1 # Copyright (C) 2015-2018 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <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)
44 # 14 - fix ghost root vivification
47 # n.b. FLAG_PURE_NOT is expensive not suitable for a public website
48 # as it could become a denial-of-service vector
49 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
53 my %bool_pfx_internal = (
54 type => 'T', # "mail" or "ghost"
55 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
58 my %bool_pfx_external = (
59 mid => 'XMID', # Message-ID (full/exact)
63 # for mairix compatibility
65 m => 'XM', # 'mid:' (bool) is exact, 'm:' (prob) can do partial
85 dfblob => 'XDFPRE XDFPOST',
88 '' => 'XM S A XNQ XQUOT XFN',
91 # not documenting m: and mid: for now, the using the URLs works w/o Xapian
93 's:' => 'match within Subject e.g. s:"a quick brown fox"',
95 date range as YYYYMMDD e.g. d:19931002..20101002
96 Open-ended ranges such as d:19931002.. and d:..20101002
99 'b:' => 'match within message body, including text attachments',
100 'nq:' => 'match non-quoted text within message body',
101 'q:' => 'match quoted text within message body',
102 'n:' => 'match filename of attachment(s)',
103 't:' => 'match within the To header',
104 'c:' => 'match within the Cc header',
105 'f:' => 'match within the From header',
106 'a:' => 'match within the To, Cc, and From headers',
107 'tc:' => 'match within the To and Cc headers',
108 'bs:' => 'match within the Subject and body',
109 'dfn:' => 'match filename from diff',
110 'dfa:' => 'match diff removed (-) lines',
111 'dfb:' => 'match diff added (+) lines',
112 'dfhh:' => 'match diff hunk header context (usually a function name)',
113 'dfctx:' => 'match diff context lines',
114 'dfpre:' => 'match pre-image git blob ID',
115 'dfpost:' => 'match post-image git blob ID',
116 'dfblob:' => 'match either pre or post-image git blob ID',
120 my $mail_query = Search::Xapian::Query->new('T' . 'mail');
124 if ($self->{version} == 1) {
125 "$self->{mainrepo}/public-inbox/xapian" . SCHEMA_VERSION;
127 my $dir = "$self->{mainrepo}/xap" . SCHEMA_VERSION;
128 my $part = $self->{partition};
129 defined $part or die "partition not given";
135 my ($class, $mainrepo, $altid) = @_;
139 $version = $ibx->{version} || 1;
140 $mainrepo = $ibx->{mainrepo};
143 mainrepo => $mainrepo,
148 my $dir = "$self->{mainrepo}/xap" . SCHEMA_VERSION;
151 foreach my $part (<$dir/*>) {
152 -d $part && $part =~ m!/\d+\z! or next;
154 my $sub = Search::Xapian::Database->new($part);
156 $xdb->add_database($sub);
161 warn "v2 repo with $parts found in $dir\n";
163 $self->{skel} = Search::Xapian::Database->new("$dir/all");
165 $self->{xdb} = Search::Xapian::Database->new($self->xdir);
170 sub reopen { $_[0]->{xdb}->reopen }
174 my ($self, $query_string, $opts) = @_;
178 unless ($query_string eq '') {
179 $query = $self->qp->parse_query($query_string, QP_FLAGS);
180 $opts->{relevance} = 1 unless exists $opts->{relevance};
183 _do_enquire($self, $query, $opts);
187 my ($self, $mid, $opts) = @_;
188 my $smsg = eval { $self->lookup_skeleton($mid) };
190 return { total => 0, msgs => [] } unless $smsg;
191 my $qtid = Search::Xapian::Query->new('G' . $smsg->thread_id);
192 my $path = $smsg->path;
193 if (defined $path && $path ne '') {
194 my $path = id_compress($smsg->path);
195 my $qsub = Search::Xapian::Query->new('XPATH' . $path);
196 $qtid = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
199 $opts->{limit} ||= 1000;
201 # always sort threads by timestamp, this makes life easier
202 # for the threading algorithm (in SearchThread.pm)
205 _do_enquire($self, $qtid, $opts);
209 my ($self, $cb) = @_;
212 eval { $ret = $cb->() };
213 return $ret unless $@;
214 # Exception: The revision being read has been discarded -
215 # you should call Xapian::Database::reopen()
216 if (ref($@) eq 'Search::Xapian::DatabaseModifiedError') {
225 my ($self, $query, $opts) = @_;
226 retry_reopen($self, sub { _enquire_once($self, $query, $opts) });
230 my ($self, $query, $opts) = @_;
231 my $enquire = $self->enquire;
232 if (defined $query) {
233 $query = Search::Xapian::Query->new(OP_AND,$query,$mail_query);
235 $query = $mail_query;
237 $enquire->set_query($query);
239 my $desc = !$opts->{asc};
240 if ($opts->{relevance}) {
241 $enquire->set_sort_by_relevance_then_value(TS, $desc);
242 } elsif ($opts->{num}) {
243 $enquire->set_sort_by_value(NUM, 0);
245 $enquire->set_sort_by_value_then_relevance(TS, $desc);
247 my $offset = $opts->{offset} || 0;
248 my $limit = $opts->{limit} || 50;
249 my $mset = $enquire->get_mset($offset, $limit);
250 return $mset if $opts->{mset};
252 PublicInbox::SearchMsg->load_doc($_->get_document);
255 { total => $mset->get_matches_estimated, msgs => \@msgs }
259 sub stemmer { Search::Xapian::Stem->new($LANG) }
265 my $qp = $self->{query_parser};
269 $qp = Search::Xapian::QueryParser->new;
270 $qp->set_default_op(OP_AND);
271 $qp->set_database($self->{xdb});
272 $qp->set_stemmer($self->stemmer);
273 $qp->set_stemming_strategy(STEM_SOME);
274 $qp->add_valuerangeprocessor(
275 Search::Xapian::NumberValueRangeProcessor->new(YYYYMMDD, 'd:'));
277 while (my ($name, $prefix) = each %bool_pfx_external) {
278 $qp->add_boolean_prefix($name, $prefix);
281 # we do not actually create AltId objects,
282 # just parse the spec to avoid the extra DB handles for now.
283 if (my $altid = $self->{altid}) {
284 my $user_pfx = $self->{-user_pfx} ||= [];
286 # $_ = 'serial:gmane:/path/to/gmane.msgmap.sqlite3'
287 /\Aserial:(\w+):/ or next;
289 push @$user_pfx, "$pfx:", <<EOF;
290 alternate serial number e.g. $pfx:12345 (boolean)
293 $qp->add_boolean_prefix($pfx, 'X'.uc($pfx));
298 while (my ($name, $prefix) = each %prob_prefix) {
299 $qp->add_prefix($name, $_) foreach split(/ /, $prefix);
302 $self->{query_parser} = $qp;
305 sub num_range_processor {
306 $_[0]->{nrp} ||= Search::Xapian::NumberValueRangeProcessor->new(NUM);
309 # only used for NNTP server
311 my ($self, $beg, $end, $offset) = @_;
312 my $qp = Search::Xapian::QueryParser->new;
313 $qp->set_database($self->{xdb});
314 $qp->add_valuerangeprocessor($self->num_range_processor);
315 my $query = $qp->parse_query("$beg..$end", QP_FLAGS);
317 _do_enquire($self, $query, {num => 1, limit => 200, offset => $offset});
320 sub lookup_skeleton {
321 my ($self, $mid) = @_;
322 my $skel = $self->{skel} or return lookup_message($self, $mid);
323 $mid = mid_clean($mid);
324 my $term = 'XMID' . $mid;
326 my $beg = $skel->postlist_begin($term);
327 if ($beg != $skel->postlist_end($term)) {
328 my $doc_id = $beg->get_docid;
329 if (defined $doc_id) {
331 my $doc = $skel->get_document($doc_id);
332 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
333 $smsg->{doc_id} = $doc_id;
340 my ($self, $mid) = @_;
341 $mid = mid_clean($mid);
343 my $doc_id = $self->find_first_doc_id('XMID' . $mid);
345 if (defined $doc_id) {
347 my $doc = $self->{xdb}->get_document($doc_id);
348 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
349 $smsg->{doc_id} = $doc_id;
354 sub lookup_mail { # no ghosts!
355 my ($self, $mid) = @_;
356 retry_reopen($self, sub {
357 my $smsg = lookup_message($self, $mid) or return;
362 sub find_unique_doc_id {
363 my ($self, $termval) = @_;
365 my ($begin, $end) = $self->find_doc_ids($termval);
367 return undef if $begin->equal($end); # not found
369 my $rv = $begin->get_docid;
373 $begin->equal($end) or die "Term '$termval' is not unique\n";
377 # returns begin and end PostingIterator
379 my ($self, $termval) = @_;
380 my $db = $self->{xdb};
382 ($db->postlist_begin($termval), $db->postlist_end($termval));
385 sub find_first_doc_id {
386 my ($self, $termval) = @_;
388 my ($begin, $end) = $self->find_doc_ids($termval);
390 return undef if $begin->equal($end); # not found
395 # normalize subjects so they are suitable as pathnames for URLs
396 # XXX: consider for removal
399 $subj = subject_normalized($subj);
400 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
404 sub subject_normalized {
406 $subj =~ s/\A\s+//s; # no leading space
407 $subj =~ s/\s+\z//s; # no trailing space
408 $subj =~ s/\s+/ /gs; # no redundant spaces
409 $subj =~ s/\.+\z//; # no trailing '.'
410 $subj =~ s/$REPLY_RE//igo; # remove reply prefix
416 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
421 $self->qp; # parse altids
423 if (my $user_pfx = $self->{-user_pfx}) {
424 push @ret, @$user_pfx;