citadel/modules/fulltext/serv_fulltext.c

   1 /*
   2  * This module handles fulltext indexing of the message base.
   3  * Copyright (c) 2005-2018 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 /*
  85  * Flush our index cache out to disk.
  86  */
  87 void ft_flush_cache(void) {
  88         int i;
  89         time_t last_update = 0;
  90
  91         for (i=0; i<65536; ++i) {
  92                 if ((time(NULL) - last_update) >= 10) {
  93                         syslog(LOG_INFO,
  94                                 "fulltext: flushing index cache to disk (%d%% complete)",
  95                                 (i * 100 / 65536)
  96                         );
  97                         last_update = time(NULL);
  98                 }
  99                 if (ftc_msgs[i] != NULL) {
 100                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
 101                                 (ftc_num_msgs[i] * sizeof(long)));
 102                         ftc_num_msgs[i] = 0;
 103                         free(ftc_msgs[i]);
 104                         ftc_msgs[i] = NULL;
 105                 }
 106         }
 107         syslog(LOG_INFO, "fulltext: flushed index cache to disk (100%% complete)");
 108 }
 109
 110
 111 /*
 112  * Index or de-index a message.  (op == 1 to index, 0 to de-index)
 113  */
 114 void ft_index_message(long msgnum, int op) {
 115         int num_tokens = 0;
 116         int *tokens = NULL;
 117         int i, j;
 118         struct cdbdata *cdb_bucket;
 119         StrBuf *msgtext;
 120         char *txt;
 121         int tok;
 122         struct CtdlMessage *msg = NULL;
 123
 124         msg = CtdlFetchMessage(msgnum, 1, 1);
 125         if (msg == NULL) {
 126                 syslog(LOG_ERR, "fulltext: ft_index_message() could not load msg %ld", msgnum);
 127                 return;
 128         }
 129
 130         if (!CM_IsEmpty(msg, eSuppressIdx)) {
 131                 syslog(LOG_DEBUG, "fulltext: ft_index_message() excluded msg %ld", msgnum);
 132                 CM_Free(msg);
 133                 return;
 134         }
 135
 136         syslog(LOG_DEBUG, "fulltext: ft_index_message() %s msg %ld", (op ? "adding" : "removing") , msgnum);
 137
 138         /* Output the message as text before indexing it, so we don't end up
 139          * indexing a bunch of encoded base64, etc.
 140          */
 141         CC->redirect_buffer = NewStrBufPlain(NULL, SIZ);
 142         CtdlOutputPreLoadedMsg(msg, MT_CITADEL, HEADERS_ALL, 0, 1, 0);
 143         CM_Free(msg);
 144         msgtext = CC->redirect_buffer;
 145         CC->redirect_buffer = NULL;
 146         if (msgtext != NULL) {
 147                 syslog(LOG_DEBUG, "fulltext: wordbreaking message %ld (%d bytes)", msgnum, StrLength(msgtext));
 148         }
 149         txt = SmashStrBuf(&msgtext);
 150         wordbreaker(txt, &num_tokens, &tokens);
 151         free(txt);
 152
 153         syslog(LOG_DEBUG, "fulltext: 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, "fulltext: invalid token %d !!", tok);
 200                         }
 201                 }
 202
 203                 free(tokens);
 204         }
 205 }
 206
 207
 208 /*
 209  * Add a message to the list of those to be indexed.
 210  */
 211 void ft_index_msg(long msgnum, void *userdata) {
 212
 213         if ((msgnum > CtdlGetConfigLong("MMfulltext")) && (msgnum <= ft_newhighest)) {
 214                 ++ft_num_msgs;
 215                 if (ft_num_msgs > ft_num_alloc) {
 216                         ft_num_alloc += 1024;
 217                         ft_newmsgs = realloc(ft_newmsgs, (ft_num_alloc * sizeof(long)));
 218                 }
 219                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 220         }
 221
 222 }
 223
 224
 225 /*
 226  * Scan a room for messages to index.
 227  */
 228 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 229 {
 230         if (server_shutting_down)
 231                 return;
 232
 233         CtdlGetRoom(&CC->room, qrbuf->QRname);
 234         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 235 }
 236
 237
 238 /*
 239  * Begin the fulltext indexing process.
 240  */
 241 void do_fulltext_indexing(void) {
 242         int i;
 243         static time_t last_progress = 0L;
 244         static int is_running = 0;
 245         if (is_running) return;         /* Concurrency check - only one can run */
 246         is_running = 1;
 247
 248         /*
 249          * Don't do this if the site doesn't have it enabled.
 250          */
 251         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 252                 return;
 253         }
 254
 255         /*
 256          * If we've switched wordbreaker modules, burn the index and start over.
 257          */
 258         begin_critical_section(S_CONTROL);
 259         if (CtdlGetConfigInt("MM_fulltext_wordbreaker") != FT_WORDBREAKER_ID) {
 260                 syslog(LOG_DEBUG, "fulltext: wb ver on disk = %d, code ver = %d",
 261                         CtdlGetConfigInt("MM_fulltext_wordbreaker"), FT_WORDBREAKER_ID
 262                 );
 263                 syslog(LOG_INFO, "fulltext: (re)initializing index");
 264                 cdb_trunc(CDB_FULLTEXT);
 265                 CtdlSetConfigLong("MMfulltext", 0);
 266         }
 267         end_critical_section(S_CONTROL);
 268
 269         /*
 270          * Silently return if our fulltext index is up to date with new messages.
 271          */
 272         if ((CtdlGetConfigLong("MMfulltext") >= CtdlGetConfigLong("MMhighest"))) {
 273                 return;         /* nothing to do! */
 274         }
 275
 276         /*
 277          * Now go through each room and find messages to index.
 278          */
 279         ft_newhighest = CtdlGetConfigLong("MMhighest");
 280         CtdlForEachRoom(ft_index_room, NULL);   /* load all msg pointers */
 281
 282         if (ft_num_msgs > 0) {
 283                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 284                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 285                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 286                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 287                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 288                                 --ft_num_msgs;
 289                                 --i;
 290                         }
 291                 }
 292
 293                 /* Here it is ... do each message! */
 294                 for (i=0; i<ft_num_msgs; ++i) {
 295                         if (time(NULL) != last_progress) {
 296                                 syslog(LOG_DEBUG,
 297                                         "fulltext: indexed %d of %d messages (%d%%)",
 298                                                 i, ft_num_msgs,
 299                                                 ((i*100) / ft_num_msgs)
 300                                 );
 301                                 last_progress = time(NULL);
 302                         }
 303                         ft_index_message(ft_newmsgs[i], 1);
 304
 305                         /* Check to see if we need to quit early */
 306                         if (server_shutting_down) {
 307                                 syslog(LOG_DEBUG, "fulltext: indexer quitting early");
 308                                 ft_newhighest = ft_newmsgs[i];
 309                                 break;
 310                         }
 311
 312                         /* Check to see if we have to maybe flush to disk */
 313                         if (i >= FT_MAX_CACHE) {
 314                                 syslog(LOG_DEBUG, "fulltext: time to flush.");
 315                                 ft_newhighest = ft_newmsgs[i];
 316                                 break;
 317                         }
 318
 319                 }
 320
 321                 free(ft_newmsgs);
 322                 ft_num_msgs = 0;
 323                 ft_num_alloc = 0;
 324                 ft_newmsgs = NULL;
 325         }
 326
 327         if (server_shutting_down) {
 328                 is_running = 0;
 329                 return;
 330         }
 331
 332         /* Save our place so we don't have to do this again */
 333         ft_flush_cache();
 334         begin_critical_section(S_CONTROL);
 335         CtdlSetConfigLong("MMfulltext", ft_newhighest);
 336         CtdlSetConfigInt("MM_fulltext_wordbreaker", FT_WORDBREAKER_ID);
 337         end_critical_section(S_CONTROL);
 338
 339         syslog(LOG_DEBUG, "fulltext: indexing finished");
 340         is_running = 0;
 341         return;
 342 }
 343
 344
 345 /*
 346  * API call to perform searches.
 347  * (This one does the "all of these words" search.)
 348  * Caller is responsible for freeing the message list.
 349  */
 350 void ft_search(int *fts_num_msgs, long **fts_msgs, const char *search_string) {
 351         int num_tokens = 0;
 352         int *tokens = NULL;
 353         int i, j;
 354         struct cdbdata *cdb_bucket;
 355         int num_all_msgs = 0;
 356         long *all_msgs = NULL;
 357         int num_ret_msgs = 0;
 358         int num_ret_alloc = 0;
 359         long *ret_msgs = NULL;
 360         int tok;
 361
 362         wordbreaker(search_string, &num_tokens, &tokens);
 363         if (num_tokens > 0) {
 364                 for (i=0; i<num_tokens; ++i) {
 365
 366                         /* search for tokens[i] */
 367                         tok = tokens[i];
 368
 369                         /* fetch the bucket, Liza */
 370                         if (ftc_msgs[tok] == NULL) {
 371                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 372                                 if (cdb_bucket != NULL) {
 373                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 374                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 375                                         cdb_bucket->ptr = NULL;
 376                                         cdb_free(cdb_bucket);
 377                                 }
 378                                 else {
 379                                         ftc_num_msgs[tok] = 0;
 380                                         ftc_msgs[tok] = malloc(sizeof(long));
 381                                 }
 382                         }
 383
 384                         num_all_msgs += ftc_num_msgs[tok];
 385                         if (num_all_msgs > 0) {
 386                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 387                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 388                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 389                         }
 390
 391                 }
 392                 free(tokens);
 393                 if (all_msgs != NULL) {
 394                         qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 395
 396                         /*
 397                          * At this point, if a message appears num_tokens times in the
 398                          * list, then it contains all of the search tokens.
 399                          */
 400                         if (num_all_msgs >= num_tokens)
 401                                 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 402                                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 403
 404                                                 ++num_ret_msgs;
 405                                                 if (num_ret_msgs > num_ret_alloc) {
 406                                                         num_ret_alloc += 64;
 407                                                         ret_msgs = realloc(ret_msgs,
 408                                                                            (num_ret_alloc*sizeof(long)) );
 409                                                 }
 410                                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 411
 412                                         }
 413                                 }
 414                         free(all_msgs);
 415                 }
 416         }
 417
 418         *fts_num_msgs = num_ret_msgs;
 419         *fts_msgs = ret_msgs;
 420 }
 421
 422
 423 /*
 424  * This search command is for diagnostic purposes and may be removed or replaced.
 425  */
 426 void cmd_srch(char *argbuf) {
 427         int num_msgs = 0;
 428         long *msgs = NULL;
 429         int i;
 430         char search_string[256];
 431
 432         if (CtdlAccessCheck(ac_logged_in)) return;
 433
 434         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 435                 cprintf("%d Full text index is not enabled on this server.\n",
 436                         ERROR + CMD_NOT_SUPPORTED);
 437                 return;
 438         }
 439
 440         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 441         ft_search(&num_msgs, &msgs, search_string);
 442
 443         cprintf("%d %d msgs match all search words:\n",
 444                 LISTING_FOLLOWS, num_msgs);
 445         if (num_msgs > 0) {
 446                 for (i=0; i<num_msgs; ++i) {
 447                         cprintf("%ld\n", msgs[i]);
 448                 }
 449         }
 450         if (msgs != NULL) free(msgs);
 451         cprintf("000\n");
 452 }
 453
 454
 455 /*
 456  * Zero out our index cache.
 457  */
 458 void initialize_ft_cache(void) {
 459         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 460         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 461 }
 462
 463
 464 void ft_delete_remove(char *room, long msgnum)
 465 {
 466         if (room) return;
 467
 468         /* Remove from fulltext index */
 469         if (CtdlGetConfigInt("c_enable_fulltext")) {
 470                 ft_index_message(msgnum, 0);
 471         }
 472 }
 473
 474
 475 /*****************************************************************************/
 476
 477 CTDL_MODULE_INIT(fulltext)
 478 {
 479         if (!threading)
 480         {
 481                 initialize_ft_cache();
 482                 initialize_noise_words();
 483                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 484                 CtdlRegisterDeleteHook(ft_delete_remove);
 485                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 486                 CtdlRegisterCleanupHook(noise_word_cleanup);
 487                 CtdlRegisterSessionHook(do_fulltext_indexing, EVT_TIMER, PRIO_CLEANUP + 300);
 488         }
 489         /* return our module name for the log */
 490         return "fulltext";
 491 }