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
  65         inptr = inputmsg;
  66         strcpy(inbuf, "");
  67         strcpy(outbuf, "");
  68         if (msglen == 0) msglen = strlen(inputmsg);
  69
  70         outptr_buffer_size = strlen(inptr) + SIZ;
  71         outptr = malloc(outptr_buffer_size);
  72         if (outptr == NULL) return NULL;
  73         strcpy(outptr, "");
  74         output_len = 0;
  75
  76         do {
  77                 /* Fill the input buffer */
  78                 if ( (done_reading == 0) && (strlen(inbuf) < (SIZ-128)) ) {
  79
  80                         ch = *inptr++;
  81                         if (ch != 0) {
  82                                 inbuf[strlen(inbuf)+1] = 0;
  83                                 inbuf[strlen(inbuf)] = ch;
  84                         }
  85                         else {
  86                                 done_reading = 1;
  87                         }
  88
  89                         ++bytes_processed;
  90                         if (bytes_processed > msglen) {
  91                                 done_reading = 1;
  92                         }
  93
  94                 }
  95
  96                 /* Do some parsing */
  97                 if (strlen(inbuf)>0) {
  98
  99                     /* Fold in all the spacing */
 100                     for (i=0; i<strlen(inbuf); ++i) {
 101                         if (inbuf[i]==10) inbuf[i]=32;
 102                         if (inbuf[i]==13) inbuf[i]=32;
 103                         if (inbuf[i]==9) inbuf[i]=32;
 104                         /*** we like foreign characters now.
 105                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
 106                                 inbuf[i] = '?';
 107                         } */
 108                     }
 109                     for (i=0; i<strlen(inbuf); ++i) {
 110                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
 111                                 strcpy(&inbuf[i], &inbuf[i+1]);
 112                     }
 113
 114                     for (i=0; i<strlen(inbuf); ++i) {
 115
 116                         ch = inbuf[i];
 117
 118                         if (ch == '<') {
 119                                 ++nest;
 120                                 strcpy(tag, "");
 121                         }
 122
 123                         else if (ch == '>') {   /* We have a tag. */
 124                                 if (nest > 0) --nest;
 125
 126                                 /* Unqualify the tag (truncate at first space) */
 127                                 if (strchr(tag, ' ') != NULL) {
 128                                         strcpy(strchr(tag, ' '), "");
 129                                 }
 130
 131                                 if (!strcasecmp(tag, "P")) {
 132                                         strcat(outbuf, "\n\n");
 133                                 }
 134
 135                                 if (!strcasecmp(tag, "/DIV")) {
 136                                         strcat(outbuf, "\n\n");
 137                                 }
 138
 139                                 if (!strcasecmp(tag, "LI")) {
 140                                         strcat(outbuf, "\n * ");
 141                                 }
 142
 143                                 else if (!strcasecmp(tag, "/UL")) {
 144                                         strcat(outbuf, "\n\n");
 145                                 }
 146
 147                                 else if (!strcasecmp(tag, "H1")) {
 148                                         strcat(outbuf, "\n\n");
 149                                 }
 150
 151                                 else if (!strcasecmp(tag, "H2")) {
 152                                         strcat(outbuf, "\n\n");
 153                                 }
 154
 155                                 else if (!strcasecmp(tag, "H3")) {
 156                                         strcat(outbuf, "\n\n");
 157                                 }
 158
 159                                 else if (!strcasecmp(tag, "H4")) {
 160                                         strcat(outbuf, "\n\n");
 161                                 }
 162
 163                                 else if (!strcasecmp(tag, "/H1")) {
 164                                         strcat(outbuf, "\n");
 165                                 }
 166
 167                                 else if (!strcasecmp(tag, "/H2")) {
 168                                         strcat(outbuf, "\n");
 169                                 }
 170
 171                                 else if (!strcasecmp(tag, "/H3")) {
 172                                         strcat(outbuf, "\n");
 173                                 }
 174
 175                                 else if (!strcasecmp(tag, "/H4")) {
 176                                         strcat(outbuf, "\n");
 177                                 }
 178
 179                                 else if (!strcasecmp(tag, "HR")) {
 180                                         strcat(outbuf, "\n ");
 181                                         for (j=0; j<screenwidth-2; ++j)
 182                                                 strcat(outbuf, "-");
 183                                         strcat(outbuf, "\n");
 184                                 }
 185
 186                                 else if (!strcasecmp(tag, "BR")) {
 187                                         strcat(outbuf, "\n");
 188                                 }
 189
 190                                 else if (!strcasecmp(tag, "TR")) {
 191                                         strcat(outbuf, "\n");
 192                                 }
 193
 194                                 else if (!strcasecmp(tag, "/TABLE")) {
 195                                         strcat(outbuf, "\n");
 196                                 }
 197
 198                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 199                                         strcat(outbuf, "\n\n <<\n");
 200                                         ++blockquote;
 201                                 }
 202
 203                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 204                                         strcat(outbuf, "\n >>\n\n");
 205                                         --blockquote;
 206                                 }
 207
 208                         }
 209
 210                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 211                                 tag[strlen(tag)+1] = 0;
 212                                 tag[strlen(tag)] = ch;
 213                         }
 214
 215                         else if (!nest) {
 216                                 outbuf[strlen(outbuf)+1] = 0;
 217                                 outbuf[strlen(outbuf)] = ch;
 218                         }
 219                     }
 220                     strcpy(inbuf, &inbuf[i]);
 221                 }
 222
 223                 /* Convert &; tags to the forbidden characters */
 224                 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
 225
 226                         /* Character entity references */
 227                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 228                                 outbuf[i] = ' ';
 229                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 230                         }
 231
 232                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 233                                 outbuf[i] = ' ';
 234                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 235                         }
 236
 237                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 238                                 outbuf[i] = ' ';
 239                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 240                         }
 241
 242                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 243                                 outbuf[i] = ' ';
 244                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 245                         }
 246
 247                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 248                                 outbuf[i] = '<';
 249                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 250                         }
 251
 252                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 253                                 outbuf[i] = '>';
 254                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 255                         }
 256
 257                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 258                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 259                         }
 260
 261                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 262                                 outbuf[i] = '\"';
 263                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 264                         }
 265
 266                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 267                                 outbuf[i] = '`';
 268                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 269                         }
 270
 271                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 272                                 outbuf[i] = '\'';
 273                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 274                         }
 275
 276                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 277                                 outbuf[i] = '(';
 278                                 outbuf[i+1] = 'c';
 279                                 outbuf[i+2] = ')';
 280                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 281                         }
 282
 283                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 284                                 outbuf[i] = '.';
 285                                 outbuf[i+1] = '.';
 286                                 outbuf[i+2] = '.';
 287                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 288                         }
 289
 290                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 291                                 outbuf[i] = '(';
 292                                 outbuf[i+1] = 't';
 293                                 outbuf[i+2] = 'm';
 294                                 outbuf[i+3] = ')';
 295                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 296                         }
 297
 298                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 299                                 outbuf[i] = '(';
 300                                 outbuf[i+1] = 'r';
 301                                 outbuf[i+2] = ')';
 302                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 303                         }
 304
 305                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 306                                 outbuf[i] = '1';
 307                                 outbuf[i+1] = '/';
 308                                 outbuf[i+2] = '4';
 309                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 310                         }
 311
 312                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 313                                 outbuf[i] = '1';
 314                                 outbuf[i+1] = '/';
 315                                 outbuf[i+2] = '2';
 316                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 317                         }
 318
 319                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 320                                 outbuf[i] = '3';
 321                                 outbuf[i+1] = '/';
 322                                 outbuf[i+2] = '4';
 323                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 324                         }
 325
 326                         else if (!strncasecmp(&outbuf[i], "&ndash;", 7)) {
 327                                 outbuf[i] = '-';
 328                                 outbuf[i+1] = '-';
 329                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 330                         }
 331
 332                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 333                                 outbuf[i] = '-';
 334                                 outbuf[i+1] = '-';
 335                                 outbuf[i+2] = '-';
 336                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 337                         }
 338
 339                         /* two-digit decimal equivalents */
 340                         else if ((!strncmp(&outbuf[i], "&#", 2))
 341                               && (outbuf[i+4] == ';') ) {
 342                                 scanch = 0;
 343                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 344                                 outbuf[i] = scanch;
 345                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 346                         }
 347
 348                         /* three-digit decimal equivalents */
 349                         else if ((!strncmp(&outbuf[i], "&#", 2))
 350                               && (outbuf[i+5] == ';') ) {
 351                                 scanch = 0;
 352                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 353                                 outbuf[i] = scanch;
 354                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 355                         }
 356
 357                 }
 358
 359                 /* Make sure the output buffer is big enough */
 360                 if ((output_len + strlen(outbuf) + SIZ)
 361                    > outptr_buffer_size) {
 362                         outptr_buffer_size += SIZ;
 363                         outptr = realloc(outptr, outptr_buffer_size);
 364                 }
 365
 366                 /* Output any lines terminated with hard line breaks */
 367                 do {
 368                         did_out = 0;
 369                         if (strlen(outbuf)>0) {
 370                             for (i = 0; i<strlen(outbuf); ++i) {
 371                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 372
 373                                         strncpy(&outptr[output_len],
 374                                                 outbuf, i+1);
 375                                         output_len += (i+1);
 376
 377                                         if (do_citaformat) {
 378                                                 strcpy(&outptr[output_len],
 379                                                         " ");
 380                                                 ++output_len;
 381                                         }
 382
 383                                         strcpy(outbuf, &outbuf[i+1]);
 384                                         i = 0;
 385                                         did_out = 1;
 386                                 }
 387                         }
 388                     }
 389                 } while (did_out);
 390
 391                 /* Add soft line breaks */
 392                 if (strlen(outbuf) > (screenwidth - 2 )) {
 393                         rb = (-1);
 394                         for (i=0; i<(screenwidth-2); ++i) {
 395                                 if (outbuf[i]==32) rb = i;
 396                         }
 397                         if (rb>=0) {
 398                                 strncpy(&outptr[output_len], outbuf, rb);
 399                                 output_len += rb;
 400                                 strcpy(&outptr[output_len], "\n");
 401                                 output_len += 1;
 402                                 if (do_citaformat) {
 403                                         strcpy(&outptr[output_len], " ");
 404                                         ++output_len;
 405                                 }
 406                                 strcpy(outbuf, &outbuf[rb+1]);
 407                         } else {
 408                                 strncpy(&outptr[output_len], outbuf,
 409                                         screenwidth-2);
 410                                 output_len += (screenwidth-2);
 411                                 strcpy(&outptr[output_len], "\n");
 412                                 output_len += 1;
 413                                 if (do_citaformat) {
 414                                         strcpy(&outptr[output_len], " ");
 415                                         ++output_len;
 416                                 }
 417                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 418                         }
 419                 }
 420
 421         } while (done_reading == 0);
 422
 423         strcpy(&outptr[output_len], outbuf);
 424         output_len += strlen(outbuf);
 425
 426         /* Strip leading/trailing whitespace.  We can't do this with
 427          * striplt() because it uses too many strlen()'s
 428          */
 429         while ((output_len > 0) && (isspace(outptr[0]))) {
 430                 strcpy(outptr, &outptr[1]);
 431                 --output_len;
 432         }
 433         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 434                 outptr[output_len-1] = 0;
 435                 --output_len;
 436         }
 437
 438         if (outptr[output_len-1] != '\n') {
 439                 strcat(outptr, "\n");
 440                 ++output_len;
 441         }
 442
 443         return outptr;
 444
 445 }