citadel/modules/fulltext/serv_fulltext.c

   1 /*
   2  * $Id$
   3  *
   4  * This module handles fulltext indexing of the message base.
   5  *
   6  *
   7  * Copyright (c) 2005-2009 by the citadel.org team
   8  *
   9  *  This program is free software; you can redistribute it and/or modify
  10  *  it under the terms of the GNU General Public License as published by
  11  *  the Free Software Foundation; either version 3 of the License, or
  12  *  (at your option) any later version.
  13  *
  14  *  This program is distributed in the hope that it will be useful,
  15  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  *  GNU General Public License for more details.
  18  *
  19  *  You should have received a copy of the GNU General Public License
  20  *  along with this program; if not, write to the Free Software
  21  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  22  */
  23
  24
  25 #include "sysdep.h"
  26 #include <stdlib.h>
  27 #include <unistd.h>
  28 #include <stdio.h>
  29 #include <fcntl.h>
  30 #include <signal.h>
  31 #include <pwd.h>
  32 #include <errno.h>
  33 #include <sys/types.h>
  34
  35 #if TIME_WITH_SYS_TIME
  36 # include <sys/time.h>
  37 # include <time.h>
  38 #else
  39 # if HAVE_SYS_TIME_H
  40 #  include <sys/time.h>
  41 # else
  42 #  include <time.h>
  43 # endif
  44 #endif
  45
  46 #include <sys/wait.h>
  47 #include <string.h>
  48 #include <limits.h>
  49 #include <libcitadel.h>
  50 #include "citadel.h"
  51 #include "server.h"
  52 #include "citserver.h"
  53 #include "support.h"
  54 #include "config.h"
  55 #include "database.h"
  56 #include "msgbase.h"
  57 #include "control.h"
  58 #include "serv_fulltext.h"
  59 #include "ft_wordbreaker.h"
  60 #include "threads.h"
  61
  62 #include "ctdl_module.h"
  63
  64
  65
  66 long ft_newhighest = 0L;
  67 long *ft_newmsgs = NULL;
  68 int ft_num_msgs = 0;
  69 int ft_num_alloc = 0;
  70
  71
  72 int ftc_num_msgs[65536];
  73 long *ftc_msgs[65536];
  74
  75
  76 /*
  77  * Compare function
  78  */
  79 int longcmp(const void *rec1, const void *rec2) {
  80         long i1, i2;
  81
  82         i1 = *(const long *)rec1;
  83         i2 = *(const long *)rec2;
  84
  85         if (i1 > i2) return(1);
  86         if (i1 < i2) return(-1);
  87         return(0);
  88 }
  89
  90 /*
  91  * Flush our index cache out to disk.
  92  */
  93 void ft_flush_cache(void) {
  94         int i;
  95         time_t last_update = 0;
  96
  97         for (i=0; i<65536; ++i) {
  98                 if ((time(NULL) - last_update) >= 10) {
  99                         CtdlLogPrintf(CTDL_INFO,
 100                                 "Flushing index cache to disk (%d%% complete)\n",
 101                                 (i * 100 / 65536)
 102                         );
 103                         last_update = time(NULL);
 104                 }
 105                 if (ftc_msgs[i] != NULL) {
 106                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
 107                                 (ftc_num_msgs[i] * sizeof(long)));
 108                         ftc_num_msgs[i] = 0;
 109                         free(ftc_msgs[i]);
 110                         ftc_msgs[i] = NULL;
 111                 }
 112         }
 113         CtdlLogPrintf(CTDL_INFO, "Flushed index cache to disk (100%% complete)\n");
 114 }
 115
 116
 117 /*
 118  * Index or de-index a message.  (op == 1 to index, 0 to de-index)
 119  */
 120 void ft_index_message(long msgnum, int op) {
 121         int num_tokens = 0;
 122         int *tokens = NULL;
 123         int i, j;
 124         struct cdbdata *cdb_bucket;
 125         char *msgtext;
 126         int tok;
 127
 128         CtdlLogPrintf(CTDL_DEBUG, "ft_index_message() %s msg %ld\n",
 129                 (op ? "adding" : "removing") , msgnum
 130         );
 131
 132         /* Output the message as text before indexing it, so we don't end up
 133          * indexing a bunch of encoded base64, etc.
 134          */
 135         CC->redirect_buffer = malloc(SIZ);
 136         CC->redirect_len = 0;
 137         CC->redirect_alloc = SIZ;
 138         CtdlOutputMsg(msgnum, MT_CITADEL, HEADERS_ALL, 0, 1, NULL, 0);
 139         msgtext = CC->redirect_buffer;
 140         CC->redirect_buffer = NULL;
 141         CC->redirect_len = 0;
 142         CC->redirect_alloc = 0;
 143         CtdlLogPrintf(CTDL_DEBUG, "Wordbreaking message %ld...\n", msgnum);
 144         wordbreaker(msgtext, &num_tokens, &tokens);
 145         free(msgtext);
 146
 147         CtdlLogPrintf(CTDL_DEBUG, "Indexing message %ld [%d tokens]\n", msgnum, num_tokens);
 148         if (num_tokens > 0) {
 149                 for (i=0; i<num_tokens; ++i) {
 150
 151                         /* Add the message to the relevant token bucket */
 152
 153                         /* search for tokens[i] */
 154                         tok = tokens[i];
 155
 156                         if ( (tok >= 0) && (tok <= 65535) ) {
 157                                 /* fetch the bucket, Liza */
 158                                 if (ftc_msgs[tok] == NULL) {
 159                                         cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 160                                         if (cdb_bucket != NULL) {
 161                                                 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 162                                                 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 163                                                 cdb_bucket->ptr = NULL;
 164                                                 cdb_free(cdb_bucket);
 165                                         }
 166                                         else {
 167                                                 ftc_num_msgs[tok] = 0;
 168                                                 ftc_msgs[tok] = malloc(sizeof(long));
 169                                         }
 170                                 }
 171
 172
 173                                 if (op == 1) {  /* add to index */
 174                                         ++ftc_num_msgs[tok];
 175                                         ftc_msgs[tok] = realloc(ftc_msgs[tok],
 176                                                                 ftc_num_msgs[tok]*sizeof(long));
 177                                         ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
 178                                 }
 179
 180                                 if (op == 0) {  /* remove from index */
 181                                         if (ftc_num_msgs[tok] >= 1) {
 182                                                 for (j=0; j<ftc_num_msgs[tok]; ++j) {
 183                                                         if (ftc_msgs[tok][j] == msgnum) {
 184                                                                 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
 185                                                                 --ftc_num_msgs[tok];
 186                                                                 --j;
 187                                                         }
 188                                                 }
 189                                         }
 190                                 }
 191                         }
 192                         else {
 193                                 CtdlLogPrintf(CTDL_ALERT, "Invalid token %d !!\n", tok);
 194                         }
 195                 }
 196
 197                 free(tokens);
 198         }
 199 }
 200
 201
 202
 203 /*
 204  * Add a message to the list of those to be indexed.
 205  */
 206 void ft_index_msg(long msgnum, void *userdata) {
 207
 208         if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
 209                 ++ft_num_msgs;
 210                 if (ft_num_msgs > ft_num_alloc) {
 211                         ft_num_alloc += 1024;
 212                         ft_newmsgs = realloc(ft_newmsgs,
 213                                 (ft_num_alloc * sizeof(long)));
 214                 }
 215                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 216         }
 217
 218 }
 219
 220 /*
 221  * Scan a room for messages to index.
 222  */
 223 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 224 {
 225         if (CtdlThreadCheckStop())
 226                 return;
 227
 228         CtdlGetRoom(&CC->room, qrbuf->QRname);
 229         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 230 }
 231
 232
 233 /*
 234  * Begin the fulltext indexing process.  (Called as an EVT_TIMER event)
 235  */
 236 void do_fulltext_indexing(void) {
 237         int i;
 238         static time_t last_index = 0L;
 239         static time_t last_progress = 0L;
 240         time_t run_time = 0L;
 241         time_t end_time = 0L;
 242
 243         /*
 244          * Don't do this if the site doesn't have it enabled.
 245          */
 246         if (!config.c_enable_fulltext) {
 247                 return;
 248         }
 249
 250         /*
 251          * Make sure we don't run the indexer too frequently.
 252          * FIXME move the setting into config
 253          */
 254 /*
 255  * The thread sleeps for 300 seconds so we don't need this here any more
 256
 257         if ( (time(NULL) - last_index) < 300L) {
 258                 return;
 259         }
 260 */
 261
 262         /*
 263          * Check to see whether the fulltext index is up to date; if there
 264          * are no messages to index, don't waste any more time trying.
 265          */
 266         if ((CitControl.MMfulltext >= CitControl.MMhighest) && (CitControl.fulltext_wordbreaker == FT_WORDBREAKER_ID)) {
 267                 return;         /* nothing to do! */
 268         }
 269
 270         run_time = time(NULL);
 271         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() started (%ld)\n", run_time);
 272
 273         /*
 274          * If we've switched wordbreaker modules, burn the index and start
 275          * over.
 276          */
 277         begin_critical_section(S_CONTROL);
 278         if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
 279                 CtdlLogPrintf(CTDL_DEBUG, "wb ver on disk = %d, code ver = %d\n",
 280                         CitControl.fulltext_wordbreaker, FT_WORDBREAKER_ID);
 281                 CtdlLogPrintf(CTDL_INFO, "(re)initializing full text index\n");
 282                 cdb_trunc(CDB_FULLTEXT);
 283                 CitControl.MMfulltext = 0L;
 284                 put_control();
 285         }
 286         end_critical_section(S_CONTROL);
 287
 288         /*
 289          * Now go through each room and find messages to index.
 290          */
 291         ft_newhighest = CitControl.MMhighest;
 292         CtdlForEachRoom(ft_index_room, NULL);   /* load all msg pointers */
 293
 294         if (ft_num_msgs > 0) {
 295                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 296                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 297                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 298                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 299                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 300                                 --ft_num_msgs;
 301                                 --i;
 302                         }
 303                 }
 304
 305                 /* Here it is ... do each message! */
 306                 for (i=0; i<ft_num_msgs; ++i) {
 307                         if (time(NULL) != last_progress) {
 308                                 CtdlLogPrintf(CTDL_DEBUG,
 309                                         "Indexed %d of %d messages (%d%%)\n",
 310                                                 i, ft_num_msgs,
 311                                                 ((i*100) / ft_num_msgs)
 312                                 );
 313                                 last_progress = time(NULL);
 314                         }
 315                         ft_index_message(ft_newmsgs[i], 1);
 316
 317                         /* Check to see if we need to quit early */
 318                         if (CtdlThreadCheckStop()) {
 319                                 CtdlLogPrintf(CTDL_DEBUG, "Indexer quitting early\n");
 320                                 ft_newhighest = ft_newmsgs[i];
 321                                 break;
 322                         }
 323
 324                         /* Check to see if we have to maybe flush to disk */
 325                         if (i >= FT_MAX_CACHE) {
 326                                 CtdlLogPrintf(CTDL_DEBUG, "Time to flush.\n");
 327                                 ft_newhighest = ft_newmsgs[i];
 328                                 break;
 329                         }
 330
 331                 }
 332
 333                 free(ft_newmsgs);
 334                 ft_num_msgs = 0;
 335                 ft_num_alloc = 0;
 336                 ft_newmsgs = NULL;
 337         }
 338         end_time = time(NULL);
 339
 340         if (CtdlThreadCheckStop())
 341                 return;
 342
 343         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() duration (%ld)\n", end_time - run_time);
 344
 345         /* Save our place so we don't have to do this again */
 346         ft_flush_cache();
 347         begin_critical_section(S_CONTROL);
 348         CitControl.MMfulltext = ft_newhighest;
 349         CitControl.fulltext_wordbreaker = FT_WORDBREAKER_ID;
 350         put_control();
 351         end_critical_section(S_CONTROL);
 352         last_index = time(NULL);
 353
 354         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() finished\n");
 355         return;
 356 }
 357
 358 /*
 359  * Main loop for the indexer thread.
 360  */
 361 void *indexer_thread(void *arg) {
 362         struct CitContext indexerCC;
 363
 364         CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() initializing\n");
 365
 366         CtdlFillSystemContext(&indexerCC, "indexer");
 367         citthread_setspecific(MyConKey, (void *)&indexerCC );
 368
 369         while (!CtdlThreadCheckStop()) {
 370                 do_fulltext_indexing();
 371                 CtdlThreadSleep(300);
 372         }
 373
 374         CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() exiting\n");
 375         return NULL;
 376 }
 377
 378
 379
 380 /*
 381  * API call to perform searches.
 382  * (This one does the "all of these words" search.)
 383  * Caller is responsible for freeing the message list.
 384  */
 385 void ft_search(int *fts_num_msgs, long **fts_msgs, char *search_string) {
 386         int num_tokens = 0;
 387         int *tokens = NULL;
 388         int i, j;
 389         struct cdbdata *cdb_bucket;
 390         int num_all_msgs = 0;
 391         long *all_msgs = NULL;
 392         int num_ret_msgs = 0;
 393         int num_ret_alloc = 0;
 394         long *ret_msgs = NULL;
 395         int tok;
 396
 397         wordbreaker(search_string, &num_tokens, &tokens);
 398         if (num_tokens > 0) {
 399                 for (i=0; i<num_tokens; ++i) {
 400
 401                         /* search for tokens[i] */
 402                         tok = tokens[i];
 403
 404                         /* fetch the bucket, Liza */
 405                         if (ftc_msgs[tok] == NULL) {
 406                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 407                                 if (cdb_bucket != NULL) {
 408                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 409                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 410                                         cdb_bucket->ptr = NULL;
 411                                         cdb_free(cdb_bucket);
 412                                 }
 413                                 else {
 414                                         ftc_num_msgs[tok] = 0;
 415                                         ftc_msgs[tok] = malloc(sizeof(long));
 416                                 }
 417                         }
 418
 419                         num_all_msgs += ftc_num_msgs[tok];
 420                         if (num_all_msgs > 0) {
 421                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 422                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 423                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 424                         }
 425
 426                 }
 427                 free(tokens);
 428                 qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 429
 430                 /*
 431                  * At this point, if a message appears num_tokens times in the
 432                  * list, then it contains all of the search tokens.
 433                  */
 434                 if (num_all_msgs >= num_tokens)
 435                    for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 436                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 437
 438                                 ++num_ret_msgs;
 439                                 if (num_ret_msgs > num_ret_alloc) {
 440                                         num_ret_alloc += 64;
 441                                         ret_msgs = realloc(ret_msgs,
 442                                                 (num_ret_alloc*sizeof(long)) );
 443                                 }
 444                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 445
 446                         }
 447                 }
 448
 449                 free(all_msgs);
 450         }
 451
 452         *fts_num_msgs = num_ret_msgs;
 453         *fts_msgs = ret_msgs;
 454 }
 455
 456
 457 /*
 458  * This search command is for diagnostic purposes and may be removed or replaced.
 459  */
 460 void cmd_srch(char *argbuf) {
 461         int num_msgs = 0;
 462         long *msgs = NULL;
 463         int i;
 464         char search_string[256];
 465
 466         if (CtdlAccessCheck(ac_logged_in)) return;
 467
 468         if (!config.c_enable_fulltext) {
 469                 cprintf("%d Full text index is not enabled on this server.\n",
 470                         ERROR + CMD_NOT_SUPPORTED);
 471                 return;
 472         }
 473
 474         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 475         ft_search(&num_msgs, &msgs, search_string);
 476
 477         cprintf("%d %d msgs match all search words:\n",
 478                 LISTING_FOLLOWS, num_msgs);
 479         if (num_msgs > 0) {
 480                 for (i=0; i<num_msgs; ++i) {
 481                         cprintf("%ld\n", msgs[i]);
 482                 }
 483         }
 484         if (msgs != NULL) free(msgs);
 485         cprintf("000\n");
 486 }
 487
 488 /*
 489  * Zero out our index cache.
 490  */
 491 void initialize_ft_cache(void) {
 492         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 493         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 494 }
 495
 496
 497 void ft_delete_remove(char *room, long msgnum)
 498 {
 499         if (room) return;
 500
 501         /* Remove from fulltext index */
 502         if (config.c_enable_fulltext) {
 503                 ft_index_message(msgnum, 0);
 504         }
 505 }
 506
 507 /*****************************************************************************/
 508
 509 CTDL_MODULE_INIT(fulltext)
 510 {
 511         if (!threading)
 512         {
 513                 initialize_ft_cache();
 514                 initialize_noise_words();
 515                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 516                 CtdlRegisterDeleteHook(ft_delete_remove);
 517                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 518                 CtdlRegisterCleanupHook(noise_word_cleanup);
 519         }
 520         else
 521         {
 522                 CtdlThreadCreate("Indexer", CTDLTHREAD_BIGSTACK, indexer_thread, NULL);
 523         }
 524         /* return our Subversion id for the Log */
 525         return "$Id$";
 526 }