1 /* query.cc - Support for searching a notmuch database
3 * Copyright © 2009 Carl Worth
5 * This program is free software: you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation, either version 3 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program. If not, see https://www.gnu.org/licenses/ .
18 * Author: Carl Worth <cworth@cworth.org>
21 #include "notmuch-private.h"
22 #include "database-private.h"
24 #include <glib.h> /* GHashTable, GPtrArray */
26 struct _notmuch_query {
27 notmuch_database_t *notmuch;
28 const char *query_string;
30 notmuch_string_list_t *exclude_terms;
31 notmuch_exclude_t omit_excluded;
33 Xapian::Query xapian_query;
34 std::set<std::string> terms;
37 typedef struct _notmuch_mset_messages {
38 notmuch_messages_t base;
39 notmuch_database_t *notmuch;
40 Xapian::MSetIterator iterator;
41 Xapian::MSetIterator iterator_end;
42 } notmuch_mset_messages_t;
44 struct _notmuch_doc_id_set {
45 unsigned char *bitmap;
49 #define DOCIDSET_WORD(bit) ((bit) / CHAR_BIT)
50 #define DOCIDSET_BIT(bit) ((bit) % CHAR_BIT)
52 struct _notmuch_threads {
53 notmuch_query_t *query;
55 /* The ordered list of doc ids matched by the query. */
57 /* Our iterator's current position in doc_ids. */
58 unsigned int doc_id_pos;
59 /* The set of matched docid's that have not been assigned to a
60 * thread. Initially, this contains every docid in doc_ids. */
61 notmuch_doc_id_set_t match_set;
64 /* We need this in the message functions so forward declare. */
66 _notmuch_doc_id_set_init (void *ctx,
67 notmuch_doc_id_set_t *doc_ids,
73 char *env = getenv ("NOTMUCH_DEBUG_QUERY");
75 return (env && strcmp (env, "") != 0);
78 /* Explicit destructor call for placement new */
80 _notmuch_query_destructor (notmuch_query_t *query)
82 query->xapian_query.~Query();
83 query->terms.~set<std::string>();
88 notmuch_query_create (notmuch_database_t *notmuch,
89 const char *query_string)
91 notmuch_query_t *query;
94 fprintf (stderr, "Query string is:\n%s\n", query_string);
96 query = talloc (notmuch, notmuch_query_t);
97 if (unlikely (query == NULL))
100 new (&query->xapian_query) Xapian::Query ();
101 new (&query->terms) std::set<std::string> ();
102 query->parsed = false;
104 talloc_set_destructor (query, _notmuch_query_destructor);
106 query->notmuch = notmuch;
108 query->query_string = talloc_strdup (query, query_string);
110 query->sort = NOTMUCH_SORT_NEWEST_FIRST;
112 query->exclude_terms = _notmuch_string_list_create (query);
114 query->omit_excluded = NOTMUCH_EXCLUDE_TRUE;
119 static notmuch_status_t
120 _notmuch_query_ensure_parsed (notmuch_query_t *query)
123 return NOTMUCH_STATUS_SUCCESS;
126 query->xapian_query =
127 query->notmuch->query_parser->
128 parse_query (query->query_string, NOTMUCH_QUERY_PARSER_FLAGS);
130 /* Xapian doesn't support skip_to on terms from a query since
131 * they are unordered, so cache a copy of all terms in
132 * something searchable.
135 for (Xapian::TermIterator t = query->xapian_query.get_terms_begin ();
136 t != query->xapian_query.get_terms_end (); ++t)
137 query->terms.insert (*t);
139 query->parsed = true;
141 } catch (const Xapian::Error &error) {
142 if (! query->notmuch->exception_reported) {
143 _notmuch_database_log (query->notmuch,
144 "A Xapian exception occurred parsing query: %s\n",
145 error.get_msg ().c_str ());
146 _notmuch_database_log_append (query->notmuch,
147 "Query string was: %s\n",
148 query->query_string);
149 query->notmuch->exception_reported = true;
152 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
154 return NOTMUCH_STATUS_SUCCESS;
158 notmuch_query_get_query_string (const notmuch_query_t *query)
160 return query->query_string;
164 notmuch_query_set_omit_excluded (notmuch_query_t *query,
165 notmuch_exclude_t omit_excluded)
167 query->omit_excluded = omit_excluded;
171 notmuch_query_set_sort (notmuch_query_t *query, notmuch_sort_t sort)
177 notmuch_query_get_sort (const notmuch_query_t *query)
183 notmuch_query_add_tag_exclude (notmuch_query_t *query, const char *tag)
185 notmuch_status_t status;
188 status = _notmuch_query_ensure_parsed (query);
192 term = talloc_asprintf (query, "%s%s", _find_prefix ("tag"), tag);
193 if (query->terms.count (term) != 0)
194 return NOTMUCH_STATUS_IGNORED;
196 _notmuch_string_list_append (query->exclude_terms, term);
197 return NOTMUCH_STATUS_SUCCESS;
200 /* We end up having to call the destructors explicitly because we had
201 * to use "placement new" in order to initialize C++ objects within a
202 * block that we allocated with talloc. So C++ is making talloc
203 * slightly less simple to use, (we wouldn't need
204 * talloc_set_destructor at all otherwise).
207 _notmuch_messages_destructor (notmuch_mset_messages_t *messages)
209 messages->iterator.~MSetIterator ();
210 messages->iterator_end.~MSetIterator ();
215 /* Return a query that matches messages with the excluded tags
216 * registered with query. The caller of this function has to combine the returned
217 * query appropriately.*/
219 _notmuch_exclude_tags (notmuch_query_t *query)
221 Xapian::Query exclude_query = Xapian::Query::MatchNothing;
223 for (notmuch_string_node_t *term = query->exclude_terms->head; term;
225 exclude_query = Xapian::Query (Xapian::Query::OP_OR,
226 exclude_query, Xapian::Query (term->string));
228 return exclude_query;
233 notmuch_query_search_messages_st (notmuch_query_t *query,
234 notmuch_messages_t **out)
236 return notmuch_query_search_messages (query, out);
240 notmuch_query_search_messages (notmuch_query_t *query,
241 notmuch_messages_t **out)
243 return _notmuch_query_search_documents (query, "mail", out);
247 _notmuch_query_search_documents (notmuch_query_t *query,
249 notmuch_messages_t **out)
251 notmuch_database_t *notmuch = query->notmuch;
252 const char *query_string = query->query_string;
253 notmuch_mset_messages_t *messages;
254 notmuch_status_t status;
256 status = _notmuch_query_ensure_parsed (query);
260 messages = talloc (query, notmuch_mset_messages_t);
261 if (unlikely (messages == NULL))
262 return NOTMUCH_STATUS_OUT_OF_MEMORY;
266 messages->base.is_of_list_type = false;
267 messages->base.iterator = NULL;
268 messages->notmuch = notmuch;
269 new (&messages->iterator) Xapian::MSetIterator ();
270 new (&messages->iterator_end) Xapian::MSetIterator ();
272 talloc_set_destructor (messages, _notmuch_messages_destructor);
274 Xapian::Enquire enquire (*notmuch->xapian_db);
275 Xapian::Query mail_query (talloc_asprintf (query, "%s%s",
276 _find_prefix ("type"),
278 Xapian::Query final_query, exclude_query;
280 Xapian::MSetIterator iterator;
282 if (strcmp (query_string, "") == 0 ||
283 strcmp (query_string, "*") == 0) {
284 final_query = mail_query;
286 final_query = Xapian::Query (Xapian::Query::OP_AND,
287 mail_query, query->xapian_query);
289 messages->base.excluded_doc_ids = NULL;
291 if ((query->omit_excluded != NOTMUCH_EXCLUDE_FALSE) && (query->exclude_terms)) {
292 exclude_query = _notmuch_exclude_tags (query);
294 if (query->omit_excluded == NOTMUCH_EXCLUDE_TRUE ||
295 query->omit_excluded == NOTMUCH_EXCLUDE_ALL) {
296 final_query = Xapian::Query (Xapian::Query::OP_AND_NOT,
297 final_query, exclude_query);
298 } else { /* NOTMUCH_EXCLUDE_FLAG */
299 exclude_query = Xapian::Query (Xapian::Query::OP_AND,
300 exclude_query, final_query);
302 enquire.set_weighting_scheme (Xapian::BoolWeight ());
303 enquire.set_query (exclude_query);
305 mset = enquire.get_mset (0, notmuch->xapian_db->get_doccount ());
307 GArray *excluded_doc_ids = g_array_new (false, false, sizeof (unsigned int));
309 for (iterator = mset.begin (); iterator != mset.end (); iterator++) {
310 unsigned int doc_id = *iterator;
311 g_array_append_val (excluded_doc_ids, doc_id);
313 messages->base.excluded_doc_ids = talloc (messages, _notmuch_doc_id_set);
314 _notmuch_doc_id_set_init (query, messages->base.excluded_doc_ids,
316 g_array_unref (excluded_doc_ids);
321 enquire.set_weighting_scheme (Xapian::BoolWeight ());
323 switch (query->sort) {
324 case NOTMUCH_SORT_OLDEST_FIRST:
325 enquire.set_sort_by_value (NOTMUCH_VALUE_TIMESTAMP, false);
327 case NOTMUCH_SORT_NEWEST_FIRST:
328 enquire.set_sort_by_value (NOTMUCH_VALUE_TIMESTAMP, true);
330 case NOTMUCH_SORT_MESSAGE_ID:
331 enquire.set_sort_by_value (NOTMUCH_VALUE_MESSAGE_ID, false);
333 case NOTMUCH_SORT_UNSORTED:
337 if (_debug_query ()) {
338 fprintf (stderr, "Exclude query is:\n%s\n",
339 exclude_query.get_description ().c_str ());
340 fprintf (stderr, "Final query is:\n%s\n",
341 final_query.get_description ().c_str ());
344 enquire.set_query (final_query);
346 mset = enquire.get_mset (0, notmuch->xapian_db->get_doccount ());
348 messages->iterator = mset.begin ();
349 messages->iterator_end = mset.end ();
351 *out = &messages->base;
352 return NOTMUCH_STATUS_SUCCESS;
354 } catch (const Xapian::Error &error) {
355 _notmuch_database_log (notmuch,
356 "A Xapian exception occurred performing query: %s\n",
357 error.get_msg ().c_str ());
358 _notmuch_database_log_append (notmuch,
359 "Query string was: %s\n",
360 query->query_string);
362 notmuch->exception_reported = true;
363 talloc_free (messages);
364 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
369 _notmuch_mset_messages_valid (notmuch_messages_t *messages)
371 notmuch_mset_messages_t *mset_messages;
373 mset_messages = (notmuch_mset_messages_t *) messages;
375 return (mset_messages->iterator != mset_messages->iterator_end);
379 _notmuch_mset_messages_get_doc_id (notmuch_messages_t *messages)
381 notmuch_mset_messages_t *mset_messages;
383 mset_messages = (notmuch_mset_messages_t *) messages;
385 if (! _notmuch_mset_messages_valid (&mset_messages->base))
388 return *mset_messages->iterator;
392 _notmuch_mset_messages_get (notmuch_messages_t *messages)
394 notmuch_message_t *message;
395 Xapian::docid doc_id;
396 notmuch_private_status_t status;
397 notmuch_mset_messages_t *mset_messages;
399 mset_messages = (notmuch_mset_messages_t *) messages;
401 if (! _notmuch_mset_messages_valid (&mset_messages->base))
404 doc_id = *mset_messages->iterator;
406 message = _notmuch_message_create (mset_messages,
407 mset_messages->notmuch, doc_id,
410 if (message == NULL &&
411 status == NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND) {
412 INTERNAL_ERROR ("a messages iterator contains a non-existent document ID.\n");
415 if (messages->excluded_doc_ids &&
416 _notmuch_doc_id_set_contains (messages->excluded_doc_ids, doc_id))
417 notmuch_message_set_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED, true);
423 _notmuch_mset_messages_move_to_next (notmuch_messages_t *messages)
425 notmuch_mset_messages_t *mset_messages;
427 mset_messages = (notmuch_mset_messages_t *) messages;
429 mset_messages->iterator++;
433 _notmuch_doc_id_set_init (void *ctx,
434 notmuch_doc_id_set_t *doc_ids,
437 unsigned int max = 0;
438 unsigned char *bitmap;
440 for (unsigned int i = 0; i < arr->len; i++)
441 max = MAX (max, g_array_index (arr, unsigned int, i));
442 bitmap = talloc_zero_array (ctx, unsigned char, DOCIDSET_WORD (max) + 1);
447 doc_ids->bitmap = bitmap;
448 doc_ids->bound = max + 1;
450 for (unsigned int i = 0; i < arr->len; i++) {
451 unsigned int doc_id = g_array_index (arr, unsigned int, i);
452 bitmap[DOCIDSET_WORD (doc_id)] |= 1 << DOCIDSET_BIT (doc_id);
459 _notmuch_doc_id_set_contains (notmuch_doc_id_set_t *doc_ids,
462 if (doc_id >= doc_ids->bound)
464 return doc_ids->bitmap[DOCIDSET_WORD (doc_id)] & (1 << DOCIDSET_BIT (doc_id));
468 _notmuch_doc_id_set_remove (notmuch_doc_id_set_t *doc_ids,
471 if (doc_id < doc_ids->bound)
472 doc_ids->bitmap[DOCIDSET_WORD (doc_id)] &= ~(1 << DOCIDSET_BIT (doc_id));
475 /* Glib objects force use to use a talloc destructor as well, (but not
476 * nearly as ugly as the for messages due to C++ objects). At
477 * this point, I'd really like to have some talloc-friendly
478 * equivalents for the few pieces of glib that I'm using. */
480 _notmuch_threads_destructor (notmuch_threads_t *threads)
482 if (threads->doc_ids)
483 g_array_unref (threads->doc_ids);
489 notmuch_query_search_threads_st (notmuch_query_t *query, notmuch_threads_t **out)
491 return notmuch_query_search_threads (query, out);
495 notmuch_query_search_threads (notmuch_query_t *query,
496 notmuch_threads_t **out)
498 notmuch_threads_t *threads;
499 notmuch_messages_t *messages;
500 notmuch_status_t status;
502 threads = talloc (query, notmuch_threads_t);
504 return NOTMUCH_STATUS_OUT_OF_MEMORY;
505 threads->doc_ids = NULL;
506 talloc_set_destructor (threads, _notmuch_threads_destructor);
508 threads->query = query;
510 status = notmuch_query_search_messages (query, &messages);
512 talloc_free (threads);
516 threads->doc_ids = g_array_new (false, false, sizeof (unsigned int));
517 while (notmuch_messages_valid (messages)) {
518 unsigned int doc_id = _notmuch_mset_messages_get_doc_id (messages);
519 g_array_append_val (threads->doc_ids, doc_id);
520 notmuch_messages_move_to_next (messages);
522 threads->doc_id_pos = 0;
524 talloc_free (messages);
526 if (! _notmuch_doc_id_set_init (threads, &threads->match_set,
528 talloc_free (threads);
529 return NOTMUCH_STATUS_OUT_OF_MEMORY;
533 return NOTMUCH_STATUS_SUCCESS;
537 notmuch_query_destroy (notmuch_query_t *query)
543 notmuch_threads_valid (notmuch_threads_t *threads)
550 while (threads->doc_id_pos < threads->doc_ids->len) {
551 doc_id = g_array_index (threads->doc_ids, unsigned int,
552 threads->doc_id_pos);
553 if (_notmuch_doc_id_set_contains (&threads->match_set, doc_id))
556 threads->doc_id_pos++;
559 return threads->doc_id_pos < threads->doc_ids->len;
563 notmuch_threads_get (notmuch_threads_t *threads)
567 if (! notmuch_threads_valid (threads))
570 doc_id = g_array_index (threads->doc_ids, unsigned int,
571 threads->doc_id_pos);
572 return _notmuch_thread_create (threads->query,
573 threads->query->notmuch,
576 threads->query->exclude_terms,
577 threads->query->omit_excluded,
578 threads->query->sort);
582 notmuch_threads_move_to_next (notmuch_threads_t *threads)
584 threads->doc_id_pos++;
588 notmuch_threads_destroy (notmuch_threads_t *threads)
590 talloc_free (threads);
594 notmuch_query_count_messages_st (notmuch_query_t *query, unsigned *count_out)
596 return notmuch_query_count_messages (query, count_out);
600 notmuch_query_count_messages (notmuch_query_t *query, unsigned *count_out)
602 return _notmuch_query_count_documents (query, "mail", count_out);
606 _notmuch_query_count_documents (notmuch_query_t *query, const char *type, unsigned *count_out)
608 notmuch_database_t *notmuch = query->notmuch;
609 const char *query_string = query->query_string;
610 Xapian::doccount count = 0;
611 notmuch_status_t status;
613 status = _notmuch_query_ensure_parsed (query);
618 Xapian::Enquire enquire (*notmuch->xapian_db);
619 Xapian::Query mail_query (talloc_asprintf (query, "%s%s",
620 _find_prefix ("type"),
622 Xapian::Query final_query, exclude_query;
625 if (strcmp (query_string, "") == 0 ||
626 strcmp (query_string, "*") == 0) {
627 final_query = mail_query;
629 final_query = Xapian::Query (Xapian::Query::OP_AND,
630 mail_query, query->xapian_query);
633 exclude_query = _notmuch_exclude_tags (query);
635 final_query = Xapian::Query (Xapian::Query::OP_AND_NOT,
636 final_query, exclude_query);
638 enquire.set_weighting_scheme (Xapian::BoolWeight ());
639 enquire.set_docid_order (Xapian::Enquire::ASCENDING);
641 if (_debug_query ()) {
642 fprintf (stderr, "Exclude query is:\n%s\n",
643 exclude_query.get_description ().c_str ());
644 fprintf (stderr, "Final query is:\n%s\n",
645 final_query.get_description ().c_str ());
648 enquire.set_query (final_query);
651 * Set the checkatleast parameter to the number of documents
652 * in the database to make get_matches_estimated() exact.
653 * Set the max parameter to 1 to avoid fetching documents we will discard.
655 mset = enquire.get_mset (0, 1,
656 notmuch->xapian_db->get_doccount ());
658 count = mset.get_matches_estimated ();
660 } catch (const Xapian::Error &error) {
661 _notmuch_database_log (notmuch,
662 "A Xapian exception occurred performing query: %s\n",
663 error.get_msg ().c_str ());
664 _notmuch_database_log_append (notmuch,
665 "Query string was: %s\n",
666 query->query_string);
667 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
671 return NOTMUCH_STATUS_SUCCESS;
675 notmuch_query_count_threads_st (notmuch_query_t *query, unsigned *count)
677 return notmuch_query_count_threads (query, count);
681 notmuch_query_count_threads (notmuch_query_t *query, unsigned *count)
683 notmuch_messages_t *messages;
686 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
689 query->sort = NOTMUCH_SORT_UNSORTED;
690 ret = notmuch_query_search_messages (query, &messages);
694 if (messages == NULL)
695 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
697 hash = g_hash_table_new_full (g_str_hash, g_str_equal, NULL, NULL);
699 talloc_free (messages);
700 return NOTMUCH_STATUS_OUT_OF_MEMORY;
703 while (notmuch_messages_valid (messages)) {
704 notmuch_message_t *message = notmuch_messages_get (messages);
705 const char *thread_id = notmuch_message_get_thread_id (message);
706 char *thread_id_copy = talloc_strdup (messages, thread_id);
707 if (unlikely (thread_id_copy == NULL)) {
708 notmuch_message_destroy (message);
709 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
712 g_hash_table_insert (hash, thread_id_copy, NULL);
713 notmuch_message_destroy (message);
714 notmuch_messages_move_to_next (messages);
717 *count = g_hash_table_size (hash);
720 g_hash_table_unref (hash);
721 talloc_free (messages);
727 notmuch_query_get_database (const notmuch_query_t *query)
729 return query->notmuch;