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 "sysdep.h"
  10 #include <stdlib.h>
  11 #include <unistd.h>
  12 #include <stdio.h>
  13 #include <fcntl.h>
  14 #include <signal.h>
  15
  16 #if TIME_WITH_SYS_TIME
  17 # include <sys/time.h>
  18 # include <time.h>
  19 #else
  20 # if HAVE_SYS_TIME_H
  21 #  include <sys/time.h>
  22 # else
  23 #  include <time.h>
  24 # endif
  25 #endif
  26
  27 #include <ctype.h>
  28 #include <string.h>
  29 #include <errno.h>
  30 #include <limits.h>
  31 #include <libcitadel.h>
  32 #include "citadel.h"
  33 #include "server.h"
  34 #include "control.h"
  35 #include "sysdep_decls.h"
  36 #include "support.h"
  37 #include "config.h"
  38 #include "msgbase.h"
  39 #include "room_ops.h"
  40 #include "html.h"
  41
  42
  43 /*
  44  * Convert HTML to plain text.
  45  *
  46  * inputmsg      = pointer to raw HTML message
  47  * screenwidth   = desired output screenwidth
  48  * do_citaformat = set to 1 to indent newlines with spaces
  49  */
  50 char *html_to_ascii(char *inputmsg, int msglen, int screenwidth, int do_citaformat) {
  51         char inbuf[SIZ];
  52         int inbuf_len = 0;
  53         char outbuf[SIZ];
  54         char tag[1024];
  55         int done_reading = 0;
  56         char *inptr;
  57         char *outptr;
  58         size_t outptr_buffer_size;
  59         size_t output_len = 0;
  60         int i, j, ch, did_out, rb, scanch;
  61         int nest = 0;           /* Bracket nesting level */
  62         int blockquote = 0;     /* BLOCKQUOTE nesting level */
  63         int styletag = 0;       /* STYLE tag nesting level */
  64         int styletag_start = 0;
  65         int bytes_processed = 0;
  66         char nl[128];
  67
  68         strcpy(nl, "\n");
  69         inptr = inputmsg;
  70         strcpy(inbuf, "");
  71         strcpy(outbuf, "");
  72         if (msglen == 0) msglen = strlen(inputmsg);
  73
  74         outptr_buffer_size = strlen(inptr) + SIZ;
  75         outptr = malloc(outptr_buffer_size);
  76         if (outptr == NULL) return NULL;
  77         strcpy(outptr, "");
  78         output_len = 0;
  79
  80         do {
  81                 /* Fill the input buffer */
  82                 inbuf_len = strlen(inbuf);
  83                 if ( (done_reading == 0) && (inbuf_len < (SIZ-128)) ) {
  84
  85                         ch = *inptr++;
  86                         if (ch != 0) {
  87                                 inbuf[inbuf_len++] = ch;
  88                                 inbuf[inbuf_len] = 0;
  89                         }
  90                         else {
  91                                 done_reading = 1;
  92                         }
  93
  94                         ++bytes_processed;
  95                         if (bytes_processed > msglen) {
  96                                 done_reading = 1;
  97                         }
  98
  99                 }
 100
 101                 /* Do some parsing */
 102                 if (!IsEmptyStr(inbuf)) {
 103
 104
 105                     /* Fold in all the spacing */
 106                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 107                         if (inbuf[i]==10) inbuf[i]=32;
 108                         if (inbuf[i]==13) inbuf[i]=32;
 109                         if (inbuf[i]==9) inbuf[i]=32;
 110                         /*** we like foreign characters now.
 111                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
 112                                 inbuf[i] = '?';
 113                         } */
 114                     }
 115                     for (i=0; !IsEmptyStr(&inbuf[i]); ++i) {
 116                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
 117                                 strcpy(&inbuf[i], &inbuf[i+1]);
 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 (!strcasecmp(tag, "BR")) {
 202                                         strcat(outbuf, nl);
 203                                 }
 204
 205                                 else if (!strcasecmp(tag, "TR")) {
 206                                         strcat(outbuf, nl);
 207                                 }
 208
 209                                 else if (!strcasecmp(tag, "/TABLE")) {
 210                                         strcat(outbuf, nl);
 211                                 }
 212
 213                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 214                                         ++blockquote;
 215                                         strcpy(nl, "\n");
 216                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 217                                         strcat(outbuf, nl);
 218                                 }
 219
 220                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 221                                         strcat(outbuf, "\n");
 222                                         --blockquote;
 223                                         strcpy(nl, "\n");
 224                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 225                                         strcat(outbuf, nl);
 226                                 }
 227
 228                                 else if (!strcasecmp(tag, "STYLE")) {
 229                                         ++styletag;
 230                                         if (styletag == 1) {
 231                                                 styletag_start = strlen(outbuf);
 232                                         }
 233                                 }
 234
 235                                 else if (!strcasecmp(tag, "/STYLE")) {
 236                                         --styletag;
 237                                         if (styletag == 0) {
 238                                                 outbuf[styletag_start] = 0;
 239                                         }
 240                                 }
 241
 242                         }
 243
 244                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 245                                 tag[strlen(tag)+1] = 0;
 246                                 tag[strlen(tag)] = ch;
 247                         }
 248
 249                         else if (!nest) {
 250                                 outbuf[strlen(outbuf)+1] = 0;
 251                                 outbuf[strlen(outbuf)] = ch;
 252                         }
 253                     }
 254                     strcpy(inbuf, &inbuf[i]);
 255                 }
 256
 257                 /* Convert &; tags to the forbidden characters */
 258                 if (!IsEmptyStr(outbuf)) for (i=0; !IsEmptyStr(&outbuf[i]); ++i) {
 259
 260                         /* Character entity references */
 261                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 262                                 outbuf[i] = ' ';
 263                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 264                         }
 265
 266                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 267                                 outbuf[i] = ' ';
 268                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 269                         }
 270
 271                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 272                                 outbuf[i] = ' ';
 273                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 274                         }
 275
 276                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 277                                 outbuf[i] = ' ';
 278                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 279                         }
 280
 281                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 282                                 outbuf[i] = '<';
 283                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 284                         }
 285
 286                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 287                                 outbuf[i] = '>';
 288                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 289                         }
 290
 291                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 292                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 293                         }
 294
 295                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 296                                 outbuf[i] = '\"';
 297                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 298                         }
 299
 300                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 301                                 outbuf[i] = '`';
 302                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 303                         }
 304
 305                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 306                                 outbuf[i] = '\'';
 307                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 308                         }
 309
 310                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 311                                 outbuf[i] = '(';
 312                                 outbuf[i+1] = 'c';
 313                                 outbuf[i+2] = ')';
 314                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 315                         }
 316
 317                         else if (!strncasecmp(&outbuf[i], "&bull;", 6)) {
 318                                 outbuf[i] = ' ';
 319                                 outbuf[i+1] = '*';
 320                                 outbuf[i+2] = ' ';
 321                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 322                         }
 323
 324                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 325                                 outbuf[i] = '.';
 326                                 outbuf[i+1] = '.';
 327                                 outbuf[i+2] = '.';
 328                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 329                         }
 330
 331                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 332                                 outbuf[i] = '(';
 333                                 outbuf[i+1] = 't';
 334                                 outbuf[i+2] = 'm';
 335                                 outbuf[i+3] = ')';
 336                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 337                         }
 338
 339                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 340                                 outbuf[i] = '(';
 341                                 outbuf[i+1] = 'r';
 342                                 outbuf[i+2] = ')';
 343                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 344                         }
 345
 346                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 347                                 outbuf[i] = '1';
 348                                 outbuf[i+1] = '/';
 349                                 outbuf[i+2] = '4';
 350                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 351                         }
 352
 353                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 354                                 outbuf[i] = '1';
 355                                 outbuf[i+1] = '/';
 356                                 outbuf[i+2] = '2';
 357                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 358                         }
 359
 360                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 361                                 outbuf[i] = '3';
 362                                 outbuf[i+1] = '/';
 363                                 outbuf[i+2] = '4';
 364                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 365                         }
 366
 367                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 368                                 outbuf[i] = '-';
 369                                 outbuf[i+1] = '-';
 370                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 371                         }
 372
 373                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 374                                 outbuf[i] = '-';
 375                                 outbuf[i+1] = '-';
 376                                 outbuf[i+2] = '-';
 377                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 378                         }
 379
 380                         else if (!strncmp(&outbuf[i], "&Ccedil;", 8)) {
 381                                 outbuf[i] = 'C';
 382                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 383                         }
 384
 385                         else if (!strncasecmp(&outbuf[i], "&ccedil;", 8)) {
 386                                 outbuf[i] = 'c';
 387                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 388                         }
 389
 390                         else if (!strncmp(&outbuf[i], "&Egrave;", 8)) {
 391                                 outbuf[i] = 'E';
 392                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 393                         }
 394
 395                         else if (!strncasecmp(&outbuf[i], "&egrave;", 8)) {
 396                                 outbuf[i] = 'e';
 397                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 398                         }
 399
 400                         else if (!strncmp(&outbuf[i], "&Ecirc;", 7)) {
 401                                 outbuf[i] = 'E';
 402                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 403                         }
 404
 405                         else if (!strncasecmp(&outbuf[i], "&ecirc;", 7)) {
 406                                 outbuf[i] = 'e';
 407                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 408                         }
 409
 410                         else if (!strncmp(&outbuf[i], "&Eacute;", 8)) {
 411                                 outbuf[i] = 'E';
 412                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 413                         }
 414
 415                         else if (!strncasecmp(&outbuf[i], "&eacute;", 8)) {
 416                                 outbuf[i] = 'e';
 417                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 418                         }
 419
 420                         else if (!strncmp(&outbuf[i], "&Agrave;", 8)) {
 421                                 outbuf[i] = 'A';
 422                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 423                         }
 424
 425                         else if (!strncasecmp(&outbuf[i], "&agrave;", 8)) {
 426                                 outbuf[i] = 'a';
 427                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 428                         }
 429
 430                         else if (!strncasecmp(&outbuf[i], "&ldquo;", 7)) {
 431                                 outbuf[i] = '\"';
 432                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 433                         }
 434
 435                         else if (!strncasecmp(&outbuf[i], "&rdquo;", 7)) {
 436                                 outbuf[i] = '\"';
 437                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 438                         }
 439
 440                         else if (!strncasecmp(&outbuf[i], "&acute;", 7)) {
 441                                 outbuf[i] = '\'';
 442                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 443                         }
 444
 445                         /* two-digit decimal equivalents */
 446                         else if ((!strncmp(&outbuf[i], "&#", 2))
 447                               && (outbuf[i+4] == ';') ) {
 448                                 scanch = 0;
 449                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 450                                 outbuf[i] = scanch;
 451                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 452                         }
 453
 454                         /* three-digit decimal equivalents */
 455                         else if ((!strncmp(&outbuf[i], "&#", 2))
 456                               && (outbuf[i+5] == ';') ) {
 457                                 scanch = 0;
 458                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 459                                 outbuf[i] = scanch;
 460                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 461                         }
 462
 463                 }
 464
 465                 /* Make sure the output buffer is big enough */
 466                 if ((output_len + strlen(outbuf) + SIZ) > outptr_buffer_size) {
 467                         outptr_buffer_size += SIZ;
 468                         outptr = realloc(outptr, outptr_buffer_size);
 469                         if (outptr == NULL) {
 470                                 abort();
 471                         }
 472                 }
 473
 474                 /* Output any lines terminated with hard line breaks */
 475                 do {
 476                         did_out = 0;
 477                         if (strlen(outbuf) > 0) {
 478                             for (i = 0; i<strlen(outbuf); ++i) {
 479                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 480
 481                                         strncpy(&outptr[output_len], outbuf, i+1);
 482                                         output_len += (i+1);
 483
 484                                         if (do_citaformat) {
 485                                                 strcpy(&outptr[output_len], " ");
 486                                                 ++output_len;
 487                                         }
 488
 489                                         strcpy(outbuf, &outbuf[i+1]);
 490                                         i = 0;
 491                                         did_out = 1;
 492                                 }
 493                         }
 494                     }
 495                 } while (did_out);
 496
 497                 /* Add soft line breaks */
 498                 if (strlen(outbuf) > (screenwidth - 2 )) {
 499                         rb = (-1);
 500                         for (i=0; i<(screenwidth-2); ++i) {
 501                                 if (outbuf[i]==32) rb = i;
 502                         }
 503                         if (rb>=0) {
 504                                 strncpy(&outptr[output_len], outbuf, rb);
 505                                 output_len += rb;
 506                                 strcpy(&outptr[output_len], nl);
 507                                 output_len += strlen(nl);
 508                                 if (do_citaformat) {
 509                                         strcpy(&outptr[output_len], " ");
 510                                         ++output_len;
 511                                 }
 512                                 strcpy(outbuf, &outbuf[rb+1]);
 513                         } else {
 514                                 strncpy(&outptr[output_len], outbuf,
 515                                         screenwidth-2);
 516                                 output_len += (screenwidth-2);
 517                                 strcpy(&outptr[output_len], nl);
 518                                 output_len += strlen(nl);
 519                                 if (do_citaformat) {
 520                                         strcpy(&outptr[output_len], " ");
 521                                         ++output_len;
 522                                 }
 523                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 524                         }
 525                 }
 526
 527         } while (done_reading == 0);
 528
 529         strcpy(&outptr[output_len], outbuf);
 530         output_len += strlen(outbuf);
 531
 532         /* Strip leading/trailing whitespace.  We can't do this with
 533          * striplt() because it uses too many strlen()'s
 534          */
 535         while ((output_len > 0) && (isspace(outptr[0]))) {
 536                 strcpy(outptr, &outptr[1]);
 537                 --output_len;
 538         }
 539         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 540                 outptr[output_len-1] = 0;
 541                 --output_len;
 542         }
 543
 544         if (outptr[output_len-1] != '\n') {
 545                 strcat(outptr, "\n");
 546                 ++output_len;
 547         }
 548
 549         return outptr;
 550
 551 }