webcit/html.c

   1 /*
   2  * $Id: html.c 6014 2008-02-04 18:38:35Z ajc $
   3  *
   4  * Functions which handle translation between HTML and plain text
   5  * Copyright (c) 2000-2005 by Art Cancro and others.   This program is
   6  * released under the terms of the GNU General Public License.
   7  */
   8
   9 #include "webcit.h"
  10
  11
  12 /*
  13  * Convert HTML to plain text.
  14  *
  15  * inputmsg      = pointer to raw HTML message
  16  * screenwidth   = desired output screenwidth
  17  * do_citaformat = set to 1 to indent newlines with spaces
  18  */
  19 char *html_to_ascii(char *inputmsg, int msglen, int screenwidth, int do_citaformat) {
  20         char inbuf[SIZ];
  21         int inbuf_len = 0;
  22         char outbuf[SIZ];
  23         char tag[1024];
  24         int done_reading = 0;
  25         char *inptr;
  26         char *outptr;
  27         size_t outptr_buffer_size;
  28         size_t output_len = 0;
  29         int i, j, ch, did_out, rb, scanch;
  30         int nest = 0;           /* Bracket nesting level */
  31         int blockquote = 0;     /* BLOCKQUOTE nesting level */
  32         int styletag = 0;       /* STYLE tag nesting level */
  33         int styletag_start = 0;
  34         int bytes_processed = 0;
  35         char nl[128];
  36
  37         strcpy(nl, "\n");
  38         inptr = inputmsg;
  39         strcpy(inbuf, "");
  40         strcpy(outbuf, "");
  41         if (msglen == 0) msglen = strlen(inputmsg);
  42
  43         outptr_buffer_size = strlen(inptr) + SIZ;
  44         outptr = malloc(outptr_buffer_size);
  45         if (outptr == NULL) return NULL;
  46         strcpy(outptr, "");
  47         output_len = 0;
  48
  49         do {
  50                 /* Fill the input buffer */
  51                 inbuf_len = strlen(inbuf);
  52                 if ( (done_reading == 0) && (inbuf_len < (SIZ-128)) ) {
  53
  54                         ch = *inptr++;
  55                         if (ch != 0) {
  56                                 inbuf[inbuf_len++] = ch;
  57                                 inbuf[inbuf_len] = 0;
  58                         }
  59                         else {
  60                                 done_reading = 1;
  61                         }
  62
  63                         ++bytes_processed;
  64                         if (bytes_processed > msglen) {
  65                                 done_reading = 1;
  66                         }
  67
  68                 }
  69
  70                 /* Do some parsing */
  71                 if (!IsEmptyStr(inbuf)) {
  72
  73
  74                     /* Fold in all the spacing */
  75                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
  76                         if (inbuf[i]==10) inbuf[i]=32;
  77                         if (inbuf[i]==13) inbuf[i]=32;
  78                         if (inbuf[i]==9) inbuf[i]=32;
  79                         /*** we like foreign characters now.
  80                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
  81                                 inbuf[i] = '?';
  82                         } */
  83                     }
  84                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
  85                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
  86                                 strcpy(&inbuf[i], &inbuf[i+1]);
  87                     }
  88
  89                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
  90
  91                         ch = inbuf[i];
  92
  93                         if (ch == '<') {
  94                                 ++nest;
  95                                 strcpy(tag, "");
  96                         }
  97
  98                         else if (ch == '>') {   /* We have a tag. */
  99                                 if (nest > 0) --nest;
 100
 101                                 /* Unqualify the tag (truncate at first space) */
 102                                 if (strchr(tag, ' ') != NULL) {
 103                                         strcpy(strchr(tag, ' '), "");
 104                                 }
 105
 106                                 if (!strcasecmp(tag, "P")) {
 107                                         strcat(outbuf, nl);
 108                                         strcat(outbuf, nl);
 109                                 }
 110
 111                                 if (!strcasecmp(tag, "/DIV")) {
 112                                         strcat(outbuf, nl);
 113                                         strcat(outbuf, nl);
 114                                 }
 115
 116                                 if (!strcasecmp(tag, "LI")) {
 117                                         strcat(outbuf, nl);
 118                                         strcat(outbuf, " * ");
 119                                 }
 120
 121                                 else if (!strcasecmp(tag, "/UL")) {
 122                                         strcat(outbuf, nl);
 123                                         strcat(outbuf, nl);
 124                                 }
 125
 126                                 else if (!strcasecmp(tag, "H1")) {
 127                                         strcat(outbuf, nl);
 128                                         strcat(outbuf, nl);
 129                                 }
 130
 131                                 else if (!strcasecmp(tag, "H2")) {
 132                                         strcat(outbuf, nl);
 133                                         strcat(outbuf, nl);
 134                                 }
 135
 136                                 else if (!strcasecmp(tag, "H3")) {
 137                                         strcat(outbuf, nl);
 138                                         strcat(outbuf, nl);
 139                                 }
 140
 141                                 else if (!strcasecmp(tag, "H4")) {
 142                                         strcat(outbuf, nl);
 143                                         strcat(outbuf, nl);
 144                                 }
 145
 146                                 else if (!strcasecmp(tag, "/H1")) {
 147                                         strcat(outbuf, nl);
 148                                 }
 149
 150                                 else if (!strcasecmp(tag, "/H2")) {
 151                                         strcat(outbuf, nl);
 152                                 }
 153
 154                                 else if (!strcasecmp(tag, "/H3")) {
 155                                         strcat(outbuf, nl);
 156                                 }
 157
 158                                 else if (!strcasecmp(tag, "/H4")) {
 159                                         strcat(outbuf, nl);
 160                                 }
 161
 162                                 else if (!strcasecmp(tag, "HR")) {
 163                                         strcat(outbuf, nl);
 164                                         strcat(outbuf, " ");
 165                                         for (j=0; j<screenwidth-2; ++j)
 166                                                 strcat(outbuf, "-");
 167                                         strcat(outbuf, nl);
 168                                 }
 169
 170                                 else if (!strcasecmp(tag, "BR")) {
 171                                         strcat(outbuf, nl);
 172                                 }
 173
 174                                 else if (!strcasecmp(tag, "TR")) {
 175                                         strcat(outbuf, nl);
 176                                 }
 177
 178                                 else if (!strcasecmp(tag, "/TABLE")) {
 179                                         strcat(outbuf, nl);
 180                                 }
 181
 182                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 183                                         ++blockquote;
 184                                         strcpy(nl, "\n");
 185                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 186                                         strcat(outbuf, nl);
 187                                 }
 188
 189                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 190                                         strcat(outbuf, "\n");
 191                                         --blockquote;
 192                                         strcpy(nl, "\n");
 193                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 194                                         strcat(outbuf, nl);
 195                                 }
 196
 197                                 else if (!strcasecmp(tag, "STYLE")) {
 198                                         ++styletag;
 199                                         if (styletag == 1) {
 200                                                 styletag_start = strlen(outbuf);
 201                                         }
 202                                 }
 203
 204                                 else if (!strcasecmp(tag, "/STYLE")) {
 205                                         --styletag;
 206                                         if (styletag == 0) {
 207                                                 outbuf[styletag_start] = 0;
 208                                         }
 209                                 }
 210
 211                         }
 212
 213                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 214                                 tag[strlen(tag)+1] = 0;
 215                                 tag[strlen(tag)] = ch;
 216                         }
 217
 218                         else if (!nest) {
 219                                 outbuf[strlen(outbuf)+1] = 0;
 220                                 outbuf[strlen(outbuf)] = ch;
 221                         }
 222                     }
 223                     strcpy(inbuf, &inbuf[i]);
 224                 }
 225
 226                 /* Convert &; tags to the forbidden characters */
 227                 if (!IsEmptyStr(outbuf)) for (i=0; !IsEmptyStr(&outbuf[i]); ++i) {
 228
 229                         /* Character entity references */
 230                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 231                                 outbuf[i] = ' ';
 232                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 233                         }
 234
 235                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 236                                 outbuf[i] = ' ';
 237                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 238                         }
 239
 240                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 241                                 outbuf[i] = ' ';
 242                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 243                         }
 244
 245                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 246                                 outbuf[i] = ' ';
 247                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 248                         }
 249
 250                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 251                                 outbuf[i] = '<';
 252                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 253                         }
 254
 255                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 256                                 outbuf[i] = '>';
 257                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 258                         }
 259
 260                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 261                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 262                         }
 263
 264                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 265                                 outbuf[i] = '\"';
 266                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 267                         }
 268
 269                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 270                                 outbuf[i] = '`';
 271                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 272                         }
 273
 274                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 275                                 outbuf[i] = '\'';
 276                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 277                         }
 278
 279                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 280                                 outbuf[i] = '(';
 281                                 outbuf[i+1] = 'c';
 282                                 outbuf[i+2] = ')';
 283                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 284                         }
 285
 286                         else if (!strncasecmp(&outbuf[i], "&bull;", 6)) {
 287                                 outbuf[i] = ' ';
 288                                 outbuf[i+1] = '*';
 289                                 outbuf[i+2] = ' ';
 290                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 291                         }
 292
 293                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 294                                 outbuf[i] = '.';
 295                                 outbuf[i+1] = '.';
 296                                 outbuf[i+2] = '.';
 297                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 298                         }
 299
 300                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 301                                 outbuf[i] = '(';
 302                                 outbuf[i+1] = 't';
 303                                 outbuf[i+2] = 'm';
 304                                 outbuf[i+3] = ')';
 305                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 306                         }
 307
 308                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 309                                 outbuf[i] = '(';
 310                                 outbuf[i+1] = 'r';
 311                                 outbuf[i+2] = ')';
 312                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 313                         }
 314
 315                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 316                                 outbuf[i] = '1';
 317                                 outbuf[i+1] = '/';
 318                                 outbuf[i+2] = '4';
 319                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 320                         }
 321
 322                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 323                                 outbuf[i] = '1';
 324                                 outbuf[i+1] = '/';
 325                                 outbuf[i+2] = '2';
 326                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 327                         }
 328
 329                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 330                                 outbuf[i] = '3';
 331                                 outbuf[i+1] = '/';
 332                                 outbuf[i+2] = '4';
 333                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 334                         }
 335
 336                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 337                                 outbuf[i] = '-';
 338                                 outbuf[i+1] = '-';
 339                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 340                         }
 341
 342                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 343                                 outbuf[i] = '-';
 344                                 outbuf[i+1] = '-';
 345                                 outbuf[i+2] = '-';
 346                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 347                         }
 348
 349                         else if (!strncmp(&outbuf[i], "&Ccedil;", 8)) {
 350                                 outbuf[i] = 'C';
 351                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 352                         }
 353
 354                         else if (!strncasecmp(&outbuf[i], "&ccedil;", 8)) {
 355                                 outbuf[i] = 'c';
 356                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 357                         }
 358
 359                         else if (!strncmp(&outbuf[i], "&Egrave;", 8)) {
 360                                 outbuf[i] = 'E';
 361                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 362                         }
 363
 364                         else if (!strncasecmp(&outbuf[i], "&egrave;", 8)) {
 365                                 outbuf[i] = 'e';
 366                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 367                         }
 368
 369                         else if (!strncmp(&outbuf[i], "&Ecirc;", 7)) {
 370                                 outbuf[i] = 'E';
 371                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 372                         }
 373
 374                         else if (!strncasecmp(&outbuf[i], "&ecirc;", 7)) {
 375                                 outbuf[i] = 'e';
 376                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 377                         }
 378
 379                         else if (!strncmp(&outbuf[i], "&Eacute;", 8)) {
 380                                 outbuf[i] = 'E';
 381                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 382                         }
 383
 384                         else if (!strncasecmp(&outbuf[i], "&eacute;", 8)) {
 385                                 outbuf[i] = 'e';
 386                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 387                         }
 388
 389                         else if (!strncmp(&outbuf[i], "&Agrave;", 8)) {
 390                                 outbuf[i] = 'A';
 391                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 392                         }
 393
 394                         else if (!strncasecmp(&outbuf[i], "&agrave;", 8)) {
 395                                 outbuf[i] = 'a';
 396                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 397                         }
 398
 399                         else if (!strncasecmp(&outbuf[i], "&ldquo;", 7)) {
 400                                 outbuf[i] = '\"';
 401                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 402                         }
 403
 404                         else if (!strncasecmp(&outbuf[i], "&rdquo;", 7)) {
 405                                 outbuf[i] = '\"';
 406                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 407                         }
 408
 409                         else if (!strncasecmp(&outbuf[i], "&acute;", 7)) {
 410                                 outbuf[i] = '\'';
 411                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 412                         }
 413
 414                         /* two-digit decimal equivalents */
 415                         else if ((!strncmp(&outbuf[i], "&#", 2))
 416                               && (outbuf[i+4] == ';') ) {
 417                                 scanch = 0;
 418                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 419                                 outbuf[i] = scanch;
 420                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 421                         }
 422
 423                         /* three-digit decimal equivalents */
 424                         else if ((!strncmp(&outbuf[i], "&#", 2))
 425                               && (outbuf[i+5] == ';') ) {
 426                                 scanch = 0;
 427                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 428                                 outbuf[i] = scanch;
 429                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 430                         }
 431
 432                 }
 433
 434                 /* Make sure the output buffer is big enough */
 435                 if ((output_len + strlen(outbuf) + SIZ) > outptr_buffer_size) {
 436                         outptr_buffer_size += SIZ;
 437                         outptr = realloc(outptr, outptr_buffer_size);
 438                         if (outptr == NULL) {
 439                                 abort();
 440                         }
 441                 }
 442
 443                 /* Output any lines terminated with hard line breaks */
 444                 do {
 445                         did_out = 0;
 446                         if (strlen(outbuf) > 0) {
 447                             for (i = 0; i<strlen(outbuf); ++i) {
 448                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 449
 450                                         strncpy(&outptr[output_len], outbuf, i+1);
 451                                         output_len += (i+1);
 452
 453                                         if (do_citaformat) {
 454                                                 strcpy(&outptr[output_len], " ");
 455                                                 ++output_len;
 456                                         }
 457
 458                                         strcpy(outbuf, &outbuf[i+1]);
 459                                         i = 0;
 460                                         did_out = 1;
 461                                 }
 462                         }
 463                     }
 464                 } while (did_out);
 465
 466                 /* Add soft line breaks */
 467                 if (strlen(outbuf) > (screenwidth - 2 )) {
 468                         rb = (-1);
 469                         for (i=0; i<(screenwidth-2); ++i) {
 470                                 if (outbuf[i]==32) rb = i;
 471                         }
 472                         if (rb>=0) {
 473                                 strncpy(&outptr[output_len], outbuf, rb);
 474                                 output_len += rb;
 475                                 strcpy(&outptr[output_len], nl);
 476                                 output_len += strlen(nl);
 477                                 if (do_citaformat) {
 478                                         strcpy(&outptr[output_len], " ");
 479                                         ++output_len;
 480                                 }
 481                                 strcpy(outbuf, &outbuf[rb+1]);
 482                         } else {
 483                                 strncpy(&outptr[output_len], outbuf,
 484                                         screenwidth-2);
 485                                 output_len += (screenwidth-2);
 486                                 strcpy(&outptr[output_len], nl);
 487                                 output_len += strlen(nl);
 488                                 if (do_citaformat) {
 489                                         strcpy(&outptr[output_len], " ");
 490                                         ++output_len;
 491                                 }
 492                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 493                         }
 494                 }
 495
 496         } while (done_reading == 0);
 497
 498         strcpy(&outptr[output_len], outbuf);
 499         output_len += strlen(outbuf);
 500
 501         /* Strip leading/trailing whitespace.  We can't do this with
 502          * striplt() because it uses too many strlen()'s
 503          */
 504         while ((output_len > 0) && (isspace(outptr[0]))) {
 505                 strcpy(outptr, &outptr[1]);
 506                 --output_len;
 507         }
 508         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 509                 outptr[output_len-1] = 0;
 510                 --output_len;
 511         }
 512
 513         if (outptr[output_len-1] != '\n') {
 514                 strcat(outptr, "\n");
 515                 ++output_len;
 516         }
 517
 518         return outptr;
 519
 520 }