libcitadel/lib/html_to_ascii.c

   1 /*
   2  * Functions which handle translation between HTML and plain text
   3  * Copyright (c) 2000-2010 by the citadel.org team
   4  *
   5  * This program is open source 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 3 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 <stdlib.h>
  21 #include <unistd.h>
  22 #include <stdio.h>
  23 #include <signal.h>
  24 #include <sys/types.h>
  25 #include <ctype.h>
  26 #include <string.h>
  27 #include <sys/stat.h>
  28 #include <errno.h>
  29 #include <limits.h>
  30
  31 #if TIME_WITH_SYS_TIME
  32 # include <sys/time.h>
  33 # include <time.h>
  34 #else
  35 # if HAVE_SYS_TIME_H
  36 #  include <sys/time.h>
  37 # else
  38 #  include <time.h>
  39 # endif
  40 #endif
  41
  42 #include "libcitadel.h"
  43
  44
  45 /*
  46  * Convert HTML to plain text.
  47  *
  48  * inputmsg      = pointer to raw HTML message
  49  * screenwidth   = desired output screenwidth
  50  * do_citaformat = set to 1 to indent newlines with spaces
  51  */
  52 char *html_to_ascii(const char *inputmsg, int msglen, int screenwidth, int do_citaformat) {
  53         char inbuf[SIZ];
  54         int inbuf_len = 0;
  55         char outbuf[SIZ];
  56         char tag[1024];
  57         int done_reading = 0;
  58         const char *inptr;
  59         char *outptr;
  60         size_t outptr_buffer_size;
  61         size_t output_len = 0;
  62         int i, j, ch, did_out, rb, scanch;
  63         int nest = 0;           /* Bracket nesting level */
  64         int blockquote = 0;     /* BLOCKQUOTE nesting level */
  65         int styletag = 0;       /* STYLE tag nesting level */
  66         int styletag_start = 0;
  67         int bytes_processed = 0;
  68         char nl[128];
  69
  70         tag[0] = '\0';
  71         strcpy(nl, "\n");
  72         inptr = inputmsg;
  73         strcpy(inbuf, "");
  74         strcpy(outbuf, "");
  75         if (msglen == 0) msglen = strlen(inputmsg);
  76
  77         outptr_buffer_size = strlen(inptr) + SIZ;
  78         outptr = malloc(outptr_buffer_size);
  79         if (outptr == NULL) return NULL;
  80         strcpy(outptr, "");
  81         output_len = 0;
  82
  83         do {
  84                 /* Fill the input buffer */
  85                 inbuf_len = strlen(inbuf);
  86                 if ( (done_reading == 0) && (inbuf_len < (SIZ-128)) ) {
  87
  88                         ch = *inptr++;
  89                         if (ch != 0) {
  90                                 inbuf[inbuf_len++] = ch;
  91                                 inbuf[inbuf_len] = 0;
  92                         }
  93                         else {
  94                                 done_reading = 1;
  95                         }
  96
  97                         ++bytes_processed;
  98                         if (bytes_processed > msglen) {
  99                                 done_reading = 1;
 100                         }
 101
 102                 }
 103
 104                 /* Do some parsing */
 105                 if (!IsEmptyStr(inbuf)) {
 106
 107
 108                     /* Fold in all the spacing */
 109                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 110                         if (inbuf[i]==10) inbuf[i]=32;
 111                         if (inbuf[i]==13) inbuf[i]=32;
 112                         if (inbuf[i]==9) inbuf[i]=32;
 113                     }
 114                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 115                         while ((inbuf[i]==32)&&(inbuf[i+1]==32)) {
 116                                 strcpy(&inbuf[i], &inbuf[i+1]);
 117                         }
 118                     }
 119
 120                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 121
 122                         ch = inbuf[i];
 123
 124                         if (ch == '<') {
 125                                 ++nest;
 126                                 strcpy(tag, "");
 127                         }
 128
 129                         else if (ch == '>') {   /* We have a tag. */
 130                                 if (nest > 0) --nest;
 131
 132                                 /* Unqualify the tag (truncate at first space) */
 133                                 if (strchr(tag, ' ') != NULL) {
 134                                         strcpy(strchr(tag, ' '), "");
 135                                 }
 136
 137                                 if (!strcasecmp(tag, "P")) {
 138                                         strcat(outbuf, nl);
 139                                         strcat(outbuf, nl);
 140                                 }
 141
 142                                 if (!strcasecmp(tag, "/DIV")) {
 143                                         strcat(outbuf, nl);
 144                                         strcat(outbuf, nl);
 145                                 }
 146
 147                                 if (!strcasecmp(tag, "LI")) {
 148                                         strcat(outbuf, nl);
 149                                         strcat(outbuf, " * ");
 150                                 }
 151
 152                                 else if (!strcasecmp(tag, "/UL")) {
 153                                         strcat(outbuf, nl);
 154                                         strcat(outbuf, nl);
 155                                 }
 156
 157                                 else if (!strcasecmp(tag, "H1")) {
 158                                         strcat(outbuf, nl);
 159                                         strcat(outbuf, nl);
 160                                 }
 161
 162                                 else if (!strcasecmp(tag, "H2")) {
 163                                         strcat(outbuf, nl);
 164                                         strcat(outbuf, nl);
 165                                 }
 166
 167                                 else if (!strcasecmp(tag, "H3")) {
 168                                         strcat(outbuf, nl);
 169                                         strcat(outbuf, nl);
 170                                 }
 171
 172                                 else if (!strcasecmp(tag, "H4")) {
 173                                         strcat(outbuf, nl);
 174                                         strcat(outbuf, nl);
 175                                 }
 176
 177                                 else if (!strcasecmp(tag, "/H1")) {
 178                                         strcat(outbuf, nl);
 179                                 }
 180
 181                                 else if (!strcasecmp(tag, "/H2")) {
 182                                         strcat(outbuf, nl);
 183                                 }
 184
 185                                 else if (!strcasecmp(tag, "/H3")) {
 186                                         strcat(outbuf, nl);
 187                                 }
 188
 189                                 else if (!strcasecmp(tag, "/H4")) {
 190                                         strcat(outbuf, nl);
 191                                 }
 192
 193                                 else if (!strcasecmp(tag, "HR")) {
 194                                         strcat(outbuf, nl);
 195                                         strcat(outbuf, " ");
 196                                         for (j=0; j<screenwidth-2; ++j)
 197                                                 strcat(outbuf, "-");
 198                                         strcat(outbuf, nl);
 199                                 }
 200
 201                                 else if (
 202                                         (!strcasecmp(tag, "B"))
 203                                         || (!strcasecmp(tag, "/B"))
 204                                         || (!strcasecmp(tag, "STRONG"))
 205                                         || (!strcasecmp(tag, "/STRONG"))
 206                                 ) {
 207                                         strcat(outbuf, "*");
 208
 209                                 }
 210
 211                                 else if (
 212                                         (!strcasecmp(tag, "I"))
 213                                         || (!strcasecmp(tag, "/I"))
 214                                         || (!strcasecmp(tag, "EM"))
 215                                         || (!strcasecmp(tag, "/EM"))
 216                                 ) {
 217                                         strcat(outbuf, "/");
 218
 219                                 }
 220
 221                                 else if (
 222                                         (!strcasecmp(tag, "U"))
 223                                         || (!strcasecmp(tag, "/U"))
 224                                 ) {
 225                                         strcat(outbuf, "_");
 226
 227                                 }
 228
 229                                 else if (!strcasecmp(tag, "BR")) {
 230                                         strcat(outbuf, nl);
 231                                 }
 232
 233                                 else if (!strcasecmp(tag, "TR")) {
 234                                         strcat(outbuf, nl);
 235                                 }
 236
 237                                 else if (!strcasecmp(tag, "/TABLE")) {
 238                                         strcat(outbuf, nl);
 239                                 }
 240
 241                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 242                                         ++blockquote;
 243                                         strcpy(nl, "\n");
 244                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 245                                         strcat(outbuf, nl);
 246                                 }
 247
 248                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 249                                         strcat(outbuf, "\n");
 250                                         --blockquote;
 251                                         strcpy(nl, "\n");
 252                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 253                                         strcat(outbuf, nl);
 254                                 }
 255
 256                                 else if (!strcasecmp(tag, "STYLE")) {
 257                                         ++styletag;
 258                                         if (styletag == 1) {
 259                                                 styletag_start = strlen(outbuf);
 260                                         }
 261                                 }
 262
 263                                 else if (!strcasecmp(tag, "/STYLE")) {
 264                                         --styletag;
 265                                         if (styletag == 0) {
 266                                                 outbuf[styletag_start] = 0;
 267                                         }
 268                                 }
 269
 270                         }
 271
 272                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 273                                 tag[strlen(tag)+1] = 0;
 274                                 tag[strlen(tag)] = ch;
 275                         }
 276
 277                         else if (!nest) {
 278                                 outbuf[strlen(outbuf)+1] = 0;
 279                                 outbuf[strlen(outbuf)] = ch;
 280                         }
 281                     }
 282                     strcpy(inbuf, &inbuf[i]);
 283                 }
 284
 285                 /* Convert &; tags to the forbidden characters */
 286                 if (!IsEmptyStr(outbuf)) for (i=0; !IsEmptyStr(&outbuf[i]); ++i) {
 287
 288                         /* Character entity references */
 289                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 290                                 outbuf[i] = ' ';
 291                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 292                         }
 293
 294                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 295                                 outbuf[i] = ' ';
 296                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 297                         }
 298
 299                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 300                                 outbuf[i] = ' ';
 301                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 302                         }
 303
 304                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 305                                 outbuf[i] = ' ';
 306                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 307                         }
 308
 309                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 310                                 outbuf[i] = '<';
 311                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 312                         }
 313
 314                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 315                                 outbuf[i] = '>';
 316                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 317                         }
 318
 319                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 320                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 321                         }
 322
 323                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 324                                 outbuf[i] = '\"';
 325                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 326                         }
 327
 328                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 329                                 outbuf[i] = '`';
 330                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 331                         }
 332
 333                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 334                                 outbuf[i] = '\'';
 335                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 336                         }
 337
 338                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 339                                 outbuf[i] = '(';
 340                                 outbuf[i+1] = 'c';
 341                                 outbuf[i+2] = ')';
 342                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 343                         }
 344
 345                         else if (!strncasecmp(&outbuf[i], "&bull;", 6)) {
 346                                 outbuf[i] = ' ';
 347                                 outbuf[i+1] = '*';
 348                                 outbuf[i+2] = ' ';
 349                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 350                         }
 351
 352                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 353                                 outbuf[i] = '.';
 354                                 outbuf[i+1] = '.';
 355                                 outbuf[i+2] = '.';
 356                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 357                         }
 358
 359                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 360                                 outbuf[i] = '(';
 361                                 outbuf[i+1] = 't';
 362                                 outbuf[i+2] = 'm';
 363                                 outbuf[i+3] = ')';
 364                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 365                         }
 366
 367                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 368                                 outbuf[i] = '(';
 369                                 outbuf[i+1] = 'r';
 370                                 outbuf[i+2] = ')';
 371                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 372                         }
 373
 374                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 375                                 outbuf[i] = '1';
 376                                 outbuf[i+1] = '/';
 377                                 outbuf[i+2] = '4';
 378                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 379                         }
 380
 381                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 382                                 outbuf[i] = '1';
 383                                 outbuf[i+1] = '/';
 384                                 outbuf[i+2] = '2';
 385                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 386                         }
 387
 388                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 389                                 outbuf[i] = '3';
 390                                 outbuf[i+1] = '/';
 391                                 outbuf[i+2] = '4';
 392                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 393                         }
 394
 395                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 396                                 outbuf[i] = '-';
 397                                 outbuf[i+1] = '-';
 398                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 399                         }
 400
 401                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 402                                 outbuf[i] = '-';
 403                                 outbuf[i+1] = '-';
 404                                 outbuf[i+2] = '-';
 405                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 406                         }
 407
 408                         else if (!strncmp(&outbuf[i], "&Ccedil;", 8)) {
 409                                 outbuf[i] = 'C';
 410                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 411                         }
 412
 413                         else if (!strncasecmp(&outbuf[i], "&ccedil;", 8)) {
 414                                 outbuf[i] = 'c';
 415                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 416                         }
 417
 418                         else if (!strncmp(&outbuf[i], "&Egrave;", 8)) {
 419                                 outbuf[i] = 'E';
 420                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 421                         }
 422
 423                         else if (!strncasecmp(&outbuf[i], "&egrave;", 8)) {
 424                                 outbuf[i] = 'e';
 425                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 426                         }
 427
 428                         else if (!strncmp(&outbuf[i], "&Ecirc;", 7)) {
 429                                 outbuf[i] = 'E';
 430                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 431                         }
 432
 433                         else if (!strncasecmp(&outbuf[i], "&ecirc;", 7)) {
 434                                 outbuf[i] = 'e';
 435                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 436                         }
 437
 438                         else if (!strncmp(&outbuf[i], "&Eacute;", 8)) {
 439                                 outbuf[i] = 'E';
 440                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 441                         }
 442
 443                         else if (!strncasecmp(&outbuf[i], "&eacute;", 8)) {
 444                                 outbuf[i] = 'e';
 445                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 446                         }
 447
 448                         else if (!strncmp(&outbuf[i], "&Agrave;", 8)) {
 449                                 outbuf[i] = 'A';
 450                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 451                         }
 452
 453                         else if (!strncasecmp(&outbuf[i], "&agrave;", 8)) {
 454                                 outbuf[i] = 'a';
 455                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 456                         }
 457
 458                         else if (!strncasecmp(&outbuf[i], "&ldquo;", 7)) {
 459                                 outbuf[i] = '\"';
 460                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 461                         }
 462
 463                         else if (!strncasecmp(&outbuf[i], "&rdquo;", 7)) {
 464                                 outbuf[i] = '\"';
 465                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 466                         }
 467
 468                         else if (!strncasecmp(&outbuf[i], "&acute;", 7)) {
 469                                 outbuf[i] = '\'';
 470                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 471                         }
 472
 473                         else if (!strncasecmp(&outbuf[i], "&#8217;", 7)) {
 474                                 outbuf[i] = '\'';
 475                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 476                         }
 477
 478                         else if (!strncasecmp(&outbuf[i], "&#8211;", 7)) {
 479                                 outbuf[i] = '-';
 480                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 481                         }
 482
 483                         /* two-digit decimal equivalents */
 484                         else if (outbuf[i] == '&'       &&
 485                                  outbuf[i + 1] == '#'   &&
 486                                  isdigit(outbuf[i + 2]) &&
 487                                  isdigit(outbuf[i + 3]) &&
 488                                  (outbuf[i+4] == ';') )
 489                         {
 490                                 scanch = 0;
 491                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 492                                 outbuf[i] = scanch;
 493                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 494                         }
 495
 496                         /* three-digit decimal equivalents */
 497                         else if (outbuf[i] == '&'       &&
 498                                  outbuf[i + 1] == '#'   &&
 499                                  isdigit(outbuf[i + 2]) &&
 500                                  isdigit(outbuf[i + 3]) &&
 501                                  isdigit(outbuf[i + 4]) &&
 502                                  (outbuf[i + 5] == ';') )
 503                         {
 504                                 scanch = 0;
 505                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 506                                 outbuf[i] = scanch;
 507                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 508                         }
 509
 510                         /* four-digit decimal equivalents */
 511                         else if (outbuf[i] == '&'       &&
 512                                  outbuf[i + 1] == '#'   &&
 513                                  isdigit(outbuf[i + 2]) &&
 514                                  isdigit(outbuf[i + 3]) &&
 515                                  isdigit(outbuf[i + 4]) &&
 516                                  isdigit(outbuf[i + 5]) &&
 517                                  (outbuf[i + 6] == ';') )
 518                         {
 519                                 scanch = 0;
 520                                 sscanf(&outbuf[i+2], "%04d", &scanch);
 521                                 outbuf[i] = scanch;
 522                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 523                         }
 524
 525                 }
 526
 527                 /* Make sure the output buffer is big enough */
 528                 if ((output_len + strlen(outbuf) + SIZ) > outptr_buffer_size) {
 529                         outptr_buffer_size += SIZ;
 530                         outptr = realloc(outptr, outptr_buffer_size);
 531                         if (outptr == NULL) {
 532                                 abort();
 533                         }
 534                 }
 535
 536                 /* Output any lines terminated with hard line breaks */
 537                 do {
 538                         did_out = 0;
 539                         if (strlen(outbuf) > 0) {
 540                             for (i = 0; i<strlen(outbuf); ++i) {
 541                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 542
 543                                         strncpy(&outptr[output_len], outbuf, i+1);
 544                                         output_len += (i+1);
 545
 546                                         if (do_citaformat) {
 547                                                 strcpy(&outptr[output_len], " ");
 548                                                 ++output_len;
 549                                         }
 550
 551                                         strcpy(outbuf, &outbuf[i+1]);
 552                                         i = 0;
 553                                         did_out = 1;
 554                                 }
 555                         }
 556                     }
 557                 } while (did_out);
 558
 559                 /* Add soft line breaks */
 560                 if (strlen(outbuf) > (screenwidth - 2 )) {
 561                         rb = (-1);
 562                         for (i=0; i<(screenwidth-2); ++i) {
 563                                 if (outbuf[i]==32) rb = i;
 564                         }
 565                         if (rb>=0) {
 566                                 strncpy(&outptr[output_len], outbuf, rb);
 567                                 output_len += rb;
 568                                 strcpy(&outptr[output_len], nl);
 569                                 output_len += strlen(nl);
 570                                 if (do_citaformat) {
 571                                         strcpy(&outptr[output_len], " ");
 572                                         ++output_len;
 573                                 }
 574                                 strcpy(outbuf, &outbuf[rb+1]);
 575                         } else {
 576                                 strncpy(&outptr[output_len], outbuf,
 577                                         screenwidth-2);
 578                                 output_len += (screenwidth-2);
 579                                 strcpy(&outptr[output_len], nl);
 580                                 output_len += strlen(nl);
 581                                 if (do_citaformat) {
 582                                         strcpy(&outptr[output_len], " ");
 583                                         ++output_len;
 584                                 }
 585                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 586                         }
 587                 }
 588
 589         } while (done_reading == 0);
 590
 591         strcpy(&outptr[output_len], outbuf);
 592         output_len += strlen(outbuf);
 593
 594         /* Strip leading/trailing whitespace.  We can't do this with
 595          * striplt() because it uses too many strlen()'s
 596          */
 597         while ((output_len > 0) && (isspace(outptr[0]))) {
 598                 strcpy(outptr, &outptr[1]);
 599                 --output_len;
 600         }
 601         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 602                 outptr[output_len-1] = 0;
 603                 --output_len;
 604         }
 605
 606         if ((output_len > 0) && (outptr[output_len-1] != '\n')) {
 607                 strcat(outptr, "\n");
 608                 ++output_len;
 609         }
 610
 611         return outptr;
 612
 613 }