X-Git-Url: https://git.notmuchmail.org/git?p=notmuch;a=blobdiff_plain;f=notmuch-index-message.cc;h=5c73825ba2f48e3d7964994d497e104c1a6464a0;hp=f7ea340f977329d5da7354a2a613ec5766259720;hb=914df660c423c078df20390a4b647edbe8fb712a;hpb=1c63ec7031926fcd06ac3293d9e402af069ea6ac diff --git a/notmuch-index-message.cc b/notmuch-index-message.cc index f7ea340f..5c73825b 100644 --- a/notmuch-index-message.cc +++ b/notmuch-index-message.cc @@ -17,11 +17,44 @@ * Author: Carl Worth */ +/* This indexer creates a Xapian mail index that is remarkably similar + * to that created by sup. The big difference, (and the thing that + * will keep a notmuch index from being used by sup directly), is that + * sup expects a serialized ruby data structure in the document's data + * field, but notmuch just puts the mail's filename there (trusting + * that the email client can get the data in needs from the filename). + * + * Note: One bug here is that sup actually merges together fields such + * as To, CC, Bcc etc. when finding multiple emails with the same + * message ID. To support something similar, notmuch should list + * multiple files in the data field. + * + * Other differences between sup and notmuch-index identified so far: + * + * o sup supports encrypted mime parts by prompting for a passphrase + * to decrypt the message. So far, notmuch doesn't support this, + * both because I'm lazy to code it, and I also think doing so + * would present a security leak. + * + * o sup and notmuch have different heuristics for identifying (and + * thus ignoring) signatures. For example, sup considers a line + * consisting of two hypens as a signature separator, while + * notmuch expects those two hyphens to be followed by a space + * character. + * + * o sup as been seen to split some numbers before indexing + * them. For example, the number 1754 in an email message was + * indexed by sup as separate terms 17 and 54. I couldn't find any + * explanation for this behavior and did not try to replicate it + * in notmuch. + */ + #include #include #include #include #include +#include #include @@ -139,11 +172,29 @@ gen_terms_address_name (Xapian::TermGenerator term_gen, const char *prefix_name) { const char *name; + int own_name = 0; name = internet_address_get_name (address); + /* In the absence of a name, we'll strip the part before the @ + * from the address. */ + if (! name) { + InternetAddressMailbox *mailbox = INTERNET_ADDRESS_MAILBOX (address); + const char *addr = internet_address_mailbox_get_addr (mailbox); + const char *at; + + at = strchr (addr, '@'); + if (at) { + name = strndup (addr, at - addr); + own_name = 1; + } + } + if (name) gen_terms (term_gen, prefix_name, name); + + if (own_name) + free ((void *) name); } static void @@ -154,6 +205,9 @@ gen_terms_address_names (Xapian::TermGenerator term_gen, int i; InternetAddress *address; + if (addresses == NULL) + return; + for (i = 0; i < internet_address_list_length (addresses); i++) { address = internet_address_list_get_address (addresses, i); gen_terms_address_name (term_gen, address, address_type); @@ -184,6 +238,9 @@ add_terms_address_addrs (Xapian::Document doc, int i; InternetAddress *address; + if (addresses == NULL) + return; + for (i = 0; i < internet_address_list_length (addresses); i++) { address = internet_address_list_get_address (addresses, i); add_term_address_addr (doc, address, address_type); @@ -196,6 +253,9 @@ skip_re_in_subject (const char *subject) { const char *s = subject; + if (subject == NULL) + return NULL; + while (*s) { while (*s && isspace (*s)) s++; @@ -251,26 +311,44 @@ static void insert_thread_id (GHashTable *thread_ids, Xapian::Document doc) { string value_string; - const char *value; + const char *value, *id, *comma; value_string = doc.get_value (NOTMUCH_VALUE_THREAD); value = value_string.c_str(); - if (strlen (value)) - g_hash_table_insert (thread_ids, strdup (value), NULL); + if (strlen (value)) { + id = value; + while (*id) { + comma = strchr (id, ','); + if (comma == NULL) + comma = id + strlen (id); + g_hash_table_insert (thread_ids, + strndup (id, comma - id), NULL); + id = comma; + if (*id) + id++; + } + } } -static char * -find_thread_id (Xapian::Database db, - GPtrArray *parents, - const char *message_id) +/* Return one or more thread_ids, (as a GPtrArray of strings), for the + * given message based on looking into the database for any messages + * referenced in parents, and also for any messages in the database + * referencing message_id. + * + * Caller should free all strings in the array and the array itself, + * (g_ptr_array_free) when done. */ +static GPtrArray * +find_thread_ids (Xapian::Database db, + GPtrArray *parents, + const char *message_id) { Xapian::PostingIterator child, children_end; Xapian::Document doc; GHashTable *thread_ids; GList *keys, *l; - GString *result = NULL; unsigned int i; const char *parent_message_id; + GPtrArray *result; thread_ids = g_hash_table_new (g_str_hash, g_str_equal); @@ -286,51 +364,37 @@ find_thread_id (Xapian::Database db, insert_thread_id (thread_ids, doc); } + result = g_ptr_array_new (); + keys = g_hash_table_get_keys (thread_ids); for (l = keys; l; l = l->next) { char *id = (char *) l->data; - if (result == NULL) { - result = g_string_new (id); - } else { - g_string_append_printf (result, ",%s", id); - } - free (id); + g_ptr_array_add (result, id); } - if (result) - return g_string_free (result, FALSE); - else - return NULL; + return result; } /* Add a term for each message-id in the References header of the * message. */ static void parse_references (GPtrArray *array, - const char *refs) + const char *refs_str) { - const char *end, *next; + GMimeReferences *refs, *r; + const char *message_id; - if (refs == NULL) + if (refs_str == NULL) return; - while (*refs) { - while (*refs && isspace (*refs)) - refs++; - if (*refs == '<') - refs++; - end = refs; - while (*end && !isspace (*end)) - end++; - next = end; - end--; - if (end > refs && *end == '>') - end--; - if (end > refs) { - g_ptr_array_add (array, g_strndup (refs, end - refs + 1)); - } - refs = next; + refs = g_mime_references_decode (refs_str); + + for (r = refs; r; r = r->next) { + message_id = g_mime_references_get_message_id (r); + g_ptr_array_add (array, g_strdup (message_id)); } + + g_mime_references_free (refs); } /* Given a string representing the body of a message, generate terms @@ -357,8 +421,8 @@ gen_terms_body_str (Xapian::TermGenerator term_gen, } line_end = next_line - 1; - /* Trim whitespace. */ - while (*next_line && isspace (*next_line)) + /* Get to the next non-blank line. */ + while (*next_line == '\n') next_line++; /* Skip lines that are quotes. */ @@ -366,9 +430,6 @@ gen_terms_body_str (Xapian::TermGenerator term_gen, continue; /* Also skip lines introducing a quote on the next line. */ - while (line_end > line && isspace (*line_end)) - line_end--; - if (*line_end == ':' && *next_line == '>') continue; @@ -387,34 +448,74 @@ gen_terms_body_str (Xapian::TermGenerator term_gen, /* Callback to generate terms for each mime part of a message. */ static void -gen_terms_part (GMimeObject *parent, - GMimeObject *part, - gpointer user_data) +gen_terms_part (Xapian::TermGenerator term_gen, + GMimeObject *part) { - Xapian::TermGenerator *term_gen = (Xapian::TermGenerator *) user_data; GMimeStream *stream; GMimeDataWrapper *wrapper; GByteArray *byte_array; + GMimeContentDisposition *disposition; char *body; + if (GMIME_IS_MULTIPART (part)) { + GMimeMultipart *multipart = GMIME_MULTIPART (part); + int i; + + for (i = 0; i < g_mime_multipart_get_count (multipart); i++) { + if (GMIME_IS_MULTIPART_SIGNED (multipart)) { + /* Don't index the signature. */ + if (i == 1) + continue; + if (i > 1) + fprintf (stderr, "Warning: Unexpected extra parts of mutlipart/signed. Indexing anyway.\n"); + } + gen_terms_part (term_gen, + g_mime_multipart_get_part (multipart, i)); + } + return; + } + if (! GMIME_IS_PART (part)) { fprintf (stderr, "Warning: Not indexing unknown mime part: %s.\n", g_type_name (G_OBJECT_TYPE (part))); return; } + disposition = g_mime_object_get_content_disposition (part); + if (disposition && + strcmp (disposition->disposition, GMIME_DISPOSITION_ATTACHMENT) == 0) + { + const char *filename = g_mime_part_get_filename (GMIME_PART (part)); + const char *extension; + + add_term (term_gen.get_document (), "label", "attachment"); + gen_terms (term_gen, "attachment", filename); + + if (filename) { + extension = strchr (filename, '.'); + if (extension) { + add_term (term_gen.get_document (), "attachment_extension", + extension + 1); + } + } + + return; + } + byte_array = g_byte_array_new (); stream = g_mime_stream_mem_new_with_byte_array (byte_array); g_mime_stream_mem_set_owner (GMIME_STREAM_MEM (stream), FALSE); wrapper = g_mime_part_get_content_object (GMIME_PART (part)); - g_mime_data_wrapper_write_to_stream (wrapper, stream); + if (wrapper) + g_mime_data_wrapper_write_to_stream (wrapper, stream); g_object_unref (stream); + g_byte_array_append (byte_array, (guint8 *) "\0", 1); body = (char *) g_byte_array_free (byte_array, FALSE); - gen_terms_body_str (*term_gen, body); + gen_terms_body_str (term_gen, body); free (body); } @@ -430,12 +531,12 @@ index_file (Xapian::WritableDatabase db, GMimeParser *parser; GMimeMessage *message; InternetAddressList *addresses; - GPtrArray *parents; + GPtrArray *parents, *thread_ids; FILE *file; const char *subject, *refs, *in_reply_to, *from; - const char *message_id, *thread_id; + const char *message_id; time_t time; struct tm gm_time_tm; @@ -475,7 +576,7 @@ index_file (Xapian::WritableDatabase db, gen_terms (term_gen, "subject", subject); gen_terms (term_gen, "body", subject); - g_mime_message_foreach (message, gen_terms_part, &term_gen); + gen_terms_part (term_gen, g_mime_message_get_mime_part (message)); parents = g_ptr_array_new (); @@ -491,7 +592,7 @@ index_file (Xapian::WritableDatabase db, message_id = g_mime_message_get_message_id (message); - thread_id = find_thread_id (db, parents, message_id); + thread_ids = find_thread_ids (db, parents, message_id); for (i = 0; i < parents->len; i++) g_free (g_ptr_array_index (parents, i)); @@ -526,10 +627,28 @@ index_file (Xapian::WritableDatabase db, add_term (doc, "msgid", message_id); doc.add_value (NOTMUCH_VALUE_MESSAGE_ID, message_id); - if (thread_id) { - add_term (doc, "thread", thread_id); - doc.add_value (NOTMUCH_VALUE_THREAD, thread_id); - free ((void *) thread_id); + if (thread_ids->len) { + unsigned int i; + GString *thread_id; + char *id; + + for (i = 0; i < thread_ids->len; i++) { + id = (char *) thread_ids->pdata[i]; + + add_term (doc, "thread", id); + + if (i == 0) + thread_id = g_string_new (id); + else + g_string_append_printf (thread_id, ",%s", id); + + free (id); + } + g_ptr_array_free (thread_ids, TRUE); + + doc.add_value (NOTMUCH_VALUE_THREAD, thread_id->str); + + g_string_free (thread_id, TRUE); } else { /* If not referenced thread, use the message ID */ add_term (doc, "thread", message_id); @@ -562,6 +681,8 @@ main (int argc, char **argv) GIOChannel *channel; GIOStatus gio_status; GError *error = NULL; + int count; + struct timeval tv_start, tv_now; if (argc < 2) { usage (argv[0]); @@ -583,6 +704,10 @@ main (int argc, char **argv) channel = g_io_channel_unix_new (fileno (stdin)); + count = 0; + + gettimeofday (&tv_start, NULL); + while (1) { gio_status = g_io_channel_read_line (channel, &filename, NULL, NULL, &error); @@ -598,6 +723,14 @@ main (int argc, char **argv) index_file (db, term_gen, filename); g_free (filename); + + count++; + if (count % 250 == 0) { + gettimeofday (&tv_now, NULL); + printf ("Indexed %d messages (%g messages/second)\n", + count, count / ((tv_now.tv_sec - tv_start.tv_sec) + + (tv_now.tv_usec - tv_start.tv_usec) / 1e6)); + } } } catch (const Xapian::Error &error) {