1 /* database.cc - The database interfaces of the notmuch mail library
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 http://www.gnu.org/licenses/ .
18 * Author: Carl Worth <cworth@cworth.org>
21 #include "database-private.h"
22 #include "parse-time-vrp.h"
23 #include "string-util.h"
32 #include <glib.h> /* g_free, GPtrArray, GHashTable */
33 #include <glib-object.h> /* g_type_init */
35 #include <gmime/gmime.h> /* g_mime_init */
39 #define ARRAY_SIZE(arr) (sizeof (arr) / sizeof (arr[0]))
46 #define NOTMUCH_DATABASE_VERSION 3
48 #define STRINGIFY(s) _SUB_STRINGIFY(s)
49 #define _SUB_STRINGIFY(s) #s
51 /* Here's the current schema for our database (for NOTMUCH_DATABASE_VERSION):
53 * We currently have three different types of documents (mail, ghost,
54 * and directory) and also some metadata.
58 * A mail document is associated with a particular email message. It
59 * is stored in one or more files on disk (though only one has its
60 * content indexed) and is uniquely identified by its "id" field
61 * (which is generally the message ID). It is indexed with the
62 * following prefixed terms which the database uses to construct
65 * Single terms of given prefix:
69 * id: Unique ID of mail. This is from the Message-ID header
70 * if present and not too long (see NOTMUCH_MESSAGE_ID_MAX).
71 * If it's present and too long, then we use
72 * "notmuch-sha1-<sha1_sum_of_message_id>".
73 * If this header is not present, we use
74 * "notmuch-sha1-<sha1_sum_of_entire_file>".
76 * thread: The ID of the thread to which the mail belongs
78 * replyto: The ID from the In-Reply-To header of the mail (if any).
80 * Multiple terms of given prefix:
82 * reference: All message IDs from In-Reply-To and References
83 * headers in the message.
85 * tag: Any tags associated with this message by the user.
87 * file-direntry: A colon-separated pair of values
88 * (INTEGER:STRING), where INTEGER is the
89 * document ID of a directory document, and
90 * STRING is the name of a file within that
91 * directory for this mail message.
93 * A mail document also has four values:
95 * TIMESTAMP: The time_t value corresponding to the message's
98 * MESSAGE_ID: The unique ID of the mail mess (see "id" above)
100 * FROM: The value of the "From" header
102 * SUBJECT: The value of the "Subject" header
104 * In addition, terms from the content of the message are added with
105 * "from", "to", "attachment", and "subject" prefixes for use by the
106 * user in searching. Similarly, terms from the path of the mail
107 * message are added with "folder" and "path" prefixes. But the
108 * database doesn't really care itself about any of these.
110 * The data portion of a mail document is empty.
112 * Ghost mail document [if NOTMUCH_FEATURE_GHOSTS]
113 * -----------------------------------------------
114 * A ghost mail document is like a mail document, but where we don't
115 * have the message content. These are used to track thread reference
116 * information for messages we haven't received.
118 * A ghost mail document has type: ghost; id and thread fields that
119 * are identical to the mail document fields; and a MESSAGE_ID value.
123 * A directory document is used by a client of the notmuch library to
124 * maintain data necessary to allow for efficient polling of mail
127 * All directory documents contain one term:
129 * directory: The directory path (relative to the database path)
130 * Or the SHA1 sum of the directory path (if the
131 * path itself is too long to fit in a Xapian
134 * And all directory documents for directories other than top-level
135 * directories also contain the following term:
137 * directory-direntry: A colon-separated pair of values
138 * (INTEGER:STRING), where INTEGER is the
139 * document ID of the parent directory
140 * document, and STRING is the name of this
141 * directory within that parent.
143 * All directory documents have a single value:
145 * TIMESTAMP: The mtime of the directory (at last scan)
147 * The data portion of a directory document contains the path of the
148 * directory (relative to the database path).
152 * Xapian allows us to store arbitrary name-value pairs as
153 * "metadata". We currently use the following metadata names with the
156 * version The database schema version, (which is distinct
157 * from both the notmuch package version (see
158 * notmuch --version) and the libnotmuch library
159 * version. The version is stored as an base-10
160 * ASCII integer. The initial database version
161 * was 1, (though a schema existed before that
162 * were no "version" database value existed at
163 * all). Successive versions are allocated as
164 * changes are made to the database (such as by
165 * indexing new fields).
167 * features The set of features supported by this
168 * database. This consists of a set of
169 * '\n'-separated lines, where each is a feature
170 * name, a '\t', and compatibility flags. If the
171 * compatibility flags contain 'w', then the
172 * opener must support this feature to safely
173 * write this database. If the compatibility
174 * flags contain 'r', then the opener must
175 * support this feature to read this database.
176 * Introduced in database version 3.
178 * last_thread_id The last thread ID generated. This is stored
179 * as a 16-byte hexadecimal ASCII representation
180 * of a 64-bit unsigned integer. The first ID
181 * generated is 1 and the value will be
182 * incremented for each thread ID.
187 * If ! NOTMUCH_FEATURE_GHOSTS, there are no ghost mail documents.
188 * Instead, the database has the following additional database
191 * thread_id_* A pre-allocated thread ID for a particular
192 * message. This is actually an arbitrarily large
193 * family of metadata name. Any particular name is
194 * formed by concatenating "thread_id_" with a message
195 * ID (or the SHA1 sum of a message ID if it is very
196 * long---see description of 'id' in the mail
197 * document). The value stored is a thread ID.
199 * These thread ID metadata values are stored
200 * whenever a message references a parent message
201 * that does not yet exist in the database. A
202 * thread ID will be allocated and stored, and if
203 * the message is later added, the stored thread
204 * ID will be used (and the metadata value will
207 * Even before a message is added, it's
208 * pre-allocated thread ID is useful so that all
209 * descendant messages that reference this common
210 * parent can be recognized as belonging to the
214 /* With these prefix values we follow the conventions published here:
216 * http://xapian.org/docs/omega/termprefixes.html
218 * as much as makes sense. Note that I took some liberty in matching
219 * the reserved prefix values to notmuch concepts, (for example, 'G'
220 * is documented as "newsGroup (or similar entity - e.g. a web forum
221 * name)", for which I think the thread is the closest analogue in
222 * notmuch. This in spite of the fact that we will eventually be
223 * storing mailing-list messages where 'G' for "mailing list name"
224 * might be even a closer analogue. I'm treating the single-character
225 * prefixes preferentially for core notmuch concepts (which will be
226 * nearly universal to all mail messages).
229 static prefix_t BOOLEAN_PREFIX_INTERNAL[] = {
231 { "reference", "XREFERENCE" },
232 { "replyto", "XREPLYTO" },
233 { "directory", "XDIRECTORY" },
234 { "file-direntry", "XFDIRENTRY" },
235 { "directory-direntry", "XDDIRENTRY" },
238 static prefix_t BOOLEAN_PREFIX_EXTERNAL[] = {
245 * Without the ":", since this is a multi-letter prefix, Xapian
246 * will add a colon itself if the first letter of the path is
247 * upper-case ASCII. Including the ":" forces there to always be a
248 * colon, which keeps our own logic simpler.
250 { "folder", "XFOLDER:" },
253 static prefix_t PROBABILISTIC_PREFIX[]= {
256 { "attachment", "XATTACHMENT" },
257 { "subject", "XSUBJECT"},
261 _find_prefix (const char *name)
265 for (i = 0; i < ARRAY_SIZE (BOOLEAN_PREFIX_INTERNAL); i++) {
266 if (strcmp (name, BOOLEAN_PREFIX_INTERNAL[i].name) == 0)
267 return BOOLEAN_PREFIX_INTERNAL[i].prefix;
270 for (i = 0; i < ARRAY_SIZE (BOOLEAN_PREFIX_EXTERNAL); i++) {
271 if (strcmp (name, BOOLEAN_PREFIX_EXTERNAL[i].name) == 0)
272 return BOOLEAN_PREFIX_EXTERNAL[i].prefix;
275 for (i = 0; i < ARRAY_SIZE (PROBABILISTIC_PREFIX); i++) {
276 if (strcmp (name, PROBABILISTIC_PREFIX[i].name) == 0)
277 return PROBABILISTIC_PREFIX[i].prefix;
280 INTERNAL_ERROR ("No prefix exists for '%s'\n", name);
285 static const struct {
286 /* NOTMUCH_FEATURE_* value. */
287 _notmuch_features value;
288 /* Feature name as it appears in the database. This name should
289 * be appropriate for displaying to the user if an older version
290 * of notmuch doesn't support this feature. */
292 /* Compatibility flags when this feature is declared. */
294 } feature_names[] = {
295 { NOTMUCH_FEATURE_FILE_TERMS,
296 "multiple paths per message", "rw" },
297 { NOTMUCH_FEATURE_DIRECTORY_DOCS,
298 "relative directory paths", "rw" },
299 /* Header values are not required for reading a database because a
300 * reader can just refer to the message file. */
301 { NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES,
302 "from/subject/message-ID in database", "w" },
303 { NOTMUCH_FEATURE_BOOL_FOLDER,
304 "exact folder:/path: search", "rw" },
305 { NOTMUCH_FEATURE_GHOSTS,
306 "mail documents for missing messages", "w"},
307 /* Knowledge of the index mime-types are not required for reading
308 * a database because a reader will just be unable to query
310 { NOTMUCH_FEATURE_INDEXED_MIMETYPES,
311 "indexed MIME types", "w"},
315 notmuch_status_to_string (notmuch_status_t status)
318 case NOTMUCH_STATUS_SUCCESS:
319 return "No error occurred";
320 case NOTMUCH_STATUS_OUT_OF_MEMORY:
321 return "Out of memory";
322 case NOTMUCH_STATUS_READ_ONLY_DATABASE:
323 return "Attempt to write to a read-only database";
324 case NOTMUCH_STATUS_XAPIAN_EXCEPTION:
325 return "A Xapian exception occurred";
326 case NOTMUCH_STATUS_FILE_ERROR:
327 return "Something went wrong trying to read or write a file";
328 case NOTMUCH_STATUS_FILE_NOT_EMAIL:
329 return "File is not an email";
330 case NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID:
331 return "Message ID is identical to a message in database";
332 case NOTMUCH_STATUS_NULL_POINTER:
333 return "Erroneous NULL pointer";
334 case NOTMUCH_STATUS_TAG_TOO_LONG:
335 return "Tag value is too long (exceeds NOTMUCH_TAG_MAX)";
336 case NOTMUCH_STATUS_UNBALANCED_FREEZE_THAW:
337 return "Unbalanced number of calls to notmuch_message_freeze/thaw";
338 case NOTMUCH_STATUS_UNBALANCED_ATOMIC:
339 return "Unbalanced number of calls to notmuch_database_begin_atomic/end_atomic";
340 case NOTMUCH_STATUS_UNSUPPORTED_OPERATION:
341 return "Unsupported operation";
342 case NOTMUCH_STATUS_UPGRADE_REQUIRED:
343 return "Operation requires a database upgrade";
345 case NOTMUCH_STATUS_LAST_STATUS:
346 return "Unknown error status value";
351 find_doc_ids_for_term (notmuch_database_t *notmuch,
353 Xapian::PostingIterator *begin,
354 Xapian::PostingIterator *end)
356 *begin = notmuch->xapian_db->postlist_begin (term);
358 *end = notmuch->xapian_db->postlist_end (term);
362 find_doc_ids (notmuch_database_t *notmuch,
363 const char *prefix_name,
365 Xapian::PostingIterator *begin,
366 Xapian::PostingIterator *end)
370 term = talloc_asprintf (notmuch, "%s%s",
371 _find_prefix (prefix_name), value);
373 find_doc_ids_for_term (notmuch, term, begin, end);
378 notmuch_private_status_t
379 _notmuch_database_find_unique_doc_id (notmuch_database_t *notmuch,
380 const char *prefix_name,
382 unsigned int *doc_id)
384 Xapian::PostingIterator i, end;
386 find_doc_ids (notmuch, prefix_name, value, &i, &end);
390 return NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND;
395 #if DEBUG_DATABASE_SANITY
399 INTERNAL_ERROR ("Term %s:%s is not unique as expected.\n",
403 return NOTMUCH_PRIVATE_STATUS_SUCCESS;
406 static Xapian::Document
407 find_document_for_doc_id (notmuch_database_t *notmuch, unsigned doc_id)
409 return notmuch->xapian_db->get_document (doc_id);
412 /* Generate a compressed version of 'message_id' of the form:
414 * notmuch-sha1-<sha1_sum_of_message_id>
417 _notmuch_message_id_compressed (void *ctx, const char *message_id)
419 char *sha1, *compressed;
421 sha1 = _notmuch_sha1_of_string (message_id);
423 compressed = talloc_asprintf (ctx, "notmuch-sha1-%s", sha1);
430 notmuch_database_find_message (notmuch_database_t *notmuch,
431 const char *message_id,
432 notmuch_message_t **message_ret)
434 notmuch_private_status_t status;
437 if (message_ret == NULL)
438 return NOTMUCH_STATUS_NULL_POINTER;
440 if (strlen (message_id) > NOTMUCH_MESSAGE_ID_MAX)
441 message_id = _notmuch_message_id_compressed (notmuch, message_id);
444 status = _notmuch_database_find_unique_doc_id (notmuch, "id",
445 message_id, &doc_id);
447 if (status == NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND)
450 *message_ret = _notmuch_message_create (notmuch, notmuch, doc_id,
452 if (*message_ret == NULL)
453 return NOTMUCH_STATUS_OUT_OF_MEMORY;
456 return NOTMUCH_STATUS_SUCCESS;
457 } catch (const Xapian::Error &error) {
458 fprintf (stderr, "A Xapian exception occurred finding message: %s.\n",
459 error.get_msg().c_str());
460 notmuch->exception_reported = TRUE;
462 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
466 /* Advance 'str' past any whitespace or RFC 822 comments. A comment is
467 * a (potentially nested) parenthesized sequence with '\' used to
468 * escape any character (including parentheses).
470 * If the sequence to be skipped continues to the end of the string,
471 * then 'str' will be left pointing at the final terminating '\0'
475 skip_space_and_comments (const char **str)
480 while (*s && (isspace (*s) || *s == '(')) {
481 while (*s && isspace (*s))
486 while (*s && nesting) {
489 } else if (*s == ')') {
491 } else if (*s == '\\') {
503 /* Parse an RFC 822 message-id, discarding whitespace, any RFC 822
504 * comments, and the '<' and '>' delimiters.
506 * If not NULL, then *next will be made to point to the first character
507 * not parsed, (possibly pointing to the final '\0' terminator.
509 * Returns a newly talloc'ed string belonging to 'ctx'.
511 * Returns NULL if there is any error parsing the message-id. */
513 _parse_message_id (void *ctx, const char *message_id, const char **next)
518 if (message_id == NULL || *message_id == '\0')
523 skip_space_and_comments (&s);
525 /* Skip any unstructured text as well. */
526 while (*s && *s != '<')
537 skip_space_and_comments (&s);
540 while (*end && *end != '>')
549 if (end > s && *end == '>')
554 result = talloc_strndup (ctx, s, end - s + 1);
556 /* Finally, collapse any whitespace that is within the message-id
562 for (r = result, len = strlen (r); *r; r++, len--)
563 if (*r == ' ' || *r == '\t')
564 memmove (r, r+1, len);
570 /* Parse a References header value, putting a (talloc'ed under 'ctx')
571 * copy of each referenced message-id into 'hash'.
573 * We explicitly avoid including any reference identical to
574 * 'message_id' in the result (to avoid mass confusion when a single
575 * message references itself cyclically---and yes, mail messages are
576 * not infrequent in the wild that do this---don't ask me why).
578 * Return the last reference parsed, if it is not equal to message_id.
581 parse_references (void *ctx,
582 const char *message_id,
586 char *ref, *last_ref = NULL;
588 if (refs == NULL || *refs == '\0')
592 ref = _parse_message_id (ctx, refs, &refs);
594 if (ref && strcmp (ref, message_id)) {
595 g_hash_table_insert (hash, ref, NULL);
600 /* The return value of this function is used to add a parent
601 * reference to the database. We should avoid making a message
602 * its own parent, thus the above check.
608 notmuch_database_create (const char *path, notmuch_database_t **database)
610 notmuch_status_t status = NOTMUCH_STATUS_SUCCESS;
611 notmuch_database_t *notmuch = NULL;
612 char *notmuch_path = NULL;
617 fprintf (stderr, "Error: Cannot create a database for a NULL path.\n");
618 status = NOTMUCH_STATUS_NULL_POINTER;
622 err = stat (path, &st);
624 fprintf (stderr, "Error: Cannot create database at %s: %s.\n",
625 path, strerror (errno));
626 status = NOTMUCH_STATUS_FILE_ERROR;
630 if (! S_ISDIR (st.st_mode)) {
631 fprintf (stderr, "Error: Cannot create database at %s: Not a directory.\n",
633 status = NOTMUCH_STATUS_FILE_ERROR;
637 notmuch_path = talloc_asprintf (NULL, "%s/%s", path, ".notmuch");
639 err = mkdir (notmuch_path, 0755);
642 fprintf (stderr, "Error: Cannot create directory %s: %s.\n",
643 notmuch_path, strerror (errno));
644 status = NOTMUCH_STATUS_FILE_ERROR;
648 status = notmuch_database_open (path,
649 NOTMUCH_DATABASE_MODE_READ_WRITE,
654 /* Upgrade doesn't add these feature to existing databases, but
655 * new databases have them. */
656 notmuch->features |= NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES;
657 notmuch->features |= NOTMUCH_FEATURE_INDEXED_MIMETYPES;
659 status = notmuch_database_upgrade (notmuch, NULL, NULL);
661 notmuch_database_close(notmuch);
667 talloc_free (notmuch_path);
672 talloc_free (notmuch);
677 _notmuch_database_ensure_writable (notmuch_database_t *notmuch)
679 if (notmuch->mode == NOTMUCH_DATABASE_MODE_READ_ONLY) {
680 fprintf (stderr, "Cannot write to a read-only database.\n");
681 return NOTMUCH_STATUS_READ_ONLY_DATABASE;
684 return NOTMUCH_STATUS_SUCCESS;
687 /* Parse a database features string from the given database version.
688 * Returns the feature bit set.
690 * For version < 3, this ignores the features string and returns a
691 * hard-coded set of features.
693 * If there are unrecognized features that are required to open the
694 * database in mode (which should be 'r' or 'w'), return a
695 * comma-separated list of unrecognized but required features in
696 * *incompat_out suitable for presenting to the user. *incompat_out
697 * will be allocated from ctx.
699 static _notmuch_features
700 _parse_features (const void *ctx, const char *features, unsigned int version,
701 char mode, char **incompat_out)
703 _notmuch_features res = static_cast<_notmuch_features>(0);
704 unsigned int namelen, i;
708 /* Prior to database version 3, features were implied by the
711 return NOTMUCH_FEATURES_V0;
712 else if (version == 1)
713 return NOTMUCH_FEATURES_V1;
714 else if (version == 2)
715 return NOTMUCH_FEATURES_V2;
717 /* Parse the features string */
718 while ((features = strtok_len_c (features + llen, "\n", &llen)) != NULL) {
719 flags = strchr (features, '\t');
720 if (! flags || flags > features + llen)
722 namelen = flags - features;
724 for (i = 0; i < ARRAY_SIZE (feature_names); ++i) {
725 if (strlen (feature_names[i].name) == namelen &&
726 strncmp (feature_names[i].name, features, namelen) == 0) {
727 res |= feature_names[i].value;
732 if (i == ARRAY_SIZE (feature_names) && incompat_out) {
733 /* Unrecognized feature */
734 const char *have = strchr (flags, mode);
735 if (have && have < features + llen) {
736 /* This feature is required to access this database in
737 * 'mode', but we don't understand it. */
739 *incompat_out = talloc_strdup (ctx, "");
740 *incompat_out = talloc_asprintf_append_buffer (
741 *incompat_out, "%s%.*s", **incompat_out ? ", " : "",
751 _print_features (const void *ctx, unsigned int features)
754 char *res = talloc_strdup (ctx, "");
756 for (i = 0; i < ARRAY_SIZE (feature_names); ++i)
757 if (features & feature_names[i].value)
758 res = talloc_asprintf_append_buffer (
759 res, "%s\t%s\n", feature_names[i].name, feature_names[i].flags);
765 notmuch_database_open (const char *path,
766 notmuch_database_mode_t mode,
767 notmuch_database_t **database)
769 notmuch_status_t status = NOTMUCH_STATUS_SUCCESS;
770 void *local = talloc_new (NULL);
771 notmuch_database_t *notmuch = NULL;
772 char *notmuch_path, *xapian_path, *incompat_features;
775 unsigned int i, version;
776 static int initialized = 0;
779 fprintf (stderr, "Error: Cannot open a database for a NULL path.\n");
780 status = NOTMUCH_STATUS_NULL_POINTER;
784 if (! (notmuch_path = talloc_asprintf (local, "%s/%s", path, ".notmuch"))) {
785 fprintf (stderr, "Out of memory\n");
786 status = NOTMUCH_STATUS_OUT_OF_MEMORY;
790 err = stat (notmuch_path, &st);
792 fprintf (stderr, "Error opening database at %s: %s\n",
793 notmuch_path, strerror (errno));
794 status = NOTMUCH_STATUS_FILE_ERROR;
798 if (! (xapian_path = talloc_asprintf (local, "%s/%s", notmuch_path, "xapian"))) {
799 fprintf (stderr, "Out of memory\n");
800 status = NOTMUCH_STATUS_OUT_OF_MEMORY;
804 /* Initialize the GLib type system and threads */
805 #if !GLIB_CHECK_VERSION(2, 35, 1)
809 /* Initialize gmime */
811 g_mime_init (GMIME_ENABLE_RFC2047_WORKAROUNDS);
815 notmuch = talloc_zero (NULL, notmuch_database_t);
816 notmuch->exception_reported = FALSE;
817 notmuch->path = talloc_strdup (notmuch, path);
819 if (notmuch->path[strlen (notmuch->path) - 1] == '/')
820 notmuch->path[strlen (notmuch->path) - 1] = '\0';
822 notmuch->mode = mode;
823 notmuch->atomic_nesting = 0;
825 string last_thread_id;
827 if (mode == NOTMUCH_DATABASE_MODE_READ_WRITE) {
828 notmuch->xapian_db = new Xapian::WritableDatabase (xapian_path,
829 Xapian::DB_CREATE_OR_OPEN);
831 notmuch->xapian_db = new Xapian::Database (xapian_path);
834 /* Check version. As of database version 3, we represent
835 * changes in terms of features, so assume a version bump
836 * means a dramatically incompatible change. */
837 version = notmuch_database_get_version (notmuch);
838 if (version > NOTMUCH_DATABASE_VERSION) {
840 "Error: Notmuch database at %s\n"
841 " has a newer database format version (%u) than supported by this\n"
842 " version of notmuch (%u).\n",
843 notmuch_path, version, NOTMUCH_DATABASE_VERSION);
844 notmuch->mode = NOTMUCH_DATABASE_MODE_READ_ONLY;
845 notmuch_database_destroy (notmuch);
847 status = NOTMUCH_STATUS_FILE_ERROR;
851 /* Check features. */
852 incompat_features = NULL;
853 notmuch->features = _parse_features (
854 local, notmuch->xapian_db->get_metadata ("features").c_str (),
855 version, mode == NOTMUCH_DATABASE_MODE_READ_WRITE ? 'w' : 'r',
857 if (incompat_features) {
859 "Error: Notmuch database at %s\n"
860 " requires features (%s)\n"
861 " not supported by this version of notmuch.\n",
862 notmuch_path, incompat_features);
863 notmuch->mode = NOTMUCH_DATABASE_MODE_READ_ONLY;
864 notmuch_database_destroy (notmuch);
866 status = NOTMUCH_STATUS_FILE_ERROR;
870 notmuch->last_doc_id = notmuch->xapian_db->get_lastdocid ();
871 last_thread_id = notmuch->xapian_db->get_metadata ("last_thread_id");
872 if (last_thread_id.empty ()) {
873 notmuch->last_thread_id = 0;
878 str = last_thread_id.c_str ();
879 notmuch->last_thread_id = strtoull (str, &end, 16);
881 INTERNAL_ERROR ("Malformed database last_thread_id: %s", str);
884 notmuch->query_parser = new Xapian::QueryParser;
885 notmuch->term_gen = new Xapian::TermGenerator;
886 notmuch->term_gen->set_stemmer (Xapian::Stem ("english"));
887 notmuch->value_range_processor = new Xapian::NumberValueRangeProcessor (NOTMUCH_VALUE_TIMESTAMP);
888 notmuch->date_range_processor = new ParseTimeValueRangeProcessor (NOTMUCH_VALUE_TIMESTAMP);
890 notmuch->query_parser->set_default_op (Xapian::Query::OP_AND);
891 notmuch->query_parser->set_database (*notmuch->xapian_db);
892 notmuch->query_parser->set_stemmer (Xapian::Stem ("english"));
893 notmuch->query_parser->set_stemming_strategy (Xapian::QueryParser::STEM_SOME);
894 notmuch->query_parser->add_valuerangeprocessor (notmuch->value_range_processor);
895 notmuch->query_parser->add_valuerangeprocessor (notmuch->date_range_processor);
897 for (i = 0; i < ARRAY_SIZE (BOOLEAN_PREFIX_EXTERNAL); i++) {
898 prefix_t *prefix = &BOOLEAN_PREFIX_EXTERNAL[i];
899 notmuch->query_parser->add_boolean_prefix (prefix->name,
903 for (i = 0; i < ARRAY_SIZE (PROBABILISTIC_PREFIX); i++) {
904 prefix_t *prefix = &PROBABILISTIC_PREFIX[i];
905 notmuch->query_parser->add_prefix (prefix->name, prefix->prefix);
907 } catch (const Xapian::Error &error) {
908 fprintf (stderr, "A Xapian exception occurred opening database: %s\n",
909 error.get_msg().c_str());
910 notmuch_database_destroy (notmuch);
912 status = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
921 talloc_free (notmuch);
926 notmuch_database_close (notmuch_database_t *notmuch)
928 notmuch_status_t status = NOTMUCH_STATUS_SUCCESS;
930 /* Many Xapian objects (and thus notmuch objects) hold references to
931 * the database, so merely deleting the database may not suffice to
932 * close it. Thus, we explicitly close it here. */
933 if (notmuch->xapian_db != NULL) {
935 /* If there's an outstanding transaction, it's unclear if
936 * closing the Xapian database commits everything up to
937 * that transaction, or may discard committed (but
938 * unflushed) transactions. To be certain, explicitly
939 * cancel any outstanding transaction before closing. */
940 if (notmuch->mode == NOTMUCH_DATABASE_MODE_READ_WRITE &&
941 notmuch->atomic_nesting)
942 (static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db))
943 ->cancel_transaction ();
945 /* Close the database. This implicitly flushes
946 * outstanding changes. */
947 notmuch->xapian_db->close();
948 } catch (const Xapian::Error &error) {
949 status = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
950 if (! notmuch->exception_reported) {
951 fprintf (stderr, "Error: A Xapian exception occurred closing database: %s\n",
952 error.get_msg().c_str());
957 delete notmuch->term_gen;
958 notmuch->term_gen = NULL;
959 delete notmuch->query_parser;
960 notmuch->query_parser = NULL;
961 delete notmuch->xapian_db;
962 notmuch->xapian_db = NULL;
963 delete notmuch->value_range_processor;
964 notmuch->value_range_processor = NULL;
965 delete notmuch->date_range_processor;
966 notmuch->date_range_processor = NULL;
971 #if HAVE_XAPIAN_COMPACT
973 unlink_cb (const char *path,
974 unused (const struct stat *sb),
976 unused (struct FTW *ftw))
978 return remove (path);
982 rmtree (const char *path)
984 return nftw (path, unlink_cb, 64, FTW_DEPTH | FTW_PHYS);
987 class NotmuchCompactor : public Xapian::Compactor
989 notmuch_compact_status_cb_t status_cb;
990 void *status_closure;
993 NotmuchCompactor(notmuch_compact_status_cb_t cb, void *closure) :
994 status_cb (cb), status_closure (closure) { }
997 set_status (const std::string &table, const std::string &status)
1001 if (status_cb == NULL)
1004 if (status.length () == 0)
1005 msg = talloc_asprintf (NULL, "compacting table %s", table.c_str());
1007 msg = talloc_asprintf (NULL, " %s", status.c_str());
1013 status_cb (msg, status_closure);
1018 /* Compacts the given database, optionally saving the original database
1019 * in backup_path. Additionally, a callback function can be provided to
1020 * give the user feedback on the progress of the (likely long-lived)
1021 * compaction process.
1023 * The backup path must point to a directory on the same volume as the
1024 * original database. Passing a NULL backup_path will result in the
1025 * uncompacted database being deleted after compaction has finished.
1026 * Note that the database write lock will be held during the
1027 * compaction process to protect data integrity.
1030 notmuch_database_compact (const char *path,
1031 const char *backup_path,
1032 notmuch_compact_status_cb_t status_cb,
1036 char *notmuch_path, *xapian_path, *compact_xapian_path;
1037 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
1038 notmuch_database_t *notmuch = NULL;
1039 struct stat statbuf;
1040 notmuch_bool_t keep_backup;
1042 local = talloc_new (NULL);
1044 return NOTMUCH_STATUS_OUT_OF_MEMORY;
1046 ret = notmuch_database_open (path, NOTMUCH_DATABASE_MODE_READ_WRITE, ¬much);
1051 if (! (notmuch_path = talloc_asprintf (local, "%s/%s", path, ".notmuch"))) {
1052 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
1056 if (! (xapian_path = talloc_asprintf (local, "%s/%s", notmuch_path, "xapian"))) {
1057 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
1061 if (! (compact_xapian_path = talloc_asprintf (local, "%s.compact", xapian_path))) {
1062 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
1066 if (backup_path == NULL) {
1067 if (! (backup_path = talloc_asprintf (local, "%s.old", xapian_path))) {
1068 ret = NOTMUCH_STATUS_OUT_OF_MEMORY;
1071 keep_backup = FALSE;
1077 if (stat (backup_path, &statbuf) != -1) {
1078 fprintf (stderr, "Path already exists: %s\n", backup_path);
1079 ret = NOTMUCH_STATUS_FILE_ERROR;
1082 if (errno != ENOENT) {
1083 fprintf (stderr, "Unknown error while stat()ing path: %s\n",
1085 ret = NOTMUCH_STATUS_FILE_ERROR;
1089 /* Unconditionally attempt to remove old work-in-progress database (if
1090 * any). This is "protected" by database lock. If this fails due to write
1091 * errors (etc), the following code will fail and provide error message.
1093 (void) rmtree (compact_xapian_path);
1096 NotmuchCompactor compactor (status_cb, closure);
1098 compactor.set_renumber (false);
1099 compactor.add_source (xapian_path);
1100 compactor.set_destdir (compact_xapian_path);
1101 compactor.compact ();
1102 } catch (const Xapian::Error &error) {
1103 fprintf (stderr, "Error while compacting: %s\n", error.get_msg().c_str());
1104 ret = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
1108 if (rename (xapian_path, backup_path)) {
1109 fprintf (stderr, "Error moving %s to %s: %s\n",
1110 xapian_path, backup_path, strerror (errno));
1111 ret = NOTMUCH_STATUS_FILE_ERROR;
1115 if (rename (compact_xapian_path, xapian_path)) {
1116 fprintf (stderr, "Error moving %s to %s: %s\n",
1117 compact_xapian_path, xapian_path, strerror (errno));
1118 ret = NOTMUCH_STATUS_FILE_ERROR;
1122 if (! keep_backup) {
1123 if (rmtree (backup_path)) {
1124 fprintf (stderr, "Error removing old database %s: %s\n",
1125 backup_path, strerror (errno));
1126 ret = NOTMUCH_STATUS_FILE_ERROR;
1133 notmuch_status_t ret2;
1135 ret2 = notmuch_database_destroy (notmuch);
1137 /* don't clobber previous error status */
1138 if (ret == NOTMUCH_STATUS_SUCCESS && ret2 != NOTMUCH_STATUS_SUCCESS)
1142 talloc_free (local);
1148 notmuch_database_compact (unused (const char *path),
1149 unused (const char *backup_path),
1150 unused (notmuch_compact_status_cb_t status_cb),
1151 unused (void *closure))
1153 fprintf (stderr, "notmuch was compiled against a xapian version lacking compaction support.\n");
1154 return NOTMUCH_STATUS_UNSUPPORTED_OPERATION;
1159 notmuch_database_destroy (notmuch_database_t *notmuch)
1161 notmuch_status_t status;
1163 status = notmuch_database_close (notmuch);
1164 talloc_free (notmuch);
1170 notmuch_database_get_path (notmuch_database_t *notmuch)
1172 return notmuch->path;
1176 notmuch_database_get_version (notmuch_database_t *notmuch)
1178 unsigned int version;
1179 string version_string;
1183 version_string = notmuch->xapian_db->get_metadata ("version");
1184 if (version_string.empty ())
1187 str = version_string.c_str ();
1188 if (str == NULL || *str == '\0')
1191 version = strtoul (str, &end, 10);
1193 INTERNAL_ERROR ("Malformed database version: %s", str);
1199 notmuch_database_needs_upgrade (notmuch_database_t *notmuch)
1201 return notmuch->mode == NOTMUCH_DATABASE_MODE_READ_WRITE &&
1202 ((NOTMUCH_FEATURES_CURRENT & ~notmuch->features) ||
1203 (notmuch_database_get_version (notmuch) < NOTMUCH_DATABASE_VERSION));
1206 static volatile sig_atomic_t do_progress_notify = 0;
1209 handle_sigalrm (unused (int signal))
1211 do_progress_notify = 1;
1214 /* Upgrade the current database.
1216 * After opening a database in read-write mode, the client should
1217 * check if an upgrade is needed (notmuch_database_needs_upgrade) and
1218 * if so, upgrade with this function before making any modifications.
1220 * The optional progress_notify callback can be used by the caller to
1221 * provide progress indication to the user. If non-NULL it will be
1222 * called periodically with 'count' as the number of messages upgraded
1223 * so far and 'total' the overall number of messages that will be
1227 notmuch_database_upgrade (notmuch_database_t *notmuch,
1228 void (*progress_notify) (void *closure,
1232 void *local = talloc_new (NULL);
1233 Xapian::TermIterator t, t_end;
1234 Xapian::WritableDatabase *db;
1235 struct sigaction action;
1236 struct itimerval timerval;
1237 notmuch_bool_t timer_is_active = FALSE;
1238 enum _notmuch_features target_features, new_features;
1239 notmuch_status_t status;
1240 notmuch_private_status_t private_status;
1241 unsigned int count = 0, total = 0;
1243 status = _notmuch_database_ensure_writable (notmuch);
1247 db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
1249 target_features = notmuch->features | NOTMUCH_FEATURES_CURRENT;
1250 new_features = NOTMUCH_FEATURES_CURRENT & ~notmuch->features;
1252 if (! notmuch_database_needs_upgrade (notmuch))
1253 return NOTMUCH_STATUS_SUCCESS;
1255 if (progress_notify) {
1256 /* Setup our handler for SIGALRM */
1257 memset (&action, 0, sizeof (struct sigaction));
1258 action.sa_handler = handle_sigalrm;
1259 sigemptyset (&action.sa_mask);
1260 action.sa_flags = SA_RESTART;
1261 sigaction (SIGALRM, &action, NULL);
1263 /* Then start a timer to send SIGALRM once per second. */
1264 timerval.it_interval.tv_sec = 1;
1265 timerval.it_interval.tv_usec = 0;
1266 timerval.it_value.tv_sec = 1;
1267 timerval.it_value.tv_usec = 0;
1268 setitimer (ITIMER_REAL, &timerval, NULL);
1270 timer_is_active = TRUE;
1273 /* Figure out how much total work we need to do. */
1275 (NOTMUCH_FEATURE_FILE_TERMS | NOTMUCH_FEATURE_BOOL_FOLDER)) {
1276 notmuch_query_t *query = notmuch_query_create (notmuch, "");
1277 total += notmuch_query_count_messages (query);
1278 notmuch_query_destroy (query);
1280 if (new_features & NOTMUCH_FEATURE_DIRECTORY_DOCS) {
1281 t_end = db->allterms_end ("XTIMESTAMP");
1282 for (t = db->allterms_begin ("XTIMESTAMP"); t != t_end; t++)
1285 if (new_features & NOTMUCH_FEATURE_GHOSTS) {
1286 /* The ghost message upgrade converts all thread_id_*
1287 * metadata values into ghost message documents. */
1288 t_end = db->metadata_keys_end ("thread_id_");
1289 for (t = db->metadata_keys_begin ("thread_id_"); t != t_end; ++t)
1293 /* Perform the upgrade in a transaction. */
1294 db->begin_transaction (true);
1296 /* Set the target features so we write out changes in the desired
1298 notmuch->features = target_features;
1300 /* Perform per-message upgrades. */
1302 (NOTMUCH_FEATURE_FILE_TERMS | NOTMUCH_FEATURE_BOOL_FOLDER)) {
1303 notmuch_query_t *query = notmuch_query_create (notmuch, "");
1304 notmuch_messages_t *messages;
1305 notmuch_message_t *message;
1308 for (messages = notmuch_query_search_messages (query);
1309 notmuch_messages_valid (messages);
1310 notmuch_messages_move_to_next (messages))
1312 if (do_progress_notify) {
1313 progress_notify (closure, (double) count / total);
1314 do_progress_notify = 0;
1317 message = notmuch_messages_get (messages);
1319 /* Before version 1, each message document had its
1320 * filename in the data field. Copy that into the new
1321 * format by calling notmuch_message_add_filename.
1323 if (new_features & NOTMUCH_FEATURE_FILE_TERMS) {
1324 filename = _notmuch_message_talloc_copy_data (message);
1325 if (filename && *filename != '\0') {
1326 _notmuch_message_add_filename (message, filename);
1327 _notmuch_message_clear_data (message);
1329 talloc_free (filename);
1332 /* Prior to version 2, the "folder:" prefix was
1333 * probabilistic and stemmed. Change it to the current
1334 * boolean prefix. Add "path:" prefixes while at it.
1336 if (new_features & NOTMUCH_FEATURE_BOOL_FOLDER)
1337 _notmuch_message_upgrade_folder (message);
1339 _notmuch_message_sync (message);
1341 notmuch_message_destroy (message);
1346 notmuch_query_destroy (query);
1349 /* Perform per-directory upgrades. */
1351 /* Before version 1 we stored directory timestamps in
1352 * XTIMESTAMP documents instead of the current XDIRECTORY
1353 * documents. So copy those as well. */
1354 if (new_features & NOTMUCH_FEATURE_DIRECTORY_DOCS) {
1355 t_end = notmuch->xapian_db->allterms_end ("XTIMESTAMP");
1357 for (t = notmuch->xapian_db->allterms_begin ("XTIMESTAMP");
1361 Xapian::PostingIterator p, p_end;
1362 std::string term = *t;
1364 p_end = notmuch->xapian_db->postlist_end (term);
1366 for (p = notmuch->xapian_db->postlist_begin (term);
1370 Xapian::Document document;
1372 notmuch_directory_t *directory;
1374 if (do_progress_notify) {
1375 progress_notify (closure, (double) count / total);
1376 do_progress_notify = 0;
1379 document = find_document_for_doc_id (notmuch, *p);
1380 mtime = Xapian::sortable_unserialise (
1381 document.get_value (NOTMUCH_VALUE_TIMESTAMP));
1383 directory = _notmuch_directory_create (notmuch, term.c_str() + 10,
1384 NOTMUCH_FIND_CREATE, &status);
1385 notmuch_directory_set_mtime (directory, mtime);
1386 notmuch_directory_destroy (directory);
1388 db->delete_document (*p);
1395 /* Perform metadata upgrades. */
1397 /* Prior to NOTMUCH_FEATURE_GHOSTS, thread IDs for missing
1398 * messages were stored as database metadata. Change these to
1401 if (new_features & NOTMUCH_FEATURE_GHOSTS) {
1402 notmuch_message_t *message;
1403 std::string message_id, thread_id;
1405 t_end = db->metadata_keys_end (NOTMUCH_METADATA_THREAD_ID_PREFIX);
1406 for (t = db->metadata_keys_begin (NOTMUCH_METADATA_THREAD_ID_PREFIX);
1408 if (do_progress_notify) {
1409 progress_notify (closure, (double) count / total);
1410 do_progress_notify = 0;
1413 message_id = (*t).substr (
1414 strlen (NOTMUCH_METADATA_THREAD_ID_PREFIX));
1415 thread_id = db->get_metadata (*t);
1417 /* Create ghost message */
1418 message = _notmuch_message_create_for_message_id (
1419 notmuch, message_id.c_str (), &private_status);
1420 if (private_status == NOTMUCH_PRIVATE_STATUS_SUCCESS) {
1421 /* Document already exists; ignore the stored thread ID */
1422 } else if (private_status ==
1423 NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND) {
1424 private_status = _notmuch_message_initialize_ghost (
1425 message, thread_id.c_str ());
1426 if (! private_status)
1427 _notmuch_message_sync (message);
1430 if (private_status) {
1432 "Upgrade failed while creating ghost messages.\n");
1433 status = COERCE_STATUS (private_status, "Unexpected status from _notmuch_message_initialize_ghost");
1437 /* Clear saved metadata thread ID */
1438 db->set_metadata (*t, "");
1444 status = NOTMUCH_STATUS_SUCCESS;
1445 db->set_metadata ("features", _print_features (local, notmuch->features));
1446 db->set_metadata ("version", STRINGIFY (NOTMUCH_DATABASE_VERSION));
1449 if (status == NOTMUCH_STATUS_SUCCESS)
1450 db->commit_transaction ();
1452 db->cancel_transaction ();
1454 if (timer_is_active) {
1455 /* Now stop the timer. */
1456 timerval.it_interval.tv_sec = 0;
1457 timerval.it_interval.tv_usec = 0;
1458 timerval.it_value.tv_sec = 0;
1459 timerval.it_value.tv_usec = 0;
1460 setitimer (ITIMER_REAL, &timerval, NULL);
1462 /* And disable the signal handler. */
1463 action.sa_handler = SIG_IGN;
1464 sigaction (SIGALRM, &action, NULL);
1467 talloc_free (local);
1472 notmuch_database_begin_atomic (notmuch_database_t *notmuch)
1474 if (notmuch->mode == NOTMUCH_DATABASE_MODE_READ_ONLY ||
1475 notmuch->atomic_nesting > 0)
1479 (static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db))->begin_transaction (false);
1480 } catch (const Xapian::Error &error) {
1481 fprintf (stderr, "A Xapian exception occurred beginning transaction: %s.\n",
1482 error.get_msg().c_str());
1483 notmuch->exception_reported = TRUE;
1484 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
1488 notmuch->atomic_nesting++;
1489 return NOTMUCH_STATUS_SUCCESS;
1493 notmuch_database_end_atomic (notmuch_database_t *notmuch)
1495 Xapian::WritableDatabase *db;
1497 if (notmuch->atomic_nesting == 0)
1498 return NOTMUCH_STATUS_UNBALANCED_ATOMIC;
1500 if (notmuch->mode == NOTMUCH_DATABASE_MODE_READ_ONLY ||
1501 notmuch->atomic_nesting > 1)
1504 db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
1506 db->commit_transaction ();
1508 /* This is a hack for testing. Xapian never flushes on a
1509 * non-flushed commit, even if the flush threshold is 1.
1510 * However, we rely on flushing to test atomicity. */
1511 const char *thresh = getenv ("XAPIAN_FLUSH_THRESHOLD");
1512 if (thresh && atoi (thresh) == 1)
1514 } catch (const Xapian::Error &error) {
1515 fprintf (stderr, "A Xapian exception occurred committing transaction: %s.\n",
1516 error.get_msg().c_str());
1517 notmuch->exception_reported = TRUE;
1518 return NOTMUCH_STATUS_XAPIAN_EXCEPTION;
1522 notmuch->atomic_nesting--;
1523 return NOTMUCH_STATUS_SUCCESS;
1526 /* We allow the user to use arbitrarily long paths for directories. But
1527 * we have a term-length limit. So if we exceed that, we'll use the
1528 * SHA-1 of the path for the database term.
1530 * Note: This function may return the original value of 'path'. If it
1531 * does not, then the caller is responsible to free() the returned
1535 _notmuch_database_get_directory_db_path (const char *path)
1537 int term_len = strlen (_find_prefix ("directory")) + strlen (path);
1539 if (term_len > NOTMUCH_TERM_MAX)
1540 return _notmuch_sha1_of_string (path);
1545 /* Given a path, split it into two parts: the directory part is all
1546 * components except for the last, and the basename is that last
1547 * component. Getting the return-value for either part is optional
1548 * (the caller can pass NULL).
1550 * The original 'path' can represent either a regular file or a
1551 * directory---the splitting will be carried out in the same way in
1552 * either case. Trailing slashes on 'path' will be ignored, and any
1553 * cases of multiple '/' characters appearing in series will be
1554 * treated as a single '/'.
1556 * Allocation (if any) will have 'ctx' as the talloc owner. But
1557 * pointers will be returned within the original path string whenever
1560 * Note: If 'path' is non-empty and contains no non-trailing slash,
1561 * (that is, consists of a filename with no parent directory), then
1562 * the directory returned will be an empty string. However, if 'path'
1563 * is an empty string, then both directory and basename will be
1567 _notmuch_database_split_path (void *ctx,
1569 const char **directory,
1570 const char **basename)
1574 if (path == NULL || *path == '\0') {
1579 return NOTMUCH_STATUS_SUCCESS;
1582 /* Find the last slash (not counting a trailing slash), if any. */
1584 slash = path + strlen (path) - 1;
1586 /* First, skip trailing slashes. */
1587 while (slash != path) {
1594 /* Then, find a slash. */
1595 while (slash != path) {
1605 /* Finally, skip multiple slashes. */
1606 while (slash != path) {
1613 if (slash == path) {
1615 *directory = talloc_strdup (ctx, "");
1620 *directory = talloc_strndup (ctx, path, slash - path + 1);
1623 return NOTMUCH_STATUS_SUCCESS;
1626 /* Find the document ID of the specified directory.
1628 * If (flags & NOTMUCH_FIND_CREATE), a new directory document will be
1629 * created if one does not exist for 'path'. Otherwise, if the
1630 * directory document does not exist, this sets *directory_id to
1631 * ((unsigned int)-1) and returns NOTMUCH_STATUS_SUCCESS.
1634 _notmuch_database_find_directory_id (notmuch_database_t *notmuch,
1636 notmuch_find_flags_t flags,
1637 unsigned int *directory_id)
1639 notmuch_directory_t *directory;
1640 notmuch_status_t status;
1644 return NOTMUCH_STATUS_SUCCESS;
1647 directory = _notmuch_directory_create (notmuch, path, flags, &status);
1648 if (status || !directory) {
1653 *directory_id = _notmuch_directory_get_document_id (directory);
1655 notmuch_directory_destroy (directory);
1657 return NOTMUCH_STATUS_SUCCESS;
1661 _notmuch_database_get_directory_path (void *ctx,
1662 notmuch_database_t *notmuch,
1663 unsigned int doc_id)
1665 Xapian::Document document;
1667 document = find_document_for_doc_id (notmuch, doc_id);
1669 return talloc_strdup (ctx, document.get_data ().c_str ());
1672 /* Given a legal 'filename' for the database, (either relative to
1673 * database path or absolute with initial components identical to
1674 * database path), return a new string (with 'ctx' as the talloc
1675 * owner) suitable for use as a direntry term value.
1677 * If (flags & NOTMUCH_FIND_CREATE), the necessary directory documents
1678 * will be created in the database as needed. Otherwise, if the
1679 * necessary directory documents do not exist, this sets
1680 * *direntry to NULL and returns NOTMUCH_STATUS_SUCCESS.
1683 _notmuch_database_filename_to_direntry (void *ctx,
1684 notmuch_database_t *notmuch,
1685 const char *filename,
1686 notmuch_find_flags_t flags,
1689 const char *relative, *directory, *basename;
1690 Xapian::docid directory_id;
1691 notmuch_status_t status;
1693 relative = _notmuch_database_relative_path (notmuch, filename);
1695 status = _notmuch_database_split_path (ctx, relative,
1696 &directory, &basename);
1700 status = _notmuch_database_find_directory_id (notmuch, directory, flags,
1702 if (status || directory_id == (unsigned int)-1) {
1707 *direntry = talloc_asprintf (ctx, "%u:%s", directory_id, basename);
1709 return NOTMUCH_STATUS_SUCCESS;
1712 /* Given a legal 'path' for the database, return the relative path.
1714 * The return value will be a pointer to the original path contents,
1715 * and will be either the original string (if 'path' was relative) or
1716 * a portion of the string (if path was absolute and begins with the
1720 _notmuch_database_relative_path (notmuch_database_t *notmuch,
1723 const char *db_path, *relative;
1724 unsigned int db_path_len;
1726 db_path = notmuch_database_get_path (notmuch);
1727 db_path_len = strlen (db_path);
1731 if (*relative == '/') {
1732 while (*relative == '/' && *(relative+1) == '/')
1735 if (strncmp (relative, db_path, db_path_len) == 0)
1737 relative += db_path_len;
1738 while (*relative == '/')
1747 notmuch_database_get_directory (notmuch_database_t *notmuch,
1749 notmuch_directory_t **directory)
1751 notmuch_status_t status;
1753 if (directory == NULL)
1754 return NOTMUCH_STATUS_NULL_POINTER;
1758 *directory = _notmuch_directory_create (notmuch, path,
1759 NOTMUCH_FIND_LOOKUP, &status);
1760 } catch (const Xapian::Error &error) {
1761 fprintf (stderr, "A Xapian exception occurred getting directory: %s.\n",
1762 error.get_msg().c_str());
1763 notmuch->exception_reported = TRUE;
1764 status = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
1769 /* Allocate a document ID that satisfies the following criteria:
1771 * 1. The ID does not exist for any document in the Xapian database
1773 * 2. The ID was not previously returned from this function
1775 * 3. The ID is the smallest integer satisfying (1) and (2)
1777 * This function will trigger an internal error if these constraints
1778 * cannot all be satisfied, (that is, the pool of available document
1779 * IDs has been exhausted).
1782 _notmuch_database_generate_doc_id (notmuch_database_t *notmuch)
1784 assert (notmuch->last_doc_id >= notmuch->xapian_db->get_lastdocid ());
1786 notmuch->last_doc_id++;
1788 if (notmuch->last_doc_id == 0)
1789 INTERNAL_ERROR ("Xapian document IDs are exhausted.\n");
1791 return notmuch->last_doc_id;
1795 _notmuch_database_generate_thread_id (notmuch_database_t *notmuch)
1797 /* 16 bytes (+ terminator) for hexadecimal representation of
1798 * a 64-bit integer. */
1799 static char thread_id[17];
1800 Xapian::WritableDatabase *db;
1802 db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
1804 notmuch->last_thread_id++;
1806 sprintf (thread_id, "%016" PRIx64, notmuch->last_thread_id);
1808 db->set_metadata ("last_thread_id", thread_id);
1814 _get_metadata_thread_id_key (void *ctx, const char *message_id)
1816 if (strlen (message_id) > NOTMUCH_MESSAGE_ID_MAX)
1817 message_id = _notmuch_message_id_compressed (ctx, message_id);
1819 return talloc_asprintf (ctx, NOTMUCH_METADATA_THREAD_ID_PREFIX "%s",
1823 static notmuch_status_t
1824 _resolve_message_id_to_thread_id_old (notmuch_database_t *notmuch,
1826 const char *message_id,
1827 const char **thread_id_ret);
1829 /* Find the thread ID to which the message with 'message_id' belongs.
1831 * Note: 'thread_id_ret' must not be NULL!
1832 * On success '*thread_id_ret' is set to a newly talloced string belonging to
1835 * Note: If there is no message in the database with the given
1836 * 'message_id' then a new thread_id will be allocated for this
1837 * message ID and stored in the database metadata so that the
1838 * thread ID can be looked up if the message is added to the database
1841 static notmuch_status_t
1842 _resolve_message_id_to_thread_id (notmuch_database_t *notmuch,
1844 const char *message_id,
1845 const char **thread_id_ret)
1847 notmuch_private_status_t status;
1848 notmuch_message_t *message;
1850 if (! (notmuch->features & NOTMUCH_FEATURE_GHOSTS))
1851 return _resolve_message_id_to_thread_id_old (notmuch, ctx, message_id,
1854 /* Look for this message (regular or ghost) */
1855 message = _notmuch_message_create_for_message_id (
1856 notmuch, message_id, &status);
1857 if (status == NOTMUCH_PRIVATE_STATUS_SUCCESS) {
1858 /* Message exists */
1859 *thread_id_ret = talloc_steal (
1860 ctx, notmuch_message_get_thread_id (message));
1861 } else if (status == NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND) {
1862 /* Message did not exist. Give it a fresh thread ID and
1863 * populate this message as a ghost message. */
1864 *thread_id_ret = talloc_strdup (
1865 ctx, _notmuch_database_generate_thread_id (notmuch));
1866 if (! *thread_id_ret) {
1867 status = NOTMUCH_PRIVATE_STATUS_OUT_OF_MEMORY;
1869 status = _notmuch_message_initialize_ghost (message, *thread_id_ret);
1871 /* Commit the new ghost message */
1872 _notmuch_message_sync (message);
1875 /* Create failed. Fall through. */
1878 notmuch_message_destroy (message);
1880 return COERCE_STATUS (status, "Error creating ghost message");
1883 /* Pre-ghost messages _resolve_message_id_to_thread_id */
1884 static notmuch_status_t
1885 _resolve_message_id_to_thread_id_old (notmuch_database_t *notmuch,
1887 const char *message_id,
1888 const char **thread_id_ret)
1890 notmuch_status_t status;
1891 notmuch_message_t *message;
1892 string thread_id_string;
1894 Xapian::WritableDatabase *db;
1896 status = notmuch_database_find_message (notmuch, message_id, &message);
1902 *thread_id_ret = talloc_steal (ctx,
1903 notmuch_message_get_thread_id (message));
1905 notmuch_message_destroy (message);
1907 return NOTMUCH_STATUS_SUCCESS;
1910 /* Message has not been seen yet.
1912 * We may have seen a reference to it already, in which case, we
1913 * can return the thread ID stored in the metadata. Otherwise, we
1914 * generate a new thread ID and store it there.
1916 db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
1917 metadata_key = _get_metadata_thread_id_key (ctx, message_id);
1918 thread_id_string = notmuch->xapian_db->get_metadata (metadata_key);
1920 if (thread_id_string.empty()) {
1921 *thread_id_ret = talloc_strdup (ctx,
1922 _notmuch_database_generate_thread_id (notmuch));
1923 db->set_metadata (metadata_key, *thread_id_ret);
1925 *thread_id_ret = talloc_strdup (ctx, thread_id_string.c_str());
1928 talloc_free (metadata_key);
1930 return NOTMUCH_STATUS_SUCCESS;
1933 static notmuch_status_t
1934 _merge_threads (notmuch_database_t *notmuch,
1935 const char *winner_thread_id,
1936 const char *loser_thread_id)
1938 Xapian::PostingIterator loser, loser_end;
1939 notmuch_message_t *message = NULL;
1940 notmuch_private_status_t private_status;
1941 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
1943 find_doc_ids (notmuch, "thread", loser_thread_id, &loser, &loser_end);
1945 for ( ; loser != loser_end; loser++) {
1946 message = _notmuch_message_create (notmuch, notmuch,
1947 *loser, &private_status);
1948 if (message == NULL) {
1949 ret = COERCE_STATUS (private_status,
1950 "Cannot find document for doc_id from query");
1954 _notmuch_message_remove_term (message, "thread", loser_thread_id);
1955 _notmuch_message_add_term (message, "thread", winner_thread_id);
1956 _notmuch_message_sync (message);
1958 notmuch_message_destroy (message);
1964 notmuch_message_destroy (message);
1970 _my_talloc_free_for_g_hash (void *ptr)
1975 static notmuch_status_t
1976 _notmuch_database_link_message_to_parents (notmuch_database_t *notmuch,
1977 notmuch_message_t *message,
1978 notmuch_message_file_t *message_file,
1979 const char **thread_id)
1981 GHashTable *parents = NULL;
1982 const char *refs, *in_reply_to, *in_reply_to_message_id;
1983 const char *last_ref_message_id, *this_message_id;
1984 GList *l, *keys = NULL;
1985 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
1987 parents = g_hash_table_new_full (g_str_hash, g_str_equal,
1988 _my_talloc_free_for_g_hash, NULL);
1989 this_message_id = notmuch_message_get_message_id (message);
1991 refs = _notmuch_message_file_get_header (message_file, "references");
1992 last_ref_message_id = parse_references (message,
1996 in_reply_to = _notmuch_message_file_get_header (message_file, "in-reply-to");
1997 in_reply_to_message_id = parse_references (message,
1999 parents, in_reply_to);
2001 /* For the parent of this message, use the last message ID of the
2002 * References header, if available. If not, fall back to the
2003 * first message ID in the In-Reply-To header. */
2004 if (last_ref_message_id) {
2005 _notmuch_message_add_term (message, "replyto",
2006 last_ref_message_id);
2007 } else if (in_reply_to_message_id) {
2008 _notmuch_message_add_term (message, "replyto",
2009 in_reply_to_message_id);
2012 keys = g_hash_table_get_keys (parents);
2013 for (l = keys; l; l = l->next) {
2014 char *parent_message_id;
2015 const char *parent_thread_id = NULL;
2017 parent_message_id = (char *) l->data;
2019 _notmuch_message_add_term (message, "reference",
2022 ret = _resolve_message_id_to_thread_id (notmuch,
2029 if (*thread_id == NULL) {
2030 *thread_id = talloc_strdup (message, parent_thread_id);
2031 _notmuch_message_add_term (message, "thread", *thread_id);
2032 } else if (strcmp (*thread_id, parent_thread_id)) {
2033 ret = _merge_threads (notmuch, *thread_id, parent_thread_id);
2043 g_hash_table_unref (parents);
2048 static notmuch_status_t
2049 _notmuch_database_link_message_to_children (notmuch_database_t *notmuch,
2050 notmuch_message_t *message,
2051 const char **thread_id)
2053 const char *message_id = notmuch_message_get_message_id (message);
2054 Xapian::PostingIterator child, children_end;
2055 notmuch_message_t *child_message = NULL;
2056 const char *child_thread_id;
2057 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS;
2058 notmuch_private_status_t private_status;
2060 find_doc_ids (notmuch, "reference", message_id, &child, &children_end);
2062 for ( ; child != children_end; child++) {
2064 child_message = _notmuch_message_create (message, notmuch,
2065 *child, &private_status);
2066 if (child_message == NULL) {
2067 ret = COERCE_STATUS (private_status,
2068 "Cannot find document for doc_id from query");
2072 child_thread_id = notmuch_message_get_thread_id (child_message);
2073 if (*thread_id == NULL) {
2074 *thread_id = talloc_strdup (message, child_thread_id);
2075 _notmuch_message_add_term (message, "thread", *thread_id);
2076 } else if (strcmp (*thread_id, child_thread_id)) {
2077 _notmuch_message_remove_term (child_message, "reference",
2079 _notmuch_message_sync (child_message);
2080 ret = _merge_threads (notmuch, *thread_id, child_thread_id);
2085 notmuch_message_destroy (child_message);
2086 child_message = NULL;
2091 notmuch_message_destroy (child_message);
2096 /* Fetch and clear the stored thread_id for message, or NULL if none. */
2098 _consume_metadata_thread_id (void *ctx, notmuch_database_t *notmuch,
2099 notmuch_message_t *message)
2101 const char *message_id;
2105 message_id = notmuch_message_get_message_id (message);
2106 metadata_key = _get_metadata_thread_id_key (ctx, message_id);
2108 /* Check if we have already seen related messages to this one.
2109 * If we have then use the thread_id that we stored at that time.
2111 stored_id = notmuch->xapian_db->get_metadata (metadata_key);
2112 if (stored_id.empty ()) {
2115 Xapian::WritableDatabase *db;
2117 db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
2119 /* Clear the metadata for this message ID. We don't need it
2121 db->set_metadata (metadata_key, "");
2123 return talloc_strdup (ctx, stored_id.c_str ());
2127 /* Given a blank or ghost 'message' and its corresponding
2128 * 'message_file' link it to existing threads in the database.
2130 * First, if is_ghost, this retrieves the thread ID already stored in
2131 * the message (which will be the case if a message was previously
2132 * added that referenced this one). If the message is blank
2133 * (!is_ghost), it doesn't have a thread ID yet (we'll generate one
2134 * later in this function). If the database does not support ghost
2135 * messages, this checks for a thread ID stored in database metadata
2136 * for this message ID.
2138 * Second, we look at 'message_file' and its link-relevant headers
2139 * (References and In-Reply-To) for message IDs.
2141 * Finally, we look in the database for existing message that
2142 * reference 'message'.
2144 * In all cases, we assign to the current message the first thread ID
2145 * found. We will also merge any existing, distinct threads where this
2146 * message belongs to both, (which is not uncommon when messages are
2147 * processed out of order).
2149 * Finally, if no thread ID has been found through referenced messages, we
2150 * call _notmuch_message_generate_thread_id to generate a new thread
2151 * ID. This should only happen for new, top-level messages, (no
2152 * References or In-Reply-To header in this message, and no previously
2153 * added message refers to this message).
2155 static notmuch_status_t
2156 _notmuch_database_link_message (notmuch_database_t *notmuch,
2157 notmuch_message_t *message,
2158 notmuch_message_file_t *message_file,
2159 notmuch_bool_t is_ghost)
2161 void *local = talloc_new (NULL);
2162 notmuch_status_t status;
2163 const char *thread_id = NULL;
2165 /* Check if the message already had a thread ID */
2166 if (notmuch->features & NOTMUCH_FEATURE_GHOSTS) {
2168 thread_id = notmuch_message_get_thread_id (message);
2170 thread_id = _consume_metadata_thread_id (local, notmuch, message);
2172 _notmuch_message_add_term (message, "thread", thread_id);
2175 status = _notmuch_database_link_message_to_parents (notmuch, message,
2181 if (! (notmuch->features & NOTMUCH_FEATURE_GHOSTS)) {
2182 /* In general, it shouldn't be necessary to link children,
2183 * since the earlier indexing of those children will have
2184 * stored a thread ID for the missing parent. However, prior
2185 * to ghost messages, these stored thread IDs were NOT
2186 * rewritten during thread merging (and there was no
2187 * performant way to do so), so if indexed children were
2188 * pulled into a different thread ID by a merge, it was
2189 * necessary to pull them *back* into the stored thread ID of
2190 * the parent. With ghost messages, we just rewrite the
2191 * stored thread IDs during merging, so this workaround isn't
2193 status = _notmuch_database_link_message_to_children (notmuch, message,
2199 /* If not part of any existing thread, generate a new thread ID. */
2200 if (thread_id == NULL) {
2201 thread_id = _notmuch_database_generate_thread_id (notmuch);
2203 _notmuch_message_add_term (message, "thread", thread_id);
2207 talloc_free (local);
2213 notmuch_database_add_message (notmuch_database_t *notmuch,
2214 const char *filename,
2215 notmuch_message_t **message_ret)
2217 notmuch_message_file_t *message_file;
2218 notmuch_message_t *message = NULL;
2219 notmuch_status_t ret = NOTMUCH_STATUS_SUCCESS, ret2;
2220 notmuch_private_status_t private_status;
2221 notmuch_bool_t is_ghost = false;
2223 const char *date, *header;
2224 const char *from, *to, *subject;
2225 char *message_id = NULL;
2228 *message_ret = NULL;
2230 ret = _notmuch_database_ensure_writable (notmuch);
2234 message_file = _notmuch_message_file_open (filename);
2235 if (message_file == NULL)
2236 return NOTMUCH_STATUS_FILE_ERROR;
2238 /* Adding a message may change many documents. Do this all
2240 ret = notmuch_database_begin_atomic (notmuch);
2244 /* Parse message up front to get better error status. */
2245 ret = _notmuch_message_file_parse (message_file);
2250 /* Before we do any real work, (especially before doing a
2251 * potential SHA-1 computation on the entire file's contents),
2252 * let's make sure that what we're looking at looks like an
2253 * actual email message.
2255 from = _notmuch_message_file_get_header (message_file, "from");
2256 subject = _notmuch_message_file_get_header (message_file, "subject");
2257 to = _notmuch_message_file_get_header (message_file, "to");
2259 if ((from == NULL || *from == '\0') &&
2260 (subject == NULL || *subject == '\0') &&
2261 (to == NULL || *to == '\0'))
2263 ret = NOTMUCH_STATUS_FILE_NOT_EMAIL;
2267 /* Now that we're sure it's mail, the first order of business
2268 * is to find a message ID (or else create one ourselves). */
2270 header = _notmuch_message_file_get_header (message_file, "message-id");
2271 if (header && *header != '\0') {
2272 message_id = _parse_message_id (message_file, header, NULL);
2274 /* So the header value isn't RFC-compliant, but it's
2275 * better than no message-id at all. */
2276 if (message_id == NULL)
2277 message_id = talloc_strdup (message_file, header);
2280 if (message_id == NULL ) {
2281 /* No message-id at all, let's generate one by taking a
2282 * hash over the file's contents. */
2283 char *sha1 = _notmuch_sha1_of_file (filename);
2285 /* If that failed too, something is really wrong. Give up. */
2287 ret = NOTMUCH_STATUS_FILE_ERROR;
2291 message_id = talloc_asprintf (message_file,
2292 "notmuch-sha1-%s", sha1);
2296 /* Now that we have a message ID, we get a message object,
2297 * (which may or may not reference an existing document in the
2300 message = _notmuch_message_create_for_message_id (notmuch,
2304 talloc_free (message_id);
2306 if (message == NULL) {
2307 ret = COERCE_STATUS (private_status,
2308 "Unexpected status value from _notmuch_message_create_for_message_id");
2312 _notmuch_message_add_filename (message, filename);
2314 /* Is this a newly created message object or a ghost
2315 * message? We have to be slightly careful: if this is a
2316 * blank message, it's not safe to call
2317 * notmuch_message_get_flag yet. */
2318 if (private_status == NOTMUCH_PRIVATE_STATUS_NO_DOCUMENT_FOUND ||
2319 (is_ghost = notmuch_message_get_flag (
2320 message, NOTMUCH_MESSAGE_FLAG_GHOST))) {
2321 _notmuch_message_add_term (message, "type", "mail");
2323 /* Convert ghost message to a regular message */
2324 _notmuch_message_remove_term (message, "type", "ghost");
2326 ret = _notmuch_database_link_message (notmuch, message,
2327 message_file, is_ghost);
2331 date = _notmuch_message_file_get_header (message_file, "date");
2332 _notmuch_message_set_header_values (message, date, from, subject);
2334 ret = _notmuch_message_index_file (message, message_file);
2338 ret = NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID;
2341 _notmuch_message_sync (message);
2342 } catch (const Xapian::Error &error) {
2343 fprintf (stderr, "A Xapian exception occurred adding message: %s.\n",
2344 error.get_msg().c_str());
2345 notmuch->exception_reported = TRUE;
2346 ret = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
2352 if ((ret == NOTMUCH_STATUS_SUCCESS ||
2353 ret == NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID) && message_ret)
2354 *message_ret = message;
2356 notmuch_message_destroy (message);
2360 _notmuch_message_file_close (message_file);
2362 ret2 = notmuch_database_end_atomic (notmuch);
2363 if ((ret == NOTMUCH_STATUS_SUCCESS ||
2364 ret == NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID) &&
2365 ret2 != NOTMUCH_STATUS_SUCCESS)
2372 notmuch_database_remove_message (notmuch_database_t *notmuch,
2373 const char *filename)
2375 notmuch_status_t status;
2376 notmuch_message_t *message;
2378 status = notmuch_database_find_message_by_filename (notmuch, filename,
2381 if (status == NOTMUCH_STATUS_SUCCESS && message) {
2382 status = _notmuch_message_remove_filename (message, filename);
2383 if (status == NOTMUCH_STATUS_SUCCESS)
2384 _notmuch_message_delete (message);
2385 else if (status == NOTMUCH_STATUS_DUPLICATE_MESSAGE_ID)
2386 _notmuch_message_sync (message);
2388 notmuch_message_destroy (message);
2395 notmuch_database_find_message_by_filename (notmuch_database_t *notmuch,
2396 const char *filename,
2397 notmuch_message_t **message_ret)
2400 const char *prefix = _find_prefix ("file-direntry");
2401 char *direntry, *term;
2402 Xapian::PostingIterator i, end;
2403 notmuch_status_t status;
2405 if (message_ret == NULL)
2406 return NOTMUCH_STATUS_NULL_POINTER;
2408 if (! (notmuch->features & NOTMUCH_FEATURE_FILE_TERMS))
2409 return NOTMUCH_STATUS_UPGRADE_REQUIRED;
2411 /* return NULL on any failure */
2412 *message_ret = NULL;
2414 local = talloc_new (notmuch);
2417 status = _notmuch_database_filename_to_direntry (
2418 local, notmuch, filename, NOTMUCH_FIND_LOOKUP, &direntry);
2419 if (status || !direntry)
2422 term = talloc_asprintf (local, "%s%s", prefix, direntry);
2424 find_doc_ids_for_term (notmuch, term, &i, &end);
2427 notmuch_private_status_t private_status;
2429 *message_ret = _notmuch_message_create (notmuch, notmuch, *i,
2431 if (*message_ret == NULL)
2432 status = NOTMUCH_STATUS_OUT_OF_MEMORY;
2434 } catch (const Xapian::Error &error) {
2435 fprintf (stderr, "Error: A Xapian exception occurred finding message by filename: %s\n",
2436 error.get_msg().c_str());
2437 notmuch->exception_reported = TRUE;
2438 status = NOTMUCH_STATUS_XAPIAN_EXCEPTION;
2442 talloc_free (local);
2444 if (status && *message_ret) {
2445 notmuch_message_destroy (*message_ret);
2446 *message_ret = NULL;
2451 notmuch_string_list_t *
2452 _notmuch_database_get_terms_with_prefix (void *ctx, Xapian::TermIterator &i,
2453 Xapian::TermIterator &end,
2456 int prefix_len = strlen (prefix);
2457 notmuch_string_list_t *list;
2459 list = _notmuch_string_list_create (ctx);
2460 if (unlikely (list == NULL))
2463 for (i.skip_to (prefix); i != end; i++) {
2464 /* Terminate loop at first term without desired prefix. */
2465 if (strncmp ((*i).c_str (), prefix, prefix_len))
2468 _notmuch_string_list_append (list, (*i).c_str () + prefix_len);
2475 notmuch_database_get_all_tags (notmuch_database_t *db)
2477 Xapian::TermIterator i, end;
2478 notmuch_string_list_t *tags;
2481 i = db->xapian_db->allterms_begin();
2482 end = db->xapian_db->allterms_end();
2483 tags = _notmuch_database_get_terms_with_prefix (db, i, end,
2484 _find_prefix ("tag"));
2485 _notmuch_string_list_sort (tags);
2486 return _notmuch_tags_create (db, tags);
2487 } catch (const Xapian::Error &error) {
2488 fprintf (stderr, "A Xapian exception occurred getting tags: %s.\n",
2489 error.get_msg().c_str());
2490 db->exception_reported = TRUE;