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>
35 #include "sysdep_decls.h"
36 #include "citserver.h"
39 #include "serv_extensions.h"
45 #include "serv_fulltext.h"
46 #include "ft_wordbreaker.h"
49 long ft_newhighest = 0L;
50 long *ft_newmsgs = NULL;
58 int longcmp(const void *rec1, const void *rec2) {
61 i1 = *(const long *)rec1;
62 i2 = *(const long *)rec2;
64 if (i1 > i2) return(1);
65 if (i1 < i2) return(-1);
73 * Index or de-index a message. (op == 1 to index, 0 to de-index)
75 void ft_index_message(long msgnum, int op) {
79 struct cdbdata *cdb_bucket;
84 lprintf(CTDL_DEBUG, "ft_index_message() %s msg %ld\n",
85 (op ? "adding" : "removing") , msgnum
88 /* Output the message as text before indexing it, so we don't end up
89 * indexing a bunch of encoded base64, etc.
91 CC->redirect_buffer = malloc(SIZ);
93 CC->redirect_alloc = SIZ;
94 CtdlOutputMsg(msgnum, MT_CITADEL, HEADERS_ALL, 0, 1);
95 msgtext = CC->redirect_buffer;
96 CC->redirect_buffer = NULL;
98 CC->redirect_alloc = 0;
99 lprintf(CTDL_DEBUG, "Wordbreaking message %ld...\n", msgnum);
100 wordbreaker(msgtext, &num_tokens, &tokens);
103 lprintf(CTDL_DEBUG, "Indexing message %ld...\n", msgnum);
104 if (num_tokens > 0) {
105 for (i=0; i<num_tokens; ++i) {
107 /* Add the message to the relevant token bucket */
109 /* FIXME do we need to lock the file here? */
110 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tokens[i],
112 if (cdb_bucket == NULL) {
113 cdb_bucket = malloc(sizeof(struct cdbdata));
115 cdb_bucket->ptr = NULL;
119 num_msgs = cdb_bucket->len / sizeof(long);
122 if (op == 1) { /* add to index */
124 cdb_bucket->ptr = realloc(cdb_bucket->ptr,
125 num_msgs*sizeof(long));
126 msgs = (long *) cdb_bucket->ptr;
127 msgs[num_msgs - 1] = msgnum;
130 if (op == 0) { /* remove from index */
132 msgs = (long *) cdb_bucket->ptr;
133 for (j=0; j<num_msgs; ++j) {
134 if (msgs[j] == msgnum) {
135 memmove(&msgs[j], &msgs[j+1], ((num_msgs - j - 1)*sizeof(long)));
143 cdb_store(CDB_FULLTEXT, &tokens[i], sizeof(int),
144 msgs, (num_msgs*sizeof(long)) );
146 cdb_free(cdb_bucket);
148 /* FIXME do we need to unlock the file here? */
158 * Add a message to the list of those to be indexed.
160 void ft_index_msg(long msgnum, void *userdata) {
162 if ((msgnum > CitControl.MMfulltext) && (msgnum <= ft_newhighest)) {
164 if (ft_num_msgs > ft_num_alloc) {
165 ft_num_alloc += 1024;
166 ft_newmsgs = realloc(ft_newmsgs,
167 (ft_num_alloc * sizeof(long)));
169 ft_newmsgs[ft_num_msgs - 1] = msgnum;
175 * Scan a room for messages to index.
177 void ft_index_room(struct ctdlroom *qrbuf, void *data)
179 getroom(&CC->room, qrbuf->QRname);
180 CtdlForEachMessage(MSGS_ALL, 0L, NULL, NULL, ft_index_msg, NULL);
185 * Begin the fulltext indexing process. (Called as an EVT_TIMER event)
187 void do_fulltext_indexing(void) {
189 static time_t last_index = 0L;
190 static time_t last_progress = 0L;
193 * Don't do this if the site doesn't have it enabled.
195 if (!config.c_enable_fulltext) {
200 * Make sure we don't run the indexer too frequently.
201 * FIXME move the setting into config
203 if ( (time(NULL) - last_index) < 300L) {
208 * Check to see whether the fulltext index is up to date; if there
209 * are no messages to index, don't waste any more time trying.
211 if (CitControl.MMfulltext >= CitControl.MMhighest) {
212 return; /* nothing to do! */
215 lprintf(CTDL_DEBUG, "do_fulltext_indexing() started\n");
218 * If we've switched wordbreaker modules, burn the index and start
221 if (CitControl.fulltext_wordbreaker != FT_WORDBREAKER_ID) {
222 lprintf(CTDL_INFO, "(re)initializing full text index\n");
223 cdb_trunc(CDB_FULLTEXT);
224 CitControl.MMfulltext = 0L;
229 * Now go through each room and find messages to index.
231 ft_newhighest = CitControl.MMhighest;
232 ForEachRoom(ft_index_room, NULL); /* load all msg pointers */
234 if (ft_num_msgs > 0) {
235 qsort(ft_newmsgs, ft_num_msgs, sizeof(long), longcmp);
236 for (i=0; i<(ft_num_msgs-1); ++i) { /* purge dups */
237 if (ft_newmsgs[i] == ft_newmsgs[i+1]) {
238 memmove(&ft_newmsgs[i], &ft_newmsgs[i+1],
239 ((ft_num_msgs - i - 1)*sizeof(long)));
245 /* Here it is ... do each message! */
246 for (i=0; i<ft_num_msgs; ++i) {
247 if ((time(NULL) - last_progress) > 10) {
249 "Indexed %d of %d messages (%d%%)\n",
251 ((i*100) / ft_num_msgs)
253 last_progress = time(NULL);
255 ft_index_message(ft_newmsgs[i], 1);
257 /* Check to see if we need to quit early */
259 lprintf(CTDL_DEBUG, "Indexer quitting early\n");
260 ft_newhighest = ft_newmsgs[i];
272 /* Save our place so we don't have to do this again */
273 CitControl.MMfulltext = ft_newhighest;
274 CitControl.fulltext_wordbreaker = FT_WORDBREAKER_ID;
276 last_index = time(NULL);
278 lprintf(CTDL_DEBUG, "do_fulltext_indexing() finished\n");
283 * Main loop for the indexer thread.
285 void *indexer_thread(void *arg) {
286 struct CitContext indexerCC;
288 lprintf(CTDL_DEBUG, "indexer_thread() initializing\n");
290 memset(&indexerCC, 0, sizeof(struct CitContext));
291 indexerCC.internal_pgm = 1;
292 indexerCC.cs_pid = 0;
293 pthread_setspecific(MyConKey, (void *)&indexerCC );
297 while (!time_to_die) {
298 do_fulltext_indexing();
302 lprintf(CTDL_DEBUG, "indexer_thread() exiting\n");
309 * API call to perform searches.
310 * (This one does the "all of these words" search.)
311 * Caller is responsible for freeing the message list.
313 void ft_search(int *fts_num_msgs, long **fts_msgs, char *search_string) {
317 struct cdbdata *cdb_bucket;
320 int num_all_msgs = 0;
321 long *all_msgs = NULL;
322 int num_ret_msgs = 0;
323 int num_ret_alloc = 0;
324 long *ret_msgs = NULL;
326 wordbreaker(search_string, &num_tokens, &tokens);
327 if (num_tokens > 0) {
328 for (i=0; i<num_tokens; ++i) {
330 /* search for tokens[i] */
331 cdb_bucket = cdb_fetch(CDB_FULLTEXT, &tokens[i],
333 if (cdb_bucket != NULL) {
334 num_msgs = cdb_bucket->len / sizeof(long);
335 msgs = (long *)cdb_bucket->ptr;
337 num_all_msgs += num_msgs;
338 if (num_all_msgs > 0) {
339 all_msgs = realloc(all_msgs,
340 num_all_msgs*sizeof(long) );
341 memcpy(&all_msgs[num_all_msgs-num_msgs],
342 msgs, num_msgs*sizeof(long) );
345 cdb_free(cdb_bucket);
350 qsort(all_msgs, num_all_msgs, sizeof(long), longcmp);
353 * At this point, if a message appears num_tokens times in the
354 * list, then it contains all of the search tokens.
356 if (num_all_msgs >= num_tokens)
357 for (j=0; j<(num_all_msgs-num_tokens+1); ++j) {
358 if (all_msgs[j] == all_msgs[j+num_tokens-1]) {
361 if (num_ret_msgs > num_ret_alloc) {
363 ret_msgs = realloc(ret_msgs,
364 (num_ret_alloc*sizeof(long)) );
366 ret_msgs[num_ret_msgs - 1] = all_msgs[j];
374 *fts_num_msgs = num_ret_msgs;
375 *fts_msgs = ret_msgs;
380 * Tentative form of a search command
382 void cmd_srch(char *argbuf) {
386 char search_string[256];
388 if (CtdlAccessCheck(ac_logged_in)) return;
390 if (!config.c_enable_fulltext) {
391 cprintf("%d Full text index is not enabled on this server.\n",
392 ERROR + CMD_NOT_SUPPORTED);
395 extract_token(search_string, argbuf, 0, '|', sizeof search_string);
396 ft_search(&num_msgs, &msgs, search_string);
398 cprintf("%d %d msgs match all search words:\n",
399 LISTING_FOLLOWS, num_msgs);
401 for (i=0; i<num_msgs; ++i) {
402 cprintf("%ld\n", msgs[i]);
405 if (msgs != NULL) free(msgs);
410 /*****************************************************************************/
412 char *serv_fulltext_init(void)
414 CtdlRegisterProtoHook(cmd_srch, "SRCH", "Full text search");