1 /* thread.cc - Results of thread-based searches from 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 <gmime/gmime.h>
25 #include <glib.h> /* GHashTable */
27 #ifdef DEBUG_THREADING
28 #define THREAD_DEBUG(format, ...) fprintf (stderr, "DT: " format " (%s).\n", ##__VA_ARGS__, \
31 #define THREAD_DEBUG(format, ...) do {} while (0) /* ignored */
34 struct _notmuch_thread {
35 notmuch_database_t *notmuch;
38 GHashTable *authors_hash;
39 GPtrArray *authors_array;
40 GHashTable *matched_authors_hash;
41 GPtrArray *matched_authors_array;
45 /* All messages, oldest first. */
46 notmuch_message_list_t *message_list;
47 /* Top-level messages, oldest first. */
48 notmuch_message_list_t *toplevel_list;
50 GHashTable *message_hash;
59 _notmuch_thread_destructor (notmuch_thread_t *thread)
61 g_hash_table_unref (thread->authors_hash);
62 g_hash_table_unref (thread->matched_authors_hash);
63 g_hash_table_unref (thread->tags);
64 g_hash_table_unref (thread->message_hash);
66 if (thread->authors_array) {
67 g_ptr_array_free (thread->authors_array, true);
68 thread->authors_array = NULL;
71 if (thread->matched_authors_array) {
72 g_ptr_array_free (thread->matched_authors_array, true);
73 thread->matched_authors_array = NULL;
79 /* Add each author of the thread to the thread's authors_hash and to
80 * the thread's authors_array. */
82 _thread_add_author (notmuch_thread_t *thread,
90 if (g_hash_table_lookup_extended (thread->authors_hash,
94 author_copy = talloc_strdup (thread, author);
96 g_hash_table_insert (thread->authors_hash, author_copy, NULL);
98 g_ptr_array_add (thread->authors_array, author_copy);
101 /* Add each matched author of the thread to the thread's
102 * matched_authors_hash and to the thread's matched_authors_array. */
104 _thread_add_matched_author (notmuch_thread_t *thread,
112 if (g_hash_table_lookup_extended (thread->matched_authors_hash,
116 author_copy = talloc_strdup (thread, author);
118 g_hash_table_insert (thread->matched_authors_hash, author_copy, NULL);
120 g_ptr_array_add (thread->matched_authors_array, author_copy);
123 /* Construct an authors string from matched_authors_array and
124 * authors_array. The string contains matched authors first, then
125 * non-matched authors (with the two groups separated by '|'). Within
126 * each group, authors are listed in date order. */
128 _resolve_thread_authors_string (notmuch_thread_t *thread)
132 int first_non_matched_author = 1;
134 /* First, list all matched authors in date order. */
135 for (i = 0; i < thread->matched_authors_array->len; i++) {
136 author = (char *) g_ptr_array_index (thread->matched_authors_array, i);
138 thread->authors = talloc_asprintf (thread, "%s, %s",
142 thread->authors = author;
145 /* Next, append any non-matched authors that haven't already appeared. */
146 for (i = 0; i < thread->authors_array->len; i++) {
147 author = (char *) g_ptr_array_index (thread->authors_array, i);
148 if (g_hash_table_lookup_extended (thread->matched_authors_hash,
151 if (first_non_matched_author) {
152 thread->authors = talloc_asprintf (thread, "%s| %s",
156 thread->authors = talloc_asprintf (thread, "%s, %s",
161 first_non_matched_author = 0;
164 g_ptr_array_free (thread->authors_array, true);
165 thread->authors_array = NULL;
166 g_ptr_array_free (thread->matched_authors_array, true);
167 thread->matched_authors_array = NULL;
169 if (! thread->authors)
170 thread->authors = talloc_strdup (thread, "");
173 /* clean up the ugly "Lastname, Firstname" format that some mail systems
174 * (most notably, Exchange) are creating to be "Firstname Lastname"
175 * To make sure that we don't change other potential situations where a
176 * comma is in the name, we check that we match one of these patterns
177 * "Last, First" <first.last@company.com>
178 * "Last, First MI" <first.mi.last@company.com>
181 _thread_cleanup_author (notmuch_thread_t *thread,
182 const char *author, const char *from)
184 char *clean_author, *test_author;
191 clean_author = talloc_strdup (thread, author);
192 if (clean_author == NULL)
194 /* check if there's a comma in the name and that there's a
195 * component of the name behind it (so the name doesn't end with
196 * the comma - in which case the string that strchr finds is just
197 * one character long ",\0").
198 * Otherwise just return the copy of the original author name that
200 comma = strchr (author, ',');
201 if (comma && strlen (comma) > 1) {
202 /* let's assemble what we think is the correct name */
203 lname = comma - author;
205 /* Skip all the spaces after the comma */
206 fname = strlen (author) - lname - 1;
208 while (*comma == ' ') {
212 strncpy (clean_author, comma, fname);
214 *(clean_author + fname) = ' ';
215 strncpy (clean_author + fname + 1, author, lname);
216 *(clean_author + fname + 1 + lname) = '\0';
217 /* make a temporary copy and see if it matches the email */
218 test_author = talloc_strdup (thread, clean_author);
220 blank = strchr (test_author, ' ');
221 while (blank != NULL) {
223 blank = strchr (test_author, ' ');
225 if (strcasestr (from, test_author) == NULL)
226 /* we didn't identify this as part of the email address
227 * so let's punt and return the original author */
228 strcpy (clean_author, author);
233 /* Add 'message' as a message that belongs to 'thread'.
235 * The 'thread' will talloc_steal the 'message' and hold onto a
239 _thread_add_message (notmuch_thread_t *thread,
240 notmuch_message_t *message,
241 notmuch_string_list_t *exclude_terms,
242 notmuch_exclude_t omit_exclude)
244 notmuch_tags_t *tags;
246 InternetAddressList *list = NULL;
247 InternetAddress *address;
248 const char *from, *author;
250 bool message_excluded = false;
252 if (omit_exclude != NOTMUCH_EXCLUDE_FALSE) {
253 for (tags = notmuch_message_get_tags (message);
254 notmuch_tags_valid (tags);
255 notmuch_tags_move_to_next (tags)) {
256 tag = notmuch_tags_get (tags);
257 /* Is message excluded? */
258 for (notmuch_string_node_t *term = exclude_terms->head;
261 /* Check for an empty string, and then ignore initial 'K'. */
262 if (*(term->string) && strcmp (tag, (term->string + 1)) == 0) {
263 message_excluded = true;
270 if (message_excluded && omit_exclude == NOTMUCH_EXCLUDE_ALL)
273 _notmuch_message_list_add_message (thread->message_list,
274 talloc_steal (thread, message));
275 thread->total_messages++;
276 thread->total_files += notmuch_message_count_files (message);
278 g_hash_table_insert (thread->message_hash,
279 xstrdup (notmuch_message_get_message_id (message)),
282 from = notmuch_message_get_header (message, "from");
284 list = internet_address_list_parse (NULL, from);
287 address = internet_address_list_get_address (list, 0);
289 author = internet_address_get_name (address);
290 /* We treat quoted empty names as if they were empty. */
291 if (author == NULL || author[0] == '\0') {
292 InternetAddressMailbox *mailbox;
293 mailbox = INTERNET_ADDRESS_MAILBOX (address);
294 author = internet_address_mailbox_get_addr (mailbox);
296 clean_author = _thread_cleanup_author (thread, author, from);
297 _thread_add_author (thread, clean_author);
298 _notmuch_message_set_author (message, clean_author);
300 g_object_unref (G_OBJECT (list));
303 if (! thread->subject) {
305 subject = notmuch_message_get_header (message, "subject");
306 thread->subject = talloc_strdup (thread, subject ? subject : "");
309 for (tags = notmuch_message_get_tags (message);
310 notmuch_tags_valid (tags);
311 notmuch_tags_move_to_next (tags)) {
312 tag = notmuch_tags_get (tags);
313 g_hash_table_insert (thread->tags, xstrdup (tag), NULL);
316 /* Mark excluded messages. */
317 if (message_excluded)
318 notmuch_message_set_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED, true);
322 _thread_set_subject_from_message (notmuch_thread_t *thread,
323 notmuch_message_t *message)
326 const char *cleaned_subject;
328 subject = notmuch_message_get_header (message, "subject");
332 if ((strncasecmp (subject, "Re: ", 4) == 0) ||
333 (strncasecmp (subject, "Aw: ", 4) == 0) ||
334 (strncasecmp (subject, "Vs: ", 4) == 0) ||
335 (strncasecmp (subject, "Sv: ", 4) == 0)) {
337 cleaned_subject = talloc_strndup (thread,
339 strlen (subject) - 4);
341 cleaned_subject = talloc_strdup (thread, subject);
344 if (! EMPTY_STRING (cleaned_subject)) {
346 talloc_free (thread->subject);
348 thread->subject = talloc_strdup (thread, cleaned_subject);
352 /* Add a message to this thread which is known to match the original
353 * search specification. The 'sort' parameter controls whether the
354 * oldest or newest matching subject is applied to the thread as a
355 * whole. Returns 0 on success.
358 _thread_add_matched_message (notmuch_thread_t *thread,
359 notmuch_message_t *message,
363 notmuch_message_t *hashed_message;
364 notmuch_bool_t is_set;
366 date = notmuch_message_get_date (message);
368 if (date < thread->oldest || ! thread->matched_messages) {
369 thread->oldest = date;
370 if (sort == NOTMUCH_SORT_OLDEST_FIRST)
371 _thread_set_subject_from_message (thread, message);
374 if (date > thread->newest || ! thread->matched_messages) {
375 thread->newest = date;
376 const char *cur_subject = notmuch_thread_get_subject (thread);
377 if (sort != NOTMUCH_SORT_OLDEST_FIRST || EMPTY_STRING (cur_subject))
378 _thread_set_subject_from_message (thread, message);
381 if (notmuch_message_get_flag_st (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED, &is_set))
384 thread->matched_messages++;
386 if (g_hash_table_lookup_extended (thread->message_hash,
387 notmuch_message_get_message_id (message), NULL,
388 (void **) &hashed_message)) {
389 notmuch_message_set_flag (hashed_message,
390 NOTMUCH_MESSAGE_FLAG_MATCH, 1);
393 _thread_add_matched_author (thread, _notmuch_message_get_author (hashed_message));
398 _parent_via_in_reply_to (notmuch_thread_t *thread, notmuch_message_t *message)
400 notmuch_message_t *parent;
401 const char *in_reply_to;
403 in_reply_to = _notmuch_message_get_in_reply_to (message);
404 THREAD_DEBUG ("checking message = %s in_reply_to=%s",
405 notmuch_message_get_message_id (message), in_reply_to);
407 if (in_reply_to && (! EMPTY_STRING (in_reply_to)) &&
408 g_hash_table_lookup_extended (thread->message_hash,
410 (void **) &parent)) {
411 _notmuch_message_add_reply (parent, message);
419 _parent_or_toplevel (notmuch_thread_t *thread, notmuch_message_t *message)
421 size_t max_depth = 0;
422 notmuch_message_t *new_parent;
423 notmuch_message_t *parent = NULL;
424 const notmuch_string_list_t *references =
425 _notmuch_message_get_references (message);
427 THREAD_DEBUG ("trying to reparent via references: %s",
428 notmuch_message_get_message_id (message));
430 for (notmuch_string_node_t *ref_node = references->head;
431 ref_node; ref_node = ref_node->next) {
432 THREAD_DEBUG ("checking reference=%s", ref_node->string);
433 if ((g_hash_table_lookup_extended (thread->message_hash,
434 ref_node->string, NULL,
435 (void **) &new_parent))) {
436 size_t new_depth = _notmuch_message_get_thread_depth (new_parent);
437 THREAD_DEBUG ("got depth %lu", new_depth);
438 if (new_depth > max_depth || ! parent) {
439 THREAD_DEBUG ("adding at depth %lu parent=%s", new_depth, ref_node->string);
440 max_depth = new_depth;
446 THREAD_DEBUG ("adding reply %s to parent=%s",
447 notmuch_message_get_message_id (message),
448 notmuch_message_get_message_id (parent));
449 _notmuch_message_add_reply (parent, message);
451 THREAD_DEBUG ("adding as toplevel %s",
452 notmuch_message_get_message_id (message));
453 _notmuch_message_list_add_message (thread->toplevel_list, message);
458 _resolve_thread_relationships (notmuch_thread_t *thread)
460 notmuch_message_node_t *node, *first_node;
461 notmuch_message_t *message;
463 notmuch_message_list_t *maybe_toplevel_list;
465 first_node = thread->message_list->head;
469 local = talloc_new (thread);
470 maybe_toplevel_list = _notmuch_message_list_create (local);
472 for (node = first_node->next; node; node = node->next) {
473 message = node->message;
474 if (! _parent_via_in_reply_to (thread, message))
475 _notmuch_message_list_add_message (maybe_toplevel_list, message);
479 * if we reach the end of the list without finding a top-level
480 * message, that means the thread is a cycle (or set of cycles)
481 * and any message can be considered top-level. Choose the oldest
482 * message, which happens to be first in our list.
485 message = first_node->message;
486 THREAD_DEBUG ("checking first message %s",
487 notmuch_message_get_message_id (message));
489 if (_notmuch_message_list_empty (maybe_toplevel_list) ||
490 ! _parent_via_in_reply_to (thread, message)) {
492 THREAD_DEBUG ("adding first message as toplevel = %s",
493 notmuch_message_get_message_id (message));
494 _notmuch_message_list_add_message (maybe_toplevel_list, message);
498 for (notmuch_messages_t *messages = _notmuch_messages_create (maybe_toplevel_list);
499 notmuch_messages_valid (messages);
500 notmuch_messages_move_to_next (messages)) {
501 notmuch_message_t *message = notmuch_messages_get (messages);
502 _notmuch_message_label_depths (message, 0);
505 for (notmuch_messages_t *roots = _notmuch_messages_create (maybe_toplevel_list);
506 notmuch_messages_valid (roots);
507 notmuch_messages_move_to_next (roots)) {
508 notmuch_message_t *message = notmuch_messages_get (roots);
509 if (_notmuch_messages_has_next (roots) || ! _notmuch_message_list_empty (
510 thread->toplevel_list))
511 _parent_or_toplevel (thread, message);
513 _notmuch_message_list_add_message (thread->toplevel_list, message);
516 /* XXX this could be made conditional on messages being inserted
517 * (out of order) in later passes
519 thread->toplevel_list = _notmuch_message_sort_subtrees (thread, thread->toplevel_list);
524 /* Create a new notmuch_thread_t object by finding the thread
525 * containing the message with the given doc ID, treating any messages
526 * contained in match_set as "matched". Remove all messages in the
527 * thread from match_set.
529 * Creating the thread will perform a database search to get all
530 * messages belonging to the thread and will get the first subject
531 * line, the total count of messages, and all authors in the thread.
532 * Each message in the thread is checked against match_set to allow
533 * for a separate count of matched messages, and to allow a viewer to
534 * display these messages differently.
536 * Here, 'ctx' is talloc context for the resulting thread object.
538 * This function returns NULL in the case of any error.
541 _notmuch_thread_create (void *ctx,
542 notmuch_database_t *notmuch,
543 unsigned int seed_doc_id,
544 notmuch_doc_id_set_t *match_set,
545 notmuch_string_list_t *exclude_terms,
546 notmuch_exclude_t omit_excluded,
549 void *local = talloc_new (ctx);
550 notmuch_thread_t *thread = NULL;
551 notmuch_message_t *seed_message;
552 const char *thread_id;
553 char *thread_id_query_string;
554 notmuch_query_t *thread_id_query;
556 notmuch_messages_t *messages;
557 notmuch_message_t *message;
558 notmuch_status_t status;
560 seed_message = _notmuch_message_create (local, notmuch, seed_doc_id, NULL);
562 INTERNAL_ERROR ("Thread seed message %u does not exist", seed_doc_id);
564 thread_id = notmuch_message_get_thread_id (seed_message);
565 thread_id_query_string = talloc_asprintf (local, "thread:%s", thread_id);
566 if (unlikely (thread_id_query_string == NULL))
569 thread_id_query = talloc_steal (
570 local, notmuch_query_create (notmuch, thread_id_query_string));
571 if (unlikely (thread_id_query == NULL))
574 thread = talloc (local, notmuch_thread_t);
575 if (unlikely (thread == NULL))
578 talloc_set_destructor (thread, _notmuch_thread_destructor);
580 thread->notmuch = notmuch;
581 thread->thread_id = talloc_strdup (thread, thread_id);
582 thread->subject = NULL;
583 thread->authors_hash = g_hash_table_new_full (g_str_hash, g_str_equal,
585 thread->authors_array = g_ptr_array_new ();
586 thread->matched_authors_hash = g_hash_table_new_full (g_str_hash,
589 thread->matched_authors_array = g_ptr_array_new ();
590 thread->authors = NULL;
591 thread->tags = g_hash_table_new_full (g_str_hash, g_str_equal,
594 thread->message_list = _notmuch_message_list_create (thread);
595 thread->toplevel_list = _notmuch_message_list_create (thread);
596 if (unlikely (thread->message_list == NULL ||
597 thread->toplevel_list == NULL)) {
602 thread->message_hash = g_hash_table_new_full (g_str_hash, g_str_equal,
605 thread->total_messages = 0;
606 thread->total_files = 0;
607 thread->matched_messages = 0;
611 /* We use oldest-first order unconditionally here to obtain the
612 * proper author ordering for the thread. The 'sort' parameter
613 * passed to this function is used only to indicate whether the
614 * oldest or newest subject is desired. */
615 notmuch_query_set_sort (thread_id_query, NOTMUCH_SORT_OLDEST_FIRST);
617 status = notmuch_query_search_messages (thread_id_query, &messages);
622 notmuch_messages_valid (messages);
623 notmuch_messages_move_to_next (messages)) {
626 message = notmuch_messages_get (messages);
627 doc_id = _notmuch_message_get_doc_id (message);
628 if (doc_id == seed_doc_id)
629 message = seed_message;
631 _thread_add_message (thread, message, exclude_terms, omit_excluded);
633 if ( _notmuch_doc_id_set_contains (match_set, doc_id)) {
634 _notmuch_doc_id_set_remove (match_set, doc_id);
635 if (_thread_add_matched_message (thread, message, sort)) {
641 _notmuch_message_close (message);
644 _resolve_thread_authors_string (thread);
646 _resolve_thread_relationships (thread);
648 /* Commit to returning thread. */
649 (void) talloc_steal (ctx, thread);
657 notmuch_thread_get_toplevel_messages (notmuch_thread_t *thread)
659 return _notmuch_messages_create (thread->toplevel_list);
663 notmuch_thread_get_messages (notmuch_thread_t *thread)
665 return _notmuch_messages_create (thread->message_list);
669 notmuch_thread_get_thread_id (notmuch_thread_t *thread)
671 return thread->thread_id;
675 notmuch_thread_get_total_messages (notmuch_thread_t *thread)
677 return thread->total_messages;
681 notmuch_thread_get_total_files (notmuch_thread_t *thread)
683 return thread->total_files;
687 notmuch_thread_get_matched_messages (notmuch_thread_t *thread)
689 return thread->matched_messages;
693 notmuch_thread_get_authors (notmuch_thread_t *thread)
695 return thread->authors;
699 notmuch_thread_get_subject (notmuch_thread_t *thread)
701 return thread->subject;
705 notmuch_thread_get_oldest_date (notmuch_thread_t *thread)
707 return thread->oldest;
711 notmuch_thread_get_newest_date (notmuch_thread_t *thread)
713 return thread->newest;
717 notmuch_thread_get_tags (notmuch_thread_t *thread)
719 notmuch_string_list_t *tags;
722 tags = _notmuch_string_list_create (thread);
723 if (unlikely (tags == NULL))
726 keys = g_hash_table_get_keys (thread->tags);
728 for (l = keys; l; l = l->next)
729 _notmuch_string_list_append (tags, (char *) l->data);
733 _notmuch_string_list_sort (tags);
735 return _notmuch_tags_create (thread, tags);
739 notmuch_thread_destroy (notmuch_thread_t *thread)
741 talloc_free (thread);