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 notmuch_query_syntax_t syntax;
34 Xapian::Query xapian_query;
35 std::set<std::string> terms;
38 typedef struct _notmuch_mset_messages {
39 notmuch_messages_t base;
40 notmuch_database_t *notmuch;
41 Xapian::MSetIterator iterator;
42 Xapian::MSetIterator iterator_end;
43 } notmuch_mset_messages_t;
45 struct _notmuch_doc_id_set {
46 unsigned char *bitmap;
50 #define DOCIDSET_WORD(bit) ((bit) / CHAR_BIT)
51 #define DOCIDSET_BIT(bit) ((bit) % CHAR_BIT)
53 struct _notmuch_threads {
54 notmuch_query_t *query;
56 /* The ordered list of doc ids matched by the query. */
58 /* Our iterator's current position in doc_ids. */
59 unsigned int doc_id_pos;
60 /* The set of matched docid's that have not been assigned to a
61 * thread. Initially, this contains every docid in doc_ids. */
62 notmuch_doc_id_set_t match_set;
65 /* We need this in the message functions so forward declare. */
67 _notmuch_doc_id_set_init (void *ctx,
68 notmuch_doc_id_set_t *doc_ids,
74 char *env = getenv ("NOTMUCH_DEBUG_QUERY");
76 return (env && strcmp (env, "") != 0);
79 /* Explicit destructor call for placement new */
81 _notmuch_query_destructor (notmuch_query_t *query)
83 query->xapian_query.~Query();
84 query->terms.~set<std::string>();
88 static notmuch_query_t *
89 _notmuch_query_constructor (notmuch_database_t *notmuch,
90 const char *query_string)
92 notmuch_query_t *query;
95 fprintf (stderr, "Query string is:\n%s\n", query_string);
97 query = talloc (notmuch, notmuch_query_t);
98 if (unlikely (query == NULL))
101 new (&query->xapian_query) Xapian::Query ();
102 new (&query->terms) std::set<std::string> ();
103 query->parsed = false;
105 talloc_set_destructor (query, _notmuch_query_destructor);
107 query->notmuch = notmuch;
110 query->query_string = talloc_strdup (query, query_string);
112 query->query_string = NULL;
114 query->sort = NOTMUCH_SORT_NEWEST_FIRST;
116 query->exclude_terms = _notmuch_string_list_create (query);
118 query->omit_excluded = NOTMUCH_EXCLUDE_TRUE;
124 notmuch_query_create (notmuch_database_t *notmuch,
125 const char *query_string)
128 notmuch_query_t *query;
129 notmuch_status_t status;
131 status = notmuch_query_create_with_syntax (notmuch, query_string,
132 NOTMUCH_QUERY_SYNTAX_XAPIAN,
141 notmuch_query_create_with_syntax (notmuch_database_t *notmuch,
142 const char *query_string,
143 notmuch_query_syntax_t syntax,
144 notmuch_query_t **output)
147 notmuch_query_t *query;
150 return NOTMUCH_STATUS_NULL_POINTER;
152 query = _notmuch_query_constructor (notmuch, query_string);
154 return NOTMUCH_STATUS_OUT_OF_MEMORY;
156 if (syntax == NOTMUCH_QUERY_SYNTAX_SEXP && ! HAVE_SFSEXP) {
157 _notmuch_database_log (notmuch, "sexp query parser not available");
158 return NOTMUCH_STATUS_ILLEGAL_ARGUMENT;
161 query->syntax = syntax;
165 return NOTMUCH_STATUS_SUCCESS;
169 _notmuch_query_cache_terms (notmuch_query_t *query)
171 /* Xapian doesn't support skip_to on terms from a query since
172 * they are unordered, so cache a copy of all terms in
173 * something searchable.
176 for (Xapian::TermIterator t = query->xapian_query.get_terms_begin ();
177 t != query->xapian_query.get_terms_end (); ++t)
178 query->terms.insert (*t);
182 _notmuch_query_string_to_xapian_query (notmuch_database_t *notmuch,
183 std::string query_string,
184 Xapian::Query &output,
188 if (query_string == "" || query_string == "*") {
189 output = Xapian::Query::MatchAll;
192 notmuch->query_parser->
193 parse_query (query_string, NOTMUCH_QUERY_PARSER_FLAGS);
195 } catch (const Xapian::Error &error) {
196 if (! notmuch->exception_reported) {
197 _notmuch_database_log (notmuch,
198 "A Xapian exception occurred parsing query: %s\n",
199 error.get_msg ().c_str ());
200 _notmuch_database_log_append (notmuch,
201 "Query string was: %s\n",
202 query_string.c_str ());
203 notmuch->exception_reported = true;
206 msg = error.get_msg ();
207 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
209 return NOTMUCH_STATUS_SUCCESS;
212 static notmuch_status_t
213 _notmuch_query_ensure_parsed_xapian (notmuch_query_t *query)
215 notmuch_status_t status;
216 std::string msg; /* ignored */
218 status = _notmuch_query_string_to_xapian_query (query->notmuch, query->query_string,
219 query->xapian_query, msg);
223 query->parsed = true;
225 _notmuch_query_cache_terms (query);
227 return NOTMUCH_STATUS_SUCCESS;
230 static notmuch_status_t
231 _notmuch_query_ensure_parsed_sexpr (notmuch_query_t *query)
233 notmuch_status_t status;
236 return NOTMUCH_STATUS_SUCCESS;
238 status = _notmuch_sexp_string_to_xapian_query (query->notmuch, query->query_string,
239 query->xapian_query);
243 _notmuch_query_cache_terms (query);
244 return NOTMUCH_STATUS_SUCCESS;
247 static notmuch_status_t
248 _notmuch_query_ensure_parsed (notmuch_query_t *query)
251 return NOTMUCH_STATUS_SUCCESS;
254 if (query->syntax == NOTMUCH_QUERY_SYNTAX_SEXP)
255 return _notmuch_query_ensure_parsed_sexpr (query);
258 return _notmuch_query_ensure_parsed_xapian (query);
262 notmuch_query_get_query_string (const notmuch_query_t *query)
264 return query->query_string;
268 notmuch_query_set_omit_excluded (notmuch_query_t *query,
269 notmuch_exclude_t omit_excluded)
271 query->omit_excluded = omit_excluded;
275 notmuch_query_set_sort (notmuch_query_t *query, notmuch_sort_t sort)
281 notmuch_query_get_sort (const notmuch_query_t *query)
287 notmuch_query_add_tag_exclude (notmuch_query_t *query, const char *tag)
289 notmuch_status_t status;
292 status = _notmuch_query_ensure_parsed (query);
296 term = talloc_asprintf (query, "%s%s", _find_prefix ("tag"), tag);
297 if (query->terms.count (term) != 0)
298 return NOTMUCH_STATUS_IGNORED;
300 _notmuch_string_list_append (query->exclude_terms, term);
301 return NOTMUCH_STATUS_SUCCESS;
304 /* We end up having to call the destructors explicitly because we had
305 * to use "placement new" in order to initialize C++ objects within a
306 * block that we allocated with talloc. So C++ is making talloc
307 * slightly less simple to use, (we wouldn't need
308 * talloc_set_destructor at all otherwise).
311 _notmuch_messages_destructor (notmuch_mset_messages_t *messages)
313 messages->iterator.~MSetIterator ();
314 messages->iterator_end.~MSetIterator ();
319 /* Return a query that matches messages with the excluded tags
320 * registered with query. The caller of this function has to combine the returned
321 * query appropriately.*/
323 _notmuch_exclude_tags (notmuch_query_t *query)
325 Xapian::Query exclude_query = Xapian::Query::MatchNothing;
327 for (notmuch_string_node_t *term = query->exclude_terms->head; term;
329 exclude_query = Xapian::Query (Xapian::Query::OP_OR,
330 exclude_query, Xapian::Query (term->string));
332 return exclude_query;
337 notmuch_query_search_messages_st (notmuch_query_t *query,
338 notmuch_messages_t **out)
340 return notmuch_query_search_messages (query, out);
344 notmuch_query_search_messages (notmuch_query_t *query,
345 notmuch_messages_t **out)
347 return _notmuch_query_search_documents (query, "mail", out);
351 _notmuch_query_search_documents (notmuch_query_t *query,
353 notmuch_messages_t **out)
355 notmuch_database_t *notmuch = query->notmuch;
356 notmuch_mset_messages_t *messages;
357 notmuch_status_t status;
359 status = _notmuch_query_ensure_parsed (query);
363 messages = talloc (query, notmuch_mset_messages_t);
364 if (unlikely (messages == NULL))
365 return NOTMUCH_STATUS_OUT_OF_MEMORY;
369 messages->base.is_of_list_type = false;
370 messages->base.iterator = NULL;
371 messages->notmuch = notmuch;
372 new (&messages->iterator) Xapian::MSetIterator ();
373 new (&messages->iterator_end) Xapian::MSetIterator ();
375 talloc_set_destructor (messages, _notmuch_messages_destructor);
377 Xapian::Enquire enquire (*notmuch->xapian_db);
378 Xapian::Query mail_query (talloc_asprintf (query, "%s%s",
379 _find_prefix ("type"),
381 Xapian::Query final_query, exclude_query;
383 Xapian::MSetIterator iterator;
385 final_query = Xapian::Query (Xapian::Query::OP_AND,
386 mail_query, query->xapian_query);
388 messages->base.excluded_doc_ids = NULL;
390 if ((query->omit_excluded != NOTMUCH_EXCLUDE_FALSE) && (query->exclude_terms)) {
391 exclude_query = _notmuch_exclude_tags (query);
393 if (query->omit_excluded == NOTMUCH_EXCLUDE_TRUE ||
394 query->omit_excluded == NOTMUCH_EXCLUDE_ALL) {
395 final_query = Xapian::Query (Xapian::Query::OP_AND_NOT,
396 final_query, exclude_query);
397 } else { /* NOTMUCH_EXCLUDE_FLAG */
398 exclude_query = Xapian::Query (Xapian::Query::OP_AND,
399 exclude_query, final_query);
401 enquire.set_weighting_scheme (Xapian::BoolWeight ());
402 enquire.set_query (exclude_query);
404 mset = enquire.get_mset (0, notmuch->xapian_db->get_doccount ());
406 GArray *excluded_doc_ids = g_array_new (false, false, sizeof (unsigned int));
408 for (iterator = mset.begin (); iterator != mset.end (); iterator++) {
409 unsigned int doc_id = *iterator;
410 g_array_append_val (excluded_doc_ids, doc_id);
412 messages->base.excluded_doc_ids = talloc (messages, _notmuch_doc_id_set);
413 _notmuch_doc_id_set_init (query, messages->base.excluded_doc_ids,
415 g_array_unref (excluded_doc_ids);
420 enquire.set_weighting_scheme (Xapian::BoolWeight ());
422 switch (query->sort) {
423 case NOTMUCH_SORT_OLDEST_FIRST:
424 enquire.set_sort_by_value (NOTMUCH_VALUE_TIMESTAMP, false);
426 case NOTMUCH_SORT_NEWEST_FIRST:
427 enquire.set_sort_by_value (NOTMUCH_VALUE_TIMESTAMP, true);
429 case NOTMUCH_SORT_MESSAGE_ID:
430 enquire.set_sort_by_value (NOTMUCH_VALUE_MESSAGE_ID, false);
432 case NOTMUCH_SORT_UNSORTED:
436 if (_debug_query ()) {
437 fprintf (stderr, "Exclude query is:\n%s\n",
438 exclude_query.get_description ().c_str ());
439 fprintf (stderr, "Final query is:\n%s\n",
440 final_query.get_description ().c_str ());
443 enquire.set_query (final_query);
445 mset = enquire.get_mset (0, notmuch->xapian_db->get_doccount ());
447 messages->iterator = mset.begin ();
448 messages->iterator_end = mset.end ();
450 *out = &messages->base;
451 return NOTMUCH_STATUS_SUCCESS;
453 } catch (const Xapian::Error &error) {
454 _notmuch_database_log (notmuch,
455 "A Xapian exception occurred performing query: %s\n",
456 error.get_msg ().c_str ());
457 _notmuch_database_log_append (notmuch,
458 "Query string was: %s\n",
459 query->query_string);
461 notmuch->exception_reported = true;
462 talloc_free (messages);
463 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
468 _notmuch_mset_messages_valid (notmuch_messages_t *messages)
470 notmuch_mset_messages_t *mset_messages;
472 mset_messages = (notmuch_mset_messages_t *) messages;
474 return (mset_messages->iterator != mset_messages->iterator_end);
478 _notmuch_mset_messages_get_doc_id (notmuch_messages_t *messages)
480 notmuch_mset_messages_t *mset_messages;
482 mset_messages = (notmuch_mset_messages_t *) messages;
484 if (! _notmuch_mset_messages_valid (&mset_messages->base))
487 return *mset_messages->iterator;
491 _notmuch_mset_messages_get (notmuch_messages_t *messages)
493 notmuch_message_t *message;
494 Xapian::docid doc_id;
495 notmuch_private_status_t status;
496 notmuch_mset_messages_t *mset_messages;
498 mset_messages = (notmuch_mset_messages_t *) messages;
500 if (! _notmuch_mset_messages_valid (&mset_messages->base))
503 doc_id = *mset_messages->iterator;
505 message = _notmuch_message_create (mset_messages,
506 mset_messages->notmuch, doc_id,
509 if (message == NULL &&
510 status == NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND) {
511 INTERNAL_ERROR ("a messages iterator contains a non-existent document ID.\n");
514 if (messages->excluded_doc_ids &&
515 _notmuch_doc_id_set_contains (messages->excluded_doc_ids, doc_id))
516 notmuch_message_set_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED, true);
522 _notmuch_mset_messages_move_to_next (notmuch_messages_t *messages)
524 notmuch_mset_messages_t *mset_messages;
526 mset_messages = (notmuch_mset_messages_t *) messages;
528 mset_messages->iterator++;
532 _notmuch_doc_id_set_init (void *ctx,
533 notmuch_doc_id_set_t *doc_ids,
536 unsigned int max = 0;
537 unsigned char *bitmap;
539 for (unsigned int i = 0; i < arr->len; i++)
540 max = MAX (max, g_array_index (arr, unsigned int, i));
541 bitmap = talloc_zero_array (ctx, unsigned char, DOCIDSET_WORD (max) + 1);
546 doc_ids->bitmap = bitmap;
547 doc_ids->bound = max + 1;
549 for (unsigned int i = 0; i < arr->len; i++) {
550 unsigned int doc_id = g_array_index (arr, unsigned int, i);
551 bitmap[DOCIDSET_WORD (doc_id)] |= 1 << DOCIDSET_BIT (doc_id);
558 _notmuch_doc_id_set_contains (notmuch_doc_id_set_t *doc_ids,
561 if (doc_id >= doc_ids->bound)
563 return doc_ids->bitmap[DOCIDSET_WORD (doc_id)] & (1 << DOCIDSET_BIT (doc_id));
567 _notmuch_doc_id_set_remove (notmuch_doc_id_set_t *doc_ids,
570 if (doc_id < doc_ids->bound)
571 doc_ids->bitmap[DOCIDSET_WORD (doc_id)] &= ~(1 << DOCIDSET_BIT (doc_id));
574 /* Glib objects force use to use a talloc destructor as well, (but not
575 * nearly as ugly as the for messages due to C++ objects). At
576 * this point, I'd really like to have some talloc-friendly
577 * equivalents for the few pieces of glib that I'm using. */
579 _notmuch_threads_destructor (notmuch_threads_t *threads)
581 if (threads->doc_ids)
582 g_array_unref (threads->doc_ids);
588 notmuch_query_search_threads_st (notmuch_query_t *query, notmuch_threads_t **out)
590 return notmuch_query_search_threads (query, out);
594 notmuch_query_search_threads (notmuch_query_t *query,
595 notmuch_threads_t **out)
597 notmuch_threads_t *threads;
598 notmuch_messages_t *messages;
599 notmuch_status_t status;
601 threads = talloc (query, notmuch_threads_t);
603 return NOTMUCH_STATUS_OUT_OF_MEMORY;
604 threads->doc_ids = NULL;
605 talloc_set_destructor (threads, _notmuch_threads_destructor);
607 threads->query = query;
609 status = notmuch_query_search_messages (query, &messages);
611 talloc_free (threads);
615 threads->doc_ids = g_array_new (false, false, sizeof (unsigned int));
616 while (notmuch_messages_valid (messages)) {
617 unsigned int doc_id = _notmuch_mset_messages_get_doc_id (messages);
618 g_array_append_val (threads->doc_ids, doc_id);
619 notmuch_messages_move_to_next (messages);
621 threads->doc_id_pos = 0;
623 talloc_free (messages);
625 if (! _notmuch_doc_id_set_init (threads, &threads->match_set,
627 talloc_free (threads);
628 return NOTMUCH_STATUS_OUT_OF_MEMORY;
632 return NOTMUCH_STATUS_SUCCESS;
636 notmuch_query_destroy (notmuch_query_t *query)
642 notmuch_threads_valid (notmuch_threads_t *threads)
649 while (threads->doc_id_pos < threads->doc_ids->len) {
650 doc_id = g_array_index (threads->doc_ids, unsigned int,
651 threads->doc_id_pos);
652 if (_notmuch_doc_id_set_contains (&threads->match_set, doc_id))
655 threads->doc_id_pos++;
658 return threads->doc_id_pos < threads->doc_ids->len;
662 notmuch_threads_get (notmuch_threads_t *threads)
666 if (! notmuch_threads_valid (threads))
669 doc_id = g_array_index (threads->doc_ids, unsigned int,
670 threads->doc_id_pos);
671 return _notmuch_thread_create (threads->query,
672 threads->query->notmuch,
675 threads->query->exclude_terms,
676 threads->query->omit_excluded,
677 threads->query->sort);
681 notmuch_threads_move_to_next (notmuch_threads_t *threads)
683 threads->doc_id_pos++;
687 notmuch_threads_destroy (notmuch_threads_t *threads)
689 talloc_free (threads);
693 notmuch_query_count_messages_st (notmuch_query_t *query, unsigned *count_out)
695 return notmuch_query_count_messages (query, count_out);
699 notmuch_query_count_messages (notmuch_query_t *query, unsigned *count_out)
701 return _notmuch_query_count_documents (query, "mail", count_out);
705 _notmuch_query_count_documents (notmuch_query_t *query, const char *type, unsigned *count_out)
707 notmuch_database_t *notmuch = query->notmuch;
708 Xapian::doccount count = 0;
709 notmuch_status_t status;
711 status = _notmuch_query_ensure_parsed (query);
716 Xapian::Enquire enquire (*notmuch->xapian_db);
717 Xapian::Query mail_query (talloc_asprintf (query, "%s%s",
718 _find_prefix ("type"),
720 Xapian::Query final_query, exclude_query;
723 final_query = Xapian::Query (Xapian::Query::OP_AND,
724 mail_query, query->xapian_query);
726 exclude_query = _notmuch_exclude_tags (query);
728 final_query = Xapian::Query (Xapian::Query::OP_AND_NOT,
729 final_query, exclude_query);
731 enquire.set_weighting_scheme (Xapian::BoolWeight ());
732 enquire.set_docid_order (Xapian::Enquire::ASCENDING);
734 if (_debug_query ()) {
735 fprintf (stderr, "Exclude query is:\n%s\n",
736 exclude_query.get_description ().c_str ());
737 fprintf (stderr, "Final query is:\n%s\n",
738 final_query.get_description ().c_str ());
741 enquire.set_query (final_query);
744 * Set the checkatleast parameter to the number of documents
745 * in the database to make get_matches_estimated() exact.
746 * Set the max parameter to 1 to avoid fetching documents we will discard.
748 mset = enquire.get_mset (0, 1,
749 notmuch->xapian_db->get_doccount ());
751 count = mset.get_matches_estimated ();
753 } catch (const Xapian::Error &error) {
754 _notmuch_database_log (notmuch,
755 "A Xapian exception occurred performing query: %s\n",
756 error.get_msg ().c_str ());
757 _notmuch_database_log_append (notmuch,
758 "Query string was: %s\n",
759 query->query_string);
760 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
764 return NOTMUCH_STATUS_SUCCESS;
768 notmuch_query_count_threads_st (notmuch_query_t *query, unsigned *count)
770 return notmuch_query_count_threads (query, count);
774 notmuch_query_count_threads (notmuch_query_t *query, unsigned *count)
776 notmuch_messages_t *messages;
779 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
782 query->sort = NOTMUCH_SORT_UNSORTED;
783 ret = notmuch_query_search_messages (query, &messages);
787 if (messages == NULL)
788 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
790 hash = g_hash_table_new_full (g_str_hash, g_str_equal, NULL, NULL);
792 talloc_free (messages);
793 return NOTMUCH_STATUS_OUT_OF_MEMORY;
796 while (notmuch_messages_valid (messages)) {
797 notmuch_message_t *message = notmuch_messages_get (messages);
798 const char *thread_id = notmuch_message_get_thread_id (message);
799 char *thread_id_copy = talloc_strdup (messages, thread_id);
800 if (unlikely (thread_id_copy == NULL)) {
801 notmuch_message_destroy (message);
802 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
805 g_hash_table_insert (hash, thread_id_copy, NULL);
806 notmuch_message_destroy (message);
807 notmuch_messages_move_to_next (messages);
810 *count = g_hash_table_size (hash);
813 g_hash_table_unref (hash);
814 talloc_free (messages);
820 notmuch_query_get_database (const notmuch_query_t *query)
822 return query->notmuch;
826 _notmuch_query_expand (notmuch_database_t *notmuch, const char *field, Xapian::Query subquery,
827 Xapian::Query &output, std::string &msg)
829 std::set<std::string> terms;
830 const std::string term_prefix = _find_prefix (field);
832 if (_debug_query ()) {
833 fprintf (stderr, "Expanding subquery:\n%s\n",
834 subquery.get_description ().c_str ());
838 Xapian::Enquire enquire (*notmuch->xapian_db);
841 enquire.set_weighting_scheme (Xapian::BoolWeight ());
842 enquire.set_query (subquery);
844 mset = enquire.get_mset (0, notmuch->xapian_db->get_doccount ());
846 for (Xapian::MSetIterator iterator = mset.begin (); iterator != mset.end (); iterator++) {
847 Xapian::docid doc_id = *iterator;
848 Xapian::Document doc = notmuch->xapian_db->get_document (doc_id);
849 Xapian::TermIterator i = doc.termlist_begin ();
851 for (i.skip_to (term_prefix);
852 i != doc.termlist_end () && ((*i).rfind (term_prefix, 0) == 0); i++) {
856 output = Xapian::Query (Xapian::Query::OP_OR, terms.begin (), terms.end ());
857 if (_debug_query ()) {
858 fprintf (stderr, "Expanded query:\n%s\n",
859 subquery.get_description ().c_str ());
862 } catch (const Xapian::Error &error) {
863 _notmuch_database_log (notmuch,
864 "A Xapian exception occurred expanding query: %s\n",
865 error.get_msg ().c_str ());
866 msg = error.get_msg ();
867 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
870 return NOTMUCH_STATUS_SUCCESS;