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 for (i = 0; i < internet_address_list_length (addresses); i++) {
208 address = internet_address_list_get_address (addresses, i);
209 gen_terms_address_name (term_gen, address, address_type);
210 gen_terms_address_name (term_gen, address, "name");
211 gen_terms_address_name (term_gen, address, "body");
216 add_term_address_addr (Xapian::Document doc,
217 InternetAddress *address,
218 const char *prefix_name)
220 InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address);
223 addr = internet_address_mailbox_get_addr (mailbox);
226 add_term (doc, prefix_name, addr);
230 add_terms_address_addrs (Xapian::Document doc,
231 InternetAddressList *addresses,
232 const char *address_type)
235 InternetAddress *address;
237 for (i = 0; i < internet_address_list_length (addresses); i++) {
238 address = internet_address_list_get_address (addresses, i);
239 add_term_address_addr (doc, address, address_type);
240 add_term_address_addr (doc, address, "email");
245 skip_re_in_subject (const char *subject)
247 const char *s = subject;
250 while (*s && isspace (*s))
252 if (strncasecmp (s, "re:", 3) == 0)
262 find_messages_by_term (Xapian::Database db,
263 const char *prefix_name,
265 Xapian::PostingIterator *begin,
266 Xapian::PostingIterator *end)
268 Xapian::PostingIterator i;
271 term = g_strdup_printf ("%s%s", find_prefix (prefix_name), value);
273 *begin = db.postlist_begin (term);
276 *end = db.postlist_end (term);
282 find_message_by_docid (Xapian::Database db, Xapian::docid docid)
284 return db.get_document (docid);
288 find_message_by_message_id (Xapian::Database db, const char *message_id)
290 Xapian::PostingIterator i, end;
292 find_messages_by_term (db, "msgid", message_id, &i, &end);
295 return find_message_by_docid (db, *i);
297 return Xapian::Document ();
301 insert_thread_id (GHashTable *thread_ids, Xapian::Document doc)
304 const char *value, *id, *comma;
306 value_string = doc.get_value (NOTMUCH_VALUE_THREAD);
307 value = value_string.c_str();
308 if (strlen (value)) {
311 comma = strchr (id, ',');
313 comma = id + strlen (id);
314 g_hash_table_insert (thread_ids,
315 strndup (id, comma - id), NULL);
323 /* Return one or more thread_ids, (as a GPtrArray of strings), for the
324 * given message based on looking into the database for any messages
325 * referenced in parents, and also for any messages in the database
326 * referencing message_id.
328 * Caller should free all strings in the array and the array itself,
329 * (g_ptr_array_free) when done. */
331 find_thread_ids (Xapian::Database db,
333 const char *message_id)
335 Xapian::PostingIterator child, children_end;
336 Xapian::Document doc;
337 GHashTable *thread_ids;
340 const char *parent_message_id;
343 thread_ids = g_hash_table_new (g_str_hash, g_str_equal);
345 find_messages_by_term (db, "ref", message_id, &child, &children_end);
346 for ( ; child != children_end; child++) {
347 doc = find_message_by_docid (db, *child);
348 insert_thread_id (thread_ids, doc);
351 for (i = 0; i < parents->len; i++) {
352 parent_message_id = (char *) g_ptr_array_index (parents, i);
353 doc = find_message_by_message_id (db, parent_message_id);
354 insert_thread_id (thread_ids, doc);
357 result = g_ptr_array_new ();
359 keys = g_hash_table_get_keys (thread_ids);
360 for (l = keys; l; l = l->next) {
361 char *id = (char *) l->data;
362 g_ptr_array_add (result, id);
368 /* Add a term for each message-id in the References header of the
371 parse_references (GPtrArray *array,
372 const char *refs_str)
374 GMimeReferences *refs, *r;
375 const char *message_id;
377 if (refs_str == NULL)
380 refs = g_mime_references_decode (refs_str);
382 for (r = refs; r; r = r->next) {
383 message_id = g_mime_references_get_message_id (r);
384 g_ptr_array_add (array, g_strdup (message_id));
387 g_mime_references_free (refs);
390 /* Given a string representing the body of a message, generate terms
391 * for it, (skipping quoted portions and signatures). */
393 gen_terms_body_str (Xapian::TermGenerator term_gen,
396 char *line, *line_end, *next_line;
408 next_line = strchr (line, '\n');
409 if (next_line == NULL) {
410 next_line = line + strlen (line);
412 line_end = next_line - 1;
414 /* Get to the next non-blank line. */
415 while (*next_line == '\n')
418 /* Skip lines that are quotes. */
422 /* Also skip lines introducing a quote on the next line. */
423 if (*line_end == ':' && *next_line == '>')
426 /* Finally, bail as soon as we see a signature. */
427 /* XXX: Should only do this if "near" the end of the message. */
428 if (strncmp (line, "-- ", 3) == 0 ||
429 strncmp (line, "----------", 10) == 0 ||
430 strncmp (line, "__________", 10) == 0)
433 *(line_end + 1) = '\0';
434 gen_terms (term_gen, "body", line);
439 /* Callback to generate terms for each mime part of a message. */
441 gen_terms_part (Xapian::TermGenerator term_gen,
445 GMimeDataWrapper *wrapper;
446 GByteArray *byte_array;
447 GMimeContentDisposition *disposition;
450 if (GMIME_IS_MULTIPART (part)) {
451 GMimeMultipart *multipart = GMIME_MULTIPART (part);
454 for (i = 0; i < g_mime_multipart_get_count (multipart); i++) {
455 if (GMIME_IS_MULTIPART_SIGNED (multipart)) {
456 /* Don't index the signature. */
460 fprintf (stderr, "Warning: Unexpected extra parts of mutlipart/signed. Indexing anyway.\n");
462 gen_terms_part (term_gen,
463 g_mime_multipart_get_part (multipart, i));
468 if (! GMIME_IS_PART (part)) {
469 fprintf (stderr, "Warning: Not indexing unknown mime part: %s.\n",
470 g_type_name (G_OBJECT_TYPE (part)));
474 disposition = g_mime_object_get_content_disposition (part);
476 strcmp (disposition->disposition, GMIME_DISPOSITION_ATTACHMENT) == 0)
478 const char *filename = g_mime_part_get_filename (GMIME_PART (part));
479 const char *extension;
481 add_term (term_gen.get_document (), "label", "attachment");
482 gen_terms (term_gen, "attachment", filename);
485 extension = strchr (filename, '.');
487 add_term (term_gen.get_document (), "attachment_extension",
495 byte_array = g_byte_array_new ();
497 stream = g_mime_stream_mem_new_with_byte_array (byte_array);
498 g_mime_stream_mem_set_owner (GMIME_STREAM_MEM (stream), FALSE);
499 wrapper = g_mime_part_get_content_object (GMIME_PART (part));
500 g_mime_data_wrapper_write_to_stream (wrapper, stream);
502 g_object_unref (stream);
504 g_byte_array_append (byte_array, (guint8 *) "\0", 1);
505 body = (char *) g_byte_array_free (byte_array, FALSE);
507 gen_terms_body_str (term_gen, body);
513 index_file (Xapian::WritableDatabase db,
514 Xapian::TermGenerator term_gen,
515 const char *filename)
517 Xapian::Document doc;
521 GMimeMessage *message;
522 InternetAddressList *addresses;
523 GPtrArray *parents, *thread_ids;
527 const char *subject, *refs, *in_reply_to, *from;
528 const char *message_id;
531 struct tm gm_time_tm;
532 char date_str[16]; /* YYYYMMDDHHMMSS + 1 for Y100k compatibility ;-) */
535 file = fopen (filename, "r");
537 fprintf (stderr, "Error opening %s: %s\n", filename, strerror (errno));
541 stream = g_mime_stream_file_new (file);
543 parser = g_mime_parser_new_with_stream (stream);
545 message = g_mime_parser_construct_message (parser);
547 doc = Xapian::Document ();
549 doc.set_data (filename);
551 term_gen.set_stemmer (Xapian::Stem ("english"));
553 term_gen.set_document (doc);
555 from = g_mime_message_get_sender (message);
556 addresses = internet_address_list_parse_string (from);
558 gen_terms_address_names (term_gen, addresses, "from_name");
560 addresses = g_mime_message_get_all_recipients (message);
561 gen_terms_address_names (term_gen, addresses, "to_name");
563 subject = g_mime_message_get_subject (message);
564 subject = skip_re_in_subject (subject);
565 gen_terms (term_gen, "subject", subject);
566 gen_terms (term_gen, "body", subject);
568 gen_terms_part (term_gen, g_mime_message_get_mime_part (message));
570 parents = g_ptr_array_new ();
572 refs = g_mime_object_get_header (GMIME_OBJECT (message), "references");
573 parse_references (parents, refs);
575 in_reply_to = g_mime_object_get_header (GMIME_OBJECT (message),
577 parse_references (parents, in_reply_to);
579 for (i = 0; i < parents->len; i++)
580 add_term (doc, "ref", (char *) g_ptr_array_index (parents, i));
582 message_id = g_mime_message_get_message_id (message);
584 thread_ids = find_thread_ids (db, parents, message_id);
586 for (i = 0; i < parents->len; i++)
587 g_free (g_ptr_array_index (parents, i));
588 g_ptr_array_free (parents, TRUE);
590 from = g_mime_message_get_sender (message);
591 addresses = internet_address_list_parse_string (from);
593 add_terms_address_addrs (doc, addresses, "from_email");
595 add_terms_address_addrs (doc,
596 g_mime_message_get_all_recipients (message),
599 g_mime_message_get_date (message, &time, NULL);
601 gmtime_r (&time, &gm_time_tm);
603 if (strftime (date_str, sizeof (date_str),
604 "%Y%m%d%H%M%S", &gm_time_tm) == 0) {
605 fprintf (stderr, "Internal error formatting time\n");
609 add_term (doc, "date", date_str);
611 add_term (doc, "label", "inbox");
612 add_term (doc, "label", "unread");
613 add_term (doc, "type", "mail");
614 add_term (doc, "source_id", "1");
616 add_term (doc, "msgid", message_id);
617 doc.add_value (NOTMUCH_VALUE_MESSAGE_ID, message_id);
619 if (thread_ids->len) {
624 for (i = 0; i < thread_ids->len; i++) {
625 id = (char *) thread_ids->pdata[i];
627 add_term (doc, "thread", id);
630 thread_id = g_string_new (id);
632 g_string_append_printf (thread_id, ",%s", id);
636 g_ptr_array_free (thread_ids, TRUE);
638 doc.add_value (NOTMUCH_VALUE_THREAD, thread_id->str);
640 g_string_free (thread_id, TRUE);
642 /* If not referenced thread, use the message ID */
643 add_term (doc, "thread", message_id);
644 doc.add_value (NOTMUCH_VALUE_THREAD, message_id);
647 doc.add_value (NOTMUCH_VALUE_DATE, Xapian::sortable_serialise (time));
649 db.add_document (doc);
651 g_object_unref (message);
652 g_object_unref (parser);
653 g_object_unref (stream);
657 usage (const char *argv0)
659 fprintf (stderr, "Usage: %s <path-to-xapian-database>\n", argv0);
660 fprintf (stderr, "\n");
661 fprintf (stderr, "Messages to be indexed are read from stdnin as absolute filenames\n");
662 fprintf (stderr, "one file per line.");
666 main (int argc, char **argv)
668 const char *database_path;
671 GIOStatus gio_status;
672 GError *error = NULL;
679 database_path = argv[1];
684 Xapian::WritableDatabase db;
685 Xapian::TermGenerator term_gen;
687 db = Xapian::WritableDatabase (database_path,
688 Xapian::DB_CREATE_OR_OPEN);
690 term_gen = Xapian::TermGenerator ();
692 channel = g_io_channel_unix_new (fileno (stdin));
695 gio_status = g_io_channel_read_line (channel, &filename,
697 if (gio_status == G_IO_STATUS_EOF)
699 if (gio_status != G_IO_STATUS_NORMAL) {
700 fprintf (stderr, "An error occurred reading from stdin: %s\n",
705 g_strchomp (filename);
706 index_file (db, term_gen, filename);
711 } catch (const Xapian::Error &error) {
712 cerr << "A Xapian exception occurred: " << error.get_msg () << endl;