libcitadel/lib/html_to_ascii.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 <stdlib.h>
  10 #include <unistd.h>
  11 #include <stdio.h>
  12 #include <signal.h>
  13 #include <sys/types.h>
  14 #include <ctype.h>
  15 #include <string.h>
  16 #include <sys/stat.h>
  17 #include <errno.h>
  18 #include <limits.h>
  19
  20 #if TIME_WITH_SYS_TIME
  21 # include <sys/time.h>
  22 # include <time.h>
  23 #else
  24 # if HAVE_SYS_TIME_H
  25 #  include <sys/time.h>
  26 # else
  27 #  include <time.h>
  28 # endif
  29 #endif
  30
  31 #include "libcitadel.h"
  32
  33
  34 /*
  35  * Convert HTML to plain text.
  36  *
  37  * inputmsg      = pointer to raw HTML message
  38  * screenwidth   = desired output screenwidth
  39  * do_citaformat = set to 1 to indent newlines with spaces
  40  */
  41 char *html_to_ascii(char *inputmsg, int msglen, int screenwidth, int do_citaformat) {
  42         char inbuf[SIZ];
  43         int inbuf_len = 0;
  44         char outbuf[SIZ];
  45         char tag[1024];
  46         int done_reading = 0;
  47         char *inptr;
  48         char *outptr;
  49         size_t outptr_buffer_size;
  50         size_t output_len = 0;
  51         int i, j, ch, did_out, rb, scanch;
  52         int nest = 0;           /* Bracket nesting level */
  53         int blockquote = 0;     /* BLOCKQUOTE nesting level */
  54         int styletag = 0;       /* STYLE tag nesting level */
  55         int styletag_start = 0;
  56         int bytes_processed = 0;
  57         char nl[128];
  58
  59         strcpy(nl, "\n");
  60         inptr = inputmsg;
  61         strcpy(inbuf, "");
  62         strcpy(outbuf, "");
  63         if (msglen == 0) msglen = strlen(inputmsg);
  64
  65         outptr_buffer_size = strlen(inptr) + SIZ;
  66         outptr = malloc(outptr_buffer_size);
  67         if (outptr == NULL) return NULL;
  68         strcpy(outptr, "");
  69         output_len = 0;
  70
  71         do {
  72                 /* Fill the input buffer */
  73                 inbuf_len = strlen(inbuf);
  74                 if ( (done_reading == 0) && (inbuf_len < (SIZ-128)) ) {
  75
  76                         ch = *inptr++;
  77                         if (ch != 0) {
  78                                 inbuf[inbuf_len++] = ch;
  79                                 inbuf[inbuf_len] = 0;
  80                         }
  81                         else {
  82                                 done_reading = 1;
  83                         }
  84
  85                         ++bytes_processed;
  86                         if (bytes_processed > msglen) {
  87                                 done_reading = 1;
  88                         }
  89
  90                 }
  91
  92                 /* Do some parsing */
  93                 if (!IsEmptyStr(inbuf)) {
  94
  95
  96                     /* Fold in all the spacing */
  97                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
  98                         if (inbuf[i]==10) inbuf[i]=32;
  99                         if (inbuf[i]==13) inbuf[i]=32;
 100                         if (inbuf[i]==9) inbuf[i]=32;
 101                     }
 102                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 103                         while ((inbuf[i]==32)&&(inbuf[i+1]==32)) {
 104                                 strcpy(&inbuf[i], &inbuf[i+1]);
 105                         }
 106                     }
 107
 108                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 109
 110                         ch = inbuf[i];
 111
 112                         if (ch == '<') {
 113                                 ++nest;
 114                                 strcpy(tag, "");
 115                         }
 116
 117                         else if (ch == '>') {   /* We have a tag. */
 118                                 if (nest > 0) --nest;
 119
 120                                 /* Unqualify the tag (truncate at first space) */
 121                                 if (strchr(tag, ' ') != NULL) {
 122                                         strcpy(strchr(tag, ' '), "");
 123                                 }
 124
 125                                 if (!strcasecmp(tag, "P")) {
 126                                         strcat(outbuf, nl);
 127                                         strcat(outbuf, nl);
 128                                 }
 129
 130                                 if (!strcasecmp(tag, "/DIV")) {
 131                                         strcat(outbuf, nl);
 132                                         strcat(outbuf, nl);
 133                                 }
 134
 135                                 if (!strcasecmp(tag, "LI")) {
 136                                         strcat(outbuf, nl);
 137                                         strcat(outbuf, " * ");
 138                                 }
 139
 140                                 else if (!strcasecmp(tag, "/UL")) {
 141                                         strcat(outbuf, nl);
 142                                         strcat(outbuf, nl);
 143                                 }
 144
 145                                 else if (!strcasecmp(tag, "H1")) {
 146                                         strcat(outbuf, nl);
 147                                         strcat(outbuf, nl);
 148                                 }
 149
 150                                 else if (!strcasecmp(tag, "H2")) {
 151                                         strcat(outbuf, nl);
 152                                         strcat(outbuf, nl);
 153                                 }
 154
 155                                 else if (!strcasecmp(tag, "H3")) {
 156                                         strcat(outbuf, nl);
 157                                         strcat(outbuf, nl);
 158                                 }
 159
 160                                 else if (!strcasecmp(tag, "H4")) {
 161                                         strcat(outbuf, nl);
 162                                         strcat(outbuf, nl);
 163                                 }
 164
 165                                 else if (!strcasecmp(tag, "/H1")) {
 166                                         strcat(outbuf, nl);
 167                                 }
 168
 169                                 else if (!strcasecmp(tag, "/H2")) {
 170                                         strcat(outbuf, nl);
 171                                 }
 172
 173                                 else if (!strcasecmp(tag, "/H3")) {
 174                                         strcat(outbuf, nl);
 175                                 }
 176
 177                                 else if (!strcasecmp(tag, "/H4")) {
 178                                         strcat(outbuf, nl);
 179                                 }
 180
 181                                 else if (!strcasecmp(tag, "HR")) {
 182                                         strcat(outbuf, nl);
 183                                         strcat(outbuf, " ");
 184                                         for (j=0; j<screenwidth-2; ++j)
 185                                                 strcat(outbuf, "-");
 186                                         strcat(outbuf, nl);
 187                                 }
 188
 189                                 else if (!strcasecmp(tag, "BR")) {
 190                                         strcat(outbuf, nl);
 191                                 }
 192
 193                                 else if (!strcasecmp(tag, "TR")) {
 194                                         strcat(outbuf, nl);
 195                                 }
 196
 197                                 else if (!strcasecmp(tag, "/TABLE")) {
 198                                         strcat(outbuf, nl);
 199                                 }
 200
 201                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 202                                         ++blockquote;
 203                                         strcpy(nl, "\n");
 204                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 205                                         strcat(outbuf, nl);
 206                                 }
 207
 208                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 209                                         strcat(outbuf, "\n");
 210                                         --blockquote;
 211                                         strcpy(nl, "\n");
 212                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 213                                         strcat(outbuf, nl);
 214                                 }
 215
 216                                 else if (!strcasecmp(tag, "STYLE")) {
 217                                         ++styletag;
 218                                         if (styletag == 1) {
 219                                                 styletag_start = strlen(outbuf);
 220                                         }
 221                                 }
 222
 223                                 else if (!strcasecmp(tag, "/STYLE")) {
 224                                         --styletag;
 225                                         if (styletag == 0) {
 226                                                 outbuf[styletag_start] = 0;
 227                                         }
 228                                 }
 229
 230                         }
 231
 232                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 233                                 tag[strlen(tag)+1] = 0;
 234                                 tag[strlen(tag)] = ch;
 235                         }
 236
 237                         else if (!nest) {
 238                                 outbuf[strlen(outbuf)+1] = 0;
 239                                 outbuf[strlen(outbuf)] = ch;
 240                         }
 241                     }
 242                     strcpy(inbuf, &inbuf[i]);
 243                 }
 244
 245                 /* Convert &; tags to the forbidden characters */
 246                 if (!IsEmptyStr(outbuf)) for (i=0; !IsEmptyStr(&outbuf[i]); ++i) {
 247
 248                         /* Character entity references */
 249                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 250                                 outbuf[i] = ' ';
 251                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 252                         }
 253
 254                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 255                                 outbuf[i] = ' ';
 256                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 257                         }
 258
 259                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 260                                 outbuf[i] = ' ';
 261                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 262                         }
 263
 264                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 265                                 outbuf[i] = ' ';
 266                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 267                         }
 268
 269                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 270                                 outbuf[i] = '<';
 271                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 272                         }
 273
 274                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 275                                 outbuf[i] = '>';
 276                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 277                         }
 278
 279                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 280                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 281                         }
 282
 283                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 284                                 outbuf[i] = '\"';
 285                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 286                         }
 287
 288                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 289                                 outbuf[i] = '`';
 290                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 291                         }
 292
 293                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 294                                 outbuf[i] = '\'';
 295                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 296                         }
 297
 298                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 299                                 outbuf[i] = '(';
 300                                 outbuf[i+1] = 'c';
 301                                 outbuf[i+2] = ')';
 302                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 303                         }
 304
 305                         else if (!strncasecmp(&outbuf[i], "&bull;", 6)) {
 306                                 outbuf[i] = ' ';
 307                                 outbuf[i+1] = '*';
 308                                 outbuf[i+2] = ' ';
 309                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 310                         }
 311
 312                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 313                                 outbuf[i] = '.';
 314                                 outbuf[i+1] = '.';
 315                                 outbuf[i+2] = '.';
 316                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 317                         }
 318
 319                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 320                                 outbuf[i] = '(';
 321                                 outbuf[i+1] = 't';
 322                                 outbuf[i+2] = 'm';
 323                                 outbuf[i+3] = ')';
 324                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 325                         }
 326
 327                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 328                                 outbuf[i] = '(';
 329                                 outbuf[i+1] = 'r';
 330                                 outbuf[i+2] = ')';
 331                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 332                         }
 333
 334                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 335                                 outbuf[i] = '1';
 336                                 outbuf[i+1] = '/';
 337                                 outbuf[i+2] = '4';
 338                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 339                         }
 340
 341                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 342                                 outbuf[i] = '1';
 343                                 outbuf[i+1] = '/';
 344                                 outbuf[i+2] = '2';
 345                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 346                         }
 347
 348                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 349                                 outbuf[i] = '3';
 350                                 outbuf[i+1] = '/';
 351                                 outbuf[i+2] = '4';
 352                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 353                         }
 354
 355                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 356                                 outbuf[i] = '-';
 357                                 outbuf[i+1] = '-';
 358                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 359                         }
 360
 361                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 362                                 outbuf[i] = '-';
 363                                 outbuf[i+1] = '-';
 364                                 outbuf[i+2] = '-';
 365                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 366                         }
 367
 368                         else if (!strncmp(&outbuf[i], "&Ccedil;", 8)) {
 369                                 outbuf[i] = 'C';
 370                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 371                         }
 372
 373                         else if (!strncasecmp(&outbuf[i], "&ccedil;", 8)) {
 374                                 outbuf[i] = 'c';
 375                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 376                         }
 377
 378                         else if (!strncmp(&outbuf[i], "&Egrave;", 8)) {
 379                                 outbuf[i] = 'E';
 380                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 381                         }
 382
 383                         else if (!strncasecmp(&outbuf[i], "&egrave;", 8)) {
 384                                 outbuf[i] = 'e';
 385                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 386                         }
 387
 388                         else if (!strncmp(&outbuf[i], "&Ecirc;", 7)) {
 389                                 outbuf[i] = 'E';
 390                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 391                         }
 392
 393                         else if (!strncasecmp(&outbuf[i], "&ecirc;", 7)) {
 394                                 outbuf[i] = 'e';
 395                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 396                         }
 397
 398                         else if (!strncmp(&outbuf[i], "&Eacute;", 8)) {
 399                                 outbuf[i] = 'E';
 400                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 401                         }
 402
 403                         else if (!strncasecmp(&outbuf[i], "&eacute;", 8)) {
 404                                 outbuf[i] = 'e';
 405                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 406                         }
 407
 408                         else if (!strncmp(&outbuf[i], "&Agrave;", 8)) {
 409                                 outbuf[i] = 'A';
 410                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 411                         }
 412
 413                         else if (!strncasecmp(&outbuf[i], "&agrave;", 8)) {
 414                                 outbuf[i] = 'a';
 415                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 416                         }
 417
 418                         else if (!strncasecmp(&outbuf[i], "&ldquo;", 7)) {
 419                                 outbuf[i] = '\"';
 420                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 421                         }
 422
 423                         else if (!strncasecmp(&outbuf[i], "&rdquo;", 7)) {
 424                                 outbuf[i] = '\"';
 425                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 426                         }
 427
 428                         else if (!strncasecmp(&outbuf[i], "&acute;", 7)) {
 429                                 outbuf[i] = '\'';
 430                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 431                         }
 432
 433                         /* two-digit decimal equivalents */
 434                         else if ((!strncmp(&outbuf[i], "&#", 2))
 435                               && (outbuf[i+4] == ';') ) {
 436                                 scanch = 0;
 437                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 438                                 outbuf[i] = scanch;
 439                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 440                         }
 441
 442                         /* three-digit decimal equivalents */
 443                         else if ((!strncmp(&outbuf[i], "&#", 2))
 444                               && (outbuf[i+5] == ';') ) {
 445                                 scanch = 0;
 446                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 447                                 outbuf[i] = scanch;
 448                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 449                         }
 450
 451                 }
 452
 453                 /* Make sure the output buffer is big enough */
 454                 if ((output_len + strlen(outbuf) + SIZ) > outptr_buffer_size) {
 455                         outptr_buffer_size += SIZ;
 456                         outptr = realloc(outptr, outptr_buffer_size);
 457                         if (outptr == NULL) {
 458                                 abort();
 459                         }
 460                 }
 461
 462                 /* Output any lines terminated with hard line breaks */
 463                 do {
 464                         did_out = 0;
 465                         if (strlen(outbuf) > 0) {
 466                             for (i = 0; i<strlen(outbuf); ++i) {
 467                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 468
 469                                         strncpy(&outptr[output_len], outbuf, i+1);
 470                                         output_len += (i+1);
 471
 472                                         if (do_citaformat) {
 473                                                 strcpy(&outptr[output_len], " ");
 474                                                 ++output_len;
 475                                         }
 476
 477                                         strcpy(outbuf, &outbuf[i+1]);
 478                                         i = 0;
 479                                         did_out = 1;
 480                                 }
 481                         }
 482                     }
 483                 } while (did_out);
 484
 485                 /* Add soft line breaks */
 486                 if (strlen(outbuf) > (screenwidth - 2 )) {
 487                         rb = (-1);
 488                         for (i=0; i<(screenwidth-2); ++i) {
 489                                 if (outbuf[i]==32) rb = i;
 490                         }
 491                         if (rb>=0) {
 492                                 strncpy(&outptr[output_len], outbuf, rb);
 493                                 output_len += rb;
 494                                 strcpy(&outptr[output_len], nl);
 495                                 output_len += strlen(nl);
 496                                 if (do_citaformat) {
 497                                         strcpy(&outptr[output_len], " ");
 498                                         ++output_len;
 499                                 }
 500                                 strcpy(outbuf, &outbuf[rb+1]);
 501                         } else {
 502                                 strncpy(&outptr[output_len], outbuf,
 503                                         screenwidth-2);
 504                                 output_len += (screenwidth-2);
 505                                 strcpy(&outptr[output_len], nl);
 506                                 output_len += strlen(nl);
 507                                 if (do_citaformat) {
 508                                         strcpy(&outptr[output_len], " ");
 509                                         ++output_len;
 510                                 }
 511                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 512                         }
 513                 }
 514
 515         } while (done_reading == 0);
 516
 517         strcpy(&outptr[output_len], outbuf);
 518         output_len += strlen(outbuf);
 519
 520         /* Strip leading/trailing whitespace.  We can't do this with
 521          * striplt() because it uses too many strlen()'s
 522          */
 523         while ((output_len > 0) && (isspace(outptr[0]))) {
 524                 strcpy(outptr, &outptr[1]);
 525                 --output_len;
 526         }
 527         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 528                 outptr[output_len-1] = 0;
 529                 --output_len;
 530         }
 531
 532         if (outptr[output_len-1] != '\n') {
 533                 strcat(outptr, "\n");
 534                 ++output_len;
 535         }
 536
 537         return outptr;
 538
 539 }