src/html.c

   1 /*
   2  * Sylpheed -- a GTK+ based, lightweight, and fast e-mail client
   3  * Copyright (C) 1999-2003 Hiroyuki Yamamoto
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  18  */
  19
  20 #include <glib.h>
  21 #include <stdio.h>
  22 #include <string.h>
  23 #include <ctype.h>
  24
  25 #include "html.h"
  26 #include "codeconv.h"
  27 #include "utils.h"
  28
  29 #define HTMLBUFSIZE     8192
  30 #define HR_STR          "------------------------------------------------"
  31
  32 typedef struct _HTMLSymbol      HTMLSymbol;
  33
  34 struct _HTMLSymbol
  35 {
  36         gchar *const key;
  37         gchar *const val;
  38 };
  39
  40 static HTMLSymbol symbol_list[] = {
  41         {"&lt;"    , "<"},
  42         {"&gt;"    , ">"},
  43         {"&amp;"   , "&"},
  44         {"&quot;"  , "\""},
  45         {"&nbsp;"  , " "},
  46         {"&trade;" , "(TM)"},
  47
  48         {"&#153;", "(TM)"},
  49 };
  50
  51 static HTMLSymbol ascii_symbol_list[] = {
  52         {"&iexcl;" , "^!"},
  53         {"&brvbar;", "|"},
  54         {"&copy;"  , "(C)"},
  55         {"&laquo;" , "<<"},
  56         {"&reg;"   , "(R)"},
  57
  58         {"&sup2;"  , "^2"},
  59         {"&sup3;"  , "^3"},
  60         {"&acute;" , "'"},
  61         {"&cedil;" , ","},
  62         {"&sup1;"  , "^1"},
  63         {"&raquo;" , ">>"},
  64         {"&frac14;", "1/4"},
  65         {"&frac12;", "1/2"},
  66         {"&frac34;", "3/4"},
  67         {"&iquest;", "^?"},
  68
  69         {"&Agrave;", "A`"},
  70         {"&Aacute;", "A'"},
  71         {"&Acirc;" , "A^"},
  72         {"&Atilde;", "A~"},
  73         {"&AElig;" , "AE"},
  74         {"&Egrave;", "E`"},
  75         {"&Eacute;", "E'"},
  76         {"&Ecirc;" , "E^"},
  77         {"&Igrave;", "I`"},
  78         {"&Iacute;", "I'"},
  79         {"&Icirc;" , "I^"},
  80
  81         {"&Ntilde;", "N~"},
  82         {"&Ograve;", "O`"},
  83         {"&Oacute;", "O'"},
  84         {"&Ocirc;" , "O^"},
  85         {"&Otilde;", "O~"},
  86         {"&Ugrave;", "U`"},
  87         {"&Uacute;", "U'"},
  88         {"&Ucirc;" , "U^"},
  89         {"&Yacute;", "Y'"},
  90
  91         {"&agrave;", "a`"},
  92         {"&aacute;", "a'"},
  93         {"&acirc;" , "a^"},
  94         {"&atilde;", "a~"},
  95         {"&aelig;" , "ae"},
  96         {"&egrave;", "e`"},
  97         {"&eacute;", "e'"},
  98         {"&ecirc;" , "e^"},
  99         {"&igrave;", "i`"},
 100         {"&iacute;", "i'"},
 101         {"&icirc;" , "i^"},
 102
 103         {"&ntilde;", "n~"},
 104         {"&ograve;", "o`"},
 105         {"&oacute;", "o'"},
 106         {"&ocirc;" , "o^"},
 107         {"&otilde;", "o~"},
 108         {"&ugrave;", "u`"},
 109         {"&uacute;", "u'"},
 110         {"&ucirc;" , "u^"},
 111         {"&yacute;", "y'"},
 112 };
 113
 114 static HTMLSymbol eucjp_symbol_list[] = {
 115         {"&iexcl;" , "^!"},
 116         {"&cent;"  , "\xa1\xf1"},
 117         {"&pound;" , "\xa1\xf2"},
 118         {"&yen;"   , "\xa1\xef"},
 119         {"&brvbar;", "|"},
 120         {"&sect;"  , "\xa1\xf8"},
 121         {"&uml;"   , "\xa1\xaf"},
 122         {"&copy;"  , "(C)"},
 123         {"&laquo;" , "<<"},
 124         {"&reg;"   , "(R)"},
 125
 126         {"&deg;"   , "\xa1\xeb"},
 127         {"&plusmn;", "\xa1\xde"},
 128         {"&sup2;"  , "^2"},
 129         {"&sup3;"  , "^3"},
 130         {"&acute;" , "'"},
 131         {"&micro;" , "\xa6\xcc"},
 132         {"&para;"  , "\xa2\xf9"},
 133         {"&middot;", "\xa1\xa6"},
 134         {"&cedil;" , ","},
 135         {"&sup1;"  , "^1"},
 136         {"&raquo;" , ">>"},
 137         {"&frac14;", "1/4"},
 138         {"&frac12;", "1/2"},
 139         {"&frac34;", "3/4"},
 140         {"&iquest;", "^?"},
 141
 142         {"&Agrave;", "A`"},
 143         {"&Aacute;", "A'"},
 144         {"&Acirc;" , "A^"},
 145         {"&Atilde;", "A~"},
 146         {"&Auml;"  , "A\xa1\xaf"},
 147         {"&Aring;" , "A\xa1\xeb"},
 148         {"&AElig;" , "AE"},
 149         {"&Egrave;", "E`"},
 150         {"&Eacute;", "E'"},
 151         {"&Ecirc;" , "E^"},
 152         {"&Euml;"  , "E\xa1\xaf"},
 153         {"&Igrave;", "I`"},
 154         {"&Iacute;", "I'"},
 155         {"&Icirc;" , "I^"},
 156         {"&Iuml;"  , "I\xa1\xaf"},
 157
 158         {"&Ntilde;", "N~"},
 159         {"&Ograve;", "O`"},
 160         {"&Oacute;", "O'"},
 161         {"&Ocirc;" , "O^"},
 162         {"&Otilde;", "O~"},
 163         {"&Ouml;"  , "O\xa1\xaf"},
 164         {"&times;" , "\xa1\xdf"},
 165         {"&Ugrave;", "U`"},
 166         {"&Uacute;", "U'"},
 167         {"&Ucirc;" , "U^"},
 168         {"&Uuml;"  , "U\xa1\xaf"},
 169         {"&Yacute;", "Y'"},
 170
 171         {"&agrave;", "a`"},
 172         {"&aacute;", "a'"},
 173         {"&acirc;" , "a^"},
 174         {"&atilde;", "a~"},
 175         {"&auml;"  , "a\xa1\xaf"},
 176         {"&aring;" , "a\xa1\xeb"},
 177         {"&aelig;" , "ae"},
 178         {"&egrave;", "e`"},
 179         {"&eacute;", "e'"},
 180         {"&ecirc;" , "e^"},
 181         {"&euml;"  , "e\xa1\xaf"},
 182         {"&igrave;", "i`"},
 183         {"&iacute;", "i'"},
 184         {"&icirc;" , "i^"},
 185         {"&iuml;"  , "i\xa1\xaf"},
 186
 187         {"&eth;"   , "\xa2\xdf"},
 188         {"&ntilde;", "n~"},
 189         {"&ograve;", "o`"},
 190         {"&oacute;", "o'"},
 191         {"&ocirc;" , "o^"},
 192         {"&otilde;", "o~"},
 193         {"&ouml;"  , "o\xa1\xaf"},
 194         {"&divide;", "\xa1\xe0"},
 195         {"&ugrave;", "u`"},
 196         {"&uacute;", "u'"},
 197         {"&ucirc;" , "u^"},
 198         {"&uuml;"  , "u\xa1\xaf"},
 199         {"&yacute;", "y'"},
 200         {"&yuml;"  , "y\xa1\xaf"},
 201 };
 202
 203 static HTMLSymbol latin_symbol_list[] = {
 204         {"&iexcl;" , "\xa1"},
 205         {"&cent;"  , "\xa2"},
 206         {"&pound;" , "\xa3"},
 207         {"&curren;", "\xa4"},
 208         {"&yen;"   , "\xa5"},
 209         {"&brvbar;", "\xa6"},
 210         {"&sect;"  , "\xa7"},
 211         {"&uml;"   , "\xa8"},
 212         {"&copy;"  , "\xa9"},
 213         {"&ordf;"  , "\xaa"},
 214         {"&laquo;" , "\xab"},
 215         {"&not;"   , "\xac"},
 216         {"&shy;"   , "\xad"},
 217         {"&reg;"   , "\xae"},
 218         {"&macr;"  , "\xaf"},
 219
 220         {"&deg;"   , "\xb0"},
 221         {"&plusmn;", "\xb1"},
 222         {"&sup2;"  , "\xb2"},
 223         {"&sup3;"  , "\xb3"},
 224         {"&acute;" , "\xb4"},
 225         {"&micro;" , "\xb5"},
 226         {"&para;"  , "\xb6"},
 227         {"&middot;", "\xb7"},
 228         {"&cedil;" , "\xb8"},
 229         {"&sup1;"  , "\xb9"},
 230         {"&ordm;"  , "\xba"},
 231         {"&raquo;" , "\xbb"},
 232         {"&frac14;", "\xbc"},
 233         {"&frac12;", "\xbd"},
 234         {"&frac34;", "\xbe"},
 235         {"&iquest;", "\xbf"},
 236
 237         {"&Agrave;", "\xc0"},
 238         {"&Aacute;", "\xc1"},
 239         {"&Acirc;" , "\xc2"},
 240         {"&Atilde;", "\xc3"},
 241         {"&Auml;"  , "\xc4"},
 242         {"&Aring;" , "\xc5"},
 243         {"&AElig;" , "\xc6"},
 244         {"&Ccedil;", "\xc7"},
 245         {"&Egrave;", "\xc8"},
 246         {"&Eacute;", "\xc9"},
 247         {"&Ecirc;" , "\xca"},
 248         {"&Euml;"  , "\xcb"},
 249         {"&Igrave;", "\xcc"},
 250         {"&Iacute;", "\xcd"},
 251         {"&Icirc;" , "\xce"},
 252         {"&Iuml;"  , "\xcf"},
 253
 254         {"&ETH;"   , "\xd0"},
 255         {"&Ntilde;", "\xd1"},
 256         {"&Ograve;", "\xd2"},
 257         {"&Oacute;", "\xd3"},
 258         {"&Ocirc;" , "\xd4"},
 259         {"&Otilde;", "\xd5"},
 260         {"&Ouml;"  , "\xd6"},
 261         {"&times;" , "\xd7"},
 262         {"&Oslash;", "\xd8"},
 263         {"&Ugrave;", "\xd9"},
 264         {"&Uacute;", "\xda"},
 265         {"&Ucirc;" , "\xdb"},
 266         {"&Uuml;"  , "\xdc"},
 267         {"&Yacute;", "\xdd"},
 268         {"&THORN;" , "\xde"},
 269         {"&szlig;" , "\xdf"},
 270
 271         {"&agrave;", "\xe0"},
 272         {"&aacute;", "\xe1"},
 273         {"&acirc;" , "\xe2"},
 274         {"&atilde;", "\xe3"},
 275         {"&auml;"  , "\xe4"},
 276         {"&aring;" , "\xe5"},
 277         {"&aelig;" , "\xe6"},
 278         {"&ccedil;", "\xe7"},
 279         {"&egrave;", "\xe8"},
 280         {"&eacute;", "\xe9"},
 281         {"&ecirc;" , "\xea"},
 282         {"&euml;"  , "\xeb"},
 283         {"&igrave;", "\xec"},
 284         {"&iacute;", "\xed"},
 285         {"&icirc;" , "\xee"},
 286         {"&iuml;"  , "\xef"},
 287
 288         {"&eth;"   , "\xf0"},
 289         {"&ntilde;", "\xf1"},
 290         {"&ograve;", "\xf2"},
 291         {"&oacute;", "\xf3"},
 292         {"&ocirc;" , "\xf4"},
 293         {"&otilde;", "\xf5"},
 294         {"&ouml;"  , "\xf6"},
 295         {"&divide;", "\xf7"},
 296         {"&oslash;", "\xf8"},
 297         {"&ugrave;", "\xf9"},
 298         {"&uacute;", "\xfa"},
 299         {"&ucirc;" , "\xfb"},
 300         {"&uuml;"  , "\xfc"},
 301         {"&yacute;", "\xfd"},
 302         {"&thorn;" , "\xfe"},
 303         {"&yuml;"  , "\xff"},
 304 };
 305
 306 static GHashTable *default_symbol_table;
 307 static GHashTable *eucjp_symbol_table;
 308 static GHashTable *latin_symbol_table;
 309
 310 static HTMLState html_read_line         (HTMLParser     *parser);
 311 static void html_append_char            (HTMLParser     *parser,
 312                                          gchar           ch);
 313 static void html_append_str             (HTMLParser     *parser,
 314                                          const gchar    *str,
 315                                          gint            len);
 316 static HTMLState html_parse_tag         (HTMLParser     *parser);
 317 static void html_parse_special          (HTMLParser     *parser);
 318 static void html_get_parenthesis        (HTMLParser     *parser,
 319                                          gchar          *buf,
 320                                          gint            len);
 321
 322
 323 HTMLParser *html_parser_new(FILE *fp, CodeConverter *conv)
 324 {
 325         HTMLParser *parser;
 326
 327         g_return_val_if_fail(fp != NULL, NULL);
 328         g_return_val_if_fail(conv != NULL, NULL);
 329
 330         parser = g_new0(HTMLParser, 1);
 331         parser->fp = fp;
 332         parser->conv = conv;
 333         parser->str = g_string_new(NULL);
 334         parser->buf = g_string_new(NULL);
 335         parser->bufp = parser->buf->str;
 336         parser->state = HTML_NORMAL;
 337         parser->href = NULL;
 338         parser->newline = TRUE;
 339         parser->empty_line = TRUE;
 340         parser->space = FALSE;
 341         parser->pre = FALSE;
 342
 343 #define SYMBOL_TABLE_ADD(table, list) \
 344 { \
 345         gint i; \
 346  \
 347         for (i = 0; i < sizeof(list) / sizeof(list[0]); i++) \
 348                 g_hash_table_insert(table, list[i].key, list[i].val); \
 349 }
 350
 351         if (!default_symbol_table) {
 352                 default_symbol_table =
 353                         g_hash_table_new(g_str_hash, g_str_equal);
 354                 SYMBOL_TABLE_ADD(default_symbol_table, symbol_list);
 355                 SYMBOL_TABLE_ADD(default_symbol_table, ascii_symbol_list);
 356         }
 357         if (!eucjp_symbol_table) {
 358                 eucjp_symbol_table =
 359                         g_hash_table_new(g_str_hash, g_str_equal);
 360                 SYMBOL_TABLE_ADD(eucjp_symbol_table, symbol_list);
 361                 SYMBOL_TABLE_ADD(eucjp_symbol_table, eucjp_symbol_list);
 362         }
 363         if (!latin_symbol_table) {
 364                 latin_symbol_table =
 365                         g_hash_table_new(g_str_hash, g_str_equal);
 366                 SYMBOL_TABLE_ADD(latin_symbol_table, symbol_list);
 367                 SYMBOL_TABLE_ADD(latin_symbol_table, latin_symbol_list);
 368         }
 369
 370 #undef SYMBOL_TABLE_ADD
 371
 372         if (conv->charset == C_ISO_8859_1)
 373                 parser->symbol_table = latin_symbol_table;
 374         else if ((conv->charset == C_ISO_2022_JP   ||
 375                   conv->charset == C_ISO_2022_JP_2 ||
 376                   conv->charset == C_EUC_JP        ||
 377                   conv->charset == C_SHIFT_JIS) &&
 378                  conv_get_current_charset() == C_EUC_JP)
 379                 parser->symbol_table = eucjp_symbol_table;
 380         else
 381                 parser->symbol_table = default_symbol_table;
 382
 383         return parser;
 384 }
 385
 386 void html_parser_destroy(HTMLParser *parser)
 387 {
 388         g_string_free(parser->str, TRUE);
 389         g_string_free(parser->buf, TRUE);
 390         g_free(parser->href);
 391         g_free(parser);
 392 }
 393
 394 gchar *html_parse(HTMLParser *parser)
 395 {
 396         parser->state = HTML_NORMAL;
 397         g_string_truncate(parser->str, 0);
 398
 399         if (*parser->bufp == '\0') {
 400                 g_string_truncate(parser->buf, 0);
 401                 parser->bufp = parser->buf->str;
 402                 if (html_read_line(parser) == HTML_EOF)
 403                         return NULL;
 404         }
 405
 406         while (*parser->bufp != '\0') {
 407                 switch (*parser->bufp) {
 408                 case '<': {
 409                         HTMLState st;
 410                         st = html_parse_tag(parser);
 411                         /* when we see an href, we need to flush the str
 412                          * buffer.  Then collect all the chars until we
 413                          * see the end anchor tag
 414                          */
 415                         if (HTML_HREF_BEG == st || HTML_HREF == st)
 416                             return parser->str->str;
 417                         } break;
 418                 case '&':
 419                         html_parse_special(parser);
 420                         break;
 421                 case ' ':
 422                 case '\t':
 423                 case '\r':
 424                 case '\n':
 425                         if (parser->bufp[0] == '\r' && parser->bufp[1] == '\n')
 426                                 parser->bufp++;
 427
 428                         if (!parser->pre) {
 429                                 if (!parser->newline)
 430                                         parser->space = TRUE;
 431
 432                                 parser->bufp++;
 433                                 break;
 434                         }
 435                         /* fallthrough */
 436                 default:
 437                         html_append_char(parser, *parser->bufp++);
 438                 }
 439         }
 440
 441         return parser->str->str;
 442 }
 443
 444 static HTMLState html_read_line(HTMLParser *parser)
 445 {
 446         gchar buf[HTMLBUFSIZE];
 447         gchar buf2[HTMLBUFSIZE];
 448         gint index;
 449
 450         if (fgets(buf, sizeof(buf), parser->fp) == NULL) {
 451                 parser->state = HTML_EOF;
 452                 return HTML_EOF;
 453         }
 454
 455         if (conv_convert(parser->conv, buf2, sizeof(buf2), buf) < 0) {
 456                 index = parser->bufp - parser->buf->str;
 457
 458                 conv_localetodisp(buf2, sizeof(buf2), buf);
 459                 g_string_append(parser->buf, buf2);
 460
 461                 parser->bufp = parser->buf->str + index;
 462
 463                 return HTML_CONV_FAILED;
 464         }
 465
 466         index = parser->bufp - parser->buf->str;
 467
 468         g_string_append(parser->buf, buf2);
 469
 470         parser->bufp = parser->buf->str + index;
 471
 472         return HTML_NORMAL;
 473 }
 474
 475 static void html_append_char(HTMLParser *parser, gchar ch)
 476 {
 477         GString *str = parser->str;
 478
 479         if (!parser->pre && parser->space) {
 480                 g_string_append_c(str, ' ');
 481                 parser->space = FALSE;
 482         }
 483
 484         g_string_append_c(str, ch);
 485
 486         parser->empty_line = FALSE;
 487         if (ch == '\n') {
 488                 parser->newline = TRUE;
 489                 if (str->len > 1 && str->str[str->len - 2] == '\n')
 490                         parser->empty_line = TRUE;
 491         } else
 492                 parser->newline = FALSE;
 493 }
 494
 495 static void html_append_str(HTMLParser *parser, const gchar *str, gint len)
 496 {
 497         GString *string = parser->str;
 498
 499         if (!parser->pre && parser->space) {
 500                 g_string_append_c(string, ' ');
 501                 parser->space = FALSE;
 502         }
 503
 504         if (len == 0) return;
 505         if (len < 0)
 506                 g_string_append(string, str);
 507         else {
 508                 gchar *s;
 509                 Xstrndup_a(s, str, len, return);
 510                 g_string_append(string, s);
 511         }
 512
 513         parser->empty_line = FALSE;
 514         if (string->len > 0 && string->str[string->len - 1] == '\n') {
 515                 parser->newline = TRUE;
 516                 if (string->len > 1 && string->str[string->len - 2] == '\n')
 517                         parser->empty_line = TRUE;
 518         } else
 519                 parser->newline = FALSE;
 520 }
 521
 522 static HTMLTag *html_get_tag(const gchar *str)
 523 {
 524         HTMLTag *tag;
 525         gchar *tmp;
 526         guchar *tmpp;
 527
 528         g_return_val_if_fail(str != NULL, NULL);
 529
 530         if (*str == '\0' || *str == '!') return NULL;
 531
 532         Xstrdup_a(tmp, str, return NULL);
 533
 534         tag = g_new0(HTMLTag, 1);
 535
 536         for (tmpp = tmp; *tmpp != '\0' && !isspace(*tmpp); tmpp++)
 537                 ;
 538
 539         if (*tmpp == '\0') {
 540                 g_strdown(tmp);
 541                 tag->name = g_strdup(tmp);
 542                 return tag;
 543         } else {
 544                 *tmpp++ = '\0';
 545                 g_strdown(tmp);
 546                 tag->name = g_strdup(tmp);
 547         }
 548
 549         while (*tmpp != '\0') {
 550                 HTMLAttr *attr;
 551                 gchar *attr_name;
 552                 gchar *attr_value;
 553                 gchar *p;
 554                 gchar quote;
 555
 556                 while (isspace(*tmpp)) tmpp++;
 557                 attr_name = tmpp;
 558
 559                 while (*tmpp != '\0' && !isspace(*tmpp) && *tmpp != '=') tmpp++;
 560                 if (*tmpp != '\0' && *tmpp != '=') {
 561                         *tmpp++ = '\0';
 562                         while (isspace(*tmpp)) tmpp++;
 563                 }
 564
 565                 if (*tmpp == '=') {
 566                         *tmpp++ = '\0';
 567                         while (isspace(*tmpp)) tmpp++;
 568
 569                         if (*tmpp == '"' || *tmpp == '\'') {
 570                                 /* name="value" */
 571                                 quote = *tmpp;
 572                                 tmpp++;
 573                                 attr_value = tmpp;
 574                                 if ((p = strchr(attr_value, quote)) == NULL) {
 575                                         g_warning("html_get_tag(): syntax error in tag: '%s'\n", str);
 576                                         return tag;
 577                                 }
 578                                 tmpp = p;
 579                                 *tmpp++ = '\0';
 580                                 while (isspace(*tmpp)) tmpp++;
 581                         } else {
 582                                 /* name=value */
 583                                 attr_value = tmpp;
 584                                 while (*tmpp != '\0' && !isspace(*tmpp)) tmpp++;
 585                                 if (*tmpp != '\0')
 586                                         *tmpp++ = '\0';
 587                         }
 588                 } else
 589                         attr_value = "";
 590
 591                 g_strchomp(attr_name);
 592                 g_strdown(attr_name);
 593                 attr = g_new(HTMLAttr, 1);
 594                 attr->name = g_strdup(attr_name);
 595                 attr->value = g_strdup(attr_value);
 596                 tag->attr = g_list_append(tag->attr, attr);
 597         }
 598
 599         return tag;
 600 }
 601
 602 static void html_free_tag(HTMLTag *tag)
 603 {
 604         if (!tag) return;
 605
 606         g_free(tag->name);
 607         while (tag->attr != NULL) {
 608                 HTMLAttr *attr = (HTMLAttr *)tag->attr->data;
 609                 g_free(attr->name);
 610                 g_free(attr->value);
 611                 g_free(attr);
 612                 tag->attr = g_list_remove(tag->attr, tag->attr->data);
 613         }
 614         g_free(tag);
 615 }
 616
 617 static HTMLState html_parse_tag(HTMLParser *parser)
 618 {
 619         gchar buf[HTMLBUFSIZE];
 620         HTMLTag *tag;
 621
 622         html_get_parenthesis(parser, buf, sizeof(buf));
 623
 624         tag = html_get_tag(buf);
 625
 626         parser->state = HTML_UNKNOWN;
 627         if (!tag) return HTML_UNKNOWN;
 628
 629         if (!strcmp(tag->name, "br")) {
 630                 parser->space = FALSE;
 631                 html_append_char(parser, '\n');
 632                 parser->state = HTML_BR;
 633         } else if (!strcmp(tag->name, "a")) {
 634                 if (tag->attr && tag->attr->data &&
 635                     !strcmp(((HTMLAttr *)tag->attr->data)->name, "href")) {
 636                         g_free(parser->href);
 637                         parser->href =
 638                                 g_strdup(((HTMLAttr *)tag->attr->data)->value);
 639                         parser->state = HTML_HREF_BEG;
 640                 }
 641         } else if (!strcmp(tag->name, "/a")) {
 642                 parser->state = HTML_HREF;
 643         } else if (!strcmp(tag->name, "p")) {
 644                 parser->space = FALSE;
 645                 if (!parser->empty_line) {
 646                         parser->space = FALSE;
 647                         if (!parser->newline) html_append_char(parser, '\n');
 648                         html_append_char(parser, '\n');
 649                 }
 650                 parser->state = HTML_PAR;
 651         } else if (!strcmp(tag->name, "pre")) {
 652                 parser->pre = TRUE;
 653                 parser->state = HTML_PRE;
 654         } else if (!strcmp(tag->name, "/pre")) {
 655                 parser->pre = FALSE;
 656                 parser->state = HTML_NORMAL;
 657         } else if (!strcmp(tag->name, "hr")) {
 658                 if (!parser->newline) {
 659                         parser->space = FALSE;
 660                         html_append_char(parser, '\n');
 661                 }
 662                 html_append_str(parser, HR_STR "\n", -1);
 663                 parser->state = HTML_HR;
 664         } else if (!strcmp(tag->name, "div")    ||
 665                    !strcmp(tag->name, "ul")     ||
 666                    !strcmp(tag->name, "li")     ||
 667                    !strcmp(tag->name, "table")  ||
 668                    !strcmp(tag->name, "tr")     ||
 669                    (tag->name[0] == 'h' && isdigit((guchar)tag->name[1]))) {
 670                 if (!parser->newline) {
 671                         parser->space = FALSE;
 672                         html_append_char(parser, '\n');
 673                 }
 674                 parser->state = HTML_NORMAL;
 675         } else if (!strcmp(tag->name, "/table") ||
 676                    (tag->name[0] == '/' &&
 677                     tag->name[1] == 'h' &&
 678                     isdigit((guchar)tag->name[1]))) {
 679                 if (!parser->empty_line) {
 680                         parser->space = FALSE;
 681                         if (!parser->newline) html_append_char(parser, '\n');
 682                         html_append_char(parser, '\n');
 683                 }
 684                 parser->state = HTML_NORMAL;
 685         } else if (!strcmp(tag->name, "/div")   ||
 686                    !strcmp(tag->name, "/ul")    ||
 687                    !strcmp(tag->name, "/li")) {
 688                 if (!parser->newline) {
 689                         parser->space = FALSE;
 690                         html_append_char(parser, '\n');
 691                 }
 692                 parser->state = HTML_NORMAL;
 693                         }
 694
 695         html_free_tag(tag);
 696
 697         return parser->state;
 698 }
 699
 700 static void html_parse_special(HTMLParser *parser)
 701 {
 702         gchar symbol_name[9];
 703         gint n;
 704         const gchar *val;
 705
 706         parser->state = HTML_UNKNOWN;
 707         g_return_if_fail(*parser->bufp == '&');
 708
 709         /* &foo; */
 710         for (n = 0; parser->bufp[n] != '\0' && parser->bufp[n] != ';'; n++)
 711                 ;
 712         if (n > 7 || parser->bufp[n] != ';') {
 713                 /* output literal `&' */
 714                 html_append_char(parser, *parser->bufp++);
 715                 parser->state = HTML_NORMAL;
 716                 return;
 717         }
 718         strncpy2(symbol_name, parser->bufp, n + 2);
 719         parser->bufp += n + 1;
 720
 721         if ((val = g_hash_table_lookup(parser->symbol_table, symbol_name))
 722             != NULL) {
 723                 html_append_str(parser, val, -1);
 724                 parser->state = HTML_NORMAL;
 725                 return;
 726         } else if (symbol_name[1] == '#' && isdigit((guchar)symbol_name[2])) {
 727                 gint ch;
 728
 729                 ch = atoi(symbol_name + 2);
 730                 if ((ch > 0 && ch <= 127) ||
 731                     (ch >= 128 && ch <= 255 &&
 732                      parser->conv->charset == C_ISO_8859_1)) {
 733                         html_append_char(parser, ch);
 734                         parser->state = HTML_NORMAL;
 735                         return;
 736                 } else {
 737                     char *symb = NULL;
 738                     switch (ch) {
 739                     /* http://www.w3schools.com/html/html_entitiesref.asp */
 740                     case 338:   /* capital ligature OE  &OElig;  */
 741                         symb = "OE";  break;
 742                     case 339:   /* small ligature OE    &oelig;  */
 743                         symb = "oe";  break;
 744                     case 352:   /* capital S w/caron    &Scaron; */
 745                     case 353:   /* small S w/caron      &scaron; */
 746                     case 376:   /* cap Y w/ diaeres     &Yuml;   */
 747                         break;
 748                     case 710:   /* circumflex accent    &circ;   */
 749                         symb = "^";  break;
 750                     case 732:   /* small tilde          &tilde;  */
 751                         symb = "~";  break;
 752                     case 8194:  /* en space             &ensp;   */
 753                     case 8195:  /* em space             &emsp;   */
 754                     case 8201:  /* thin space           &thinsp; */
 755                         symb = " ";  break;
 756                     case 8204:  /* zero width non-joiner &zwnj;  */
 757                     case 8205:  /* zero width joiner    &zwj;    */
 758                     case 8206:  /* l-t-r mark           &lrm;    */
 759                     case 8207:  /* r-t-l mark           &rlm     */
 760                         break;
 761                     case 8211:  /* en dash              &ndash;  */
 762                         symb = "-";  break;
 763                     case 8212:  /* em dash              &mdash;  */
 764                         symb = "--";  break;
 765                     case 8216:  /* l single quot mark   &lsquo;  */
 766                         symb = "`";  break;
 767                     case 8217:  /* r single quot mark   &rsquo;  */
 768                         symb = "'";  break;
 769                     case 8218:  /* single low-9 quot    &sbquo;  */
 770                         symb = ",";  break;
 771                     case 8220:  /* l double quot mark   &ldquo;  */
 772                         symb = "``";  break;
 773                     case 8221:  /* r double quot mark   &rdquo;  */
 774                         symb = "''";  break;
 775                     case 8222:  /* double low-9 quot    &bdquo;  */
 776                         symb = ",,";  break;
 777                     case 8224:  /* dagger               &dagger; */
 778                     case 8225:  /* double dagger        &Dagger; */
 779                         break;
 780                     case 8230:  /* horizontal ellipsis  &hellip; */
 781                         symb = "...";  break;
 782                     case 8240:  /* per mile             &permil; */
 783                         symb = "\%o";  break;
 784                     case 8249:  /* l-pointing angle quot &lsaquo; */
 785                         symb = "<";  break;
 786                     case 8250:  /* r-pointing angle quot &rsaquo; */
 787                         symb = ">";  break;
 788                     case 8364:  /* euro                 &euro;   */
 789                         symb = "&euro";  break;
 790                     case 8482:  /* trademark            &trade;  */
 791                         symb  = "(TM)";  break;
 792                     default: break;
 793                     }
 794                     if (symb) {
 795                         html_append_str(parser, symb, -1);
 796                         parser->state = HTML_NORMAL;
 797                         return;
 798                     }
 799                 }
 800         }
 801
 802         html_append_str(parser, symbol_name, -1);
 803 }
 804
 805 static void html_get_parenthesis(HTMLParser *parser, gchar *buf, gint len)
 806 {
 807         gchar *p;
 808
 809         buf[0] = '\0';
 810         g_return_if_fail(*parser->bufp == '<');
 811
 812         /* ignore comment / CSS / script stuff */
 813         if (!strncmp(parser->bufp, "<!--", 4)) {
 814                 parser->bufp += 4;
 815                 while ((p = strstr(parser->bufp, "-->")) == NULL)
 816                         if (html_read_line(parser) == HTML_EOF) return;
 817                 parser->bufp = p + 3;
 818                 return;
 819         }
 820         if (!g_ascii_strncasecmp(parser->bufp, "<style", 6)) {
 821                 parser->bufp += 6;
 822                 while ((p = strcasestr(parser->bufp, "</style>")) == NULL)
 823                         if (html_read_line(parser) == HTML_EOF) return;
 824                 parser->bufp = p + 8;
 825                 return;
 826         }
 827         if (!g_ascii_strncasecmp(parser->bufp, "<script", 7)) {
 828                 parser->bufp += 7;
 829                 while ((p = strcasestr(parser->bufp, "</script>")) == NULL)
 830                         if (html_read_line(parser) == HTML_EOF) return;
 831                 parser->bufp = p + 9;
 832                 return;
 833         }
 834
 835         parser->bufp++;
 836         while ((p = strchr(parser->bufp, '>')) == NULL)
 837                 if (html_read_line(parser) == HTML_EOF) return;
 838
 839         strncpy2(buf, parser->bufp, MIN(p - parser->bufp + 1, len));
 840         g_strstrip(buf);
 841         parser->bufp = p + 1;
 842 }