4 * This module handles fulltext indexing of the message base.
17 #include <sys/types.h>
19 #if TIME_WITH_SYS_TIME
20 # include <sys/time.h>
24 # include <sys/time.h>
33 #include <libcitadel.h>
36 #include "citserver.h"
43 #include "serv_fulltext.h"
44 #include "ft_wordbreaker.h"
47 #include "ctdl_module.h"
51 long ft_newhighest = 0L;
52 long *ft_newmsgs = NULL;
57 int ftc_num_msgs[65536];
58 long *ftc_msgs[65536];
64 int longcmp(const void *rec1, const void *rec2) {
67 i1 = *(const long *)rec1;
68 i2 = *(const long *)rec2;
70 if (i1 > i2) return(1);
71 if (i1 < i2) return(-1);
76 * Flush our index cache out to disk.
78 void ft_flush_cache(void) {
80 time_t last_update = 0;
82 for (i=0; i<65536; ++i) {
83 if ((time(NULL) - last_update) >= 10) {
84 CtdlLogPrintf(CTDL_INFO,
85 "Flushing index cache to disk (%d%% complete)\n",
88 last_update = time(NULL);
90 if (ftc_msgs[i] != NULL) {
91 cdb_store(CDB_FULLTEXT, &i, sizeof(int), ftc_msgs[i],
92 (ftc_num_msgs[i] * sizeof(long)));
98 CtdlLogPrintf(CTDL_INFO, "Flushed index cache to disk (100%% complete)\n");
103 * Index or de-index a message. (op == 1 to index, 0 to de-index)
105 void ft_index_message(long msgnum, int op) {
109 struct cdbdata *cdb_bucket;
113 CtdlLogPrintf(CTDL_DEBUG, "ft_index_message() %s msg %ld\n",
114 (op ? "adding" : "removing") , msgnum
117 /* Output the message as text before indexing it, so we don't end up
118 * indexing a bunch of encoded base64, etc.
120 CC->redirect_buffer = malloc(SIZ);
121 CC->redirect_len = 0;
122 CC->redirect_alloc = SIZ;
123 CtdlOutputMsg(msgnum, MT_CITADEL, HEADERS_ALL, 0, 1, NULL, 0);
124 msgtext = CC->redirect_buffer;
125 CC->redirect_buffer = NULL;
126 CC->redirect_len = 0;
127 CC->redirect_alloc = 0;
128 CtdlLogPrintf(CTDL_DEBUG, "Wordbreaking message %ld...\n", msgnum);
129 wordbreaker(msgtext, &num_tokens, &tokens);
132 CtdlLogPrintf(CTDL_DEBUG, "Indexing message %ld [%d tokens]\n", msgnum, num_tokens);
133 if (num_tokens > 0) {
134 for (i=0; i<num_tokens; ++i) {
136 /* Add the message to the relevant token bucket */
138 /* search for tokens[i] */
141 if ( (tok >= 0) && (tok <= 65535) ) {
142 /* fetch the bucket, Liza */
143 if (ftc_msgs[tok] == NULL) {
144 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
145 if (cdb_bucket != NULL) {
146 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
147 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
148 cdb_bucket->ptr = NULL;
149 cdb_free(cdb_bucket);
152 ftc_num_msgs[tok] = 0;
153 ftc_msgs[tok] = malloc(sizeof(long));
158 if (op == 1) { /* add to index */
160 ftc_msgs[tok] = realloc(ftc_msgs[tok],
161 ftc_num_msgs[tok]*sizeof(long));
162 ftc_msgs[tok][ftc_num_msgs[tok] - 1] = msgnum;
165 if (op == 0) { /* remove from index */
166 if (ftc_num_msgs[tok] >= 1) {
167 for (j=0; j<ftc_num_msgs[tok]; ++j) {
168 if (ftc_msgs[tok][j] == msgnum) {
169 memmove(&ftc_msgs[tok][j], &ftc_msgs[tok][j+1], ((ftc_num_msgs[tok] - j - 1)*sizeof(long)));
178 CtdlLogPrintf(CTDL_ALERT, "Invalid token %d !!\n", tok);
189 * Add a message to the list of those to be indexed.
191 void ft_index_msg(long msgnum, void *userdata) {
193 if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
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)));
200 ft_newmsgs[ft_num_msgs - 1] = msgnum;
206 * Scan a room for messages to index.
208 void ft_index_room(struct ctdlroom *qrbuf, void *data)
210 if (CtdlThreadCheckStop())
213 getroom(&CC->room, qrbuf->QRname);
214 CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, NULL, ft_index_msg, NULL);
219 * Begin the fulltext indexing process. (Called as an EVT_TIMER event)
221 void do_fulltext_indexing(void) {
223 static time_t last_index = 0L;
224 static time_t last_progress = 0L;
225 time_t run_time = 0L;
226 time_t end_time = 0L;
229 * Don't do this if the site doesn't have it enabled.
231 if (!config.c_enable_fulltext) {
236 * Make sure we don't run the indexer too frequently.
237 * FIXME move the setting into config
240 * The thread sleeps for 300 seconds so we don't need this here any more
242 if ( (time(NULL) - last_index) < 300L) {
248 * Check to see whether the fulltext index is up to date; if there
249 * are no messages to index, don't waste any more time trying.
251 if ((CitControl.MMfulltext >= CitControl.MMhighest) && (CitControl.fulltext_wordbreaker == FT_WORDBREAKER_ID)) {
252 return; /* nothing to do! */
255 run_time = time(NULL);
256 CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() started (%ld)\n", run_time);
259 * If we've switched wordbreaker modules, burn the index and start
262 begin_critical_section(S_CONTROL);
263 if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
264 CtdlLogPrintf(CTDL_DEBUG, "wb ver on disk = %d, code ver = %d\n",
265 CitControl.fulltext_wordbreaker, FT_WORDBREAKER_ID);
266 CtdlLogPrintf(CTDL_INFO, "(re)initializing full text index\n");
267 cdb_trunc(CDB_FULLTEXT);
268 CitControl.MMfulltext = 0L;
271 end_critical_section(S_CONTROL);
274 * Now go through each room and find messages to index.
276 ft_newhighest = CitControl.MMhighest;
277 ForEachRoom(ft_index_room, NULL); /* load all msg pointers */
279 if (ft_num_msgs > 0) {
280 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
281 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
282 if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
283 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
284 ((ft_num_msgs - i - 1)*sizeof(long)));
290 /* Here it is ... do each message! */
291 for (i=0; i<ft_num_msgs; ++i) {
292 if (time(NULL) != last_progress) {
293 CtdlLogPrintf(CTDL_DEBUG,
294 "Indexed %d of %d messages (%d%%)\n",
296 ((i*100) / ft_num_msgs)
298 last_progress = time(NULL);
300 ft_index_message(ft_newmsgs[i], 1);
302 /* Check to see if we need to quit early */
303 if (CtdlThreadCheckStop()) {
304 CtdlLogPrintf(CTDL_DEBUG, "Indexer quitting early\n");
305 ft_newhighest = ft_newmsgs[i];
309 /* Check to see if we have to maybe flush to disk */
310 if (i >= FT_MAX_CACHE) {
311 CtdlLogPrintf(CTDL_DEBUG, "Time to flush.\n");
312 ft_newhighest = ft_newmsgs[i];
323 end_time = time(NULL);
325 if (CtdlThreadCheckStop())
328 CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() duration (%ld)\n", end_time - run_time);
330 /* Save our place so we don't have to do this again */
332 begin_critical_section(S_CONTROL);
333 CitControl.MMfulltext = ft_newhighest;
334 CitControl.fulltext_wordbreaker = FT_WORDBREAKER_ID;
336 end_critical_section(S_CONTROL);
337 last_index = time(NULL);
339 CtdlLogPrintf(CTDL_DEBUG, "do_fulltext_indexing() finished\n");
344 * Main loop for the indexer thread.
346 void *indexer_thread(void *arg) {
347 struct CitContext indexerCC;
349 CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() initializing\n");
351 CtdlFillPrivateContext(&indexerCC, "indexer");
352 citthread_setspecific(MyConKey, (void *)&indexerCC );
354 while (!CtdlThreadCheckStop()) {
355 do_fulltext_indexing();
356 CtdlThreadSleep(300);
359 CtdlLogPrintf(CTDL_DEBUG, "indexer_thread() exiting\n");
366 * API call to perform searches.
367 * (This one does the "all of these words" search.)
368 * Caller is responsible for freeing the message list.
370 void ft_search(int *fts_num_msgs, long **fts_msgs, char *search_string) {
374 struct cdbdata *cdb_bucket;
375 int num_all_msgs = 0;
376 long *all_msgs = NULL;
377 int num_ret_msgs = 0;
378 int num_ret_alloc = 0;
379 long *ret_msgs = NULL;
382 wordbreaker(search_string, &num_tokens, &tokens);
383 if (num_tokens > 0) {
384 for (i=0; i<num_tokens; ++i) {
386 /* search for tokens[i] */
389 /* fetch the bucket, Liza */
390 if (ftc_msgs[tok] == NULL) {
391 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tok, sizeof(int));
392 if (cdb_bucket != NULL) {
393 ftc_num_msgs[tok] = cdb_bucket->len / sizeof(long);
394 ftc_msgs[tok] = (long *)cdb_bucket->ptr;
395 cdb_bucket->ptr = NULL;
396 cdb_free(cdb_bucket);
399 ftc_num_msgs[tok] = 0;
400 ftc_msgs[tok] = malloc(sizeof(long));
404 num_all_msgs += ftc_num_msgs[tok];
405 if (num_all_msgs > 0) {
406 all_msgs = realloc(all_msgs, num_all_msgs*sizeof(long) );
407 memcpy(&all_msgs[num_all_msgs-ftc_num_msgs[tok]],
408 ftc_msgs[tok], ftc_num_msgs[tok]*sizeof(long) );
413 qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
416 * At this point, if a message appears num_tokens times in the
417 * list, then it contains all of the search tokens.
419 if (num_all_msgs >= num_tokens)
420 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
421 if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
424 if (num_ret_msgs > num_ret_alloc) {
426 ret_msgs = realloc(ret_msgs,
427 (num_ret_alloc*sizeof(long)) );
429 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
437 *fts_num_msgs = num_ret_msgs;
438 *fts_msgs = ret_msgs;
443 * This search command is for diagnostic purposes and may be removed or replaced.
445 void cmd_srch(char *argbuf) {
449 char search_string[256];
451 if (CtdlAccessCheck(ac_logged_in)) return;
453 if (!config.c_enable_fulltext) {
454 cprintf("%d Full text index is not enabled on this server.\n",
455 ERROR + CMD_NOT_SUPPORTED);
459 extract_token(search_string, argbuf, 0, '|', sizeof search_string);
460 ft_search(&num_msgs, &msgs, search_string);
462 cprintf("%d %d msgs match all search words:\n",
463 LISTING_FOLLOWS, num_msgs);
465 for (i=0; i<num_msgs; ++i) {
466 cprintf("%ld\n", msgs[i]);
469 if (msgs != NULL) free(msgs);
474 * Zero out our index cache.
476 void initialize_ft_cache(void) {
477 memset(ftc_num_msgs, 0, (65536 * sizeof(int)));
478 memset(ftc_msgs, 0, (65536 * sizeof(long *)));
482 void ft_delete_remove(char *room, long msgnum)
486 /* Remove from fulltext index */
487 if (config.c_enable_fulltext) {
488 ft_index_message(msgnum, 0);
492 /*****************************************************************************/
494 CTDL_MODULE_INIT(fulltext)
498 initialize_ft_cache();
499 initialize_noise_words();
500 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");
501 CtdlRegisterDeleteHook(ft_delete_remove);
502 CtdlRegisterSearchFuncHook(ft_search, "fulltext");
503 CtdlRegisterCleanupHook(noise_word_cleanup);
507 CtdlThreadCreate("Indexer", CTDLTHREAD_BIGSTACK, indexer_thread, NULL);
509 /* return our Subversion id for the Log */