )
 851                 {
 852                     /* Not well formed -- kill everything */
 853                     free( psz_html_start );
 854                     psz_html_start = NULL;
 855                     break;
 856                 }
 857                 *psz_lastTag = '\0';
 858                 if( !b_ignore )
 859                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 860
 861                 psz_subtitle += i_len;
 862             }
 863             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 864                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 865             {
 866                 /* We have a single < */
 867                 HtmlPut( &psz_html, "&lt;" );
 868                 psz_subtitle++;
 869             }
 870             else
 871             {
 872                 /* We have an unknown tag or a single < */
 873
 874                 /* Search for the next tag or end of tag or end of string */
 875                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 876                 char *psz_closing = strstr( psz_subtitle, "/>" );
 877
 878                 if( psz_closing && psz_closing < psz_stop )
 879                 {
 880                     /* We have a self closed tag, remove it */
 881                     psz_subtitle = &psz_closing[2];
 882                 }
 883                 else if( *psz_stop == '>' )
 884                 {
 885                     char psz_match[256];
 886
 887                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 888                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 889
 890                     if( strstr( psz_subtitle, psz_match ) )
 891                     {
 892                         /* We have the closing tag, ignore it TODO */
 893                         psz_subtitle = &psz_stop[1];
 894                         strcat( psz_tag, "I" );
 895                     }
 896                     else
 897                     {
 898                         int i_len = psz_stop + 1 - psz_subtitle;
 899
 900                         /* Copy the whole data */
 901                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 902                         {
 903                             if( *psz_subtitle == '<' )
 904                                 HtmlPut( &psz_html, "&lt;" );
 905                             else if( *psz_subtitle == '>' )
 906                                 HtmlPut( &psz_html, "&gt;" );
 907                             else
 908                                 *psz_html++ = *psz_subtitle;
 909                         }
 910                     }
 911                 }
 912                 else
 913                 {
 914                     /* We have a single < */
 915                     HtmlPut( &psz_html, "&lt;" );
 916                     psz_subtitle++;
 917                 }
 918             }
 919         }
 920         else if( *psz_subtitle == '&' )
 921         {
 922             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 923             {
 924                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 925             }
 926             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 927             {
 928                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 929             }
 930             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 931             {
 932                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 933             }
 934             else
 935             {
 936                 HtmlPut( &psz_html, "&amp;" );
 937                 psz_subtitle++;
 938             }
 939         }
 940         else if( *psz_subtitle == '>' )
 941         {
 942             HtmlPut( &psz_html, "&gt;" );
 943             psz_subtitle++;
 944         }
 945         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 946                  strchr( psz_subtitle, '}' ) )
 947         {
 948             /* Check for forced alignment */
 949             if( !b_has_align &&
 950                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 951             {
 952                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 953                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 954                 const int i_id = psz_subtitle[4] - '1';
 955
 956                 b_has_align = true;
 957                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 958             }
 959             /* TODO fr -> rotation */
 960
 961             /* Hide {\stupidity} */
 962             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 963         }
 964         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == 'Y'
 965                 && psz_subtitle[2] == ':' && strchr( psz_subtitle, '}' ) )
 966         {
 967             /* Hide {Y:stupidity} */
 968             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 969         }
 970         else if( psz_subtitle[0] == '\\' && psz_subtitle[1] )
 971         {
 972             if( psz_subtitle[1] == 'N' || psz_subtitle[1] == 'n' )
 973             {
 974                 HtmlPut( &psz_html, "<br/>" );
 975                 psz_subtitle += 2;
 976             }
 977             else if( psz_subtitle[1] == 'h' )
 978             {
 979                 /* Non breakable space */
 980                 HtmlPut( &psz_html, NO_BREAKING_SPACE );
 981                 psz_subtitle += 2;
 982             }
 983             else
 984             {
 985                 HtmlPut( &psz_html, "\\" );
 986                 psz_subtitle++;
 987             }
 988         }
 989         else
 990         {
 991             *psz_html = *psz_subtitle;
 992             if( psz_html > psz_html_start )
 993             {
 994                 /* Check for double whitespace */
 995                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
 996                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
 997                 {
 998                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
 999                     psz_html--;
1000                 }
1001             }
1002             psz_html++;
1003             psz_subtitle++;
1004         }
1005
1006         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
1007         {
1008             const int i_len = psz_html - psz_html_start;
1009
1010             i_buf_size += 200;
1011             char *psz_new = realloc( psz_html_start, i_buf_size );
1012             if( !psz_new )
1013                 break;
1014             psz_html_start = psz_new;
1015             psz_html = &psz_new[i_len];
1016         }
1017     }
1018     if( psz_html_start )
1019     {
1020         static const char *psz_text_close = "</text>";
1021         static const char *psz_tag_long = "/font>";
1022
1023         /* Realloc for closing tags and shrink memory */
1024         const size_t i_length = (size_t)( psz_html - psz_html_start );
1025
1026         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
1027         char *psz_new = realloc( psz_html_start, i_size );
1028         if( psz_new )
1029         {
1030             psz_html_start = psz_new;
1031             psz_html = &psz_new[i_length];
1032
1033             /* Close not well formed subtitle */
1034             while( *psz_tag )
1035             {
1036                 /* */
1037                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
1038                 switch( *psz_last )
1039                 {
1040                 case 'b':
1041                     HtmlPut( &psz_html, "</b>" );
1042                     break;
1043                 case 'i':
1044                     HtmlPut( &psz_html, "</i>" );
1045                     break;
1046                 case 'u':
1047                     HtmlPut( &psz_html, "</u>" );
1048                     break;
1049                 case 's':
1050                     HtmlPut( &psz_html, "</s>" );
1051                     break;
1052                 case 'f':
1053                     HtmlPut( &psz_html, "/font>" );
1054                     break;
1055                 case 'I':
1056                     break;
1057                 }
1058
1059                 *psz_last = '\0';
1060             }
1061             HtmlPut( &psz_html, psz_text_close );
1062         }
1063     }
1064     free( psz_tag );
1065
1066     return psz_html_start;
1067 }
1068