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