1 package Koha::SearchEngine::Elasticsearch::QueryBuilder;
3 # This file is part of Koha.
5 # Copyright 2014 Catalyst IT Ltd.
7 # Koha is free software; you can redistribute it and/or modify it
8 # under the terms of the GNU General Public License as published by
9 # the Free Software Foundation; either version 3 of the License, or
10 # (at your option) any later version.
12 # Koha is distributed in the hope that it will be useful, but
13 # WITHOUT ANY WARRANTY; without even the implied warranty of
14 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 # GNU General Public License for more details.
17 # You should have received a copy of the GNU General Public License
18 # along with Koha; if not, see <http://www.gnu.org/licenses>.
22 Koha::SearchEngine::Elasticsearch::QueryBuilder - constructs elasticsearch
23 query objects from user-supplied queries
27 This provides the functions that take a user-supplied search query, and
28 provides something that can be given to elasticsearch to get answers.
32 use Koha::SearchEngine::Elasticsearch;
33 $builder = Koha::SearchEngine::Elasticsearch->new();
34 my $simple_query = $builder->build_query("hello");
35 # This is currently undocumented because the original code is undocumented
36 my $adv_query = $builder->build_advanced_query($indexes, $operands, $operators);
42 use base qw(Class::Accessor);
44 use List::MoreUtils qw/ each_array /;
48 use Data::Dumper; # TODO remove
52 my $simple_query = $builder->build_query("hello", %options)
54 This will build a query that can be issued to elasticsearch from the provided
55 string input. This expects a lucene style search form (see
56 L<http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-query-string-query.html#query-string-syntax>
59 It'll make an attempt to respect the various query options.
61 Additional options can be provided with the C<%options> hash.
67 This should be an arrayref of hashrefs, each containing a C<field> and an
68 C<direction> (optional, defaults to C<asc>.) The results will be sorted
69 according to these values. Valid values for C<direction> are 'asc' and 'desc'.
76 my ( $self, $query, %options ) = @_;
78 my $stemming = C4::Context->preference("QueryStemming") || 0;
79 my $auto_truncation = C4::Context->preference("QueryAutoTruncate") || 0;
80 my $weight_fields = C4::Context->preference("QueryWeightFields") || 0;
81 my $fuzzy_enabled = C4::Context->preference("QueryFuzzy") || 0;
83 $query = '*' unless defined $query;
89 fuzziness => $fuzzy_enabled ? 'auto' : '0',
90 default_operator => "AND",
91 default_field => "_all",
95 if ( $options{sort} ) {
96 foreach my $sort ( @{ $options{sort} } ) {
97 my ( $f, $d ) = @$sort{qw/ field direction /};
98 die "Invalid sort direction, $d"
99 if $d && ( $d ne 'asc' && $d ne 'desc' );
100 $d = 'asc' unless $d;
102 # TODO account for fields that don't have a 'phrase' type
103 push @{ $res->{sort} }, { "$f.phrase" => { order => $d } };
107 # See _convert_facets in Search.pm for how these get turned into
108 # things that Koha can use.
110 author => { terms => { field => "author__facet" } },
111 subject => { terms => { field => "subject__facet" } },
112 itype => { terms => { field => "itype__facet" } },
117 =head2 build_browse_query
119 my $browse_query = $builder->build_browse_query($field, $query);
121 This performs a "starts with" style query on a particular field. The field
122 to be searched must have been indexed with an appropriate mapping as a
126 # XXX this isn't really a browse query like we want in the end
127 sub build_browse_query {
128 my ( $self, $field, $query ) = @_;
130 my $fuzzy_enabled = C4::Context->preference("QueryFuzzy") || 0;
132 return { query => '*' } if !defined $query;
134 # TODO this should come from Koha::Elasticsearch
135 my %field_whitelist = (
139 $field = 'title' if !exists $field_whitelist{$field};
143 match_phrase_prefix => {
147 fuzziness => $fuzzy_enabled ? 'auto' : '0',
151 sort => [ { "$field.phrase" => { order => "asc" } } ],
155 =head2 build_query_compat
158 $error, $query, $simple_query, $query_cgi,
159 $query_desc, $limit, $limit_cgi, $limit_desc,
160 $stopwords_removed, $query_type
162 = $builder->build_query_compat( \@operators, \@operands, \@indexes,
163 \@limits, \@sort_by, $scan, $lang );
165 This handles a search using the same api as L<C4::Search::buildQuery> does.
167 A very simple query will go in with C<$operands> set to ['query'], and
168 C<$sort_by> set to ['pubdate_dsc']. This simple case will return with
169 C<$query> set to something that can perform the search, C<$simple_query>
170 set to just the search term, C<$query_cgi> set to something that can
171 reproduce this search, and C<$query_desc> set to something else.
175 sub build_query_compat {
176 my ( $self, $operators, $operands, $indexes, $orig_limits, $sort_by, $scan,
180 #die Dumper ( $self, $operators, $operands, $indexes, $limits, $sort_by, $scan, $lang );
181 my @sort_params = $self->_convert_sort_fields(@$sort_by);
182 my @index_params = $self->_convert_index_fields(@$indexes);
183 my $limits = $self->_fix_limit_special_cases($orig_limits);
185 # Merge the indexes in with the search terms and the operands so that
186 # each search thing is a handy unit.
187 unshift @$operators, undef; # The first one can't have an op
189 my $ea = each_array( @$operands, @$operators, @index_params );
190 while ( my ( $oand, $otor, $index ) = $ea->() ) {
191 next if ( !defined($oand) || $oand eq '' );
192 push @search_params, {
193 operand => $self->_clean_search_term($oand), # the search terms
194 operator => defined($otor) ? uc $otor : undef, # AND and so on
195 $index ? %$index : (),
199 # We build a string query from limits and the queries. An alternative
200 # would be to pass them separately into build_query and let it build
201 # them into a structured ES query itself. Maybe later, though that'd be
203 my $query_str = join( ' AND ',
204 join( ' ', $self->_create_query_string(@search_params) ),
205 $self->_join_queries( $self->_convert_index_strings(@$limits) ) );
207 # If there's no query on the left, let's remove the junk left behind
208 $query_str =~ s/^ AND //;
210 $options{sort} = \@sort_params;
211 my $query = $self->build_query( $query_str, %options );
214 # We roughly emulate the CGI parameters of the zebra query builder
215 my $query_cgi = 'idx=kw&q=' . uri_escape( $operands->[0] ) if @$operands;
216 my $simple_query = $operands->[0] if @$operands == 1;
217 my $query_desc = $simple_query;
218 my $limit = 'and ' . join( ' and ', @$limits );
220 '&limit=' . join( '&limit=', map { uri_escape($_) } @$orig_limits );
221 my $limit_desc = "@$limits";
224 undef, $query, $simple_query, $query_cgi, $query_desc,
225 $limit, $limit_cgi, $limit_desc, undef, undef
229 =head2 build_authorities_query
231 my $query = $builder->build_authorities_query(\%search);
233 This takes a nice description of an authority search and turns it into a black-box
234 query that can then be passed to the appropriate searcher.
236 The search description is a hashref that looks something like:
241 where => 'Heading', # search the main entry
242 operator => 'exact', # require an exact match
243 value => 'frogs', # the search string
246 where => '', # search all entries
247 operator => '', # default keyword, right truncation
255 authtypecode => 'TOPIC_TERM',
260 sub build_authorities_query {
261 my ($self, $search) = @_;
263 # Start by making the query parts
266 foreach my $s ( @{ $search->{searches} } ) {
267 my ($wh, $op, $val) = $s->{'where', 'operator', 'value'};
269 if ($op eq 'is' || $op eq '=') {
270 # look for something that matches completely
271 # note, '=' is about numerical vals. May need special handling.
272 push @filter_parts, { filter => { term => { $wh => $val }} };
273 } elsif ($op eq 'exact') {
274 # left and right truncation, otherwise an exact phrase
275 push @query_parts, { match_phrase => { $wh => $val }};
277 # regular wordlist stuff
279 push @query_parts, { match => { $wh => $val }};
282 # Merge the query and filter parts appropriately
287 =head2 build_authorities_query_compat
290 $builder->build_authorities_query_compat( \@marclist, \@and_or,
291 \@excluding, \@operator, \@value, $authtypecode, $orderby );
293 This builds a query for searching for authorities.
301 An arrayref containing where the particular term should be searched for.
302 Options are: mainmainentry, mainentry, match, match-heading, see-from, and
303 thesaurus. If left blank, any field is used.
307 Totally ignored. It is never used in L<C4::AuthoritiesMarc::SearchAuthorities>.
315 What form of search to do. Options are: is (phrase, no trunction, whole field
316 must match), = (number exact match), exact (phrase, but with left and right
317 truncation). If left blank, then word list, right truncted, anywhere is used.
321 The actual user-provided string value to search for.
325 The authority type code to search within. If blank, then all will be searched.
329 The order to sort the results by. Options are Relevance, HeadingAsc,
330 HeadingDsc, AuthidAsc, AuthidDsc.
334 marclist, operator, and value must be the same length, and the values at
335 index /i/ all relate to each other.
337 This returns a query, which is a black box object that can be passed to the
338 appropriate search object.
342 sub build_authorities_query_compat {
343 my ( $self, $marclist, $and_or, $excluding, $operator, $value,
344 $authtypecode, $orderby )
347 # This turns the old-style many-options argument form into a more
348 # extensible hash form that is understood by L<build_authorities_query>.
351 my %koha_to_index_name = (
352 mainmainentry => 'Heading-Main',
353 mainentry => 'Heading',
355 'match-heading' => 'Match-heading',
356 'see-from' => 'Match-heading-see-from',
357 thesaurus => 'Subject-heading-thesaurus',
361 # Make sure everything exists
362 foreach my $m (@$marclist) {
363 confess "Invalid marclist field provided: $m" unless exists $koha_to_index_name{$m};
365 for ( my $i = 0 ; $i < @$value ; $i++ ) {
368 where => $marclist->[$i],
369 operator => $operator->[$i],
370 value => $value->[$i],
376 ( $orderby =~ /^Heading/ ) ? 'Heading'
377 : ( $orderby =~ /^Auth/ ) ? 'Local-Number'
380 my $sort_order = ( $orderby =~ /Asc$/ ) ? 'asc' : 'desc';
381 %sort = ( $orderby => $sort_order, );
384 searches => \@searches,
385 authtypecode => $authtypecode,
387 $search{sort} = \%sort if %sort;
388 my $query = $self->build_authorities_query( \%search );
392 =head2 _convert_sort_fields
394 my @sort_params = _convert_sort_fields(@sort_by)
396 Converts the zebra-style sort index information into elasticsearch-style.
398 C<@sort_by> is the same as presented to L<build_query_compat>, and it returns
399 something that can be sent to L<build_query>.
403 sub _convert_sort_fields {
404 my ( $self, @sort_by ) = @_;
406 # Turn the sorting into something we care about.
407 my %sort_field_convert = (
408 acqdate => 'acqdate',
410 call_number => 'callnum',
411 popularity => 'issues',
412 relevance => undef, # default
414 pubdate => 'pubdate',
416 my %sort_order_convert =
417 ( qw( dsc desc ), qw( asc asc ), qw( az asc ), qw( za desc ) );
419 # Convert the fields and orders, drop anything we don't know about.
420 grep { $_->{field} } map {
421 my ( $f, $d ) = split /_/;
423 field => $sort_field_convert{$f},
424 direction => $sort_order_convert{$d}
429 =head2 _convert_index_fields
431 my @index_params = $self->_convert_index_fields(@indexes);
433 Converts zebra-style search index notation into elasticsearch-style.
435 C<@indexes> is an array of index names, as presented to L<build_query_compat>,
436 and it returns something that can be sent to L<build_query>.
438 B<TODO>: this will pull from the elasticsearch mappings table to figure out
443 our %index_field_convert = (
449 'se' => 'title-series',
450 'callnum' => 'callnum',
451 'mc-itype' => 'itype',
453 'branch' => 'homebranch',
459 sub _convert_index_fields {
460 my ( $self, @indexes ) = @_;
462 my %index_type_convert =
463 ( __default => undef, phr => 'phrase', rtrn => 'right-truncate' );
465 # Convert according to our table, drop anything that doesn't convert
466 grep { $_->{field} } map {
467 my ( $f, $t ) = split /,/;
469 field => $index_field_convert{$f},
470 type => $index_type_convert{ $t // '__default' }
475 =head2 _convert_index_strings
477 my @searches = $self->_convert_index_strings(@searches);
479 Similar to L<_convert_index_fields>, this takes strings of the form
480 B<field:search term> and rewrites the field from zebra-style to
481 elasticsearch-style. Anything it doesn't understand is returned verbatim.
485 sub _convert_index_strings {
486 my ( $self, @searches ) = @_;
489 foreach my $s (@searches) {
491 my ( $field, $term ) = $s =~ /^\s*([\w,-]*?):(.*)/;
492 unless ( defined($field) && defined($term) ) {
496 my ($conv) = $self->_convert_index_fields($field);
497 unless ( defined($conv) ) {
501 push @res, $conv->{field} . ":"
502 . $self->_modify_string_by_type( %$conv, operand => $term );
507 =head2 _modify_string_by_type
509 my $str = $self->_modify_string_by_type(%index_field);
511 If you have a search term (operand) and a type (phrase, right-truncated), this
512 will convert the string to have the function in lucene search terms, e.g.
513 wrapping quotes around it.
517 sub _modify_string_by_type {
518 my ( $self, %idx ) = @_;
520 my $type = $idx{type} || '';
521 my $str = $idx{operand};
522 return $str unless $str; # Empty or undef, we can't use it.
524 $str .= '*' if $type eq 'right-truncate';
525 $str = '"' . $str . '"' if $type eq 'phrase';
529 =head2 _convert_index_strings_freeform
531 my $search = $self->_convert_index_strings_freeform($search);
533 This is similar to L<_convert_index_strings>, however it'll search out the
534 things to change within the string. So it can handle strings such as
535 C<(su:foo) AND (su:bar)>, converting the C<su> appropriately.
539 sub _convert_index_strings_freeform {
540 my ( $self, $search ) = @_;
542 while ( my ( $zeb, $es ) = each %index_field_convert ) {
543 $search =~ s/\b$zeb:/$es:/g;
550 my $query_str = $self->_join_queries(@query_parts);
552 This takes a list of query parts, that might be search terms on their own, or
553 booleaned together, or specifying fields, or whatever, wraps them in
554 parentheses, and ANDs them all together. Suitable for feeding to the ES
560 my ( $self, @parts ) = @_;
562 @parts = grep { defined($_) && $_ ne '' } @parts;
563 return () unless @parts;
564 return $parts[0] if @parts < 2;
565 join ' AND ', map { "($_)" } @parts;
570 my @phrased_queries = $self->_make_phrases(@query_parts);
572 This takes the supplied queries and forces them to be phrases by wrapping
573 quotes around them. It understands field prefixes, e.g. 'subject:' and puts
574 the quotes outside of them if they're there.
579 my ( $self, @parts ) = @_;
580 map { s/^\s*(\w*?:)(.*)$/$1"$2"/r } @parts;
583 =head2 _create_query_string
585 my @query_strings = $self->_create_query_string(@queries);
587 Given a list of hashrefs, it will turn them into a lucene-style query string.
588 The hash should contain field, type (both for the indexes), operator, and
593 sub _create_query_string {
594 my ( $self, @queries ) = @_;
597 my $otor = $_->{operator} ? $_->{operator} . ' ' : '';
598 my $field = $_->{field} ? $_->{field} . ':' : '';
600 my $oand = $self->_modify_string_by_type(%$_);
601 "$otor($field$oand)";
605 =head2 _clean_search_term
607 my $term = $self->_clean_search_term($term);
609 This cleans a search term by removing any funny characters that may upset
610 ES and give us an error. It also calls L<_convert_index_strings_freeform>
611 to ensure those parts are correct.
615 sub _clean_search_term {
616 my ( $self, $term ) = @_;
618 $term = $self->_convert_index_strings_freeform($term);
623 =head2 _fix_limit_special_cases
625 my $limits = $self->_fix_limit_special_cases($limits);
627 This converts any special cases that the limit specifications have into things
628 that are more readily processable by the rest of the code.
630 The argument should be an arrayref, and it'll return an arrayref.
634 sub _fix_limit_special_cases {
635 my ( $self, $limits ) = @_;
638 foreach my $l (@$limits) {
640 # This is set up by opac-search.pl
641 if ( $l =~ /^yr,st-numeric,ge=/ ) {
642 my ( $start, $end ) =
643 ( $l =~ /^yr,st-numeric,ge=(.*) and yr,st-numeric,le=(.*)$/ );
644 next unless defined($start) && defined($end);
645 push @new_lim, "copydate:[$start TO $end]";
647 elsif ( $l =~ /^yr,st-numeric=/ ) {
648 my ($date) = ( $l =~ /^yr,st-numeric=(.*)$/ );
649 next unless defined($date);
650 push @new_lim, "copydate:$date";
652 elsif ( $l =~ /^available$/ ) {
653 push @new_lim, 'onloan:false';