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
60 #include <gmime/gmime.h>
66 #define ARRAY_SIZE(arr) (sizeof (arr) / sizeof (arr[0]))
68 /* Xapian complains if we provide a term longer than this. */
69 #define NOTMUCH_MAX_TERM 245
71 /* These prefix values are specifically chosen to be compatible
72 * with sup, (http://sup.rubyforge.org), written by
73 * William Morgan <wmorgan-sup@masanjin.net>, and released
74 * under the GNU GPL v2.
82 prefix_t NORMAL_PREFIX[] = {
85 { "from_name", "FN" },
91 prefix_t BOOLEAN_PREFIX[] = {
93 { "from_email", "FE" },
99 { "attachment_extension", "O" },
105 /* Similarly, these value numbers are also chosen to be sup
109 NOTMUCH_VALUE_MESSAGE_ID = 0,
110 NOTMUCH_VALUE_THREAD = 1,
111 NOTMUCH_VALUE_DATE = 2
115 find_prefix (const char *name)
119 for (i = 0; i < ARRAY_SIZE (NORMAL_PREFIX); i++)
120 if (strcmp (name, NORMAL_PREFIX[i].name) == 0)
121 return NORMAL_PREFIX[i].prefix;
123 for (i = 0; i < ARRAY_SIZE (BOOLEAN_PREFIX); i++)
124 if (strcmp (name, BOOLEAN_PREFIX[i].name) == 0)
125 return BOOLEAN_PREFIX[i].prefix;
130 int TERM_COMBINED = 0;
133 add_term (Xapian::Document doc,
134 const char *prefix_name,
143 prefix = find_prefix (prefix_name);
145 term = g_strdup_printf ("%s%s", prefix, value);
147 if (strlen (term) <= NOTMUCH_MAX_TERM)
154 gen_terms (Xapian::TermGenerator term_gen,
155 const char *prefix_name,
163 prefix = find_prefix (prefix_name);
165 term_gen.index_text (text, 1, prefix);
169 gen_terms_address_name (Xapian::TermGenerator term_gen,
170 InternetAddress *address,
171 const char *prefix_name)
176 name = internet_address_get_name (address);
178 /* In the absence of a name, we'll strip the part before the @
179 * from the address. */
181 InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address);
182 const char *addr = internet_address_mailbox_get_addr (mailbox);
185 at = strchr (addr, '@');
187 name = strndup (addr, at - addr);
193 gen_terms (term_gen, prefix_name, name);
196 free ((void *) name);
200 gen_terms_address_names (Xapian::TermGenerator term_gen,
201 InternetAddressList *addresses,
202 const char *address_type)
205 InternetAddress *address;
207 if (addresses == NULL)
210 for (i = 0; i < internet_address_list_length (addresses); i++) {
211 address = internet_address_list_get_address (addresses, i);
212 gen_terms_address_name (term_gen, address, address_type);
213 gen_terms_address_name (term_gen, address, "name");
214 gen_terms_address_name (term_gen, address, "body");
219 add_term_address_addr (Xapian::Document doc,
220 InternetAddress *address,
221 const char *prefix_name)
223 InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address);
226 addr = internet_address_mailbox_get_addr (mailbox);
229 add_term (doc, prefix_name, addr);
233 add_terms_address_addrs (Xapian::Document doc,
234 InternetAddressList *addresses,
235 const char *address_type)
238 InternetAddress *address;
240 if (addresses == NULL)
243 for (i = 0; i < internet_address_list_length (addresses); i++) {
244 address = internet_address_list_get_address (addresses, i);
245 add_term_address_addr (doc, address, address_type);
246 add_term_address_addr (doc, address, "email");
251 skip_re_in_subject (const char *subject)
253 const char *s = subject;
256 while (*s && isspace (*s))
258 if (strncasecmp (s, "re:", 3) == 0)
268 find_messages_by_term (Xapian::Database db,
269 const char *prefix_name,
271 Xapian::PostingIterator *begin,
272 Xapian::PostingIterator *end)
274 Xapian::PostingIterator i;
277 term = g_strdup_printf ("%s%s", find_prefix (prefix_name), value);
279 *begin = db.postlist_begin (term);
282 *end = db.postlist_end (term);
288 find_message_by_docid (Xapian::Database db, Xapian::docid docid)
290 return db.get_document (docid);
294 find_message_by_message_id (Xapian::Database db, const char *message_id)
296 Xapian::PostingIterator i, end;
298 find_messages_by_term (db, "msgid", message_id, &i, &end);
301 return find_message_by_docid (db, *i);
303 return Xapian::Document ();
307 insert_thread_id (GHashTable *thread_ids, Xapian::Document doc)
310 const char *value, *id, *comma;
312 value_string = doc.get_value (NOTMUCH_VALUE_THREAD);
313 value = value_string.c_str();
314 if (strlen (value)) {
317 comma = strchr (id, ',');
319 comma = id + strlen (id);
320 g_hash_table_insert (thread_ids,
321 strndup (id, comma - id), NULL);
329 /* Return one or more thread_ids, (as a GPtrArray of strings), for the
330 * given message based on looking into the database for any messages
331 * referenced in parents, and also for any messages in the database
332 * referencing message_id.
334 * Caller should free all strings in the array and the array itself,
335 * (g_ptr_array_free) when done. */
337 find_thread_ids (Xapian::Database db,
339 const char *message_id)
341 Xapian::PostingIterator child, children_end;
342 Xapian::Document doc;
343 GHashTable *thread_ids;
346 const char *parent_message_id;
349 thread_ids = g_hash_table_new (g_str_hash, g_str_equal);
351 find_messages_by_term (db, "ref", message_id, &child, &children_end);
352 for ( ; child != children_end; child++) {
353 doc = find_message_by_docid (db, *child);
354 insert_thread_id (thread_ids, doc);
357 for (i = 0; i < parents->len; i++) {
358 parent_message_id = (char *) g_ptr_array_index (parents, i);
359 doc = find_message_by_message_id (db, parent_message_id);
360 insert_thread_id (thread_ids, doc);
363 result = g_ptr_array_new ();
365 keys = g_hash_table_get_keys (thread_ids);
366 for (l = keys; l; l = l->next) {
367 char *id = (char *) l->data;
368 g_ptr_array_add (result, id);
374 /* Add a term for each message-id in the References header of the
377 parse_references (GPtrArray *array,
378 const char *refs_str)
380 GMimeReferences *refs, *r;
381 const char *message_id;
383 if (refs_str == NULL)
386 refs = g_mime_references_decode (refs_str);
388 for (r = refs; r; r = r->next) {
389 message_id = g_mime_references_get_message_id (r);
390 g_ptr_array_add (array, g_strdup (message_id));
393 g_mime_references_free (refs);
396 /* Given a string representing the body of a message, generate terms
397 * for it, (skipping quoted portions and signatures). */
399 gen_terms_body_str (Xapian::TermGenerator term_gen,
402 char *line, *line_end, *next_line;
414 next_line = strchr (line, '\n');
415 if (next_line == NULL) {
416 next_line = line + strlen (line);
418 line_end = next_line - 1;
420 /* Get to the next non-blank line. */
421 while (*next_line == '\n')
424 /* Skip lines that are quotes. */
428 /* Also skip lines introducing a quote on the next line. */
429 if (*line_end == ':' && *next_line == '>')
432 /* Finally, bail as soon as we see a signature. */
433 /* XXX: Should only do this if "near" the end of the message. */
434 if (strncmp (line, "-- ", 3) == 0 ||
435 strncmp (line, "----------", 10) == 0 ||
436 strncmp (line, "__________", 10) == 0)
439 *(line_end + 1) = '\0';
440 gen_terms (term_gen, "body", line);
445 /* Callback to generate terms for each mime part of a message. */
447 gen_terms_part (Xapian::TermGenerator term_gen,
451 GMimeDataWrapper *wrapper;
452 GByteArray *byte_array;
453 GMimeContentDisposition *disposition;
456 if (GMIME_IS_MULTIPART (part)) {
457 GMimeMultipart *multipart = GMIME_MULTIPART (part);
460 for (i = 0; i < g_mime_multipart_get_count (multipart); i++) {
461 if (GMIME_IS_MULTIPART_SIGNED (multipart)) {
462 /* Don't index the signature. */
466 fprintf (stderr, "Warning: Unexpected extra parts of mutlipart/signed. Indexing anyway.\n");
468 gen_terms_part (term_gen,
469 g_mime_multipart_get_part (multipart, i));
474 if (! GMIME_IS_PART (part)) {
475 fprintf (stderr, "Warning: Not indexing unknown mime part: %s.\n",
476 g_type_name (G_OBJECT_TYPE (part)));
480 disposition = g_mime_object_get_content_disposition (part);
482 strcmp (disposition->disposition, GMIME_DISPOSITION_ATTACHMENT) == 0)
484 const char *filename = g_mime_part_get_filename (GMIME_PART (part));
485 const char *extension;
487 add_term (term_gen.get_document (), "label", "attachment");
488 gen_terms (term_gen, "attachment", filename);
491 extension = strchr (filename, '.');
493 add_term (term_gen.get_document (), "attachment_extension",
501 byte_array = g_byte_array_new ();
503 stream = g_mime_stream_mem_new_with_byte_array (byte_array);
504 g_mime_stream_mem_set_owner (GMIME_STREAM_MEM (stream), FALSE);
505 wrapper = g_mime_part_get_content_object (GMIME_PART (part));
507 g_mime_data_wrapper_write_to_stream (wrapper, stream);
509 g_object_unref (stream);
511 g_byte_array_append (byte_array, (guint8 *) "\0", 1);
512 body = (char *) g_byte_array_free (byte_array, FALSE);
514 gen_terms_body_str (term_gen, body);
520 index_file (Xapian::WritableDatabase db,
521 Xapian::TermGenerator term_gen,
522 const char *filename)
524 Xapian::Document doc;
528 GMimeMessage *message;
529 InternetAddressList *addresses;
530 GPtrArray *parents, *thread_ids;
534 const char *subject, *refs, *in_reply_to, *from;
535 const char *message_id;
538 struct tm gm_time_tm;
539 char date_str[16]; /* YYYYMMDDHHMMSS + 1 for Y100k compatibility ;-) */
542 file = fopen (filename, "r");
544 fprintf (stderr, "Error opening %s: %s\n", filename, strerror (errno));
548 stream = g_mime_stream_file_new (file);
550 parser = g_mime_parser_new_with_stream (stream);
552 message = g_mime_parser_construct_message (parser);
554 doc = Xapian::Document ();
556 doc.set_data (filename);
558 term_gen.set_stemmer (Xapian::Stem ("english"));
560 term_gen.set_document (doc);
562 from = g_mime_message_get_sender (message);
563 addresses = internet_address_list_parse_string (from);
565 gen_terms_address_names (term_gen, addresses, "from_name");
567 addresses = g_mime_message_get_all_recipients (message);
568 gen_terms_address_names (term_gen, addresses, "to_name");
570 subject = g_mime_message_get_subject (message);
571 subject = skip_re_in_subject (subject);
572 gen_terms (term_gen, "subject", subject);
573 gen_terms (term_gen, "body", subject);
575 gen_terms_part (term_gen, g_mime_message_get_mime_part (message));
577 parents = g_ptr_array_new ();
579 refs = g_mime_object_get_header (GMIME_OBJECT (message), "references");
580 parse_references (parents, refs);
582 in_reply_to = g_mime_object_get_header (GMIME_OBJECT (message),
584 parse_references (parents, in_reply_to);
586 for (i = 0; i < parents->len; i++)
587 add_term (doc, "ref", (char *) g_ptr_array_index (parents, i));
589 message_id = g_mime_message_get_message_id (message);
591 thread_ids = find_thread_ids (db, parents, message_id);
593 for (i = 0; i < parents->len; i++)
594 g_free (g_ptr_array_index (parents, i));
595 g_ptr_array_free (parents, TRUE);
597 from = g_mime_message_get_sender (message);
598 addresses = internet_address_list_parse_string (from);
600 add_terms_address_addrs (doc, addresses, "from_email");
602 add_terms_address_addrs (doc,
603 g_mime_message_get_all_recipients (message),
606 g_mime_message_get_date (message, &time, NULL);
608 gmtime_r (&time, &gm_time_tm);
610 if (strftime (date_str, sizeof (date_str),
611 "%Y%m%d%H%M%S", &gm_time_tm) == 0) {
612 fprintf (stderr, "Internal error formatting time\n");
616 add_term (doc, "date", date_str);
618 add_term (doc, "label", "inbox");
619 add_term (doc, "label", "unread");
620 add_term (doc, "type", "mail");
621 add_term (doc, "source_id", "1");
623 add_term (doc, "msgid", message_id);
624 doc.add_value (NOTMUCH_VALUE_MESSAGE_ID, message_id);
626 if (thread_ids->len) {
631 for (i = 0; i < thread_ids->len; i++) {
632 id = (char *) thread_ids->pdata[i];
634 add_term (doc, "thread", id);
637 thread_id = g_string_new (id);
639 g_string_append_printf (thread_id, ",%s", id);
643 g_ptr_array_free (thread_ids, TRUE);
645 doc.add_value (NOTMUCH_VALUE_THREAD, thread_id->str);
647 g_string_free (thread_id, TRUE);
649 /* If not referenced thread, use the message ID */
650 add_term (doc, "thread", message_id);
651 doc.add_value (NOTMUCH_VALUE_THREAD, message_id);
654 doc.add_value (NOTMUCH_VALUE_DATE, Xapian::sortable_serialise (time));
656 db.add_document (doc);
658 g_object_unref (message);
659 g_object_unref (parser);
660 g_object_unref (stream);
664 usage (const char *argv0)
666 fprintf (stderr, "Usage: %s <path-to-xapian-database>\n", argv0);
667 fprintf (stderr, "\n");
668 fprintf (stderr, "Messages to be indexed are read from stdnin as absolute filenames\n");
669 fprintf (stderr, "one file per line.");
673 main (int argc, char **argv)
675 const char *database_path;
678 GIOStatus gio_status;
679 GError *error = NULL;
686 database_path = argv[1];
691 Xapian::WritableDatabase db;
692 Xapian::TermGenerator term_gen;
694 db = Xapian::WritableDatabase (database_path,
695 Xapian::DB_CREATE_OR_OPEN);
697 term_gen = Xapian::TermGenerator ();
699 channel = g_io_channel_unix_new (fileno (stdin));
702 gio_status = g_io_channel_read_line (channel, &filename,
704 if (gio_status == G_IO_STATUS_EOF)
706 if (gio_status != G_IO_STATUS_NORMAL) {
707 fprintf (stderr, "An error occurred reading from stdin: %s\n",
712 g_strchomp (filename);
713 index_file (db, term_gen, filename);
718 } catch (const Xapian::Error &error) {
719 cerr << "A Xapian exception occurred: " << error.get_msg () << endl;