2 * Copyright © 2009 Carl Worth
4 * This program is free software: you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation, either version 3 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program. If not, see http://www.gnu.org/licenses/ .
17 * Author: Carl Worth <cworth@cworth.org>
20 /* This indexer creates a Xapian mail index that is remarkably similar
21 * to that created by sup. The big difference, (and the thing that
22 * will keep a notmuch index from being used by sup directly), is that
23 * sup expects a serialized ruby data structure in the document's data
24 * field, but notmuch just puts the mail's filename there (trusting
25 * that the email client can get the data in needs from the filename).
27 * Note: One bug here is that sup actually merges together fields such
28 * as To, CC, Bcc etc. when finding multiple emails with the same
29 * message ID. To support something similar, notmuch should list
30 * multiple files in the data field.
32 * Other differences between sup and notmuch-index identified so far:
34 * o sup supports encrypted mime parts by prompting for a passphrase
35 * to decrypt the message. So far, notmuch doesn't support this,
36 * both because I'm lazy to code it, and I also think doing so
37 * would present a security leak.
39 * o sup and notmuch have different heuristics for identifying (and
40 * thus ignoring) signatures. For example, sup considers a line
41 * consisting of two hypens as a signature separator, while
42 * notmuch expects those two hyphens to be followed by a space
45 * o sup as been seen to split some numbers before indexing
46 * them. For example, the number 1754 in an email message was
47 * indexed by sup as separate terms 17 and 54. I couldn't find any
48 * explanation for this behavior and did not try to replicate it
61 #include <gmime/gmime.h>
67 #define ARRAY_SIZE(arr) (sizeof (arr) / sizeof (arr[0]))
69 /* Xapian complains if we provide a term longer than this. */
70 #define NOTMUCH_MAX_TERM 245
72 /* These prefix values are specifically chosen to be compatible
73 * with sup, (http://sup.rubyforge.org), written by
74 * William Morgan <wmorgan-sup@masanjin.net>, and released
75 * under the GNU GPL v2.
83 prefix_t NORMAL_PREFIX[] = {
86 { "from_name", "FN" },
92 prefix_t BOOLEAN_PREFIX[] = {
94 { "from_email", "FE" },
100 { "attachment_extension", "O" },
106 /* Similarly, these value numbers are also chosen to be sup
110 NOTMUCH_VALUE_MESSAGE_ID = 0,
111 NOTMUCH_VALUE_THREAD = 1,
112 NOTMUCH_VALUE_DATE = 2
116 find_prefix (const char *name)
120 for (i = 0; i < ARRAY_SIZE (NORMAL_PREFIX); i++)
121 if (strcmp (name, NORMAL_PREFIX[i].name) == 0)
122 return NORMAL_PREFIX[i].prefix;
124 for (i = 0; i < ARRAY_SIZE (BOOLEAN_PREFIX); i++)
125 if (strcmp (name, BOOLEAN_PREFIX[i].name) == 0)
126 return BOOLEAN_PREFIX[i].prefix;
131 /* "128 bits of thread-id ought to be enough for anybody" */
132 #define NOTMUCH_THREAD_ID_BITS 128
133 #define NOTMUCH_THREAD_ID_DIGITS (NOTMUCH_THREAD_ID_BITS / 4)
134 typedef struct _thread_id {
135 char str[NOTMUCH_THREAD_ID_DIGITS + 1];
139 thread_id_generate (thread_id_t *thread_id)
146 urandom = fopen ("/dev/urandom", "r");
147 if (urandom == NULL) {
148 fprintf (stderr, "Error opening /dev/urandom: %s\n",
150 fprintf (stderr, "Perhaps notmuch needs some portability fixes for your platform?\n");
155 for (i = 0; i < NOTMUCH_THREAD_ID_DIGITS; i += 8) {
156 fread ((void *) &value, sizeof (value), 1, urandom);
157 sprintf (s, "%08x", value);
163 printf ("Generated thread id: %s\n", thread_id->str);
167 add_term (Xapian::Document doc,
168 const char *prefix_name,
177 prefix = find_prefix (prefix_name);
179 term = g_strdup_printf ("%s%s", prefix, value);
181 if (strlen (term) <= NOTMUCH_MAX_TERM)
188 gen_terms (Xapian::TermGenerator term_gen,
189 const char *prefix_name,
197 prefix = find_prefix (prefix_name);
199 term_gen.index_text (text, 1, prefix);
203 gen_terms_address_name (Xapian::TermGenerator term_gen,
204 InternetAddress *address,
205 const char *prefix_name)
207 if (INTERNET_ADDRESS_IS_MAILBOX(address)) {
211 name = internet_address_get_name (address);
213 /* In the absence of a name, we'll strip the part before the @
214 * from the address. */
216 InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address);
217 const char *addr = internet_address_mailbox_get_addr (mailbox);
220 at = strchr (addr, '@');
222 name = strndup (addr, at - addr);
228 gen_terms (term_gen, prefix_name, name);
231 free ((void *) name);
232 } else if (INTERNET_ADDRESS_IS_GROUP (address)) {
233 InternetAddressGroup *group = INTERNET_ADDRESS_GROUP (address);
234 InternetAddressList *list = internet_address_group_get_members(group);
236 int length = internet_address_list_length(list);
239 for (i = 0; i < length; i++)
240 gen_terms_address_name(term_gen,
241 internet_address_list_get_address(list, i),
248 gen_terms_address_names (Xapian::TermGenerator term_gen,
249 InternetAddressList *addresses,
250 const char *address_type)
253 InternetAddress *address;
255 if (addresses == NULL)
258 for (i = 0; i < internet_address_list_length (addresses); i++) {
259 address = internet_address_list_get_address (addresses, i);
260 gen_terms_address_name (term_gen, address, address_type);
261 gen_terms_address_name (term_gen, address, "name");
262 gen_terms_address_name (term_gen, address, "body");
267 add_term_address_addr (Xapian::Document doc,
268 InternetAddress *address,
269 const char *prefix_name)
271 if (INTERNET_ADDRESS_IS_MAILBOX(address)) {
272 InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address);
275 addr = internet_address_mailbox_get_addr (mailbox);
278 add_term (doc, prefix_name, addr);
279 } else if (INTERNET_ADDRESS_IS_GROUP (address)) {
280 InternetAddressGroup *group = INTERNET_ADDRESS_GROUP (address);
281 InternetAddressList *list = internet_address_group_get_members(group);
283 int length = internet_address_list_length(list);
286 for (i = 0; i < length; i++)
287 add_term_address_addr(doc,
288 internet_address_list_get_address(list, i),
295 add_terms_address_addrs (Xapian::Document doc,
296 InternetAddressList *addresses,
297 const char *address_type)
300 InternetAddress *address;
302 if (addresses == NULL)
305 for (i = 0; i < internet_address_list_length (addresses); i++) {
306 address = internet_address_list_get_address (addresses, i);
307 add_term_address_addr (doc, address, address_type);
308 add_term_address_addr (doc, address, "email");
313 skip_re_in_subject (const char *subject)
315 const char *s = subject;
321 while (*s && isspace (*s))
323 if (strncasecmp (s, "re:", 3) == 0)
333 find_messages_by_term (Xapian::Database db,
334 const char *prefix_name,
336 Xapian::PostingIterator *begin,
337 Xapian::PostingIterator *end)
339 Xapian::PostingIterator i;
342 term = g_strdup_printf ("%s%s", find_prefix (prefix_name), value);
344 *begin = db.postlist_begin (term);
347 *end = db.postlist_end (term);
353 find_message_by_docid (Xapian::Database db, Xapian::docid docid)
355 return db.get_document (docid);
359 find_message_by_message_id (Xapian::Database db, const char *message_id)
361 Xapian::PostingIterator i, end;
363 find_messages_by_term (db, "msgid", message_id, &i, &end);
366 return find_message_by_docid (db, *i);
368 return Xapian::Document ();
372 insert_thread_id (GHashTable *thread_ids, Xapian::Document doc)
375 const char *value, *id, *comma;
377 value_string = doc.get_value (NOTMUCH_VALUE_THREAD);
378 value = value_string.c_str();
379 if (strlen (value)) {
382 comma = strchr (id, ',');
384 comma = id + strlen (id);
385 g_hash_table_insert (thread_ids,
386 strndup (id, comma - id), NULL);
394 /* Return one or more thread_ids, (as a GPtrArray of strings), for the
395 * given message based on looking into the database for any messages
396 * referenced in parents, and also for any messages in the database
397 * referencing message_id.
399 * Caller should free all strings in the array and the array itself,
400 * (g_ptr_array_free) when done. */
402 find_thread_ids (Xapian::Database db,
404 const char *message_id)
406 Xapian::PostingIterator child, children_end;
407 Xapian::Document doc;
408 GHashTable *thread_ids;
411 const char *parent_message_id;
414 thread_ids = g_hash_table_new_full (g_str_hash, g_str_equal,
417 find_messages_by_term (db, "ref", message_id, &child, &children_end);
418 for ( ; child != children_end; child++) {
419 doc = find_message_by_docid (db, *child);
420 insert_thread_id (thread_ids, doc);
423 for (i = 0; i < parents->len; i++) {
424 parent_message_id = (char *) g_ptr_array_index (parents, i);
425 doc = find_message_by_message_id (db, parent_message_id);
426 insert_thread_id (thread_ids, doc);
429 result = g_ptr_array_new ();
431 keys = g_hash_table_get_keys (thread_ids);
432 for (l = keys; l; l = l->next) {
433 char *id = (char *) l->data;
434 g_ptr_array_add (result, id);
438 /* We're done with the hash table, but we've taken the pointers to
439 * the allocated strings and put them into our result array, so
440 * tell the hash not to free them on its way out. */
441 g_hash_table_steal_all (thread_ids);
442 g_hash_table_unref (thread_ids);
447 /* Add a term for each message-id in the References header of the
450 parse_references (GPtrArray *array,
451 const char *refs_str)
453 GMimeReferences *refs, *r;
454 const char *message_id;
456 if (refs_str == NULL)
459 refs = g_mime_references_decode (refs_str);
461 for (r = refs; r; r = r->next) {
462 message_id = g_mime_references_get_message_id (r);
463 g_ptr_array_add (array, g_strdup (message_id));
466 g_mime_references_free (refs);
469 /* Given a string representing the body of a message, generate terms
470 * for it, (skipping quoted portions and signatures). */
472 gen_terms_body_str (Xapian::TermGenerator term_gen,
475 char *line, *line_end, *next_line;
487 next_line = strchr (line, '\n');
488 if (next_line == NULL) {
489 next_line = line + strlen (line);
491 line_end = next_line - 1;
493 /* Get to the next non-blank line. */
494 while (*next_line == '\n')
497 /* Skip blank lines. */
501 /* Skip lines that are quotes. */
505 /* Also skip lines introducing a quote on the next line. */
506 if (*line_end == ':' && *next_line == '>')
509 /* Finally, bail as soon as we see a signature. */
510 /* XXX: Should only do this if "near" the end of the message. */
511 if (strncmp (line, "-- ", 3) == 0 ||
512 strncmp (line, "----------", 10) == 0 ||
513 strncmp (line, "__________", 10) == 0)
516 *(line_end + 1) = '\0';
517 gen_terms (term_gen, "body", line);
522 /* Callback to generate terms for each mime part of a message. */
524 gen_terms_part (Xapian::TermGenerator term_gen,
528 GMimeDataWrapper *wrapper;
529 GByteArray *byte_array;
530 GMimeContentDisposition *disposition;
533 if (GMIME_IS_MULTIPART (part)) {
534 GMimeMultipart *multipart = GMIME_MULTIPART (part);
537 for (i = 0; i < g_mime_multipart_get_count (multipart); i++) {
538 if (GMIME_IS_MULTIPART_SIGNED (multipart)) {
539 /* Don't index the signature. */
543 fprintf (stderr, "Warning: Unexpected extra parts of mutlipart/signed. Indexing anyway.\n");
545 gen_terms_part (term_gen,
546 g_mime_multipart_get_part (multipart, i));
551 if (GMIME_IS_MESSAGE_PART (part)) {
552 GMimeMessage *message;
554 message = g_mime_message_part_get_message (GMIME_MESSAGE_PART (part));
556 gen_terms_part (term_gen, g_mime_message_get_mime_part (message));
561 if (! (GMIME_IS_PART (part))) {
562 fprintf (stderr, "Warning: Not indexing unknown mime part: %s.\n",
563 g_type_name (G_OBJECT_TYPE (part)));
567 disposition = g_mime_object_get_content_disposition (part);
569 strcmp (disposition->disposition, GMIME_DISPOSITION_ATTACHMENT) == 0)
571 const char *filename = g_mime_part_get_filename (GMIME_PART (part));
572 const char *extension;
574 add_term (term_gen.get_document (), "label", "attachment");
575 gen_terms (term_gen, "attachment", filename);
578 extension = strchr (filename, '.');
580 add_term (term_gen.get_document (), "attachment_extension",
588 byte_array = g_byte_array_new ();
590 stream = g_mime_stream_mem_new_with_byte_array (byte_array);
591 g_mime_stream_mem_set_owner (GMIME_STREAM_MEM (stream), FALSE);
592 wrapper = g_mime_part_get_content_object (GMIME_PART (part));
594 g_mime_data_wrapper_write_to_stream (wrapper, stream);
596 g_object_unref (stream);
598 g_byte_array_append (byte_array, (guint8 *) "\0", 1);
599 body = (char *) g_byte_array_free (byte_array, FALSE);
601 gen_terms_body_str (term_gen, body);
607 index_file (Xapian::WritableDatabase db,
608 Xapian::TermGenerator term_gen,
609 const char *filename)
611 Xapian::Document doc;
615 GMimeMessage *message;
616 InternetAddressList *addresses;
617 GPtrArray *parents, *thread_ids;
621 const char *subject, *refs, *in_reply_to, *from;
622 const char *message_id;
625 struct tm gm_time_tm;
626 char date_str[16]; /* YYYYMMDDHHMMSS + 1 for Y100k compatibility ;-) */
629 file = fopen (filename, "r");
631 fprintf (stderr, "Error opening %s: %s\n", filename, strerror (errno));
635 stream = g_mime_stream_file_new (file);
637 parser = g_mime_parser_new_with_stream (stream);
639 message = g_mime_parser_construct_message (parser);
641 doc = Xapian::Document ();
643 doc.set_data (filename);
645 term_gen.set_stemmer (Xapian::Stem ("english"));
647 term_gen.set_document (doc);
649 from = g_mime_message_get_sender (message);
650 addresses = internet_address_list_parse_string (from);
652 gen_terms_address_names (term_gen, addresses, "from_name");
654 addresses = g_mime_message_get_all_recipients (message);
655 gen_terms_address_names (term_gen, addresses, "to_name");
657 subject = g_mime_message_get_subject (message);
658 subject = skip_re_in_subject (subject);
659 gen_terms (term_gen, "subject", subject);
660 gen_terms (term_gen, "body", subject);
662 gen_terms_part (term_gen, g_mime_message_get_mime_part (message));
664 parents = g_ptr_array_new ();
666 refs = g_mime_object_get_header (GMIME_OBJECT (message), "references");
667 parse_references (parents, refs);
669 in_reply_to = g_mime_object_get_header (GMIME_OBJECT (message),
671 parse_references (parents, in_reply_to);
673 for (i = 0; i < parents->len; i++)
674 add_term (doc, "ref", (char *) g_ptr_array_index (parents, i));
676 message_id = g_mime_message_get_message_id (message);
678 thread_ids = find_thread_ids (db, parents, message_id);
680 for (i = 0; i < parents->len; i++)
681 g_free (g_ptr_array_index (parents, i));
682 g_ptr_array_free (parents, TRUE);
684 from = g_mime_message_get_sender (message);
685 addresses = internet_address_list_parse_string (from);
687 add_terms_address_addrs (doc, addresses, "from_email");
689 add_terms_address_addrs (doc,
690 g_mime_message_get_all_recipients (message),
693 g_mime_message_get_date (message, &time, NULL);
695 gmtime_r (&time, &gm_time_tm);
697 if (strftime (date_str, sizeof (date_str),
698 "%Y%m%d%H%M%S", &gm_time_tm) == 0) {
699 fprintf (stderr, "Internal error formatting time\n");
703 add_term (doc, "date", date_str);
705 add_term (doc, "label", "inbox");
706 add_term (doc, "label", "unread");
707 add_term (doc, "type", "mail");
708 add_term (doc, "source_id", "1");
711 add_term (doc, "msgid", message_id);
712 doc.add_value (NOTMUCH_VALUE_MESSAGE_ID, message_id);
715 if (thread_ids->len) {
720 for (i = 0; i < thread_ids->len; i++) {
721 id = (char *) thread_ids->pdata[i];
723 add_term (doc, "thread", id);
726 thread_id = g_string_new (id);
728 g_string_append_printf (thread_id, ",%s", id);
732 g_ptr_array_free (thread_ids, TRUE);
734 doc.add_value (NOTMUCH_VALUE_THREAD, thread_id->str);
736 g_string_free (thread_id, TRUE);
737 } else if (message_id) {
738 /* If not part of any existing thread, generate a new thread_id. */
739 thread_id_t thread_id;
741 thread_id_generate (&thread_id);
743 add_term (doc, "thread", thread_id.str);
744 doc.add_value (NOTMUCH_VALUE_THREAD, thread_id.str);
747 doc.add_value (NOTMUCH_VALUE_DATE, Xapian::sortable_serialise (time));
749 db.add_document (doc);
751 g_object_unref (message);
752 g_object_unref (parser);
753 g_object_unref (stream);
757 usage (const char *argv0)
759 fprintf (stderr, "Usage: %s <path-to-xapian-database>\n", argv0);
760 fprintf (stderr, "\n");
761 fprintf (stderr, "Messages to be indexed are read from stdnin as absolute filenames\n");
762 fprintf (stderr, "one file per line.");
766 main (int argc, char **argv)
768 const char *database_path;
771 GIOStatus gio_status;
772 GError *error = NULL;
774 struct timeval tv_start, tv_last, tv_now;
782 database_path = argv[1];
787 Xapian::WritableDatabase db;
788 Xapian::TermGenerator term_gen;
790 db = Xapian::WritableDatabase (database_path,
791 Xapian::DB_CREATE_OR_OPEN);
793 term_gen = Xapian::TermGenerator ();
795 channel = g_io_channel_unix_new (fileno (stdin));
799 gettimeofday (&tv_start, NULL);
803 gio_status = g_io_channel_read_line (channel, &filename,
805 if (gio_status == G_IO_STATUS_EOF)
807 if (gio_status != G_IO_STATUS_NORMAL) {
808 fprintf (stderr, "An error occurred reading from stdin: %s\n",
813 g_strchomp (filename);
814 index_file (db, term_gen, filename);
819 if (count % 1000 == 0) {
820 gettimeofday (&tv_now, NULL);
821 printf ("Indexed %d messages (%g messages/second)\n",
822 count, 1000 / ((tv_now.tv_sec - tv_last.tv_sec) +
823 (tv_now.tv_usec - tv_last.tv_usec) / 1e6));
828 g_io_channel_unref (channel);
830 gettimeofday (&tv_now, NULL);
831 elapsed = (tv_now.tv_sec - tv_start.tv_sec +
832 (tv_now.tv_usec - tv_start.tv_usec) / 1e6);
833 printf ("Completed indexing of %d messages in %g seconds (%g messages/second)\n",
834 count, elapsed, count / elapsed);
836 } catch (const Xapian::Error &error) {
837 cerr << "A Xapian exception occurred: " << error.get_msg () << endl;