1 # Copyright (C) 2015 all contributors <meta@public-inbox.org>
2 # License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
3 # based on notmuch, but with no concept of folders, files or flags
5 # Read-only search interface for use by the web and NNTP interfaces
6 package PublicInbox::Search;
10 # values for searching
11 use constant TS => 0; # timestamp
12 use constant NUM => 1; # NNTP article number
13 use constant BYTES => 2; # :bytes as defined in RFC 3977
14 use constant LINES => 3; # :lines as defined in RFC 3977
15 use constant YYYYMMDD => 4; # for searching in the WWW UI
17 use Search::Xapian qw/:standard/;
18 use PublicInbox::SearchMsg;
19 use PublicInbox::MIME;
20 use PublicInbox::MID qw/mid_clean id_compress/;
22 # This is English-only, everything else is non-standard and may be confused as
23 # a prefix common in patch emails
24 our $REPLY_RE = qr/^re:\s+/i;
25 our $LANG = 'english';
28 # SCHEMA_VERSION history
30 # 1 - subject_path is lower-cased
31 # 2 - subject_path is id_compress in the index, only
32 # 3 - message-ID is compressed if it includes '%' (hack!)
33 # 4 - change "Re: " normalization, avoid circular Reference ghosts
34 # 5 - subject_path drops trailing '.'
35 # 6 - preserve References: order in document data
36 # 7 - remove references and inreplyto terms
37 # 8 - remove redundant/unneeded document data
38 # 9 - disable Message-ID compression (SHA-1)
39 # 10 - optimize doc for NNTP overviews
40 # 11 - merge threads when vivifying ghosts
41 # 12 - change YYYYMMDD value column to numeric
42 # 13 - fix threading for empty References/In-Reply-To
43 # (commit 83425ef12e4b65cdcecd11ddcb38175d4a91d5a0)
46 # n.b. FLAG_PURE_NOT is expensive not suitable for a public website
47 # as it could become a denial-of-service vector
48 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
52 my %bool_pfx_internal = (
53 type => 'T', # "mail" or "ghost"
54 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
57 my %bool_pfx_external = (
58 # do we still need these? probably not..
60 mid => 'Q', # uniQue id (Message-ID)
64 # for mairix compatibility
66 m => 'XMID', # 'mid:' (bool) is exact, 'm:' (prob) can do partial
81 '' => 'XMID S A XNQ XQUOT XFN',
84 # not documenting m: and mid: for now, the using the URLs works w/o Xapian
86 's:' => 'match within Subject e.g. s:"a quick brown fox"',
88 date range as YYYYMMDD e.g. d:19931002..20101002
89 Open-ended ranges such as d:19931002.. and d:..20101002
92 'b:' => 'match within message body, including text attachments',
93 'nq:' => 'match non-quoted text within message body',
94 'quot:' => 'match quoted text within message body',
95 'n:' => 'match filename of attachment(s)',
96 't:' => 'match within the To header',
97 'c:' => 'match within the Cc header',
98 'f:' => 'match within the From header',
99 'a:' => 'match within the To, Cc, and From headers',
100 'tc:' => 'match within the To and Cc headers',
101 'bs:' => 'match within the Subject and body',
105 # df (filenames from diff)
106 # da (diff a/ removed lines)
107 # db (diff b/ added lines)
109 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
111 sub xpfx { $all_pfx{$_[0]} }
113 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
116 my (undef, $git_dir) = @_;
117 "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
121 my ($class, $git_dir, $altid) = @_;
122 my $dir = $class->xdir($git_dir);
123 my $db = Search::Xapian::Database->new($dir);
124 bless { xdb => $db, git_dir => $git_dir, altid => $altid }, $class;
127 sub reopen { $_[0]->{xdb}->reopen }
131 my ($self, $query_string, $opts) = @_;
135 unless ($query_string eq '') {
136 $query = $self->qp->parse_query($query_string, QP_FLAGS);
137 $opts->{relevance} = 1 unless exists $opts->{relevance};
140 _do_enquire($self, $query, $opts);
144 my ($self, $mid, $opts) = @_;
145 my $smsg = eval { $self->lookup_message($mid) };
147 return { total => 0, msgs => [] } unless $smsg;
148 my $qtid = Search::Xapian::Query->new(xpfx('thread').$smsg->thread_id);
149 my $path = $smsg->path;
150 if (defined $path && $path ne '') {
151 my $path = id_compress($smsg->path);
152 my $qsub = Search::Xapian::Query->new(xpfx('path').$path);
153 $qtid = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
156 $opts->{limit} ||= 1000;
158 # always sort threads by timestamp, this makes life easier
159 # for the threading algorithm (in SearchThread.pm)
162 _do_enquire($self, $qtid, $opts);
166 my ($self, $cb) = @_;
169 eval { $ret = $cb->() };
170 return $ret unless $@;
171 # Exception: The revision being read has been discarded -
172 # you should call Xapian::Database::reopen()
173 if (ref($@) eq 'Search::Xapian::DatabaseModifiedError') {
182 my ($self, $query, $opts) = @_;
183 retry_reopen($self, sub { _enquire_once($self, $query, $opts) });
187 my ($self, $query, $opts) = @_;
188 my $enquire = $self->enquire;
189 if (defined $query) {
190 $query = Search::Xapian::Query->new(OP_AND,$query,$mail_query);
192 $query = $mail_query;
194 $enquire->set_query($query);
196 my $desc = !$opts->{asc};
197 if ($opts->{relevance}) {
198 $enquire->set_sort_by_relevance_then_value(TS, $desc);
199 } elsif ($opts->{num}) {
200 $enquire->set_sort_by_value(NUM, 0);
202 $enquire->set_sort_by_value_then_relevance(TS, $desc);
204 my $offset = $opts->{offset} || 0;
205 my $limit = $opts->{limit} || 50;
206 my $mset = $enquire->get_mset($offset, $limit);
207 return $mset if $opts->{mset};
209 PublicInbox::SearchMsg->load_doc($_->get_document);
212 { total => $mset->get_matches_estimated, msgs => \@msgs }
216 sub stemmer { Search::Xapian::Stem->new($LANG) }
222 my $qp = $self->{query_parser};
226 $qp = Search::Xapian::QueryParser->new;
227 $qp->set_default_op(OP_AND);
228 $qp->set_database($self->{xdb});
229 $qp->set_stemmer($self->stemmer);
230 $qp->set_stemming_strategy(STEM_SOME);
231 $qp->add_valuerangeprocessor(
232 Search::Xapian::NumberValueRangeProcessor->new(YYYYMMDD, 'd:'));
234 while (my ($name, $prefix) = each %bool_pfx_external) {
235 $qp->add_boolean_prefix($name, $prefix);
238 # we do not actually create AltId objects,
239 # just parse the spec to avoid the extra DB handles for now.
240 if (my $altid = $self->{altid}) {
241 my $user_pfx = $self->{-user_pfx} ||= [];
243 # $_ = 'serial:gmane:/path/to/gmane.msgmap.sqlite3'
244 /\Aserial:(\w+):/ or next;
246 push @$user_pfx, "$pfx:", <<EOF;
247 alternate serial number e.g. $pfx:12345 (boolean)
250 $qp->add_boolean_prefix($pfx, 'X'.uc($pfx));
255 while (my ($name, $prefix) = each %prob_prefix) {
256 $qp->add_prefix($name, $_) foreach split(/ /, $prefix);
259 $self->{query_parser} = $qp;
262 sub num_range_processor {
263 $_[0]->{nrp} ||= Search::Xapian::NumberValueRangeProcessor->new(NUM);
266 # only used for NNTP server
268 my ($self, $beg, $end, $offset) = @_;
269 my $qp = Search::Xapian::QueryParser->new;
270 $qp->set_database($self->{xdb});
271 $qp->add_valuerangeprocessor($self->num_range_processor);
272 my $query = $qp->parse_query("$beg..$end", QP_FLAGS);
274 _do_enquire($self, $query, {num => 1, limit => 200, offset => $offset});
278 my ($self, $mid) = @_;
279 $mid = mid_clean($mid);
281 my $doc_id = $self->find_unique_doc_id('mid', $mid);
283 if (defined $doc_id) {
285 my $doc = $self->{xdb}->get_document($doc_id);
286 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
287 $smsg->{doc_id} = $doc_id;
292 sub lookup_mail { # no ghosts!
293 my ($self, $mid) = @_;
294 retry_reopen($self, sub {
295 my $smsg = lookup_message($self, $mid) or return;
296 PublicInbox::SearchMsg->load_doc($smsg->{doc});
300 sub find_unique_doc_id {
301 my ($self, $term, $value) = @_;
303 my ($begin, $end) = $self->find_doc_ids($term, $value);
305 return undef if $begin->equal($end); # not found
307 my $rv = $begin->get_docid;
311 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
315 # returns begin and end PostingIterator
317 my ($self, $term, $value) = @_;
319 $self->find_doc_ids_for_term(xpfx($term) . $value);
322 # returns begin and end PostingIterator
323 sub find_doc_ids_for_term {
324 my ($self, $term) = @_;
325 my $db = $self->{xdb};
327 ($db->postlist_begin($term), $db->postlist_end($term));
330 # normalize subjects so they are suitable as pathnames for URLs
331 # XXX: consider for removal
334 $subj = subject_normalized($subj);
335 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
339 sub subject_normalized {
341 $subj =~ s/\A\s+//s; # no leading space
342 $subj =~ s/\s+\z//s; # no trailing space
343 $subj =~ s/\s+/ /gs; # no redundant spaces
344 $subj =~ s/\.+\z//; # no trailing '.'
345 $subj =~ s/$REPLY_RE//igo; # remove reply prefix
351 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
356 $self->qp; # parse altids
358 if (my $user_pfx = $self->{-user_pfx}) {
359 push @ret, @$user_pfx;