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 msglen, 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         int bytes_processed = 0;
  68
  69         inptr = inputmsg;
  70         strcpy(inbuf, "");
  71         strcpy(outbuf, "");
  72         if (msglen == 0) msglen = strlen(inputmsg);
  73
  74         outptr_buffer_size = strlen(inptr) + SIZ;
  75         outptr = malloc(outptr_buffer_size);
  76         if (outptr == NULL) return NULL;
  77         strcpy(outptr, "");
  78         output_len = 0;
  79
  80         do {
  81                 /* Fill the input buffer */
  82                 if ( (done_reading == 0) && (strlen(inbuf) < (SIZ-128)) ) {
  83
  84                         ch = *inptr++;
  85                         if (ch != 0) {
  86                                 inbuf[strlen(inbuf)+1] = 0;
  87                                 inbuf[strlen(inbuf)] = ch;
  88                         }
  89                         else {
  90                                 done_reading = 1;
  91                         }
  92
  93                         ++bytes_processed;
  94                         if (bytes_processed > msglen) {
  95                                 done_reading = 1;
  96                         }
  97
  98                 }
  99
 100                 /* Do some parsing */
 101                 if (strlen(inbuf)>0) {
 102
 103                     /* Fold in all the spacing */
 104                     for (i=0; i<strlen(inbuf); ++i) {
 105                         if (inbuf[i]==10) inbuf[i]=32;
 106                         if (inbuf[i]==13) inbuf[i]=32;
 107                         if (inbuf[i]==9) inbuf[i]=32;
 108                         /*** we like foreign characters now.
 109                         if ((inbuf[i]<32) || (inbuf[i]>126)) {
 110                                 inbuf[i] = '?';
 111                         } */
 112                     }
 113                     for (i=0; i<strlen(inbuf); ++i) {
 114                         while ((inbuf[i]==32)&&(inbuf[i+1]==32))
 115                                 strcpy(&inbuf[i], &inbuf[i+1]);
 116                     }
 117
 118                     for (i=0; i<strlen(inbuf); ++i) {
 119
 120                         ch = inbuf[i];
 121
 122                         if (ch == '<') {
 123                                 ++nest;
 124                                 strcpy(tag, "");
 125                         }
 126
 127                         else if (ch == '>') {   /* We have a tag. */
 128                                 if (nest > 0) --nest;
 129
 130                                 /* Unqualify the tag (truncate at first space) */
 131                                 if (strchr(tag, ' ') != NULL) {
 132                                         strcpy(strchr(tag, ' '), "");
 133                                 }
 134
 135                                 if (!strcasecmp(tag, "P")) {
 136                                         strcat(outbuf, "\n\n");
 137                                 }
 138
 139                                 if (!strcasecmp(tag, "/DIV")) {
 140                                         strcat(outbuf, "\n\n");
 141                                 }
 142
 143                                 if (!strcasecmp(tag, "LI")) {
 144                                         strcat(outbuf, "\n * ");
 145                                 }
 146
 147                                 else if (!strcasecmp(tag, "/UL")) {
 148                                         strcat(outbuf, "\n\n");
 149                                 }
 150
 151                                 else if (!strcasecmp(tag, "H1")) {
 152                                         strcat(outbuf, "\n\n");
 153                                 }
 154
 155                                 else if (!strcasecmp(tag, "H2")) {
 156                                         strcat(outbuf, "\n\n");
 157                                 }
 158
 159                                 else if (!strcasecmp(tag, "H3")) {
 160                                         strcat(outbuf, "\n\n");
 161                                 }
 162
 163                                 else if (!strcasecmp(tag, "H4")) {
 164                                         strcat(outbuf, "\n\n");
 165                                 }
 166
 167                                 else if (!strcasecmp(tag, "/H1")) {
 168                                         strcat(outbuf, "\n");
 169                                 }
 170
 171                                 else if (!strcasecmp(tag, "/H2")) {
 172                                         strcat(outbuf, "\n");
 173                                 }
 174
 175                                 else if (!strcasecmp(tag, "/H3")) {
 176                                         strcat(outbuf, "\n");
 177                                 }
 178
 179                                 else if (!strcasecmp(tag, "/H4")) {
 180                                         strcat(outbuf, "\n");
 181                                 }
 182
 183                                 else if (!strcasecmp(tag, "HR")) {
 184                                         strcat(outbuf, "\n ");
 185                                         for (j=0; j<screenwidth-2; ++j)
 186                                                 strcat(outbuf, "-");
 187                                         strcat(outbuf, "\n");
 188                                 }
 189
 190                                 else if (!strcasecmp(tag, "BR")) {
 191                                         strcat(outbuf, "\n");
 192                                 }
 193
 194                                 else if (!strcasecmp(tag, "TR")) {
 195                                         strcat(outbuf, "\n");
 196                                 }
 197
 198                                 else if (!strcasecmp(tag, "/TABLE")) {
 199                                         strcat(outbuf, "\n");
 200                                 }
 201
 202                                 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
 203                                         strcat(outbuf, "\n\n <<\n");
 204                                         ++blockquote;
 205                                 }
 206
 207                                 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
 208                                         strcat(outbuf, "\n >>\n\n");
 209                                         --blockquote;
 210                                 }
 211
 212                         }
 213
 214                         else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
 215                                 tag[strlen(tag)+1] = 0;
 216                                 tag[strlen(tag)] = ch;
 217                         }
 218
 219                         else if (!nest) {
 220                                 outbuf[strlen(outbuf)+1] = 0;
 221                                 outbuf[strlen(outbuf)] = ch;
 222                         }
 223                     }
 224                     strcpy(inbuf, &inbuf[i]);
 225                 }
 226
 227                 /* Convert &; tags to the forbidden characters */
 228                 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
 229
 230                         /* Character entity references */
 231                         if (!strncasecmp(&outbuf[i], "&nbsp;", 6)) {
 232                                 outbuf[i] = ' ';
 233                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 234                         }
 235
 236                         if (!strncasecmp(&outbuf[i], "&ensp;", 6)) {
 237                                 outbuf[i] = ' ';
 238                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 239                         }
 240
 241                         if (!strncasecmp(&outbuf[i], "&emsp;", 6)) {
 242                                 outbuf[i] = ' ';
 243                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 244                         }
 245
 246                         if (!strncasecmp(&outbuf[i], "&thinsp;", 8)) {
 247                                 outbuf[i] = ' ';
 248                                 strcpy(&outbuf[i+1], &outbuf[i+8]);
 249                         }
 250
 251                         else if (!strncasecmp(&outbuf[i], "&lt;", 4)) {
 252                                 outbuf[i] = '<';
 253                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 254                         }
 255
 256                         else if (!strncasecmp(&outbuf[i], "&gt;", 4)) {
 257                                 outbuf[i] = '>';
 258                                 strcpy(&outbuf[i+1], &outbuf[i+4]);
 259                         }
 260
 261                         else if (!strncasecmp(&outbuf[i], "&amp;", 5)) {
 262                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 263                         }
 264
 265                         else if (!strncasecmp(&outbuf[i], "&quot;", 6)) {
 266                                 outbuf[i] = '\"';
 267                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 268                         }
 269
 270                         else if (!strncasecmp(&outbuf[i], "&lsquo;", 7)) {
 271                                 outbuf[i] = '`';
 272                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 273                         }
 274
 275                         else if (!strncasecmp(&outbuf[i], "&rsquo;", 7)) {
 276                                 outbuf[i] = '\'';
 277                                 strcpy(&outbuf[i+1], &outbuf[i+7]);
 278                         }
 279
 280                         else if (!strncasecmp(&outbuf[i], "&copy;", 6)) {
 281                                 outbuf[i] = '(';
 282                                 outbuf[i+1] = 'c';
 283                                 outbuf[i+2] = ')';
 284                                 strcpy(&outbuf[i+3], &outbuf[i+6]);
 285                         }
 286
 287                         else if (!strncasecmp(&outbuf[i], "&hellip;", 8)) {
 288                                 outbuf[i] = '.';
 289                                 outbuf[i+1] = '.';
 290                                 outbuf[i+2] = '.';
 291                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 292                         }
 293
 294                         else if (!strncasecmp(&outbuf[i], "&trade;", 7)) {
 295                                 outbuf[i] = '(';
 296                                 outbuf[i+1] = 't';
 297                                 outbuf[i+2] = 'm';
 298                                 outbuf[i+3] = ')';
 299                                 strcpy(&outbuf[i+4], &outbuf[i+7]);
 300                         }
 301
 302                         else if (!strncasecmp(&outbuf[i], "&reg;", 5)) {
 303                                 outbuf[i] = '(';
 304                                 outbuf[i+1] = 'r';
 305                                 outbuf[i+2] = ')';
 306                                 strcpy(&outbuf[i+3], &outbuf[i+5]);
 307                         }
 308
 309                         else if (!strncasecmp(&outbuf[i], "&frac14;", 8)) {
 310                                 outbuf[i] = '1';
 311                                 outbuf[i+1] = '/';
 312                                 outbuf[i+2] = '4';
 313                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 314                         }
 315
 316                         else if (!strncasecmp(&outbuf[i], "&frac12;", 8)) {
 317                                 outbuf[i] = '1';
 318                                 outbuf[i+1] = '/';
 319                                 outbuf[i+2] = '2';
 320                                 strcpy(&outbuf[i+3], &outbuf[i+8]);
 321                         }
 322
 323                         else if (!strncasecmp(&outbuf[i], "&frac34;", 8)) {
 324                                 outbuf[i] = '3';
 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], "&ndash;", 7)) {
 331                                 outbuf[i] = '-';
 332                                 outbuf[i+1] = '-';
 333                                 strcpy(&outbuf[i+2], &outbuf[i+7]);
 334                         }
 335
 336                         else if (!strncasecmp(&outbuf[i], "&mdash;", 7)) {
 337                                 outbuf[i] = '-';
 338                                 outbuf[i+1] = '-';
 339                                 outbuf[i+2] = '-';
 340                                 strcpy(&outbuf[i+3], &outbuf[i+7]);
 341                         }
 342
 343                         /* two-digit decimal equivalents */
 344                         else if ((!strncmp(&outbuf[i], "&#", 2))
 345                               && (outbuf[i+4] == ';') ) {
 346                                 scanch = 0;
 347                                 sscanf(&outbuf[i+2], "%02d", &scanch);
 348                                 outbuf[i] = scanch;
 349                                 strcpy(&outbuf[i+1], &outbuf[i+5]);
 350                         }
 351
 352                         /* three-digit decimal equivalents */
 353                         else if ((!strncmp(&outbuf[i], "&#", 2))
 354                               && (outbuf[i+5] == ';') ) {
 355                                 scanch = 0;
 356                                 sscanf(&outbuf[i+2], "%03d", &scanch);
 357                                 outbuf[i] = scanch;
 358                                 strcpy(&outbuf[i+1], &outbuf[i+6]);
 359                         }
 360
 361                 }
 362
 363                 /* Make sure the output buffer is big enough */
 364                 if ((output_len + strlen(outbuf) + SIZ)
 365                    > outptr_buffer_size) {
 366                         outptr_buffer_size += SIZ;
 367                         outptr = realloc(outptr, outptr_buffer_size);
 368                 }
 369
 370                 /* Output any lines terminated with hard line breaks */
 371                 do {
 372                         did_out = 0;
 373                         if (strlen(outbuf)>0) {
 374                             for (i = 0; i<strlen(outbuf); ++i) {
 375                                 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
 376
 377                                         strncpy(&outptr[output_len],
 378                                                 outbuf, i+1);
 379                                         output_len += (i+1);
 380
 381                                         if (do_citaformat) {
 382                                                 strcpy(&outptr[output_len],
 383                                                         " ");
 384                                                 ++output_len;
 385                                         }
 386
 387                                         strcpy(outbuf, &outbuf[i+1]);
 388                                         i = 0;
 389                                         did_out = 1;
 390                                 }
 391                         }
 392                     }
 393                 } while (did_out);
 394
 395                 /* Add soft line breaks */
 396                 if (strlen(outbuf) > (screenwidth - 2 )) {
 397                         rb = (-1);
 398                         for (i=0; i<(screenwidth-2); ++i) {
 399                                 if (outbuf[i]==32) rb = i;
 400                         }
 401                         if (rb>=0) {
 402                                 strncpy(&outptr[output_len], outbuf, rb);
 403                                 output_len += rb;
 404                                 strcpy(&outptr[output_len], "\n");
 405                                 output_len += 1;
 406                                 if (do_citaformat) {
 407                                         strcpy(&outptr[output_len], " ");
 408                                         ++output_len;
 409                                 }
 410                                 strcpy(outbuf, &outbuf[rb+1]);
 411                         } else {
 412                                 strncpy(&outptr[output_len], outbuf,
 413                                         screenwidth-2);
 414                                 output_len += (screenwidth-2);
 415                                 strcpy(&outptr[output_len], "\n");
 416                                 output_len += 1;
 417                                 if (do_citaformat) {
 418                                         strcpy(&outptr[output_len], " ");
 419                                         ++output_len;
 420                                 }
 421                                 strcpy(outbuf, &outbuf[screenwidth-2]);
 422                         }
 423                 }
 424
 425         } while (done_reading == 0);
 426
 427         strcpy(&outptr[output_len], outbuf);
 428         output_len += strlen(outbuf);
 429
 430         /* Strip leading/trailing whitespace.  We can't do this with
 431          * striplt() because it uses too many strlen()'s
 432          */
 433         while ((output_len > 0) && (isspace(outptr[0]))) {
 434                 strcpy(outptr, &outptr[1]);
 435                 --output_len;
 436         }
 437         while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
 438                 outptr[output_len-1] = 0;
 439                 --output_len;
 440         }
 441
 442         if (outptr[output_len-1] != '\n') {
 443                 strcat(outptr, "\n");
 444                 ++output_len;
 445         }
 446
 447         return outptr;
 448
 449 }