citadel/html.c

   1 /*
   2  * $Id$
   3  *
   4  * Functions which handle translation between HTML and plain text
   5  * Copyright (c) 2000-2001 by Art Cancro and others.   This program is
   6  * released under the terms of the GNU General Public License.
   7  */
   8
   9 #ifdef DLL_EXPORT
  10 #define IN_LIBCIT
  11 #endif
  12
  13 #include "sysdep.h"
  14 #include <stdlib.h>
  15 #include <unistd.h>
  16 #include <stdio.h>
  17 #include <fcntl.h>
  18 #include <signal.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 <ctype.h>
  32 #include <string.h>
  33 #include <errno.h>
  34 #include <limits.h>
  35 #include "citadel.h"
  36 #include "server.h"
  37 #include "serv_extensions.h"
  38 #include "control.h"
  39 #include "sysdep_decls.h"
  40 #include "support.h"
  41 #include "config.h"
  42 #include "msgbase.h"
  43 #include "tools.h"
  44 #include "room_ops.h"
  45 #include "html.h"
  46
  47
  48 /*
  49  * Convert HTML to plain text.
  50  *
  51  * inputmsg      = pointer to raw HTML message
  52  * screenwidth   = desired output screenwidth
  53  * do_citaformat = set to 1 to indent newlines with spaces
  54  */
  55 char *html_to_ascii(char *inputmsg, int screenwidth, int do_citaformat) {
  56         char inbuf[SIZ];
  57         char outbuf[SIZ];
  58         char tag[1024];
  59         int done_reading = 0;
  60         char *inptr;
  61         char *outptr;
  62         size_t outptr_buffer_size;
  63         size_t output_len = 0;
  64         int i, j, ch, did_out, rb, scanch;
  65         int nest = 0;           /* Bracket nesting level */
  66         int blockquote = 0;     /* BLOCKQUOTE nesting level */
  67
  68         inptr = inputmsg;
  69         strcpy(inbuf, "");
  70         strcpy(outbuf, "");
  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                 }
  92
  93                 /* Do some parsing */
  94                 if (strlen(inbuf)>0) {
  95
  96                     /* Fold in all the spacing */
  97                     for (i=0; i<strlen(inbuf); ++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                         /*** we like foreign characters now.
 102                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
 103                                 inbuf[i] = '?';
 104                         } */
 105                     }
 106                     for (i=0; i<strlen(inbuf); ++i) {
 107                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
 108                                 strcpy(&inbuf[i], &inbuf[i+1]);
 109                     }
 110
 111                     for (i=0; i<strlen(inbuf); ++i) {
 112
 113                         ch = inbuf[i];
 114
 115                         if (ch == '<') {
 116                                 ++nest;
 117                                 strcpy(tag, "");
 118                         }
 119
 120                         else if (ch == '>') {   /* We have a tag. */
 121                                 if (nest > 0) --nest;
 122
 123                                 /* Unqualify the tag (truncate at first space) */
 124                                 if (strchr(tag, ' ') != NULL) {
 125                                         strcpy(strchr(tag, ' '), "");
 126                                 }
 127
 128                                 if (!strcasecmp(tag, "P")) {
 129                                         strcat(outbuf, "\n\n");
 130                                 }
 131
 132                                 if (!strcasecmp(tag, "/DIV")) {
 133                                         strcat(outbuf, "\n\n");
 134                                 }
 135
 136                                 if (!strcasecmp(tag, "LI")) {
 137                                         strcat(outbuf, "\n * ");
 138                                 }
 139
 140                                 else if (!strcasecmp(tag, "/UL")) {
 141                                         strcat(outbuf, "\n\n");
 142                                 }
 143
 144                                 else if (!strcasecmp(tag, "H1")) {
 145                                         strcat(outbuf, "\n\n");
 146                                 }
 147
 148                                 else if (!strcasecmp(tag, "H2")) {
 149                                         strcat(outbuf, "\n\n");
 150                                 }
 151
 152                                 else if (!strcasecmp(tag, "H3")) {
 153                                         strcat(outbuf, "\n\n");
 154                                 }
 155
 156                                 else if (!strcasecmp(tag, "H4")) {
 157                                         strcat(outbuf, "\n\n");
 158                                 }
 159
 160                                 else if (!strcasecmp(tag, "/H1")) {
 161                                         strcat(outbuf, "\n");
 162                                 }
 163
 164                                 else if (!strcasecmp(tag, "/H2")) {
 165                                         strcat(outbuf, "\n");
 166                                 }
 167
 168                                 else if (!strcasecmp(tag, "/H3")) {
 169                                         strcat(outbuf, "\n");
 170                                 }
 171
 172                                 else if (!strcasecmp(tag, "/H4")) {
 173                                         strcat(outbuf, "\n");
 174                                 }
 175
 176                                 else if (!strcasecmp(tag, "HR")) {
 177                                         strcat(outbuf, "\n ");
 178                                         for (j=0; j<screenwidth-2; ++j)
 179                                                 strcat(outbuf, "-");
 180                                         strcat(outbuf, "\n");
 181                                 }
 182
 183                                 else if (!strcasecmp(tag, "BR")) {
 184                                         strcat(outbuf, "\n");
 185                                 }
 186
 187                                 else if (!strcasecmp(tag, "TR")) {
 188                                         strcat(outbuf, "\n");
 189                                 }
 190
 191                                 else if (!strcasecmp(tag, "/TABLE")) {
 192                                         strcat(outbuf, "\n");
 193                                 }
 194
 195                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 196                                         strcat(outbuf, "\n\n <<\n");
 197                                         ++blockquote;
 198                                 }
 199
 200                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 201                                         strcat(outbuf, "\n >>\n\n");
 202                                         --blockquote;
 203                                 }
 204
 205                         }
 206
 207                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 208                                 tag[strlen(tag)+1] = 0;
 209                                 tag[strlen(tag)] = ch;
 210                         }
 211
 212                         else if (!nest) {
 213                                 outbuf[strlen(outbuf)+1] = 0;
 214                                 outbuf[strlen(outbuf)] = ch;
 215                         }
 216                     }
 217                     strcpy(inbuf, &inbuf[i]);
 218                 }
 219
 220                 /* Convert &; tags to the forbidden characters */
 221                 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
 222
 223                         /* Character entity references */
 224                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 225                                 outbuf[i] = ' ';
 226                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 227                         }
 228
 229                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 230                                 outbuf[i] = ' ';
 231                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 232                         }
 233
 234                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 235                                 outbuf[i] = ' ';
 236                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 237                         }
 238
 239                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 240                                 outbuf[i] = ' ';
 241                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 242                         }
 243
 244                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 245                                 outbuf[i] = '<';
 246                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 247                         }
 248
 249                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 250                                 outbuf[i] = '>';
 251                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 252                         }
 253
 254                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 255                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 256                         }
 257
 258                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 259                                 outbuf[i] = '\"';
 260                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 261                         }
 262
 263                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 264                                 outbuf[i] = '`';
 265                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 266                         }
 267
 268                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 269                                 outbuf[i] = '\'';
 270                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 271                         }
 272
 273                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 274                                 outbuf[i] = '(';
 275                                 outbuf[i+1] = 'c';
 276                                 outbuf[i+2] = ')';
 277                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 278                         }
 279
 280                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 281                                 outbuf[i] = '.';
 282                                 outbuf[i+1] = '.';
 283                                 outbuf[i+2] = '.';
 284                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 285                         }
 286
 287                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 288                                 outbuf[i] = '(';
 289                                 outbuf[i+1] = 't';
 290                                 outbuf[i+2] = 'm';
 291                                 outbuf[i+3] = ')';
 292                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 293                         }
 294
 295                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 296                                 outbuf[i] = '(';
 297                                 outbuf[i+1] = 'r';
 298                                 outbuf[i+2] = ')';
 299                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 300                         }
 301
 302                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 303                                 outbuf[i] = '1';
 304                                 outbuf[i+1] = '/';
 305                                 outbuf[i+2] = '4';
 306                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 307                         }
 308
 309                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 310                                 outbuf[i] = '1';
 311                                 outbuf[i+1] = '/';
 312                                 outbuf[i+2] = '2';
 313                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 314                         }
 315
 316                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 317                                 outbuf[i] = '3';
 318                                 outbuf[i+1] = '/';
 319                                 outbuf[i+2] = '4';
 320                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 321                         }
 322
 323                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 324                                 outbuf[i] = '-';
 325                                 outbuf[i+1] = '-';
 326                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 327                         }
 328
 329                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 330                                 outbuf[i] = '-';
 331                                 outbuf[i+1] = '-';
 332                                 outbuf[i+2] = '-';
 333                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 334                         }
 335
 336                         /* two-digit decimal equivalents */
 337                         else if ((!strncmp(&outbuf[i], "&#", 2))
 338                               && (outbuf[i+4] == ';') ) {
 339                                 scanch = 0;
 340                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 341                                 outbuf[i] = scanch;
 342                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 343                         }
 344
 345                         /* three-digit decimal equivalents */
 346                         else if ((!strncmp(&outbuf[i], "&#", 2))
 347                               && (outbuf[i+5] == ';') ) {
 348                                 scanch = 0;
 349                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 350                                 outbuf[i] = scanch;
 351                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 352                         }
 353
 354                 }
 355
 356                 /* Make sure the output buffer is big enough */
 357                 if ((output_len + strlen(outbuf) + SIZ)
 358                    > outptr_buffer_size) {
 359                         outptr_buffer_size += SIZ;
 360                         outptr = realloc(outptr, outptr_buffer_size);
 361                 }
 362
 363                 /* Output any lines terminated with hard line breaks */
 364                 do {
 365                         did_out = 0;
 366                         if (strlen(outbuf)>0) {
 367                             for (i = 0; i<strlen(outbuf); ++i) {
 368                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 369
 370                                         strncpy(&outptr[output_len],
 371                                                 outbuf, i+1);
 372                                         output_len += (i+1);
 373
 374                                         if (do_citaformat) {
 375                                                 strcpy(&outptr[output_len],
 376                                                         " ");
 377                                                 ++output_len;
 378                                         }
 379
 380                                         strcpy(outbuf, &outbuf[i+1]);
 381                                         i = 0;
 382                                         did_out = 1;
 383                                 }
 384                         }
 385                     }
 386                 } while (did_out);
 387
 388                 /* Add soft line breaks */
 389                 if (strlen(outbuf) > (screenwidth - 2 )) {
 390                         rb = (-1);
 391                         for (i=0; i<(screenwidth-2); ++i) {
 392                                 if (outbuf[i]==32) rb = i;
 393                         }
 394                         if (rb>=0) {
 395                                 strncpy(&outptr[output_len], outbuf, rb);
 396                                 output_len += rb;
 397                                 strcpy(&outptr[output_len], "\n");
 398                                 output_len += 1;
 399                                 if (do_citaformat) {
 400                                         strcpy(&outptr[output_len], " ");
 401                                         ++output_len;
 402                                 }
 403                                 strcpy(outbuf, &outbuf[rb+1]);
 404                         } else {
 405                                 strncpy(&outptr[output_len], outbuf,
 406                                         screenwidth-2);
 407                                 output_len += (screenwidth-2);
 408                                 strcpy(&outptr[output_len], "\n");
 409                                 output_len += 1;
 410                                 if (do_citaformat) {
 411                                         strcpy(&outptr[output_len], " ");
 412                                         ++output_len;
 413                                 }
 414                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 415                         }
 416                 }
 417
 418         } while (done_reading == 0);
 419
 420         strcpy(&outptr[output_len], outbuf);
 421         output_len += strlen(outbuf);
 422
 423         /* Strip leading/trailing whitespace.  We can't do this with
 424          * striplt() because it uses too many strlen()'s
 425          */
 426         while ((output_len > 0) && (isspace(outptr[0]))) {
 427                 strcpy(outptr, &outptr[1]);
 428                 --output_len;
 429         }
 430         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 431                 outptr[output_len-1] = 0;
 432                 --output_len;
 433         }
 434
 435         if (outptr[output_len-1] != '\n') {
 436                 strcat(outptr, "\n");
 437                 ++output_len;
 438         }
 439
 440         return outptr;
 441
 442 }