src/html.c

   1 /*
   2  * Sylpheed -- a GTK+ based, lightweight, and fast e-mail client
   3  * Copyright (C) 1999,2000 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 #if 0
 323 static gint g_str_case_equal            (gconstpointer   v,
 324                                          gconstpointer   v2);
 325 static guint g_str_case_hash            (gconstpointer   key);
 326 #endif
 327
 328 HTMLParser *html_parser_new(FILE *fp, CodeConverter *conv)
 329 {
 330         HTMLParser *parser;
 331
 332         g_return_val_if_fail(fp != NULL, NULL);
 333         g_return_val_if_fail(conv != NULL, NULL);
 334
 335         parser = g_new0(HTMLParser, 1);
 336         parser->fp = fp;
 337         parser->conv = conv;
 338         parser->str = g_string_new(NULL);
 339         parser->buf = g_string_new(NULL);
 340         parser->bufp = parser->buf->str;
 341         parser->newline = TRUE;
 342         parser->empty_line = TRUE;
 343         parser->space = FALSE;
 344         parser->pre = FALSE;
 345
 346 #define SYMBOL_TABLE_ADD(table, list) \
 347 { \
 348         gint i; \
 349  \
 350         for (i = 0; i < sizeof(list) / sizeof(list[0]); i++) \
 351                 g_hash_table_insert(table, list[i].key, list[i].val); \
 352 }
 353
 354         if (!default_symbol_table) {
 355                 default_symbol_table =
 356                         g_hash_table_new(g_str_hash, g_str_equal);
 357                 SYMBOL_TABLE_ADD(default_symbol_table, symbol_list);
 358                 SYMBOL_TABLE_ADD(default_symbol_table, ascii_symbol_list);
 359         }
 360         if (!eucjp_symbol_table) {
 361                 eucjp_symbol_table =
 362                         g_hash_table_new(g_str_hash, g_str_equal);
 363                 SYMBOL_TABLE_ADD(eucjp_symbol_table, symbol_list);
 364                 SYMBOL_TABLE_ADD(eucjp_symbol_table, eucjp_symbol_list);
 365         }
 366         if (!latin_symbol_table) {
 367                 latin_symbol_table =
 368                         g_hash_table_new(g_str_hash, g_str_equal);
 369                 SYMBOL_TABLE_ADD(latin_symbol_table, symbol_list);
 370                 SYMBOL_TABLE_ADD(latin_symbol_table, latin_symbol_list);
 371         }
 372
 373 #undef SYMBOL_TABLE_ADD
 374
 375         if (conv->charset == C_ISO_8859_1)
 376                 parser->symbol_table = latin_symbol_table;
 377         else if ((conv->charset == C_ISO_2022_JP   ||
 378                   conv->charset == C_ISO_2022_JP_2 ||
 379                   conv->charset == C_EUC_JP        ||
 380                   conv->charset == C_SHIFT_JIS) &&
 381                  conv_get_current_charset() == C_EUC_JP)
 382                 parser->symbol_table = eucjp_symbol_table;
 383         else
 384                 parser->symbol_table = default_symbol_table;
 385
 386         return parser;
 387 }
 388
 389 void html_parser_destroy(HTMLParser *parser)
 390 {
 391         g_string_free(parser->str, TRUE);
 392         g_string_free(parser->buf, TRUE);
 393         g_free(parser);
 394 }
 395
 396 gchar *html_parse(HTMLParser *parser)
 397 {
 398         parser->state = HTML_NORMAL;
 399         g_string_truncate(parser->str, 0);
 400
 401         if (*parser->bufp == '\0') {
 402                 g_string_truncate(parser->buf, 0);
 403                 parser->bufp = parser->buf->str;
 404                 if (html_read_line(parser) == HTML_EOF)
 405                         return NULL;
 406         }
 407
 408         while (*parser->bufp != '\0') {
 409                 switch (*parser->bufp) {
 410                 case '<':
 411                         if (parser->str->len == 0)
 412                                 html_parse_tag(parser);
 413                         else
 414                                 return parser->str->str;
 415                         break;
 416                 case '&':
 417                         html_parse_special(parser);
 418                         break;
 419                 case ' ':
 420                 case '\t':
 421                 case '\r':
 422                 case '\n':
 423                         if (parser->bufp[0] == '\r' && parser->bufp[1] == '\n')
 424                                 parser->bufp++;
 425
 426                         if (!parser->pre) {
 427                                 if (!parser->newline)
 428                                         parser->space = TRUE;
 429
 430                                 parser->bufp++;
 431                                 break;
 432                         }
 433                         /* fallthrough */
 434                 default:
 435                         html_append_char(parser, *parser->bufp++);
 436                 }
 437         }
 438
 439         return parser->str->str;
 440 }
 441
 442 static HTMLState html_read_line(HTMLParser *parser)
 443 {
 444         gchar buf[HTMLBUFSIZE];
 445         gchar buf2[HTMLBUFSIZE];
 446         gint index;
 447
 448         if (fgets(buf, sizeof(buf), parser->fp) == NULL) {
 449                 parser->state = HTML_EOF;
 450                 return HTML_EOF;
 451         }
 452
 453         if (conv_convert(parser->conv, buf2, sizeof(buf2), buf) < 0) {
 454                 index = parser->bufp - parser->buf->str;
 455
 456                 conv_localetodisp(buf2, sizeof(buf2), buf);
 457                 g_string_append(parser->buf, buf2);
 458
 459                 parser->bufp = parser->buf->str + index;
 460
 461                 return HTML_CONV_FAILED;
 462         }
 463
 464         index = parser->bufp - parser->buf->str;
 465
 466         g_string_append(parser->buf, buf2);
 467
 468         parser->bufp = parser->buf->str + index;
 469
 470         return HTML_NORMAL;
 471 }
 472
 473 static void html_append_char(HTMLParser *parser, gchar ch)
 474 {
 475         GString *str = parser->str;
 476
 477         if (!parser->pre && parser->space) {
 478                 g_string_append_c(str, ' ');
 479                 parser->space = FALSE;
 480         }
 481
 482         g_string_append_c(str, ch);
 483
 484         parser->empty_line = FALSE;
 485         if (ch == '\n') {
 486                 parser->newline = TRUE;
 487                 if (str->len > 1 && str->str[str->len - 2] == '\n')
 488                         parser->empty_line = TRUE;
 489         } else
 490                 parser->newline = FALSE;
 491 }
 492
 493 static void html_append_str(HTMLParser *parser, const gchar *str, gint len)
 494 {
 495         GString *string = parser->str;
 496
 497         if (!parser->pre && parser->space) {
 498                 g_string_append_c(string, ' ');
 499                 parser->space = FALSE;
 500         }
 501
 502         if (len == 0) return;
 503         if (len < 0)
 504                 g_string_append(string, str);
 505         else {
 506                 gchar *s;
 507                 Xstrndup_a(s, str, len, return);
 508                 g_string_append(string, s);
 509         }
 510
 511         parser->empty_line = FALSE;
 512         if (string->len > 0 && string->str[string->len - 1] == '\n') {
 513                 parser->newline = TRUE;
 514                 if (string->len > 1 && string->str[string->len - 2] == '\n')
 515                         parser->empty_line = TRUE;
 516         } else
 517                 parser->newline = FALSE;
 518 }
 519
 520 static HTMLState html_parse_tag(HTMLParser *parser)
 521 {
 522         gchar buf[HTMLBUFSIZE];
 523         gchar *p;
 524         static gboolean is_in_href = FALSE;
 525
 526         html_get_parenthesis(parser, buf, sizeof(buf));
 527
 528         for (p = buf; *p != '\0'; p++) {
 529                 if (isspace(*p)) {
 530                         *p = '\0';
 531                         break;
 532                 }
 533         }
 534
 535         parser->state = HTML_UNKNOWN;
 536         if (buf[0] == '\0') return parser->state;
 537
 538         g_strdown(buf);
 539
 540         if (!strcmp(buf, "br")) {
 541                 parser->space = FALSE;
 542                 html_append_char(parser, '\n');
 543                 parser->state = HTML_BR;
 544         } else if (!strcmp(buf, "a")) {
 545                 /* look for tokens separated by space or = */
 546                 char* href_token = strtok(++p, " =");
 547                 parser->state = HTML_NORMAL;
 548                 while (href_token != NULL) {
 549                         /* look for href */
 550                         if (!strcmp(href_token, "href")) {
 551                                 /* the next token is the url, between double
 552                                          * quotes */
 553                                 char* url = strtok(NULL, "\"");
 554                                         if (url && url[0] == '\'')
 555                                           url = strtok(url,"\'");
 556
 557                                 if (!url) break;
 558                                 html_append_str(parser, url, strlen(url));
 559                                 html_append_char(parser, ' ');
 560                                 /* start enforcing html link */
 561                                 parser->state = HTML_HREF;
 562                                 is_in_href = TRUE;
 563                                 break;
 564                         }
 565                         /* or get next token */
 566                         href_token = strtok(NULL, " =");
 567                 }
 568         } else if (!strcmp(buf, "/a")) {
 569                 /* stop enforcing html link */
 570                 parser->state = HTML_NORMAL;
 571                 is_in_href = FALSE;
 572         } else if (!strcmp(buf, "p")) {
 573                 parser->space = FALSE;
 574                 if (!parser->empty_line) {
 575                         parser->space = FALSE;
 576                         if (!parser->newline) html_append_char(parser, '\n');
 577                         html_append_char(parser, '\n');
 578                 }
 579                 parser->state = HTML_PAR;
 580         } else if (!strcmp(buf, "pre")) {
 581                 parser->pre = TRUE;
 582                 parser->state = HTML_PRE;
 583         } else if (!strcmp(buf, "/pre")) {
 584                 parser->pre = FALSE;
 585                 parser->state = HTML_NORMAL;
 586         } else if (!strcmp(buf, "hr")) {
 587                 if (!parser->newline) {
 588                         parser->space = FALSE;
 589                         html_append_char(parser, '\n');
 590                 }
 591                 html_append_str(parser, HR_STR "\n", -1);
 592                 parser->state = HTML_HR;
 593         } else if (!strcmp(buf, "div")    ||
 594                    !strcmp(buf, "ul")     ||
 595                    !strcmp(buf, "li")     ||
 596                    !strcmp(buf, "table")  ||
 597                    !strcmp(buf, "tr")     ||
 598                    (buf[0] == 'h' && isdigit(buf[1]))) {
 599                 if (!parser->newline) {
 600                         parser->space = FALSE;
 601                         html_append_char(parser, '\n');
 602                 }
 603                 parser->state = HTML_NORMAL;
 604         } else if (!strcmp(buf, "/table") ||
 605                    (buf[0] == '/' && buf[1] == 'h' && isdigit(buf[1]))) {
 606                 if (!parser->empty_line) {
 607                         parser->space = FALSE;
 608                         if (!parser->newline) html_append_char(parser, '\n');
 609                         html_append_char(parser, '\n');
 610                 }
 611                 parser->state = HTML_NORMAL;
 612         } else if (!strcmp(buf, "/div")   ||
 613                    !strcmp(buf, "/ul")    ||
 614                    !strcmp(buf, "/li")) {
 615                 if (!parser->newline) {
 616                         parser->space = FALSE;
 617                         html_append_char(parser, '\n');
 618                 }
 619                 parser->state = HTML_NORMAL;
 620         }
 621
 622         if (is_in_href == TRUE) {
 623                 /* when inside a link, everything will be written as
 624                  * clickable (see textview_show_thml in textview.c) */
 625                 parser->state = HTML_HREF;
 626         }
 627
 628         return parser->state;
 629 }
 630
 631 static void html_parse_special(HTMLParser *parser)
 632 {
 633         gchar symbol_name[9];
 634         gint n;
 635         const gchar *val;
 636
 637         parser->state = HTML_UNKNOWN;
 638         g_return_if_fail(*parser->bufp == '&');
 639
 640         /* &foo; */
 641         for (n = 0; parser->bufp[n] != '\0' && parser->bufp[n] != ';'; n++)
 642                 ;
 643         if (n > 7 || parser->bufp[n] != ';') {
 644                 /* output literal `&' */
 645                 html_append_char(parser, *parser->bufp++);
 646                 parser->state = HTML_NORMAL;
 647                 return;
 648         }
 649         strncpy2(symbol_name, parser->bufp, n + 2);
 650         parser->bufp += n + 1;
 651
 652         if ((val = g_hash_table_lookup(parser->symbol_table, symbol_name))
 653             != NULL) {
 654                 html_append_str(parser, val, -1);
 655                 parser->state = HTML_NORMAL;
 656                 return;
 657         } else if (symbol_name[1] == '#' && isdigit(symbol_name[2])) {
 658                 gint ch;
 659
 660                 ch = atoi(symbol_name + 2);
 661                 if ((ch > 0 && ch <= 127) ||
 662                     (ch >= 128 && ch <= 255 &&
 663                      parser->conv->charset == C_ISO_8859_1)) {
 664                         html_append_char(parser, ch);
 665                         parser->state = HTML_NORMAL;
 666                         return;
 667                 }
 668         }
 669
 670         html_append_str(parser, symbol_name, -1);
 671 }
 672
 673 static void html_get_parenthesis(HTMLParser *parser, gchar *buf, gint len)
 674 {
 675         gchar *p;
 676
 677         buf[0] = '\0';
 678         g_return_if_fail(*parser->bufp == '<');
 679
 680         /* ignore comment / CSS / script stuff */
 681         if (!strncmp(parser->bufp, "<!--", 4)) {
 682                 parser->bufp += 4;
 683                 while ((p = strstr(parser->bufp, "-->")) == NULL)
 684                         if (html_read_line(parser) == HTML_EOF) return;
 685                 parser->bufp = p + 3;
 686                 return;
 687         }
 688         if (!g_strncasecmp(parser->bufp, "<style", 6)) {
 689                 parser->bufp += 6;
 690                 while ((p = strcasestr(parser->bufp, "</style>")) == NULL)
 691                         if (html_read_line(parser) == HTML_EOF) return;
 692                 parser->bufp = p + 8;
 693                 return;
 694         }
 695         if (!g_strncasecmp(parser->bufp, "<script", 7)) {
 696                 parser->bufp += 7;
 697                 while ((p = strcasestr(parser->bufp, "</script>")) == NULL)
 698                         if (html_read_line(parser) == HTML_EOF) return;
 699                 parser->bufp = p + 9;
 700                 return;
 701         }
 702
 703         parser->bufp++;
 704         while ((p = strchr(parser->bufp, '>')) == NULL)
 705                 if (html_read_line(parser) == HTML_EOF) return;
 706
 707         strncpy2(buf, parser->bufp, MIN(p - parser->bufp + 1, len));
 708         parser->bufp = p + 1;
 709 }
 710
 711 /* these hash functions were taken from gstring.c in glib */
 712 #if 0
 713 static gint g_str_case_equal(gconstpointer v, gconstpointer v2)
 714 {
 715         return strcasecmp((const gchar *)v, (const gchar *)v2) == 0;
 716 }
 717
 718 static guint g_str_case_hash(gconstpointer key)
 719 {
 720         const gchar *p = key;
 721         guint h = *p;
 722
 723         if (h) {
 724                 h = tolower(h);
 725                 for (p += 1; *p != '\0'; p++)
 726                         h = (h << 5) - h + tolower(*p);
 727         }
 728
 729         return h;
 730 }
 731 #endif