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_locale_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                         }
 418                         break;
 419                 case '&':
 420                         html_parse_special(parser);
 421                         break;
 422                 case ' ':
 423                 case '\t':
 424                 case '\r':
 425                 case '\n':
 426                         if (parser->bufp[0] == '\r' && parser->bufp[1] == '\n')
 427                                 parser->bufp++;
 428
 429                         if (!parser->pre) {
 430                                 if (!parser->newline)
 431                                         parser->space = TRUE;
 432
 433                                 parser->bufp++;
 434                                 break;
 435                         }
 436                         /* fallthrough */
 437                 default:
 438                         html_append_char(parser, *parser->bufp++);
 439                 }
 440         }
 441
 442         return parser->str->str;
 443 }
 444
 445 static HTMLState html_read_line(HTMLParser *parser)
 446 {
 447         gchar buf[HTMLBUFSIZE];
 448         gchar buf2[HTMLBUFSIZE];
 449         gint index;
 450
 451         if (fgets(buf, sizeof(buf), parser->fp) == NULL) {
 452                 parser->state = HTML_EOF;
 453                 return HTML_EOF;
 454         }
 455
 456         if (conv_convert(parser->conv, buf2, sizeof(buf2), buf) < 0) {
 457                 index = parser->bufp - parser->buf->str;
 458
 459                 conv_utf8todisp(buf2, sizeof(buf2), buf);
 460                 g_string_append(parser->buf, buf2);
 461
 462                 parser->bufp = parser->buf->str + index;
 463
 464                 return HTML_CONV_FAILED;
 465         }
 466
 467         index = parser->bufp - parser->buf->str;
 468
 469         g_string_append(parser->buf, buf2);
 470
 471         parser->bufp = parser->buf->str + index;
 472
 473         return HTML_NORMAL;
 474 }
 475
 476 static void html_append_char(HTMLParser *parser, gchar ch)
 477 {
 478         GString *str = parser->str;
 479
 480         if (!parser->pre && parser->space) {
 481                 g_string_append_c(str, ' ');
 482                 parser->space = FALSE;
 483         }
 484
 485         g_string_append_c(str, ch);
 486
 487         parser->empty_line = FALSE;
 488         if (ch == '\n') {
 489                 parser->newline = TRUE;
 490                 if (str->len > 1 && str->str[str->len - 2] == '\n')
 491                         parser->empty_line = TRUE;
 492         } else
 493                 parser->newline = FALSE;
 494 }
 495
 496 static void html_append_str(HTMLParser *parser, const gchar *str, gint len)
 497 {
 498         GString *string = parser->str;
 499
 500         if (!parser->pre && parser->space) {
 501                 g_string_append_c(string, ' ');
 502                 parser->space = FALSE;
 503         }
 504
 505         if (len == 0) return;
 506         if (len < 0)
 507                 g_string_append(string, str);
 508         else {
 509                 gchar *s;
 510                 Xstrndup_a(s, str, len, return);
 511                 g_string_append(string, s);
 512         }
 513
 514         parser->empty_line = FALSE;
 515         if (string->len > 0 && string->str[string->len - 1] == '\n') {
 516                 parser->newline = TRUE;
 517                 if (string->len > 1 && string->str[string->len - 2] == '\n')
 518                         parser->empty_line = TRUE;
 519         } else
 520                 parser->newline = FALSE;
 521 }
 522
 523 static HTMLTag *html_get_tag(const gchar *str)
 524 {
 525         HTMLTag *tag;
 526         gchar *tmp;
 527         guchar *tmpp;
 528
 529         g_return_val_if_fail(str != NULL, NULL);
 530
 531         if (*str == '\0' || *str == '!') return NULL;
 532
 533         Xstrdup_a(tmp, str, return NULL);
 534
 535         tag = g_new0(HTMLTag, 1);
 536
 537         for (tmpp = tmp; *tmpp != '\0' && !isspace(*tmpp); tmpp++)
 538                 ;
 539
 540         if (*tmpp == '\0') {
 541                 g_strdown(tmp);
 542                 tag->name = g_strdup(tmp);
 543                 return tag;
 544         } else {
 545                 *tmpp++ = '\0';
 546                 g_strdown(tmp);
 547                 tag->name = g_strdup(tmp);
 548         }
 549
 550         while (*tmpp != '\0') {
 551                 HTMLAttr *attr;
 552                 gchar *attr_name;
 553                 gchar *attr_value;
 554                 gchar *p;
 555                 gchar quote;
 556
 557                 while (isspace(*tmpp)) tmpp++;
 558                 attr_name = tmpp;
 559
 560                 while (*tmpp != '\0' && !isspace(*tmpp) && *tmpp != '=') tmpp++;
 561                 if (*tmpp != '\0' && *tmpp != '=') {
 562                         *tmpp++ = '\0';
 563                         while (isspace(*tmpp)) tmpp++;
 564                 }
 565
 566                 if (*tmpp == '=') {
 567                         *tmpp++ = '\0';
 568                         while (isspace(*tmpp)) tmpp++;
 569
 570                         if (*tmpp == '"' || *tmpp == '\'') {
 571                                 /* name="value" */
 572                                 quote = *tmpp;
 573                                 tmpp++;
 574                                 attr_value = tmpp;
 575                                 if ((p = strchr(attr_value, quote)) == NULL) {
 576                                         g_warning("html_get_tag(): syntax error in tag: '%s'\n", str);
 577                                         return tag;
 578                                 }
 579                                 tmpp = p;
 580                                 *tmpp++ = '\0';
 581                                 while (isspace(*tmpp)) tmpp++;
 582                         } else {
 583                                 /* name=value */
 584                                 attr_value = tmpp;
 585                                 while (*tmpp != '\0' && !isspace(*tmpp)) tmpp++;
 586                                 if (*tmpp != '\0')
 587                                         *tmpp++ = '\0';
 588                         }
 589                 } else
 590                         attr_value = "";
 591
 592                 g_strchomp(attr_name);
 593                 g_strdown(attr_name);
 594                 attr = g_new(HTMLAttr, 1);
 595                 attr->name = g_strdup(attr_name);
 596                 attr->value = g_strdup(attr_value);
 597                 tag->attr = g_list_append(tag->attr, attr);
 598         }
 599
 600         return tag;
 601 }
 602
 603 static void html_free_tag(HTMLTag *tag)
 604 {
 605         if (!tag) return;
 606
 607         g_free(tag->name);
 608         while (tag->attr != NULL) {
 609                 HTMLAttr *attr = (HTMLAttr *)tag->attr->data;
 610                 g_free(attr->name);
 611                 g_free(attr->value);
 612                 g_free(attr);
 613                 tag->attr = g_list_remove(tag->attr, tag->attr->data);
 614         }
 615         g_free(tag);
 616 }
 617
 618 static HTMLState html_parse_tag(HTMLParser *parser)
 619 {
 620         gchar buf[HTMLBUFSIZE];
 621         HTMLTag *tag;
 622
 623         html_get_parenthesis(parser, buf, sizeof(buf));
 624
 625         tag = html_get_tag(buf);
 626
 627         parser->state = HTML_UNKNOWN;
 628         if (!tag) return HTML_UNKNOWN;
 629
 630         if (!strcmp(tag->name, "br")) {
 631                 parser->space = FALSE;
 632                 html_append_char(parser, '\n');
 633                 parser->state = HTML_BR;
 634         } else if (!strcmp(tag->name, "a")) {
 635                 if (tag->attr && tag->attr->data &&
 636                     !strcmp(((HTMLAttr *)tag->attr->data)->name, "href")) {
 637                         g_free(parser->href);
 638                         parser->href =
 639                                 g_strdup(((HTMLAttr *)tag->attr->data)->value);
 640                         parser->state = HTML_HREF_BEG;
 641                 }
 642         } else if (!strcmp(tag->name, "/a")) {
 643                 parser->state = HTML_HREF;
 644         } else if (!strcmp(tag->name, "p")) {
 645                 parser->space = FALSE;
 646                 if (!parser->empty_line) {
 647                         parser->space = FALSE;
 648                         if (!parser->newline) html_append_char(parser, '\n');
 649                         html_append_char(parser, '\n');
 650                 }
 651                 parser->state = HTML_PAR;
 652         } else if (!strcmp(tag->name, "pre")) {
 653                 parser->pre = TRUE;
 654                 parser->state = HTML_PRE;
 655         } else if (!strcmp(tag->name, "/pre")) {
 656                 parser->pre = FALSE;
 657                 parser->state = HTML_NORMAL;
 658         } else if (!strcmp(tag->name, "hr")) {
 659                 if (!parser->newline) {
 660                         parser->space = FALSE;
 661                         html_append_char(parser, '\n');
 662                 }
 663                 html_append_str(parser, HR_STR "\n", -1);
 664                 parser->state = HTML_HR;
 665         } else if (!strcmp(tag->name, "div")    ||
 666                    !strcmp(tag->name, "ul")     ||
 667                    !strcmp(tag->name, "li")     ||
 668                    !strcmp(tag->name, "table")  ||
 669                    !strcmp(tag->name, "tr")     ||
 670                    (tag->name[0] == 'h' && isdigit((guchar)tag->name[1]))) {
 671                 if (!parser->newline) {
 672                         parser->space = FALSE;
 673                         html_append_char(parser, '\n');
 674                 }
 675                 parser->state = HTML_NORMAL;
 676         } else if (!strcmp(tag->name, "/table") ||
 677                    (tag->name[0] == '/' &&
 678                     tag->name[1] == 'h' &&
 679                     isdigit((guchar)tag->name[1]))) {
 680                 if (!parser->empty_line) {
 681                         parser->space = FALSE;
 682                         if (!parser->newline) html_append_char(parser, '\n');
 683                         html_append_char(parser, '\n');
 684                 }
 685                 parser->state = HTML_NORMAL;
 686         } else if (!strcmp(tag->name, "/div")   ||
 687                    !strcmp(tag->name, "/ul")    ||
 688                    !strcmp(tag->name, "/li")) {
 689                 if (!parser->newline) {
 690                         parser->space = FALSE;
 691                         html_append_char(parser, '\n');
 692                 }
 693                 parser->state = HTML_NORMAL;
 694                         }
 695
 696         html_free_tag(tag);
 697
 698         return parser->state;
 699 }
 700
 701 static void html_parse_special(HTMLParser *parser)
 702 {
 703         gchar symbol_name[9];
 704         gint n;
 705         const gchar *val;
 706
 707         parser->state = HTML_UNKNOWN;
 708         g_return_if_fail(*parser->bufp == '&');
 709
 710         /* &foo; */
 711         for (n = 0; parser->bufp[n] != '\0' && parser->bufp[n] != ';'; n++)
 712                 ;
 713         if (n > 7 || parser->bufp[n] != ';') {
 714                 /* output literal `&' */
 715                 html_append_char(parser, *parser->bufp++);
 716                 parser->state = HTML_NORMAL;
 717                 return;
 718         }
 719         strncpy2(symbol_name, parser->bufp, n + 2);
 720         parser->bufp += n + 1;
 721
 722         if ((val = g_hash_table_lookup(parser->symbol_table, symbol_name))
 723             != NULL) {
 724                 html_append_str(parser, val, -1);
 725                 parser->state = HTML_NORMAL;
 726                 return;
 727         } else if (symbol_name[1] == '#' && isdigit((guchar)symbol_name[2])) {
 728                 gint ch;
 729
 730                 ch = atoi(symbol_name + 2);
 731                 if ((ch > 0 && ch <= 127) ||
 732                     (ch >= 128 && ch <= 255 &&
 733                      parser->conv->charset == C_ISO_8859_1)) {
 734                         html_append_char(parser, ch);
 735                         parser->state = HTML_NORMAL;
 736                         return;
 737                 } else {
 738                         char *symb = NULL;
 739                         switch (ch) {
 740                         /* http://www.w3schools.com/html/html_entitiesref.asp */
 741                         case 338:       /* capital ligature OE  &OElig;  */
 742                                 symb = "OE";
 743                                 break;
 744                         case 339:       /* small ligature OE    &oelig;  */
 745                                 symb = "oe";
 746                                 break;
 747                         case 352:       /* capital S w/caron    &Scaron; */
 748                         case 353:       /* small S w/caron      &scaron; */
 749                         case 376:       /* cap Y w/ diaeres     &Yuml;   */
 750                                 break;
 751                         case 710:       /* circumflex accent    &circ;   */
 752                                 symb = "^";
 753                                 break;
 754                         case 732:       /* small tilde          &tilde;  */
 755                                 symb = "~";
 756                                 break;
 757                         case 8194:      /* en space             &ensp;   */
 758                         case 8195:      /* em space             &emsp;   */
 759                         case 8201:      /* thin space           &thinsp; */
 760                                 symb = " ";
 761                                 break;
 762                         case 8204:      /* zero width non-joiner &zwnj;  */
 763                         case 8205:      /* zero width joiner    &zwj;   */
 764                         case 8206:      /* l-t-r mark           &lrm;   */
 765                         case 8207:      /* r-t-l mark           &rlm     */
 766                                 break;
 767                         case 8211:      /* en dash              &ndash;  */
 768                                 symb = "-";
 769                                 break;
 770                         case 8212:      /* em dash              &mdash;  */
 771                                 symb = "--";
 772                                 break;
 773                         case 8216:      /* l single quot mark   &lsquo;  */
 774                                 symb = "`";
 775                                 break;
 776                         case 8217:      /* r single quot mark   &rsquo;  */
 777                                 symb = "'";
 778                                 break;
 779                         case 8218:      /* single low-9 quot    &sbquo;  */
 780                                 symb = ",";
 781                                 break;
 782                         case 8220:      /* l double quot mark   &ldquo;  */
 783                                 symb = "``";
 784                                 break;
 785                         case 8221:      /* r double quot mark   &rdquo;  */
 786                                 symb = "''";
 787                                 break;
 788                         case 8222:      /* double low-9 quot    &bdquo;  */
 789                                 symb = ",,";
 790                                 break;
 791                         case 8224:      /* dagger               &dagger; */
 792                         case 8225:      /* double dagger        &Dagger; */
 793                                 break;
 794                         case 8230:      /* horizontal ellipsis  &hellip; */
 795                                 symb = "...";
 796                                 break;
 797                         case 8240:      /* per mile             &permil; */
 798                                 symb = "\%o";
 799                                 break;
 800                         case 8249:      /* l-pointing angle quot &lsaquo; */
 801                                 symb = "<";
 802                                 break;
 803                         case 8250:      /* r-pointing angle quot &rsaquo; */
 804                                 symb = ">";
 805                                 break;
 806                         case 8364:      /* euro                 &euro;   */
 807                                 symb = "EUR";
 808                                 break;
 809                         case 8482:      /* trademark            &trade;  */
 810                                 symb  = "(TM)";
 811                                 break;
 812                         default:
 813                                 break;
 814                         }
 815                         if (symb) {
 816                                 html_append_str(parser, symb, -1);
 817                                 parser->state = HTML_NORMAL;
 818                                 return;
 819                         }
 820                 }
 821         }
 822
 823         html_append_str(parser, symbol_name, -1);
 824 }
 825
 826 static void html_get_parenthesis(HTMLParser *parser, gchar *buf, gint len)
 827 {
 828         gchar *p;
 829
 830         buf[0] = '\0';
 831         g_return_if_fail(*parser->bufp == '<');
 832
 833         /* ignore comment / CSS / script stuff */
 834         if (!strncmp(parser->bufp, "<!--", 4)) {
 835                 parser->bufp += 4;
 836                 while ((p = strstr(parser->bufp, "-->")) == NULL)
 837                         if (html_read_line(parser) == HTML_EOF) return;
 838                 parser->bufp = p + 3;
 839                 return;
 840         }
 841         if (!g_ascii_strncasecmp(parser->bufp, "<style", 6)) {
 842                 parser->bufp += 6;
 843                 while ((p = strcasestr(parser->bufp, "</style>")) == NULL)
 844                         if (html_read_line(parser) == HTML_EOF) return;
 845                 parser->bufp = p + 8;
 846                 return;
 847         }
 848         if (!g_ascii_strncasecmp(parser->bufp, "<script", 7)) {
 849                 parser->bufp += 7;
 850                 while ((p = strcasestr(parser->bufp, "</script>")) == NULL)
 851                         if (html_read_line(parser) == HTML_EOF) return;
 852                 parser->bufp = p + 9;
 853                 return;
 854         }
 855
 856         parser->bufp++;
 857         while ((p = strchr(parser->bufp, '>')) == NULL)
 858                 if (html_read_line(parser) == HTML_EOF) return;
 859
 860         strncpy2(buf, parser->bufp, MIN(p - parser->bufp + 1, len));
 861         g_strstrip(buf);
 862         parser->bufp = p + 1;
 863 }