citadel/server/modules/fulltext/serv_fulltext.c

   1 // This module handles fulltext indexing of the message base.
   2 //
   3 // Copyright (c) 2005-2023 by the citadel.org team
   4 //
   5 // This program is open source software.  Use, duplication, or disclosure
   6 // is subject to the terms of the GNU General Public License, version 3.
   7
   8 #include "../../sysdep.h"
   9 #include <stdlib.h>
  10 #include <unistd.h>
  11 #include <stdio.h>
  12 #include <fcntl.h>
  13 #include <signal.h>
  14 #include <pwd.h>
  15 #include <errno.h>
  16 #include <sys/types.h>
  17 #include <time.h>
  18 #include <sys/wait.h>
  19 #include <string.h>
  20 #include <limits.h>
  21 #include <libcitadel.h>
  22 #include "../../citadel_defs.h"
  23 #include "../../server.h"
  24 #include "../../citserver.h"
  25 #include "../../support.h"
  26 #include "../../config.h"
  27 #include "../../database.h"
  28 #include "../../msgbase.h"
  29 #include "../../control.h"
  30 #include "serv_fulltext.h"
  31 #include "ft_wordbreaker.h"
  32 #include "../../threads.h"
  33 #include "../../context.h"
  34 #include "../../ctdl_module.h"
  35
  36 long ft_newhighest = 0L;
  37 long *ft_newmsgs = NULL;
  38 int ft_num_msgs = 0;
  39 int ft_num_alloc = 0;
  40
  41 int ftc_num_msgs[65536];
  42 long *ftc_msgs[65536];
  43
  44
  45 // Compare function
  46 int longcmp(const void *rec1, const void *rec2) {
  47         long i1, i2;
  48
  49         i1 = *(const long *)rec1;
  50         i2 = *(const long *)rec2;
  51
  52         if (i1 > i2) return(1);
  53         if (i1 < i2) return(-1);
  54         return(0);
  55 }
  56
  57
  58 // Flush our index cache out to disk.
  59 void ft_flush_cache(void) {
  60         int i;
  61         time_t last_update = 0;
  62
  63         for (i=0; i<65536; ++i) {
  64                 if ((time(NULL) - last_update) >= 10) {
  65                         syslog(LOG_INFO, "fulltext: flushing index cache to disk (%d%% complete)", (i * 100 / 65536));
  66                         last_update = time(NULL);
  67                 }
  68                 if (ftc_msgs[i] != NULL) {
  69                         cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i], (ftc_num_msgs[i] * sizeof(long)));
  70                         ftc_num_msgs[i] = 0;
  71                         free(ftc_msgs[i]);
  72                         ftc_msgs[i] = NULL;
  73                 }
  74         }
  75         syslog(LOG_INFO, "fulltext: flushed index cache to disk (100%% complete)");
  76 }
  77
  78
  79 // Index or de-index a message.  (op == 1 to index, 0 to de-index)
  80 void ft_index_message(long msgnum, int op) {
  81         int num_tokens = 0;
  82         int *tokens = NULL;
  83         int i, j;
  84         struct cdbdata cdb_bucket;
  85         StrBuf *msgtext;
  86         char *txt;
  87         int tok;
  88         struct CtdlMessage *msg = NULL;
  89
  90         msg = CtdlFetchMessage(msgnum, 1);
  91         if (msg == NULL) {
  92                 syslog(LOG_ERR, "fulltext: ft_index_message() could not load msg %ld", msgnum);
  93                 return;
  94         }
  95
  96         if (!CM_IsEmpty(msg, eSuppressIdx)) {
  97                 syslog(LOG_DEBUG, "fulltext: ft_index_message() excluded msg %ld", msgnum);
  98                 CM_Free(msg);
  99                 return;
 100         }
 101
 102         syslog(LOG_DEBUG, "fulltext: ft_index_message() %s msg %ld", (op ? "adding" : "removing") , msgnum);
 103
 104         // Output the message as text before indexing it, so we don't end up
 105         // indexing a bunch of encoded base64, etc.
 106         CC->redirect_buffer = NewStrBufPlain(NULL, SIZ);
 107         CtdlOutputPreLoadedMsg(msg, MT_CITADEL, HEADERS_ALL, 0, 1, 0);
 108         CM_Free(msg);
 109         msgtext = CC->redirect_buffer;
 110         CC->redirect_buffer = NULL;
 111         if (msgtext != NULL) {
 112                 syslog(LOG_DEBUG, "fulltext: wordbreaking message %ld (%d bytes)", msgnum, StrLength(msgtext));
 113         }
 114         txt = SmashStrBuf(&msgtext);
 115         wordbreaker(txt, &num_tokens, &tokens);
 116         free(txt);
 117
 118         syslog(LOG_DEBUG, "fulltext: indexing message %ld [%d tokens]", msgnum, num_tokens);
 119         if (num_tokens > 0) {
 120                 for (i=0; i<num_tokens; ++i) {
 121
 122                         // Add the message to the relevant token bucket
 123
 124                         // search for tokens[i]
 125                         tok = tokens[i];
 126
 127                         if ( (tok >= 0) && (tok <= 65535) ) {
 128                                 // fetch the bucket, Liza
 129                                 if (ftc_msgs[tok] == NULL) {
 130                                         cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 131                                         if (cdb_bucket.ptr != NULL) {
 132                                                 ftc_num_msgs[tok] = cdb_bucket.len / sizeof(long);
 133                                                 ftc_msgs[tok] = malloc(cdb_bucket.len);
 134                                                 memcpy(ftc_msgs[tok], cdb_bucket.ptr, cdb_bucket.len);
 135                                         }
 136                                         else {
 137                                                 ftc_num_msgs[tok] = 0;
 138                                                 ftc_msgs[tok] = malloc(sizeof(long));
 139                                         }
 140                                 }
 141
 142                                 if (op == 1) {  // add to index
 143                                         ++ftc_num_msgs[tok];
 144                                         ftc_msgs[tok] = realloc(ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long));
 145                                         ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
 146                                 }
 147
 148                                 if (op == 0) {  // remove from index
 149                                         if (ftc_num_msgs[tok] >= 1) {
 150                                                 for (j=0; j<ftc_num_msgs[tok]; ++j) {
 151                                                         if (ftc_msgs[tok][j] == msgnum) {
 152                                                                 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
 153                                                                 --ftc_num_msgs[tok];
 154                                                                 --j;
 155                                                         }
 156                                                 }
 157                                         }
 158                                 }
 159                         }
 160                         else {
 161                                 syslog(LOG_ALERT, "fulltext: invalid token %d !!", tok);
 162                         }
 163                 }
 164
 165                 free(tokens);
 166         }
 167 }
 168
 169
 170 // Add a message to the list of those to be indexed.
 171 void ft_index_msg(long msgnum, void *userdata) {
 172
 173         if ((msgnum > CtdlGetConfigLong("MMfulltext")) && (msgnum <= ft_newhighest)) {
 174                 ++ft_num_msgs;
 175                 if (ft_num_msgs > ft_num_alloc) {
 176                         ft_num_alloc += 1024;
 177                         ft_newmsgs = realloc(ft_newmsgs, (ft_num_alloc * sizeof(long)));
 178                 }
 179                 ft_newmsgs[ft_num_msgs - 1] = msgnum;
 180         }
 181
 182 }
 183
 184
 185 // Scan a room for messages to index.
 186 void ft_index_room(struct ctdlroom *qrbuf, void *data) {
 187         if (server_shutting_down)
 188                 return;
 189
 190         CtdlGetRoom(&CC->room, qrbuf->QRname);
 191         CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
 192 }
 193
 194
 195 // Begin the fulltext indexing process.
 196 void do_fulltext_indexing(void) {
 197         int i;
 198         static time_t last_progress = 0L;
 199         static int is_running = 0;
 200         if (is_running) return;         // Concurrency check - only one can run
 201         is_running = 1;
 202
 203         // Don't do this if the site doesn't have it enabled.
 204         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 205                 return;
 206         }
 207
 208         // If we've switched wordbreaker modules, burn the index and start over.
 209         begin_critical_section(S_CONTROL);
 210         if (CtdlGetConfigInt("MM_fulltext_wordbreaker") != FT_WORDBREAKER_ID) {
 211                 syslog(LOG_DEBUG, "fulltext: wb ver on disk = %d, code ver = %d",
 212                         CtdlGetConfigInt("MM_fulltext_wordbreaker"), FT_WORDBREAKER_ID
 213                 );
 214                 syslog(LOG_INFO, "fulltext: (re)initializing index");
 215                 cdb_trunc(CDB_FULLTEXT);
 216                 CtdlSetConfigLong("MMfulltext", 0);
 217         }
 218         end_critical_section(S_CONTROL);
 219
 220         // Silently return if our fulltext index is up to date with new messages.
 221         if ((CtdlGetConfigLong("MMfulltext") >= CtdlGetConfigLong("MMhighest"))) {
 222                 return;         // nothing to do!
 223         }
 224
 225         // Now go through each room and find messages to index.
 226         ft_newhighest = CtdlGetConfigLong("MMhighest");
 227         CtdlForEachRoom(ft_index_room, NULL);   // load all msg pointers
 228
 229         if (ft_num_msgs > 0) {
 230                 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
 231                 for (i=0; i<(ft_num_msgs-1); ++i) { // purge dups
 232                         if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
 233                                 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1], ((ft_num_msgs - i - 1)*sizeof(long)));
 234                                 --ft_num_msgs;
 235                                 --i;
 236                         }
 237                 }
 238
 239                 // Here it is ... do each message!
 240                 for (i=0; i<ft_num_msgs; ++i) {
 241                         if (time(NULL) != last_progress) {
 242                                 syslog(LOG_DEBUG,
 243                                         "fulltext: indexed %d of %d messages (%d%%)", i, ft_num_msgs, ((i*100) / ft_num_msgs));
 244                                 last_progress = time(NULL);
 245                         }
 246                         ft_index_message(ft_newmsgs[i], 1);
 247
 248                         // Check to see if we need to quit early
 249                         if (server_shutting_down) {
 250                                 syslog(LOG_DEBUG, "fulltext: indexer quitting early");
 251                                 ft_newhighest = ft_newmsgs[i];
 252                                 break;
 253                         }
 254
 255                         // Check to see if we have to maybe flush to disk
 256                         if (i >= FT_MAX_CACHE) {
 257                                 syslog(LOG_DEBUG, "fulltext: time to flush.");
 258                                 ft_newhighest = ft_newmsgs[i];
 259                                 break;
 260                         }
 261
 262                 }
 263
 264                 free(ft_newmsgs);
 265                 ft_num_msgs = 0;
 266                 ft_num_alloc = 0;
 267                 ft_newmsgs = NULL;
 268         }
 269
 270         if (server_shutting_down) {
 271                 is_running = 0;
 272                 return;
 273         }
 274
 275         // Save our place so we don't have to do this again
 276         ft_flush_cache();
 277         begin_critical_section(S_CONTROL);
 278         CtdlSetConfigLong("MMfulltext", ft_newhighest);
 279         CtdlSetConfigInt("MM_fulltext_wordbreaker", FT_WORDBREAKER_ID);
 280         end_critical_section(S_CONTROL);
 281
 282         syslog(LOG_DEBUG, "fulltext: indexing finished");
 283         is_running = 0;
 284         return;
 285 }
 286
 287
 288 // API call to perform searches.
 289 // (This one does the "all of these words" search.)
 290 // Caller is responsible for freeing the message list.
 291 void ft_search(int *fts_num_msgs, long **fts_msgs, const char *search_string) {
 292         int num_tokens = 0;
 293         int *tokens = NULL;
 294         int i, j;
 295         struct cdbdata cdb_bucket;
 296         int num_all_msgs = 0;
 297         long *all_msgs = NULL;
 298         int num_ret_msgs = 0;
 299         int num_ret_alloc = 0;
 300         long *ret_msgs = NULL;
 301         int tok;
 302
 303         wordbreaker(search_string, &num_tokens, &tokens);
 304         if (num_tokens > 0) {
 305                 for (i=0; i<num_tokens; ++i) {
 306
 307                         // search for tokens[i]
 308                         tok = tokens[i];
 309
 310                         // fetch the bucket, Liza
 311                         if (ftc_msgs[tok] == NULL) {
 312                                 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
 313                                 if (cdb_bucket.ptr != NULL) {
 314                                         ftc_num_msgs[tok] = cdb_bucket.len / sizeof(long);
 315                                         ftc_msgs[tok] = (long *) malloc(cdb_bucket.len);
 316                                         memcpy(ftc_msgs[tok], cdb_bucket.ptr, cdb_bucket.len);
 317                                 }
 318                                 else {
 319                                         ftc_num_msgs[tok] = 0;
 320                                         ftc_msgs[tok] = malloc(sizeof(long));
 321                                 }
 322                         }
 323
 324                         num_all_msgs += ftc_num_msgs[tok];
 325                         if (num_all_msgs > 0) {
 326                                 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
 327                                 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]], ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
 328                         }
 329
 330                 }
 331                 free(tokens);
 332                 if (all_msgs != NULL) {
 333                         qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
 334
 335                         // At this point, if a message appears num_tokens times in the
 336                         // list, then it contains all of the search tokens.
 337                         if (num_all_msgs >= num_tokens)
 338                                 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
 339                                         if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
 340                                                 ++num_ret_msgs;
 341                                                 if (num_ret_msgs > num_ret_alloc) {
 342                                                         num_ret_alloc += 64;
 343                                                         ret_msgs = realloc(ret_msgs, (num_ret_alloc*sizeof(long)) );
 344                                                 }
 345                                                 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
 346
 347                                         }
 348                                 }
 349                         free(all_msgs);
 350                 }
 351         }
 352
 353         *fts_num_msgs = num_ret_msgs;
 354         *fts_msgs = ret_msgs;
 355 }
 356
 357
 358 // This search command is for diagnostic purposes and may be removed or replaced.
 359 void cmd_srch(char *argbuf) {
 360         int num_msgs = 0;
 361         long *msgs = NULL;
 362         int i;
 363         char search_string[256];
 364
 365         if (CtdlAccessCheck(ac_logged_in)) return;
 366
 367         if (!CtdlGetConfigInt("c_enable_fulltext")) {
 368                 cprintf("%d Full text index is not enabled on this server.\n", ERROR + CMD_NOT_SUPPORTED);
 369                 return;
 370         }
 371
 372         extract_token(search_string, argbuf, 0, '|', sizeof search_string);
 373         ft_search(&num_msgs, &msgs, search_string);
 374
 375         cprintf("%d %d msgs match all search words:\n",
 376                 LISTING_FOLLOWS, num_msgs);
 377         if (num_msgs > 0) {
 378                 for (i=0; i<num_msgs; ++i) {
 379                         cprintf("%ld\n", msgs[i]);
 380                 }
 381         }
 382         if (msgs != NULL) free(msgs);
 383         cprintf("000\n");
 384 }
 385
 386
 387 // Zero out our index cache.
 388 void initialize_ft_cache(void) {
 389         memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
 390         memset(ftc_msgs, 0, (65536 * sizeof(long *)));
 391 }
 392
 393
 394 void ft_delete_remove(char *room, long msgnum) {
 395         if (room) return;
 396
 397         // Remove from fulltext index
 398         if (CtdlGetConfigInt("c_enable_fulltext")) {
 399                 ft_index_message(msgnum, 0);
 400         }
 401 }
 402
 403
 404 // Initialization function, called from modules_init.c
 405 char *ctdl_module_init_fulltext(void) {
 406         if (!threading) {
 407                 initialize_ft_cache();
 408                 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
 409                 CtdlRegisterDeleteHook(ft_delete_remove);
 410                 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
 411                 CtdlRegisterSessionHook(do_fulltext_indexing, EVT_TIMER, PRIO_CLEANUP + 300);
 412         }
 413         // return our module name for the log
 414         return "fulltext";
 415 }