citadel/modules/fulltext/serv_fulltext.c

   1 /*
   2  * This module handles fulltext indexing of the message base.
   3  * Copyright (c) 2005-2011 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
  30 #if TIME_WITH_SYS_TIME
  31 # include <sys/time.h>
  32 # include <time.h>
  33 #else
  34 # if HAVE_SYS_TIME_H
  35 #  include <sys/time.h>
  36 # else
  37 #  include <time.h>
  38 # endif
  39 #endif
  40
  41 #include <sys/wait.h>
  42 #include <string.h>
  43 #include <limits.h>
  44 #include <libcitadel.h>
  45 #include "citadel.h"
  46 #include "server.h"
  47 #include "citserver.h"
  48 #include "support.h"
  49 #include "config.h"
  50 #include "database.h"
  51 #include "msgbase.h"
  52 #include "control.h"
  53 #include "serv_fulltext.h"
  54 #include "ft_wordbreaker.h"
  55 #include "threads.h"
  56 #include "context.h"
  57
  58 #include "ctdl_module.h"
  59
  60 long ft_newhighest = 0L;
  61 long *ft_newmsgs = NULL;
  62 int ft_num_msgs = 0;
  63 int ft_num_alloc = 0;
  64
  65 int ftc_num_msgs[65536];
  66 long *ftc_msgs[65536];
  67
  68
  69 /*
  70  * Compare function
  71  */
  72 int longcmp(const void *rec1, const void *rec2) {
  73         long i1, i2;
  74
  75         i1 = *(const long *)rec1;
  76         i2 = *(const long *)rec2;
  77
  78         if (i1 > i2) return(1);
  79         if (i1 < i2) return(-1);
  80         return(0);
  81 }
  82
  83 /*
  84  * Flush our index cache out to disk.
  85  */
  86 void ft_flush_cache(void) {
  87         int i;
  88         time_t last_update = 0;
  89
  90         for (i=0; i<65536; ++i) {
  91                 if ((time(NULL) - last_update) >= 10) {
  92                         syslog(LOG_INFO,
  93                                 "Flushing index cache to disk (%d%% complete)",
  94                                 (i * 100 / 65536)
  95                         );
  96                         last_update = time(NULL);
  97                 }
  98                 if (ftc_msgs[i] != NULL) {
  99                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
 100                                 (ftc_num_msgs[i] * sizeof(long)));
 101                         ftc_num_msgs[i] = 0;
 102                         free(ftc_msgs[i]);
 103                         ftc_msgs[i] = NULL;
 104                 }
 105         }
 106         syslog(LOG_INFO, "Flushed index cache to disk (100%% complete)");
 107 }
 108
 109
 110 /*
 111  * Index or de-index a message.  (op == 1 to index, 0 to de-index)
 112  */
 113 void ft_index_message(long msgnum, int op) {
 114         int num_tokens = 0;
 115         int *tokens = NULL;
 116         int i, j;
 117         struct cdbdata *cdb_bucket;
 118         StrBuf *msgtext;
 119         char *txt;
 120         int tok;
 121         struct CtdlMessage *msg = NULL;
 122
 123         msg = CtdlFetchMessage(msgnum, 1);
 124         if (msg == NULL) {
 125                 syslog(LOG_ERR, "ft_index_message() could not load msg %ld", msgnum);
 126                 return;
 127         }
 128
 129         if (msg->cm_fields['1'] != NULL) {
 130                 syslog(LOG_DEBUG, "ft_index_message() excluded msg %ld", msgnum);
 131                 CtdlFreeMessage(msg);
 132                 return;
 133         }
 134
 135         syslog(LOG_DEBUG, "ft_index_message() %s msg %ld", (op ? "adding" : "removing") , msgnum);
 136
 137         /* Output the message as text before indexing it, so we don't end up
 138          * indexing a bunch of encoded base64, etc.
 139          */
 140         CC->redirect_buffer = NewStrBufPlain(NULL, SIZ);
 141         CtdlOutputPreLoadedMsg(msg, MT_CITADEL, HEADERS_ALL, 0, 1, 0);
 142         CtdlFreeMessage(msg);
 143         msgtext = CC->redirect_buffer;
 144         CC->redirect_buffer = NULL;
 145         syslog(LOG_DEBUG, "Wordbreaking message %ld...", msgnum);
 146         if ((msgtext == NULL) || (StrLength(msgtext) == 0)) {
 147                 syslog(LOG_ALERT, "This message has a zero length.  Probable data corruption.");
 148         }
 149         txt = SmashStrBuf(&msgtext);
 150         wordbreaker(txt, &num_tokens, &tokens);
 151         free(txt);
 152
 153         syslog(LOG_DEBUG, "Indexing message %ld [%d tokens]", msgnum, num_tokens);
 154         if (num_tokens > 0) {
 155                 for (i=0; i<num_tokens; ++i) {
 156
 157                         /* Add the message to the relevant token bucket */
 158
 159                         /* search for tokens[i] */
 160                         tok = tokens[i];
 161
 162                         if ( (tok >= 0) && (tok <= 65535) ) {
 163                                 /* fetch the bucket, Liza */
 164                                 if (ftc_msgs[tok] == NULL) {
 165                                         cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 166                                         if (cdb_bucket != NULL) {
 167                                                 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 168                                                 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 169                                                 cdb_bucket->ptr = NULL;
 170                                                 cdb_free(cdb_bucket);
 171                                         }
 172                                         else {
 173                                                 ftc_num_msgs[tok] = 0;
 174                                                 ftc_msgs[tok] = malloc(sizeof(long));
 175                                         }
 176                                 }
 177
 178
 179                                 if (op == 1) {  /* add to index */
 180                                         ++ftc_num_msgs[tok];
 181                                         ftc_msgs[tok] = realloc(ftc_msgs[tok],
 182                                                                 ftc_num_msgs[tok]*sizeof(long));
 183                                         ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
 184                                 }
 185
 186                                 if (op == 0) {  /* remove from index */
 187                                         if (ftc_num_msgs[tok] >= 1) {
 188                                                 for (j=0; j<ftc_num_msgs[tok]; ++j) {
 189                                                         if (ftc_msgs[tok][j] == msgnum) {
 190                                                                 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
 191                                                                 --ftc_num_msgs[tok];
 192                                                                 --j;
 193                                                         }
 194                                                 }
 195                                         }
 196                                 }
 197                         }
 198                         else {
 199                                 syslog(LOG_ALERT, "Invalid token %d !!", tok);
 200                         }
 201                 }
 202
 203                 free(tokens);
 204         }
 205 }
 206
 207
 208
 209 /*
 210  * Add a message to the list of those to be indexed.
 211  */
 212 void ft_index_msg(long msgnum, void *userdata) {
 213
 214         if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
 215                 ++ft_num_msgs;
 216                 if (ft_num_msgs > ft_num_alloc) {
 217                         ft_num_alloc += 1024;
 218                         ft_newmsgs = realloc(ft_newmsgs,
 219                                 (ft_num_alloc * sizeof(long)));
 220                 }
 221                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 222         }
 223
 224 }
 225
 226 /*
 227  * Scan a room for messages to index.
 228  */
 229 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 230 {
 231         if (server_shutting_down)
 232                 return;
 233
 234         CtdlGetRoom(&CC->room, qrbuf->QRname);
 235         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 236 }
 237
 238
 239 /*
 240  * Begin the fulltext indexing process.
 241  */
 242 void do_fulltext_indexing(void) {
 243         int i;
 244         static time_t last_index = 0L;
 245         static time_t last_progress = 0L;
 246         time_t run_time = 0L;
 247         time_t end_time = 0L;
 248         static int is_running = 0;
 249         if (is_running) return;         /* Concurrency check - only one can run */
 250         is_running = 1;
 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         if ( (time(NULL) - last_index) < 300L) {
 265                 return;
 266         }
 267
 268         /*
 269          * Check to see whether the fulltext index is up to date; if there
 270          * are no messages to index, don't waste any more time trying.
 271          */
 272         if ((CitControl.MMfulltext >= CitControl.MMhighest) && (CitControl.fulltext_wordbreaker == FT_WORDBREAKER_ID)) {
 273                 return;         /* nothing to do! */
 274         }
 275
 276         run_time = time(NULL);
 277         syslog(LOG_DEBUG, "do_fulltext_indexing() started (%ld)", run_time);
 278
 279         /*
 280          * If we've switched wordbreaker modules, burn the index and start
 281          * over.
 282          */
 283         begin_critical_section(S_CONTROL);
 284         if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
 285                 syslog(LOG_DEBUG, "wb ver on disk = %d, code ver = %d",
 286                         CitControl.fulltext_wordbreaker, FT_WORDBREAKER_ID
 287                 );
 288                 syslog(LOG_INFO, "(re)initializing full text index");
 289                 cdb_trunc(CDB_FULLTEXT);
 290                 CitControl.MMfulltext = 0L;
 291                 put_control();
 292         }
 293         end_critical_section(S_CONTROL);
 294
 295         /*
 296          * Now go through each room and find messages to index.
 297          */
 298         ft_newhighest = CitControl.MMhighest;
 299         CtdlForEachRoom(ft_index_room, NULL);   /* load all msg pointers */
 300
 301         if (ft_num_msgs > 0) {
 302                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 303                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 304                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 305                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 306                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 307                                 --ft_num_msgs;
 308                                 --i;
 309                         }
 310                 }
 311
 312                 /* Here it is ... do each message! */
 313                 for (i=0; i<ft_num_msgs; ++i) {
 314                         if (time(NULL) != last_progress) {
 315                                 syslog(LOG_DEBUG,
 316                                         "Indexed %d of %d messages (%d%%)",
 317                                                 i, ft_num_msgs,
 318                                                 ((i*100) / ft_num_msgs)
 319                                 );
 320                                 last_progress = time(NULL);
 321                         }
 322                         ft_index_message(ft_newmsgs[i], 1);
 323
 324                         /* Check to see if we need to quit early */
 325                         if (server_shutting_down) {
 326                                 syslog(LOG_DEBUG, "Indexer quitting early");
 327                                 ft_newhighest = ft_newmsgs[i];
 328                                 break;
 329                         }
 330
 331                         /* Check to see if we have to maybe flush to disk */
 332                         if (i >= FT_MAX_CACHE) {
 333                                 syslog(LOG_DEBUG, "Time to flush.");
 334                                 ft_newhighest = ft_newmsgs[i];
 335                                 break;
 336                         }
 337
 338                 }
 339
 340                 free(ft_newmsgs);
 341                 ft_num_msgs = 0;
 342                 ft_num_alloc = 0;
 343                 ft_newmsgs = NULL;
 344         }
 345         end_time = time(NULL);
 346
 347         if (server_shutting_down) {
 348                 is_running = 0;
 349                 return;
 350         }
 351
 352         syslog(LOG_DEBUG, "do_fulltext_indexing() duration (%ld)", 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         syslog(LOG_DEBUG, "do_fulltext_indexing() finished");
 364         is_running = 0;
 365         return;
 366 }
 367
 368
 369
 370 /*
 371  * API call to perform searches.
 372  * (This one does the "all of these words" search.)
 373  * Caller is responsible for freeing the message list.
 374  */
 375 void ft_search(int *fts_num_msgs, long **fts_msgs, const char *search_string) {
 376         int num_tokens = 0;
 377         int *tokens = NULL;
 378         int i, j;
 379         struct cdbdata *cdb_bucket;
 380         int num_all_msgs = 0;
 381         long *all_msgs = NULL;
 382         int num_ret_msgs = 0;
 383         int num_ret_alloc = 0;
 384         long *ret_msgs = NULL;
 385         int tok;
 386
 387         wordbreaker(search_string, &num_tokens, &tokens);
 388         if (num_tokens > 0) {
 389                 for (i=0; i<num_tokens; ++i) {
 390
 391                         /* search for tokens[i] */
 392                         tok = tokens[i];
 393
 394                         /* fetch the bucket, Liza */
 395                         if (ftc_msgs[tok] == NULL) {
 396                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 397                                 if (cdb_bucket != NULL) {
 398                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 399                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 400                                         cdb_bucket->ptr = NULL;
 401                                         cdb_free(cdb_bucket);
 402                                 }
 403                                 else {
 404                                         ftc_num_msgs[tok] = 0;
 405                                         ftc_msgs[tok] = malloc(sizeof(long));
 406                                 }
 407                         }
 408
 409                         num_all_msgs += ftc_num_msgs[tok];
 410                         if (num_all_msgs > 0) {
 411                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 412                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 413                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 414                         }
 415
 416                 }
 417                 free(tokens);
 418                 if (all_msgs != NULL) {
 419                         qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 420
 421                         /*
 422                          * At this point, if a message appears num_tokens times in the
 423                          * list, then it contains all of the search tokens.
 424                          */
 425                         if (num_all_msgs >= num_tokens)
 426                                 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 427                                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 428
 429                                                 ++num_ret_msgs;
 430                                                 if (num_ret_msgs > num_ret_alloc) {
 431                                                         num_ret_alloc += 64;
 432                                                         ret_msgs = realloc(ret_msgs,
 433                                                                            (num_ret_alloc*sizeof(long)) );
 434                                                 }
 435                                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 436
 437                                         }
 438                                 }
 439                         free(all_msgs);
 440                 }
 441         }
 442
 443         *fts_num_msgs = num_ret_msgs;
 444         *fts_msgs = ret_msgs;
 445 }
 446
 447
 448 /*
 449  * This search command is for diagnostic purposes and may be removed or replaced.
 450  */
 451 void cmd_srch(char *argbuf) {
 452         int num_msgs = 0;
 453         long *msgs = NULL;
 454         int i;
 455         char search_string[256];
 456
 457         if (CtdlAccessCheck(ac_logged_in)) return;
 458
 459         if (!config.c_enable_fulltext) {
 460                 cprintf("%d Full text index is not enabled on this server.\n",
 461                         ERROR + CMD_NOT_SUPPORTED);
 462                 return;
 463         }
 464
 465         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 466         ft_search(&num_msgs, &msgs, search_string);
 467
 468         cprintf("%d %d msgs match all search words:\n",
 469                 LISTING_FOLLOWS, num_msgs);
 470         if (num_msgs > 0) {
 471                 for (i=0; i<num_msgs; ++i) {
 472                         cprintf("%ld\n", msgs[i]);
 473                 }
 474         }
 475         if (msgs != NULL) free(msgs);
 476         cprintf("000\n");
 477 }
 478
 479 /*
 480  * Zero out our index cache.
 481  */
 482 void initialize_ft_cache(void) {
 483         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 484         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 485 }
 486
 487
 488 void ft_delete_remove(char *room, long msgnum)
 489 {
 490         if (room) return;
 491
 492         /* Remove from fulltext index */
 493         if (config.c_enable_fulltext) {
 494                 ft_index_message(msgnum, 0);
 495         }
 496 }
 497
 498 /*****************************************************************************/
 499
 500 CTDL_MODULE_INIT(fulltext)
 501 {
 502         if (!threading)
 503         {
 504                 initialize_ft_cache();
 505                 initialize_noise_words();
 506                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 507                 CtdlRegisterDeleteHook(ft_delete_remove);
 508                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 509                 CtdlRegisterCleanupHook(noise_word_cleanup);
 510                 CtdlRegisterSessionHook(do_fulltext_indexing, EVT_TIMER, PRIO_CLEANUP + 300);
 511         }
 512         /* return our module name for the log */
 513         return "fulltext";
 514 }