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 "serv_extensions.h"
  34 #include "control.h"
  35 #include "sysdep_decls.h"
  36 #include "support.h"
  37 #include "config.h"
  38 #include "msgbase.h"
  39 #include "tools.h"
  40 #include "room_ops.h"
  41 #include "html.h"
  42
  43
  44 /*
  45  * Convert HTML to plain text.
  46  *
  47  * inputmsg      = pointer to raw HTML message
  48  * screenwidth   = desired output screenwidth
  49  * do_citaformat = set to 1 to indent newlines with spaces
  50  */
  51 char *html_to_ascii(char *inputmsg, int msglen, int screenwidth, int do_citaformat) {
  52         char inbuf[SIZ];
  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 bytes_processed = 0;
  64         char nl[128];
  65
  66         strcpy(nl, "\n");
  67         inptr = inputmsg;
  68         strcpy(inbuf, "");
  69         strcpy(outbuf, "");
  70         if (msglen == 0) msglen = strlen(inputmsg);
  71
  72         outptr_buffer_size = strlen(inptr) + SIZ;
  73         outptr = malloc(outptr_buffer_size);
  74         if (outptr == NULL) return NULL;
  75         strcpy(outptr, "");
  76         output_len = 0;
  77
  78         do {
  79                 /* Fill the input buffer */
  80                 if ( (done_reading == 0) && (strlen(inbuf) < (SIZ-128)) ) {
  81
  82                         ch = *inptr++;
  83                         if (ch != 0) {
  84                                 inbuf[strlen(inbuf)+1] = 0;
  85                                 inbuf[strlen(inbuf)] = ch;
  86                         }
  87                         else {
  88                                 done_reading = 1;
  89                         }
  90
  91                         ++bytes_processed;
  92                         if (bytes_processed > msglen) {
  93                                 done_reading = 1;
  94                         }
  95
  96                 }
  97
  98                 /* Do some parsing */
  99                 if (strlen(inbuf)>0) {
 100
 101                     /* Fold in all the spacing */
 102                     for (i=0; i<strlen(inbuf); ++i) {
 103                         if (inbuf[i]==10) inbuf[i]=32;
 104                         if (inbuf[i]==13) inbuf[i]=32;
 105                         if (inbuf[i]==9) inbuf[i]=32;
 106                         /*** we like foreign characters now.
 107                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
 108                                 inbuf[i] = '?';
 109                         } */
 110                     }
 111                     for (i=0; i<strlen(inbuf); ++i) {
 112                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
 113                                 strcpy(&inbuf[i], &inbuf[i+1]);
 114                     }
 115
 116                     for (i=0; i<strlen(inbuf); ++i) {
 117
 118                         ch = inbuf[i];
 119
 120                         if (ch == '<') {
 121                                 ++nest;
 122                                 strcpy(tag, "");
 123                         }
 124
 125                         else if (ch == '>') {   /* We have a tag. */
 126                                 if (nest > 0) --nest;
 127
 128                                 /* Unqualify the tag (truncate at first space) */
 129                                 if (strchr(tag, ' ') != NULL) {
 130                                         strcpy(strchr(tag, ' '), "");
 131                                 }
 132
 133                                 if (!strcasecmp(tag, "P")) {
 134                                         strcat(outbuf, nl);
 135                                         strcat(outbuf, nl);
 136                                 }
 137
 138                                 if (!strcasecmp(tag, "/DIV")) {
 139                                         strcat(outbuf, nl);
 140                                         strcat(outbuf, nl);
 141                                 }
 142
 143                                 if (!strcasecmp(tag, "LI")) {
 144                                         strcat(outbuf, nl);
 145                                         strcat(outbuf, " * ");
 146                                 }
 147
 148                                 else if (!strcasecmp(tag, "/UL")) {
 149                                         strcat(outbuf, nl);
 150                                         strcat(outbuf, nl);
 151                                 }
 152
 153                                 else if (!strcasecmp(tag, "H1")) {
 154                                         strcat(outbuf, nl);
 155                                         strcat(outbuf, nl);
 156                                 }
 157
 158                                 else if (!strcasecmp(tag, "H2")) {
 159                                         strcat(outbuf, nl);
 160                                         strcat(outbuf, nl);
 161                                 }
 162
 163                                 else if (!strcasecmp(tag, "H3")) {
 164                                         strcat(outbuf, nl);
 165                                         strcat(outbuf, nl);
 166                                 }
 167
 168                                 else if (!strcasecmp(tag, "H4")) {
 169                                         strcat(outbuf, nl);
 170                                         strcat(outbuf, nl);
 171                                 }
 172
 173                                 else if (!strcasecmp(tag, "/H1")) {
 174                                         strcat(outbuf, nl);
 175                                 }
 176
 177                                 else if (!strcasecmp(tag, "/H2")) {
 178                                         strcat(outbuf, nl);
 179                                 }
 180
 181                                 else if (!strcasecmp(tag, "/H3")) {
 182                                         strcat(outbuf, nl);
 183                                 }
 184
 185                                 else if (!strcasecmp(tag, "/H4")) {
 186                                         strcat(outbuf, nl);
 187                                 }
 188
 189                                 else if (!strcasecmp(tag, "HR")) {
 190                                         strcat(outbuf, nl);
 191                                         strcat(outbuf, " ");
 192                                         for (j=0; j<screenwidth-2; ++j)
 193                                                 strcat(outbuf, "-");
 194                                         strcat(outbuf, nl);
 195                                 }
 196
 197                                 else if (!strcasecmp(tag, "BR")) {
 198                                         strcat(outbuf, nl);
 199                                 }
 200
 201                                 else if (!strcasecmp(tag, "TR")) {
 202                                         strcat(outbuf, nl);
 203                                 }
 204
 205                                 else if (!strcasecmp(tag, "/TABLE")) {
 206                                         strcat(outbuf, nl);
 207                                 }
 208
 209                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 210                                         strcat(outbuf, nl);
 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                                         strcat(outbuf, nl);
 224                                 }
 225
 226                         }
 227
 228                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 229                                 tag[strlen(tag)+1] = 0;
 230                                 tag[strlen(tag)] = ch;
 231                         }
 232
 233                         else if (!nest) {
 234                                 outbuf[strlen(outbuf)+1] = 0;
 235                                 outbuf[strlen(outbuf)] = ch;
 236                         }
 237                     }
 238                     strcpy(inbuf, &inbuf[i]);
 239                 }
 240
 241                 /* Convert &; tags to the forbidden characters */
 242                 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
 243
 244                         /* Character entity references */
 245                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 246                                 outbuf[i] = ' ';
 247                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 248                         }
 249
 250                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 251                                 outbuf[i] = ' ';
 252                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 253                         }
 254
 255                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 256                                 outbuf[i] = ' ';
 257                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 258                         }
 259
 260                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 261                                 outbuf[i] = ' ';
 262                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 263                         }
 264
 265                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 266                                 outbuf[i] = '<';
 267                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 268                         }
 269
 270                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 271                                 outbuf[i] = '>';
 272                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 273                         }
 274
 275                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 276                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 277                         }
 278
 279                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 280                                 outbuf[i] = '\"';
 281                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 282                         }
 283
 284                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 285                                 outbuf[i] = '`';
 286                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 287                         }
 288
 289                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 290                                 outbuf[i] = '\'';
 291                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 292                         }
 293
 294                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 295                                 outbuf[i] = '(';
 296                                 outbuf[i+1] = 'c';
 297                                 outbuf[i+2] = ')';
 298                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 299                         }
 300
 301                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 302                                 outbuf[i] = '.';
 303                                 outbuf[i+1] = '.';
 304                                 outbuf[i+2] = '.';
 305                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 306                         }
 307
 308                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 309                                 outbuf[i] = '(';
 310                                 outbuf[i+1] = 't';
 311                                 outbuf[i+2] = 'm';
 312                                 outbuf[i+3] = ')';
 313                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 314                         }
 315
 316                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 317                                 outbuf[i] = '(';
 318                                 outbuf[i+1] = 'r';
 319                                 outbuf[i+2] = ')';
 320                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 321                         }
 322
 323                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 324                                 outbuf[i] = '1';
 325                                 outbuf[i+1] = '/';
 326                                 outbuf[i+2] = '4';
 327                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 328                         }
 329
 330                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 331                                 outbuf[i] = '1';
 332                                 outbuf[i+1] = '/';
 333                                 outbuf[i+2] = '2';
 334                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 335                         }
 336
 337                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 338                                 outbuf[i] = '3';
 339                                 outbuf[i+1] = '/';
 340                                 outbuf[i+2] = '4';
 341                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 342                         }
 343
 344                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 345                                 outbuf[i] = '-';
 346                                 outbuf[i+1] = '-';
 347                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 348                         }
 349
 350                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 351                                 outbuf[i] = '-';
 352                                 outbuf[i+1] = '-';
 353                                 outbuf[i+2] = '-';
 354                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 355                         }
 356
 357                         /* two-digit decimal equivalents */
 358                         else if ((!strncmp(&outbuf[i], "&#", 2))
 359                               && (outbuf[i+4] == ';') ) {
 360                                 scanch = 0;
 361                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 362                                 outbuf[i] = scanch;
 363                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 364                         }
 365
 366                         /* three-digit decimal equivalents */
 367                         else if ((!strncmp(&outbuf[i], "&#", 2))
 368                               && (outbuf[i+5] == ';') ) {
 369                                 scanch = 0;
 370                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 371                                 outbuf[i] = scanch;
 372                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 373                         }
 374
 375                 }
 376
 377                 /* Make sure the output buffer is big enough */
 378                 if ((output_len + strlen(outbuf) + SIZ)
 379                    > outptr_buffer_size) {
 380                         outptr_buffer_size += SIZ;
 381                         outptr = realloc(outptr, outptr_buffer_size);
 382                 }
 383
 384                 /* Output any lines terminated with hard line breaks */
 385                 do {
 386                         did_out = 0;
 387                         if (strlen(outbuf)>0) {
 388                             for (i = 0; i<strlen(outbuf); ++i) {
 389                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 390
 391                                         strncpy(&outptr[output_len],
 392                                                 outbuf, i+1);
 393                                         output_len += (i+1);
 394
 395                                         if (do_citaformat) {
 396                                                 strcpy(&outptr[output_len],
 397                                                         " ");
 398                                                 ++output_len;
 399                                         }
 400
 401                                         strcpy(outbuf, &outbuf[i+1]);
 402                                         i = 0;
 403                                         did_out = 1;
 404                                 }
 405                         }
 406                     }
 407                 } while (did_out);
 408
 409                 /* Add soft line breaks */
 410                 if (strlen(outbuf) > (screenwidth - 2 )) {
 411                         rb = (-1);
 412                         for (i=0; i<(screenwidth-2); ++i) {
 413                                 if (outbuf[i]==32) rb = i;
 414                         }
 415                         if (rb>=0) {
 416                                 strncpy(&outptr[output_len], outbuf, rb);
 417                                 output_len += rb;
 418                                 strcpy(&outptr[output_len], nl);
 419                                 output_len += strlen(nl);
 420                                 if (do_citaformat) {
 421                                         strcpy(&outptr[output_len], " ");
 422                                         ++output_len;
 423                                 }
 424                                 strcpy(outbuf, &outbuf[rb+1]);
 425                         } else {
 426                                 strncpy(&outptr[output_len], outbuf,
 427                                         screenwidth-2);
 428                                 output_len += (screenwidth-2);
 429                                 strcpy(&outptr[output_len], nl);
 430                                 output_len += strlen(nl);
 431                                 if (do_citaformat) {
 432                                         strcpy(&outptr[output_len], " ");
 433                                         ++output_len;
 434                                 }
 435                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 436                         }
 437                 }
 438
 439         } while (done_reading == 0);
 440
 441         strcpy(&outptr[output_len], outbuf);
 442         output_len += strlen(outbuf);
 443
 444         /* Strip leading/trailing whitespace.  We can't do this with
 445          * striplt() because it uses too many strlen()'s
 446          */
 447         while ((output_len > 0) && (isspace(outptr[0]))) {
 448                 strcpy(outptr, &outptr[1]);
 449                 --output_len;
 450         }
 451         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 452                 outptr[output_len-1] = 0;
 453                 --output_len;
 454         }
 455
 456         if (outptr[output_len-1] != '\n') {
 457                 strcat(outptr, "\n");
 458                 ++output_len;
 459         }
 460
 461         return outptr;
 462
 463 }