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
 180                 free(tokens);
 181         }
 182 }
 183
 184
 185
 186 /*
 187  * Add a message to the list of those to be indexed.
 188  */
 189 void ft_index_msg(long msgnum, void *userdata) {
 190
 191         if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
 192                 ++ft_num_msgs;
 193                 if (ft_num_msgs > ft_num_alloc) {
 194                         ft_num_alloc += 1024;
 195                         ft_newmsgs = realloc(ft_newmsgs,
 196                                 (ft_num_alloc * sizeof(long)));
 197                 }
 198                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 199         }
 200
 201 }
 202
 203 /*
 204  * Scan a room for messages to index.
 205  */
 206 void ft_index_room(struct ctdlroom *qrbuf, void *data)
 207 {
 208         getroom(&CC->room, qrbuf->QRname);
 209         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 210 }
 211
 212
 213 /*
 214  * Begin the fulltext indexing process.  (Called as an EVT_TIMER event)
 215  */
 216 void do_fulltext_indexing(void) {
 217         int i;
 218         static time_t last_index = 0L;
 219         static time_t last_progress = 0L;
 220
 221         /*
 222          * Don't do this if the site doesn't have it enabled.
 223          */
 224         if (!config.c_enable_fulltext) {
 225                 return;
 226         }
 227
 228         /*
 229          * Make sure we don't run the indexer too frequently.
 230          * FIXME move the setting into config
 231          */
 232         if ( (time(NULL) - last_index) < 300L) {
 233                 return;
 234         }
 235
 236         /*
 237          * Check to see whether the fulltext index is up to date; if there
 238          * are no messages to index, don't waste any more time trying.
 239          */
 240         if (CitControl.MMfulltext >= CitControl.MMhighest) {
 241                 return;         /* nothing to do! */
 242         }
 243
 244         lprintf(CTDL_DEBUG, "do_fulltext_indexing() started\n");
 245
 246         /*
 247          * If we've switched wordbreaker modules, burn the index and start
 248          * over.
 249          */
 250         begin_critical_section(S_CONTROL);
 251         lprintf(CTDL_DEBUG, "wb ver on disk = %d, code ver = %d\n",
 252                         CitControl.fulltext_wordbreaker, FT_WORDBREAKER_ID);
 253         if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
 254                 lprintf(CTDL_INFO, "(re)initializing full text index\n");
 255                 cdb_trunc(CDB_FULLTEXT);
 256                 CitControl.MMfulltext = 0L;
 257                 put_control();
 258         }
 259         end_critical_section(S_CONTROL);
 260
 261         /*
 262          * Now go through each room and find messages to index.
 263          */
 264         ft_newhighest = CitControl.MMhighest;
 265         ForEachRoom(ft_index_room, NULL);       /* load all msg pointers */
 266
 267         if (ft_num_msgs > 0) {
 268                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 269                 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
 270                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 271                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
 272                                         ((ft_num_msgs - i - 1)*sizeof(long)));
 273                                 --ft_num_msgs;
 274                                 --i;
 275                         }
 276                 }
 277
 278                 /* Here it is ... do each message! */
 279                 for (i=0; i<ft_num_msgs; ++i) {
 280                         if (time(NULL) != last_progress) {
 281                                 lprintf(CTDL_DEBUG,
 282                                         "Indexed %d of %d messages (%d%%)\n",
 283                                                 i, ft_num_msgs,
 284                                                 ((i*100) / ft_num_msgs)
 285                                 );
 286                                 last_progress = time(NULL);
 287                         }
 288                         ft_index_message(ft_newmsgs[i], 1);
 289
 290                         /* Check to see if we need to quit early */
 291                         if (time_to_die) {
 292                                 lprintf(CTDL_DEBUG, "Indexer quitting early\n");
 293                                 ft_newhighest = ft_newmsgs[i];
 294                                 break;
 295                         }
 296
 297                         /* Check to see if we have to maybe flush to disk */
 298                         if (i >= FT_MAX_CACHE) {
 299                                 lprintf(CTDL_DEBUG, "Time to flush.\n");
 300                                 ft_newhighest = ft_newmsgs[i];
 301                                 break;
 302                         }
 303
 304                 }
 305
 306                 free(ft_newmsgs);
 307                 ft_num_msgs = 0;
 308                 ft_num_alloc = 0;
 309                 ft_newmsgs = NULL;
 310         }
 311
 312         /* Save our place so we don't have to do this again */
 313         ft_flush_cache();
 314         begin_critical_section(S_CONTROL);
 315         CitControl.MMfulltext = ft_newhighest;
 316         CitControl.fulltext_wordbreaker = FT_WORDBREAKER_ID;
 317         put_control();
 318         end_critical_section(S_CONTROL);
 319         last_index = time(NULL);
 320
 321         lprintf(CTDL_DEBUG, "do_fulltext_indexing() finished\n");
 322         return;
 323 }
 324
 325 /*
 326  * Main loop for the indexer thread.
 327  */
 328 void *indexer_thread(void *arg) {
 329         struct CitContext indexerCC;
 330
 331         lprintf(CTDL_DEBUG, "indexer_thread() initializing\n");
 332
 333         memset(&indexerCC, 0, sizeof(struct CitContext));
 334         indexerCC.internal_pgm = 1;
 335         indexerCC.cs_pid = 0;
 336         pthread_setspecific(MyConKey, (void *)&indexerCC );
 337
 338         cdb_allocate_tsd();
 339
 340         while (!time_to_die) {
 341                 do_fulltext_indexing();
 342                 sleep(1);
 343         }
 344
 345         lprintf(CTDL_DEBUG, "indexer_thread() exiting\n");
 346         pthread_exit(NULL);
 347 }
 348
 349
 350
 351 /*
 352  * API call to perform searches.
 353  * (This one does the "all of these words" search.)
 354  * Caller is responsible for freeing the message list.
 355  */
 356 void ft_search(int *fts_num_msgs, long **fts_msgs, char *search_string) {
 357         int num_tokens = 0;
 358         int *tokens = NULL;
 359         int i, j;
 360         struct cdbdata *cdb_bucket;
 361         int num_all_msgs = 0;
 362         long *all_msgs = NULL;
 363         int num_ret_msgs = 0;
 364         int num_ret_alloc = 0;
 365         long *ret_msgs = NULL;
 366         int tok;
 367
 368         wordbreaker(search_string, &num_tokens, &tokens);
 369         if (num_tokens > 0) {
 370                 for (i=0; i<num_tokens; ++i) {
 371
 372                         /* search for tokens[i] */
 373                         tok = tokens[i];
 374
 375                         /* fetch the bucket, Liza */
 376                         if (ftc_msgs[tok] == NULL) {
 377                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 378                                 if (cdb_bucket != NULL) {
 379                                         ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
 380                                         ftc_msgs[tok] = (long *)cdb_bucket->ptr;
 381                                         cdb_bucket->ptr = NULL;
 382                                         cdb_free(cdb_bucket);
 383                                 }
 384                                 else {
 385                                         ftc_num_msgs[tok] = 0;
 386                                         ftc_msgs[tok] = malloc(sizeof(long));
 387                                 }
 388                         }
 389
 390                         num_all_msgs += ftc_num_msgs[tok];
 391                         if (num_all_msgs > 0) {
 392                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 393                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
 394                                         ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 395                         }
 396
 397                 }
 398                 free(tokens);
 399                 qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 400
 401                 /*
 402                  * At this point, if a message appears num_tokens times in the
 403                  * list, then it contains all of the search tokens.
 404                  */
 405                 if (num_all_msgs >= num_tokens)
 406                    for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 407                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 408
 409                                 ++num_ret_msgs;
 410                                 if (num_ret_msgs > num_ret_alloc) {
 411                                         num_ret_alloc += 64;
 412                                         ret_msgs = realloc(ret_msgs,
 413                                                 (num_ret_alloc*sizeof(long)) );
 414                                 }
 415                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 416
 417                         }
 418                 }
 419
 420                 free(all_msgs);
 421         }
 422
 423         *fts_num_msgs = num_ret_msgs;
 424         *fts_msgs = ret_msgs;
 425 }
 426
 427
 428 /*
 429  * This search command is for diagnostic purposes and may be removed or replaced.
 430  */
 431 void cmd_srch(char *argbuf) {
 432         int num_msgs = 0;
 433         long *msgs = NULL;
 434         int i;
 435         char search_string[256];
 436
 437         if (CtdlAccessCheck(ac_logged_in)) return;
 438
 439         if (!config.c_enable_fulltext) {
 440                 cprintf("%d Full text index is not enabled on this server.\n",
 441                         ERROR + CMD_NOT_SUPPORTED);
 442                 return;
 443         }
 444
 445         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 446         ft_search(&num_msgs, &msgs, search_string);
 447
 448         cprintf("%d %d msgs match all search words:\n",
 449                 LISTING_FOLLOWS, num_msgs);
 450         if (num_msgs > 0) {
 451                 for (i=0; i<num_msgs; ++i) {
 452                         cprintf("%ld\n", msgs[i]);
 453                 }
 454         }
 455         if (msgs != NULL) free(msgs);
 456         cprintf("000\n");
 457 }
 458
 459 /*
 460  * Zero out our index cache.
 461  */
 462 void initialize_ft_cache(void) {
 463         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 464         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 465 }
 466
 467
 468 /*****************************************************************************/
 469
 470 char *serv_fulltext_init(void)
 471 {
 472         initialize_ft_cache();
 473         CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 474
 475         /* return our Subversion id for the Log */
 476         return "$Id$";
 477 }