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                                         ++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, "/BLOCKQUOTE")) {
 217                                         strcat(outbuf, "\n");
 218                                         --blockquote;
 219                                         strcpy(nl, "\n");
 220                                         for (j=0; j<blockquote; ++j) strcat(nl, ">");
 221                                         strcat(outbuf, nl);
 222                                 }
 223
 224                         }
 225
 226                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 227                                 tag[strlen(tag)+1] = 0;
 228                                 tag[strlen(tag)] = ch;
 229                         }
 230
 231                         else if (!nest) {
 232                                 outbuf[strlen(outbuf)+1] = 0;
 233                                 outbuf[strlen(outbuf)] = ch;
 234                         }
 235                     }
 236                     strcpy(inbuf, &inbuf[i]);
 237                 }
 238
 239                 /* Convert &; tags to the forbidden characters */
 240                 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
 241
 242                         /* Character entity references */
 243                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 244                                 outbuf[i] = ' ';
 245                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 246                         }
 247
 248                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 249                                 outbuf[i] = ' ';
 250                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 251                         }
 252
 253                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 254                                 outbuf[i] = ' ';
 255                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 256                         }
 257
 258                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 259                                 outbuf[i] = ' ';
 260                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 261                         }
 262
 263                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 264                                 outbuf[i] = '<';
 265                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 266                         }
 267
 268                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 269                                 outbuf[i] = '>';
 270                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 271                         }
 272
 273                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 274                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 275                         }
 276
 277                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 278                                 outbuf[i] = '\"';
 279                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 280                         }
 281
 282                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 283                                 outbuf[i] = '`';
 284                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 285                         }
 286
 287                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 288                                 outbuf[i] = '\'';
 289                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 290                         }
 291
 292                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 293                                 outbuf[i] = '(';
 294                                 outbuf[i+1] = 'c';
 295                                 outbuf[i+2] = ')';
 296                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 297                         }
 298
 299                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 300                                 outbuf[i] = '.';
 301                                 outbuf[i+1] = '.';
 302                                 outbuf[i+2] = '.';
 303                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 304                         }
 305
 306                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 307                                 outbuf[i] = '(';
 308                                 outbuf[i+1] = 't';
 309                                 outbuf[i+2] = 'm';
 310                                 outbuf[i+3] = ')';
 311                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 312                         }
 313
 314                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 315                                 outbuf[i] = '(';
 316                                 outbuf[i+1] = 'r';
 317                                 outbuf[i+2] = ')';
 318                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 319                         }
 320
 321                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 322                                 outbuf[i] = '1';
 323                                 outbuf[i+1] = '/';
 324                                 outbuf[i+2] = '4';
 325                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 326                         }
 327
 328                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 329                                 outbuf[i] = '1';
 330                                 outbuf[i+1] = '/';
 331                                 outbuf[i+2] = '2';
 332                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 333                         }
 334
 335                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 336                                 outbuf[i] = '3';
 337                                 outbuf[i+1] = '/';
 338                                 outbuf[i+2] = '4';
 339                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 340                         }
 341
 342                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 343                                 outbuf[i] = '-';
 344                                 outbuf[i+1] = '-';
 345                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 346                         }
 347
 348                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 349                                 outbuf[i] = '-';
 350                                 outbuf[i+1] = '-';
 351                                 outbuf[i+2] = '-';
 352                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 353                         }
 354
 355                         /* two-digit decimal equivalents */
 356                         else if ((!strncmp(&outbuf[i], "&#", 2))
 357                               && (outbuf[i+4] == ';') ) {
 358                                 scanch = 0;
 359                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 360                                 outbuf[i] = scanch;
 361                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 362                         }
 363
 364                         /* three-digit decimal equivalents */
 365                         else if ((!strncmp(&outbuf[i], "&#", 2))
 366                               && (outbuf[i+5] == ';') ) {
 367                                 scanch = 0;
 368                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 369                                 outbuf[i] = scanch;
 370                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 371                         }
 372
 373                 }
 374
 375                 /* Make sure the output buffer is big enough */
 376                 if ((output_len + strlen(outbuf) + SIZ)
 377                    > outptr_buffer_size) {
 378                         outptr_buffer_size += SIZ;
 379                         outptr = realloc(outptr, outptr_buffer_size);
 380                 }
 381
 382                 /* Output any lines terminated with hard line breaks */
 383                 do {
 384                         did_out = 0;
 385                         if (strlen(outbuf)>0) {
 386                             for (i = 0; i<strlen(outbuf); ++i) {
 387                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 388
 389                                         strncpy(&outptr[output_len],
 390                                                 outbuf, i+1);
 391                                         output_len += (i+1);
 392
 393                                         if (do_citaformat) {
 394                                                 strcpy(&outptr[output_len],
 395                                                         " ");
 396                                                 ++output_len;
 397                                         }
 398
 399                                         strcpy(outbuf, &outbuf[i+1]);
 400                                         i = 0;
 401                                         did_out = 1;
 402                                 }
 403                         }
 404                     }
 405                 } while (did_out);
 406
 407                 /* Add soft line breaks */
 408                 if (strlen(outbuf) > (screenwidth - 2 )) {
 409                         rb = (-1);
 410                         for (i=0; i<(screenwidth-2); ++i) {
 411                                 if (outbuf[i]==32) rb = i;
 412                         }
 413                         if (rb>=0) {
 414                                 strncpy(&outptr[output_len], outbuf, rb);
 415                                 output_len += rb;
 416                                 strcpy(&outptr[output_len], nl);
 417                                 output_len += strlen(nl);
 418                                 if (do_citaformat) {
 419                                         strcpy(&outptr[output_len], " ");
 420                                         ++output_len;
 421                                 }
 422                                 strcpy(outbuf, &outbuf[rb+1]);
 423                         } else {
 424                                 strncpy(&outptr[output_len], outbuf,
 425                                         screenwidth-2);
 426                                 output_len += (screenwidth-2);
 427                                 strcpy(&outptr[output_len], nl);
 428                                 output_len += strlen(nl);
 429                                 if (do_citaformat) {
 430                                         strcpy(&outptr[output_len], " ");
 431                                         ++output_len;
 432                                 }
 433                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 434                         }
 435                 }
 436
 437         } while (done_reading == 0);
 438
 439         strcpy(&outptr[output_len], outbuf);
 440         output_len += strlen(outbuf);
 441
 442         /* Strip leading/trailing whitespace.  We can't do this with
 443          * striplt() because it uses too many strlen()'s
 444          */
 445         while ((output_len > 0) && (isspace(outptr[0]))) {
 446                 strcpy(outptr, &outptr[1]);
 447                 --output_len;
 448         }
 449         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 450                 outptr[output_len-1] = 0;
 451                 --output_len;
 452         }
 453
 454         if (outptr[output_len-1] != '\n') {
 455                 strcat(outptr, "\n");
 456                 ++output_len;
 457         }
 458
 459         return outptr;
 460
 461 }