citadel/server/modules/fulltext/serv_fulltext.c

   1 /*
   2  * This module handles fulltext indexing of the message base.
   3  * Copyright (c) 2005-2022 by the citadel.org team
   4  *
   5  * This program is open source software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License as published
   7  * by the Free Software Foundation; either version 3 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  18  */
  19
  20 #include "../../sysdep.h"
  21 #include <stdlib.h>
  22 #include <unistd.h>
  23 #include <stdio.h>
  24 #include <fcntl.h>
  25 #include <signal.h>
  26 #include <pwd.h>
  27 #include <errno.h>
  28 #include <sys/types.h>
  29 #include <time.h>
  30 #include <sys/wait.h>
  31 #include <string.h>
  32 #include <limits.h>
  33 #include <libcitadel.h>
  34 #include "../../citadel.h"
  35 #include "../../server.h"
  36 #include "../../citserver.h"
  37 #include "../../support.h"
  38 #include "../../config.h"
  39 #include "../../database.h"
  40 #include "../../msgbase.h"
  41 #include "../../control.h"
  42 #include "serv_fulltext.h"
  43 #include "ft_wordbreaker.h"
  44 #include "../../threads.h"
  45 #include "../../context.h"
  46 #include "../../ctdl_module.h"
  47
  48 long ft_newhighest = 0L;
  49 long *ft_newmsgs = NULL;
  50 int ft_num_msgs = 0;
  51 int ft_num_alloc = 0;
  52
  53 int ftc_num_msgs[65536];
  54 long *ftc_msgs[65536];
  55
  56
  57 /*
  58  * Compare function
  59  */
  60 int longcmp(const void *rec1, const void *rec2) {
  61         long i1, i2;
  62
  63         i1 = *(const long *)rec1;
  64         i2 = *(const long *)rec2;
  65
  66         if (i1 > i2) return(1);
  67         if (i1 < i2) return(-1);
  68         return(0);
  69 }
  70
  71
  72 /*
  73  * Flush our index cache out to disk.
  74  */
  75 void ft_flush_cache(void) {
  76         int i;
  77         time_t last_update = 0;
  78
  79         for (i=0; i<65536; ++i) {
  80                 if ((time(NULL) - last_update) >= 10) {
  81                         syslog(LOG_INFO,
  82                                 "fulltext: flushing index cache to disk (%d%% complete)",
  83                                 (i * 100 / 65536)
  84                         );
  85                         last_update = time(NULL);
  86                 }
  87                 if (ftc_msgs[i] != NULL) {
  88                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
  89                                 (ftc_num_msgs[i] * sizeof(long)));
  90                         ftc_num_msgs[i] = 0;
  91                         free(ftc_msgs[i]);
  92                         ftc_msgs[i] = NULL;
  93                 }
  94         }
  95         syslog(LOG_INFO, "fulltext: flushed index cache to disk (100%% complete)");
  96 }
  97
  98
  99 /*
 100  * Index or de-index a message.  (op == 1 to index, 0 to de-index)
 101  */
 102 void ft_index_message(long msgnum, int op) {
 103         int num_tokens = 0;
 104         int *tokens = NULL;
 105         int i, j;
 106         struct cdbdata *cdb_bucket;
 107         StrBuf *msgtext;
 108         char *txt;
 109         int tok;
 110         struct CtdlMessage *msg = NULL;
 111
 112         msg = CtdlFetchMessage(msgnum, 1);
 113         if (msg == NULL) {
 114                 syslog(LOG_ERR, "fulltext: ft_index_message() could not load msg %ld", msgnum);
 115                 return;
 116         }
 117
 118         if (!CM_IsEmpty(msg, eSuppressIdx)) {
 119                 syslog(LOG_DEBUG, "fulltext: ft_index_message() excluded msg %ld", msgnum);
 120                 CM_Free(msg);
 121                 return;
 122         }
 123
 124         syslog(LOG_DEBUG, "fulltext: ft_index_message() %s msg %ld", (op ? "adding" : "removing") , msgnum);
 125
 126         /* Output the message as text before indexing it, so we don't end up
 127          * indexing a bunch of encoded base64, etc.
 128          */
 129         CC->redirect_buffer = NewStrBufPlain(NULL, SIZ);
 130         CtdlOutputPreLoadedMsg(msg, MT_CITADEL, HEADERS_ALL, 0, 1, 0);
 131         CM_Free(msg);
 132         msgtext = CC->redirect_buffer;
 133         CC->redirect_buffer = NULL;
 134         if (msgtext != NULL) {
 135                 syslog(LOG_DEBUG, "fulltext: wordbreaking message %ld (%d bytes)", msgnum, StrLength(msgtext));
 136         }
 137         txt = SmashStrBuf(&msgtext);
 138         wordbreaker(txt, &num_tokens, &tokens);
 139         free(txt);
 140
 141         syslog(LOG_DEBUG, "fulltext: indexing message %ld [%d tokens]", msgnum, num_tokens);
 142         if (num_tokens > 0) {
 143                 for (i=0; i<num_tokens; ++i) {
 144
 145                         /* Add the message to the relevant token bucket */
 146
 147                         /* search for tokens[i] */
 148                         tok = tokens[i];
 149
 150                         if ( (tok >= 0) && (tok <= 65535) ) {
 151                                 /* fetch the bucket, Liza */
 152                                 if (ftc_msgs[tok] == NULL) {
 153                                         cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 154                                         if (cdb_bucket != NULL) {
 155                                                 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 156                                                 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 157                                                 cdb_bucket->ptr = NULL;
 158                                                 cdb_free(cdb_bucket);
 159                                         }
 160                                         else {
 161                                                 ftc_num_msgs[tok] = 0;
 162                                                 ftc_msgs[tok] = malloc(sizeof(long));
 163                                         }
 164                                 }
 165
 166
 167                                 if (op == 1) {  /* add to index */
 168                                         ++ftc_num_msgs[tok];
 169                                         ftc_msgs[tok] = realloc(ftc_msgs[tok],
 170                                                                 ftc_num_msgs[tok]*sizeof(long));
 171                                         ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
 172                                 }
 173
 174                                 if (op == 0) {  /* remove from index */
 175                                         if (ftc_num_msgs[tok] >= 1) {
 176                                                 for (j=0; j<ftc_num_msgs[tok]; ++j) {
 177                                                         if (ftc_msgs[tok][j] == msgnum) {
 178                                                                 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
 179                                                                 --ftc_num_msgs[tok];
 180                                                                 --j;
 181                                                         }
 182                                                 }
 183                                         }
 184                                 }
 185                         }
 186                         else {
 187                                 syslog(LOG_ALERT, "fulltext: invalid token %d !!", tok);
 188                         }
 189                 }
 190
 191                 free(tokens);
 192         }
 193 }
 194
 195
 196 /*
 197  * Add a message to the list of those to be indexed.
 198  */
 199 void ft_index_msg(long msgnum, void *userdata) {
 200
 201         if ((msgnum > CtdlGetConfigLong("MMfulltext")) && (msgnum <= ft_newhighest)) {
 202                 ++ft_num_msgs;
 203                 if (ft_num_msgs > ft_num_alloc) {
 204                         ft_num_alloc += 1024;
 205                         ft_newmsgs = realloc(ft_newmsgs, (ft_num_alloc * sizeof(long)));
 206                 }
 207                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 208         }
 209
 210 }
 211
 212
 213 /*
 214  * Scan a room for messages to index.
 215  */
 216 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 217 {
 218         if (server_shutting_down)
 219                 return;
 220
 221         CtdlGetRoom(&CC->room, qrbuf->QRname);
 222         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 223 }
 224
 225
 226 /*
 227  * Begin the fulltext indexing process.
 228  */
 229 void do_fulltext_indexing(void) {
 230         int i;
 231         static time_t last_progress = 0L;
 232         static int is_running = 0;
 233         if (is_running) return;         /* Concurrency check - only one can run */
 234         is_running = 1;
 235
 236         /*
 237          * Don't do this if the site doesn't have it enabled.
 238          */
 239         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 240                 return;
 241         }
 242
 243         /*
 244          * If we've switched wordbreaker modules, burn the index and start over.
 245          */
 246         begin_critical_section(S_CONTROL);
 247         if (CtdlGetConfigInt("MM_fulltext_wordbreaker") != FT_WORDBREAKER_ID) {
 248                 syslog(LOG_DEBUG, "fulltext: wb ver on disk = %d, code ver = %d",
 249                         CtdlGetConfigInt("MM_fulltext_wordbreaker"), FT_WORDBREAKER_ID
 250                 );
 251                 syslog(LOG_INFO, "fulltext: (re)initializing index");
 252                 cdb_trunc(CDB_FULLTEXT);
 253                 CtdlSetConfigLong("MMfulltext", 0);
 254         }
 255         end_critical_section(S_CONTROL);
 256
 257         /*
 258          * Silently return if our fulltext index is up to date with new messages.
 259          */
 260         if ((CtdlGetConfigLong("MMfulltext") >= CtdlGetConfigLong("MMhighest"))) {
 261                 return;         /* nothing to do! */
 262         }
 263
 264         /*
 265          * Now go through each room and find messages to index.
 266          */
 267         ft_newhighest = CtdlGetConfigLong("MMhighest");
 268         CtdlForEachRoom(ft_index_room, NULL);   /* load all msg pointers */
 269
 270         if (ft_num_msgs > 0) {
 271                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 272                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 273                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 274                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 275                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 276                                 --ft_num_msgs;
 277                                 --i;
 278                         }
 279                 }
 280
 281                 /* Here it is ... do each message! */
 282                 for (i=0; i<ft_num_msgs; ++i) {
 283                         if (time(NULL) != last_progress) {
 284                                 syslog(LOG_DEBUG,
 285                                         "fulltext: indexed %d of %d messages (%d%%)",
 286                                                 i, ft_num_msgs,
 287                                                 ((i*100) / ft_num_msgs)
 288                                 );
 289                                 last_progress = time(NULL);
 290                         }
 291                         ft_index_message(ft_newmsgs[i], 1);
 292
 293                         /* Check to see if we need to quit early */
 294                         if (server_shutting_down) {
 295                                 syslog(LOG_DEBUG, "fulltext: indexer quitting early");
 296                                 ft_newhighest = ft_newmsgs[i];
 297                                 break;
 298                         }
 299
 300                         /* Check to see if we have to maybe flush to disk */
 301                         if (i >= FT_MAX_CACHE) {
 302                                 syslog(LOG_DEBUG, "fulltext: time to flush.");
 303                                 ft_newhighest = ft_newmsgs[i];
 304                                 break;
 305                         }
 306
 307                 }
 308
 309                 free(ft_newmsgs);
 310                 ft_num_msgs = 0;
 311                 ft_num_alloc = 0;
 312                 ft_newmsgs = NULL;
 313         }
 314
 315         if (server_shutting_down) {
 316                 is_running = 0;
 317                 return;
 318         }
 319
 320         /* Save our place so we don't have to do this again */
 321         ft_flush_cache();
 322         begin_critical_section(S_CONTROL);
 323         CtdlSetConfigLong("MMfulltext", ft_newhighest);
 324         CtdlSetConfigInt("MM_fulltext_wordbreaker", FT_WORDBREAKER_ID);
 325         end_critical_section(S_CONTROL);
 326
 327         syslog(LOG_DEBUG, "fulltext: indexing finished");
 328         is_running = 0;
 329         return;
 330 }
 331
 332
 333 /*
 334  * API call to perform searches.
 335  * (This one does the "all of these words" search.)
 336  * Caller is responsible for freeing the message list.
 337  */
 338 void ft_search(int *fts_num_msgs, long **fts_msgs, const char *search_string) {
 339         int num_tokens = 0;
 340         int *tokens = NULL;
 341         int i, j;
 342         struct cdbdata *cdb_bucket;
 343         int num_all_msgs = 0;
 344         long *all_msgs = NULL;
 345         int num_ret_msgs = 0;
 346         int num_ret_alloc = 0;
 347         long *ret_msgs = NULL;
 348         int tok;
 349
 350         wordbreaker(search_string, &num_tokens, &tokens);
 351         if (num_tokens > 0) {
 352                 for (i=0; i<num_tokens; ++i) {
 353
 354                         /* search for tokens[i] */
 355                         tok = tokens[i];
 356
 357                         /* fetch the bucket, Liza */
 358                         if (ftc_msgs[tok] == NULL) {
 359                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 360                                 if (cdb_bucket != NULL) {
 361                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 362                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 363                                         cdb_bucket->ptr = NULL;
 364                                         cdb_free(cdb_bucket);
 365                                 }
 366                                 else {
 367                                         ftc_num_msgs[tok] = 0;
 368                                         ftc_msgs[tok] = malloc(sizeof(long));
 369                                 }
 370                         }
 371
 372                         num_all_msgs += ftc_num_msgs[tok];
 373                         if (num_all_msgs > 0) {
 374                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 375                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 376                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 377                         }
 378
 379                 }
 380                 free(tokens);
 381                 if (all_msgs != NULL) {
 382                         qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 383
 384                         /*
 385                          * At this point, if a message appears num_tokens times in the
 386                          * list, then it contains all of the search tokens.
 387                          */
 388                         if (num_all_msgs >= num_tokens)
 389                                 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 390                                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 391
 392                                                 ++num_ret_msgs;
 393                                                 if (num_ret_msgs > num_ret_alloc) {
 394                                                         num_ret_alloc += 64;
 395                                                         ret_msgs = realloc(ret_msgs,
 396                                                                            (num_ret_alloc*sizeof(long)) );
 397                                                 }
 398                                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 399
 400                                         }
 401                                 }
 402                         free(all_msgs);
 403                 }
 404         }
 405
 406         *fts_num_msgs = num_ret_msgs;
 407         *fts_msgs = ret_msgs;
 408 }
 409
 410
 411 /*
 412  * This search command is for diagnostic purposes and may be removed or replaced.
 413  */
 414 void cmd_srch(char *argbuf) {
 415         int num_msgs = 0;
 416         long *msgs = NULL;
 417         int i;
 418         char search_string[256];
 419
 420         if (CtdlAccessCheck(ac_logged_in)) return;
 421
 422         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 423                 cprintf("%d Full text index is not enabled on this server.\n",
 424                         ERROR + CMD_NOT_SUPPORTED);
 425                 return;
 426         }
 427
 428         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 429         ft_search(&num_msgs, &msgs, search_string);
 430
 431         cprintf("%d %d msgs match all search words:\n",
 432                 LISTING_FOLLOWS, num_msgs);
 433         if (num_msgs > 0) {
 434                 for (i=0; i<num_msgs; ++i) {
 435                         cprintf("%ld\n", msgs[i]);
 436                 }
 437         }
 438         if (msgs != NULL) free(msgs);
 439         cprintf("000\n");
 440 }
 441
 442
 443 /*
 444  * Zero out our index cache.
 445  */
 446 void initialize_ft_cache(void) {
 447         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 448         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 449 }
 450
 451
 452 void ft_delete_remove(char *room, long msgnum)
 453 {
 454         if (room) return;
 455
 456         /* Remove from fulltext index */
 457         if (CtdlGetConfigInt("c_enable_fulltext")) {
 458                 ft_index_message(msgnum, 0);
 459         }
 460 }
 461
 462
 463 /*****************************************************************************/
 464
 465 char *ctdl_module_init_fulltext(void) {
 466         if (!threading) {
 467                 initialize_ft_cache();
 468                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 469                 CtdlRegisterDeleteHook(ft_delete_remove);
 470                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 471                 CtdlRegisterSessionHook(do_fulltext_indexing, EVT_TIMER, PRIO_CLEANUP + 300);
 472         }
 473         /* return our module name for the log */
 474         return "fulltext";
 475 }