citadel/serv_fulltext.c

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