citadel/modules/fulltext/serv_fulltext.c

   1 /*
   2  * This module handles fulltext indexing of the message base.
   3  * Copyright (c) 2005-2009 by the citadel.org team
   4  *
   5  *  This program is free software; you can redistribute it and/or modify
   6  *  it under the terms of the GNU General Public License as published by
   7  *  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
  21 #include "sysdep.h"
  22 #include <stdlib.h>
  23 #include <unistd.h>
  24 #include <stdio.h>
  25 #include <fcntl.h>
  26 #include <signal.h>
  27 #include <pwd.h>
  28 #include <errno.h>
  29 #include <sys/types.h>
  30
  31 #if TIME_WITH_SYS_TIME
  32 # include <sys/time.h>
  33 # include <time.h>
  34 #else
  35 # if HAVE_SYS_TIME_H
  36 #  include <sys/time.h>
  37 # else
  38 #  include <time.h>
  39 # endif
  40 #endif
  41
  42 #include <sys/wait.h>
  43 #include <string.h>
  44 #include <limits.h>
  45 #include <libcitadel.h>
  46 #include "citadel.h"
  47 #include "server.h"
  48 #include "citserver.h"
  49 #include "support.h"
  50 #include "config.h"
  51 #include "database.h"
  52 #include "msgbase.h"
  53 #include "control.h"
  54 #include "serv_fulltext.h"
  55 #include "ft_wordbreaker.h"
  56 #include "threads.h"
  57 #include "context.h"
  58
  59 #include "ctdl_module.h"
  60
  61
  62
  63 long ft_newhighest = 0L;
  64 long *ft_newmsgs = NULL;
  65 int ft_num_msgs = 0;
  66 int ft_num_alloc = 0;
  67
  68
  69 int ftc_num_msgs[65536];
  70 long *ftc_msgs[65536];
  71
  72
  73 /*
  74  * Compare function
  75  */
  76 int longcmp(const void *rec1, const void *rec2) {
  77         long i1, i2;
  78
  79         i1 = *(const long *)rec1;
  80         i2 = *(const long *)rec2;
  81
  82         if (i1 > i2) return(1);
  83         if (i1 < i2) return(-1);
  84         return(0);
  85 }
  86
  87 /*
  88  * Flush our index cache out to disk.
  89  */
  90 void ft_flush_cache(void) {
  91         int i;
  92         time_t last_update = 0;
  93
  94         for (i=0; i<65536; ++i) {
  95                 if ((time(NULL) - last_update) >= 10) {
  96                         CtdlLogPrintf(CTDL_INFO,
  97                                 "Flushing index cache to disk (%d%% complete)\n",
  98                                 (i * 100 / 65536)
  99                         );
 100                         last_update = time(NULL);
 101                 }
 102                 if (ftc_msgs[i] != NULL) {
 103                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
 104                                 (ftc_num_msgs[i] * sizeof(long)));
 105                         ftc_num_msgs[i] = 0;
 106                         free(ftc_msgs[i]);
 107                         ftc_msgs[i] = NULL;
 108                 }
 109         }
 110         CtdlLogPrintf(CTDL_INFO, "Flushed index cache to disk (100%% complete)\n");
 111 }
 112
 113
 114 /*
 115  * Index or de-index a message.  (op == 1 to index, 0 to de-index)
 116  */
 117 void ft_index_message(long msgnum, int op) {
 118         int num_tokens = 0;
 119         int *tokens = NULL;
 120         int i, j;
 121         struct cdbdata *cdb_bucket;
 122         StrBuf *msgtext;
 123         char *txt;
 124         int tok;
 125         struct CtdlMessage *msg = NULL;
 126
 127         msg = CtdlFetchMessage(msgnum, 1);
 128         if (msg == NULL) {
 129                 CtdlLogPrintf(CTDL_ERR, "ft_index_message() could not load msg %ld\n", msgnum);
 130                 return;
 131         }
 132
 133         if (msg->cm_fields['1'] != NULL) {
 134                 CtdlLogPrintf(CTDL_DEBUG, "ft_index_message() excluded msg %ld\n", msgnum);
 135                 CtdlFreeMessage(msg);
 136                 return;
 137         }
 138
 139         CtdlLogPrintf(CTDL_DEBUG, "ft_index_message() %s msg %ld\n",
 140                 (op ? "adding" : "removing") , msgnum
 141         );
 142
 143         /* Output the message as text before indexing it, so we don't end up
 144          * indexing a bunch of encoded base64, etc.
 145          */
 146         CC->redirect_buffer = NewStrBufPlain(NULL, SIZ);
 147         CtdlOutputPreLoadedMsg(msg, MT_CITADEL, HEADERS_ALL, 0, 1, 0);
 148         CtdlFreeMessage(msg);
 149         msgtext = CC->redirect_buffer;
 150         CC->redirect_buffer = NULL;
 151         CtdlLogPrintf(CTDL_DEBUG, "Wordbreaking message %ld...\n", msgnum);
 152         txt = SmashStrBuf(&msgtext);
 153         wordbreaker(txt, &num_tokens, &tokens);
 154         free(txt);
 155
 156         CtdlLogPrintf(CTDL_DEBUG, "Indexing message %ld [%d tokens]\n", msgnum, num_tokens);
 157         if (num_tokens > 0) {
 158                 for (i=0; i<num_tokens; ++i) {
 159
 160                         /* Add the message to the relevant token bucket */
 161
 162                         /* search for tokens[i] */
 163                         tok = tokens[i];
 164
 165                         if ( (tok >= 0) && (tok <= 65535) ) {
 166                                 /* fetch the bucket, Liza */
 167                                 if (ftc_msgs[tok] == NULL) {
 168                                         cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 169                                         if (cdb_bucket != NULL) {
 170                                                 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 171                                                 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 172                                                 cdb_bucket->ptr = NULL;
 173                                                 cdb_free(cdb_bucket);
 174                                         }
 175                                         else {
 176                                                 ftc_num_msgs[tok] = 0;
 177                                                 ftc_msgs[tok] = malloc(sizeof(long));
 178                                         }
 179                                 }
 180
 181
 182                                 if (op == 1) {  /* add to index */
 183                                         ++ftc_num_msgs[tok];
 184                                         ftc_msgs[tok] = realloc(ftc_msgs[tok],
 185                                                                 ftc_num_msgs[tok]*sizeof(long));
 186                                         ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
 187                                 }
 188
 189                                 if (op == 0) {  /* remove from index */
 190                                         if (ftc_num_msgs[tok] >= 1) {
 191                                                 for (j=0; j<ftc_num_msgs[tok]; ++j) {
 192                                                         if (ftc_msgs[tok][j] == msgnum) {
 193                                                                 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
 194                                                                 --ftc_num_msgs[tok];
 195                                                                 --j;
 196                                                         }
 197                                                 }
 198                                         }
 199                                 }
 200                         }
 201                         else {
 202                                 CtdlLogPrintf(CTDL_ALERT, "Invalid token %d !!\n", tok);
 203                         }
 204                 }
 205
 206                 free(tokens);
 207         }
 208 }
 209
 210
 211
 212 /*
 213  * Add a message to the list of those to be indexed.
 214  */
 215 void ft_index_msg(long msgnum, void *userdata) {
 216
 217         if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
 218                 ++ft_num_msgs;
 219                 if (ft_num_msgs > ft_num_alloc) {
 220                         ft_num_alloc += 1024;
 221                         ft_newmsgs = realloc(ft_newmsgs,
 222                                 (ft_num_alloc * sizeof(long)));
 223                 }
 224                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 225         }
 226
 227 }
 228
 229 /*
 230  * Scan a room for messages to index.
 231  */
 232 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 233 {
 234         if (CtdlThreadCheckStop())
 235                 return;
 236
 237         CtdlGetRoom(&CC->room, qrbuf->QRname);
 238         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 239 }
 240
 241
 242 /*
 243  * Begin the fulltext indexing process.
 244  */
 245 void do_fulltext_indexing(void) {
 246         int i;
 247         static time_t last_index = 0L;
 248         static time_t last_progress = 0L;
 249         time_t run_time = 0L;
 250         time_t end_time = 0L;
 251
 252         /*
 253          * Don't do this if the site doesn't have it enabled.
 254          */
 255         if (!config.c_enable_fulltext) {
 256                 return;
 257         }
 258
 259         /*
 260          * Make sure we don't run the indexer too frequently.
 261          * FIXME move the setting into config
 262          */
 263 /*
 264  * The thread sleeps for 300 seconds so we don't need this here any more
 265
 266         if ( (time(NULL) - last_index) < 300L) {
 267                 return;
 268         }
 269 */
 270
 271         /*
 272          * Check to see whether the fulltext index is up to date; if there
 273          * are no messages to index, don't waste any more time trying.
 274          */
 275         if ((CitControl.MMfulltext >= CitControl.MMhighest) && (CitControl.fulltext_wordbreaker == FT_WORDBREAKER_ID)) {
 276                 return;         /* nothing to do! */
 277         }
 278
 279         run_time = time(NULL);
 280         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() started (%ld)\n", run_time);
 281
 282         /*
 283          * If we've switched wordbreaker modules, burn the index and start
 284          * over.
 285          */
 286         begin_critical_section(S_CONTROL);
 287         if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
 288                 CtdlLogPrintf(CTDL_DEBUG, "wb ver on disk = %d, code ver = %d\n",
 289                         CitControl.fulltext_wordbreaker, FT_WORDBREAKER_ID);
 290                 CtdlLogPrintf(CTDL_INFO, "(re)initializing full text index\n");
 291                 cdb_trunc(CDB_FULLTEXT);
 292                 CitControl.MMfulltext = 0L;
 293                 put_control();
 294         }
 295         end_critical_section(S_CONTROL);
 296
 297         /*
 298          * Now go through each room and find messages to index.
 299          */
 300         ft_newhighest = CitControl.MMhighest;
 301         CtdlForEachRoom(ft_index_room, NULL);   /* load all msg pointers */
 302
 303         if (ft_num_msgs > 0) {
 304                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 305                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 306                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 307                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 308                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 309                                 --ft_num_msgs;
 310                                 --i;
 311                         }
 312                 }
 313
 314                 /* Here it is ... do each message! */
 315                 for (i=0; i<ft_num_msgs; ++i) {
 316                         if (time(NULL) != last_progress) {
 317                                 CtdlLogPrintf(CTDL_DEBUG,
 318                                         "Indexed %d of %d messages (%d%%)\n",
 319                                                 i, ft_num_msgs,
 320                                                 ((i*100) / ft_num_msgs)
 321                                 );
 322                                 last_progress = time(NULL);
 323                         }
 324                         ft_index_message(ft_newmsgs[i], 1);
 325
 326                         /* Check to see if we need to quit early */
 327                         if (CtdlThreadCheckStop()) {
 328                                 CtdlLogPrintf(CTDL_DEBUG, "Indexer quitting early\n");
 329                                 ft_newhighest = ft_newmsgs[i];
 330                                 break;
 331                         }
 332
 333                         /* Check to see if we have to maybe flush to disk */
 334                         if (i >= FT_MAX_CACHE) {
 335                                 CtdlLogPrintf(CTDL_DEBUG, "Time to flush.\n");
 336                                 ft_newhighest = ft_newmsgs[i];
 337                                 break;
 338                         }
 339
 340                 }
 341
 342                 free(ft_newmsgs);
 343                 ft_num_msgs = 0;
 344                 ft_num_alloc = 0;
 345                 ft_newmsgs = NULL;
 346         }
 347         end_time = time(NULL);
 348
 349         if (CtdlThreadCheckStop())
 350                 return;
 351
 352         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() duration (%ld)\n", end_time - run_time);
 353
 354         /* Save our place so we don't have to do this again */
 355         ft_flush_cache();
 356         begin_critical_section(S_CONTROL);
 357         CitControl.MMfulltext = ft_newhighest;
 358         CitControl.fulltext_wordbreaker = FT_WORDBREAKER_ID;
 359         put_control();
 360         end_critical_section(S_CONTROL);
 361         last_index = time(NULL);
 362
 363         CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() finished\n");
 364         return;
 365 }
 366
 367 /*
 368  * Main loop for the indexer thread.
 369  */
 370 void *indexer_thread(void *arg) {
 371         struct CitContext indexerCC;
 372
 373
 374         CtdlFillSystemContext(&indexerCC, "indexer");
 375         citthread_setspecific(MyConKey, (void *)&indexerCC );
 376         CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() initializing\n");
 377
 378         while (!CtdlThreadCheckStop()) {
 379                 do_fulltext_indexing();
 380                 CtdlThreadSleep(300);
 381         }
 382
 383         CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() exiting\n");
 384         CtdlClearSystemContext();
 385         return NULL;
 386 }
 387
 388
 389
 390 /*
 391  * API call to perform searches.
 392  * (This one does the "all of these words" search.)
 393  * Caller is responsible for freeing the message list.
 394  */
 395 void ft_search(int *fts_num_msgs, long **fts_msgs, const char *search_string) {
 396         int num_tokens = 0;
 397         int *tokens = NULL;
 398         int i, j;
 399         struct cdbdata *cdb_bucket;
 400         int num_all_msgs = 0;
 401         long *all_msgs = NULL;
 402         int num_ret_msgs = 0;
 403         int num_ret_alloc = 0;
 404         long *ret_msgs = NULL;
 405         int tok;
 406
 407         wordbreaker(search_string, &num_tokens, &tokens);
 408         if (num_tokens > 0) {
 409                 for (i=0; i<num_tokens; ++i) {
 410
 411                         /* search for tokens[i] */
 412                         tok = tokens[i];
 413
 414                         /* fetch the bucket, Liza */
 415                         if (ftc_msgs[tok] == NULL) {
 416                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 417                                 if (cdb_bucket != NULL) {
 418                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 419                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 420                                         cdb_bucket->ptr = NULL;
 421                                         cdb_free(cdb_bucket);
 422                                 }
 423                                 else {
 424                                         ftc_num_msgs[tok] = 0;
 425                                         ftc_msgs[tok] = malloc(sizeof(long));
 426                                 }
 427                         }
 428
 429                         num_all_msgs += ftc_num_msgs[tok];
 430                         if (num_all_msgs > 0) {
 431                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 432                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 433                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 434                         }
 435
 436                 }
 437                 free(tokens);
 438                 qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 439
 440                 /*
 441                  * At this point, if a message appears num_tokens times in the
 442                  * list, then it contains all of the search tokens.
 443                  */
 444                 if (num_all_msgs >= num_tokens)
 445                    for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 446                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 447
 448                                 ++num_ret_msgs;
 449                                 if (num_ret_msgs > num_ret_alloc) {
 450                                         num_ret_alloc += 64;
 451                                         ret_msgs = realloc(ret_msgs,
 452                                                 (num_ret_alloc*sizeof(long)) );
 453                                 }
 454                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 455
 456                         }
 457                 }
 458
 459                 free(all_msgs);
 460         }
 461
 462         *fts_num_msgs = num_ret_msgs;
 463         *fts_msgs = ret_msgs;
 464 }
 465
 466
 467 /*
 468  * This search command is for diagnostic purposes and may be removed or replaced.
 469  */
 470 void cmd_srch(char *argbuf) {
 471         int num_msgs = 0;
 472         long *msgs = NULL;
 473         int i;
 474         char search_string[256];
 475
 476         if (CtdlAccessCheck(ac_logged_in)) return;
 477
 478         if (!config.c_enable_fulltext) {
 479                 cprintf("%d Full text index is not enabled on this server.\n",
 480                         ERROR + CMD_NOT_SUPPORTED);
 481                 return;
 482         }
 483
 484         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 485         ft_search(&num_msgs, &msgs, search_string);
 486
 487         cprintf("%d %d msgs match all search words:\n",
 488                 LISTING_FOLLOWS, num_msgs);
 489         if (num_msgs > 0) {
 490                 for (i=0; i<num_msgs; ++i) {
 491                         cprintf("%ld\n", msgs[i]);
 492                 }
 493         }
 494         if (msgs != NULL) free(msgs);
 495         cprintf("000\n");
 496 }
 497
 498 /*
 499  * Zero out our index cache.
 500  */
 501 void initialize_ft_cache(void) {
 502         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 503         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 504 }
 505
 506
 507 void ft_delete_remove(char *room, long msgnum)
 508 {
 509         if (room) return;
 510
 511         /* Remove from fulltext index */
 512         if (config.c_enable_fulltext) {
 513                 ft_index_message(msgnum, 0);
 514         }
 515 }
 516
 517 /*****************************************************************************/
 518
 519 CTDL_MODULE_INIT(fulltext)
 520 {
 521         if (!threading)
 522         {
 523                 initialize_ft_cache();
 524                 initialize_noise_words();
 525                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 526                 CtdlRegisterDeleteHook(ft_delete_remove);
 527                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 528                 CtdlRegisterCleanupHook(noise_word_cleanup);
 529         }
 530         else
 531         {
 532                 CtdlThreadCreate("Indexer", CTDLTHREAD_BIGSTACK, indexer_thread, NULL);
 533         }
 534         /* return our Subversion id for the Log */
 535         return "fulltext";
 536 }