citadel/html.c

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