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::QueryBuilder;
33 $builder = Koha::SearchEngine::Elasticsearch->new({ index => $index });
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(Koha::SearchEngine::Elasticsearch);
45 use List::MoreUtils qw/ each_array /;
54 my $simple_query = $builder->build_query("hello", %options)
56 This will build a query that can be issued to elasticsearch from the provided
57 string input. This expects a lucene style search form (see
58 L<http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-query-string-query.html#query-string-syntax>
61 It'll make an attempt to respect the various query options.
63 Additional options can be provided with the C<%options> hash.
69 This should be an arrayref of hashrefs, each containing a C<field> and an
70 C<direction> (optional, defaults to C<asc>.) The results will be sorted
71 according to these values. Valid values for C<direction> are 'asc' and 'desc'.
78 my ( $self, $query, %options ) = @_;
80 my $stemming = C4::Context->preference("QueryStemming") || 0;
81 my $auto_truncation = C4::Context->preference("QueryAutoTruncate") || 0;
82 my $weight_fields = C4::Context->preference("QueryWeightFields") || 0;
83 my $fuzzy_enabled = C4::Context->preference("QueryFuzzy") || 0;
85 $query = '*' unless defined $query;
91 fuzziness => $fuzzy_enabled ? 'auto' : '0',
92 default_operator => 'AND',
93 default_field => '_all',
94 lenient => JSON::true,
95 fields => $options{fields} || [],
99 if ( $options{sort} ) {
100 foreach my $sort ( @{ $options{sort} } ) {
101 my ( $f, $d ) = @$sort{qw/ field direction /};
102 die "Invalid sort direction, $d"
103 if $d && ( $d ne 'asc' && $d ne 'desc' );
104 $d = 'asc' unless $d;
106 $f = $self->_sort_field($f);
107 push @{ $res->{sort} }, { $f => { order => $d } };
111 # See _convert_facets in Search.pm for how these get turned into
112 # things that Koha can use.
113 $res->{aggregations} = {
114 author => { terms => { field => "author__facet" } },
115 subject => { terms => { field => "subject__facet" } },
116 itype => { terms => { field => "itype__facet" } },
117 location => { terms => { field => "location__facet" } },
118 'su-geo' => { terms => { field => "su-geo__facet" } },
119 'title-series' => { terms => { field => "title-series__facet" } },
120 ccode => { terms => { field => "ccode__facet" } },
121 ln => { terms => { field => "ln__facet" } },
124 my $display_library_facets = C4::Context->preference('DisplayLibraryFacets');
125 if ( $display_library_facets eq 'both'
126 or $display_library_facets eq 'home' ) {
127 $res->{aggregations}{homebranch} = { terms => { field => "homebranch__facet" } };
129 if ( $display_library_facets eq 'both'
130 or $display_library_facets eq 'holding' ) {
131 $res->{aggregations}{holdingbranch} = { terms => { field => "holdingbranch__facet" } };
133 if ( my $ef = $options{expanded_facet} ) {
134 $res->{aggregations}{$ef}{terms}{size} = C4::Context->preference('FacetMaxCount');
139 =head2 build_browse_query
141 my $browse_query = $builder->build_browse_query($field, $query);
143 This performs a "starts with" style query on a particular field. The field
144 to be searched must have been indexed with an appropriate mapping as a
145 "phrase" subfield, which pretty much everything has.
149 # XXX this isn't really a browse query like we want in the end
150 sub build_browse_query {
151 my ( $self, $field, $query ) = @_;
153 my $fuzzy_enabled = C4::Context->preference("QueryFuzzy") || 0;
155 return { query => '*' } if !defined $query;
157 # TODO this should come from Koha::SearchEngine::Elasticsearch
158 my %field_whitelist = (
162 $field = 'title' if !exists $field_whitelist{$field};
163 my $sort = $self->_sort_field($field);
166 match_phrase_prefix => {
170 fuzziness => $fuzzy_enabled ? 'auto' : '0',
174 sort => [ { $sort => { order => "asc" } } ],
178 =head2 build_query_compat
181 $error, $query, $simple_query, $query_cgi,
182 $query_desc, $limit, $limit_cgi, $limit_desc,
183 $stopwords_removed, $query_type
185 = $builder->build_query_compat( \@operators, \@operands, \@indexes,
186 \@limits, \@sort_by, $scan, $lang );
188 This handles a search using the same api as L<C4::Search::buildQuery> does.
190 A very simple query will go in with C<$operands> set to ['query'], and
191 C<$sort_by> set to ['pubdate_dsc']. This simple case will return with
192 C<$query> set to something that can perform the search, C<$simple_query>
193 set to just the search term, C<$query_cgi> set to something that can
194 reproduce this search, and C<$query_desc> set to something else.
198 sub build_query_compat {
199 my ( $self, $operators, $operands, $indexes, $orig_limits, $sort_by, $scan,
203 #die Dumper ( $self, $operators, $operands, $indexes, $orig_limits, $sort_by, $scan, $lang );
204 my @sort_params = $self->_convert_sort_fields(@$sort_by);
205 my @index_params = $self->_convert_index_fields(@$indexes);
206 my $limits = $self->_fix_limit_special_cases($orig_limits);
207 if ( $params->{suppress} ) { push @$limits, "suppress:0"; }
209 # Merge the indexes in with the search terms and the operands so that
210 # each search thing is a handy unit.
211 unshift @$operators, undef; # The first one can't have an op
213 my $truncate = C4::Context->preference("QueryAutoTruncate") || 0;
214 my $ea = each_array( @$operands, @$operators, @index_params );
215 while ( my ( $oand, $otor, $index ) = $ea->() ) {
216 next if ( !defined($oand) || $oand eq '' );
217 $oand = $self->_clean_search_term($oand);
218 $oand = $self->_truncate_terms($oand) if ($truncate);
219 push @search_params, {
220 operand => $oand, # the search terms
221 operator => defined($otor) ? uc $otor : undef, # AND and so on
222 $index ? %$index : (),
226 # We build a string query from limits and the queries. An alternative
227 # would be to pass them separately into build_query and let it build
228 # them into a structured ES query itself. Maybe later, though that'd be
230 my $query_str = join( ' AND ',
231 join( ' ', $self->_create_query_string(@search_params) ) || (),
232 $self->_join_queries( $self->_convert_index_strings(@$limits) ) || () );
235 if ( defined($params->{weighted_fields}) && $params->{weighted_fields} ) {
236 push @fields, sprintf("%s^%s", $_->name, $_->weight) for Koha::SearchFields->weighted_fields;
239 # If there's no query on the left, let's remove the junk left behind
240 $query_str =~ s/^ AND //;
242 $options{fields} = \@fields;
243 $options{sort} = \@sort_params;
244 $options{expanded_facet} = $params->{expanded_facet};
245 my $query = $self->build_query( $query_str, %options );
248 # We roughly emulate the CGI parameters of the zebra query builder
250 $query_cgi = 'q=' . uri_escape_utf8( $operands->[0] ) if @$operands;
252 $simple_query = $operands->[0] if @$operands == 1;
253 my $query_desc = $simple_query;
254 my $limit = $self->_join_queries( $self->_convert_index_strings(@$limits));
255 my $limit_cgi = ( $orig_limits and @$orig_limits )
256 ? '&limit=' . join( '&limit=', map { uri_escape_utf8($_) } @$orig_limits )
259 $limit_desc = "$limit" if $limit;
261 undef, $query, $simple_query, $query_cgi, $query_desc,
262 $limit, $limit_cgi, $limit_desc, undef, undef
266 =head2 build_authorities_query
268 my $query = $builder->build_authorities_query(\%search);
270 This takes a nice description of an authority search and turns it into a black-box
271 query that can then be passed to the appropriate searcher.
273 The search description is a hashref that looks something like:
278 where => 'Heading', # search the main entry
279 operator => 'exact', # require an exact match
280 value => 'frogs', # the search string
283 where => '', # search all entries
284 operator => '', # default keyword, right truncation
292 authtypecode => 'TOPIC_TERM',
297 sub build_authorities_query {
298 my ( $self, $search ) = @_;
300 # Start by making the query parts
303 foreach my $s ( @{ $search->{searches} } ) {
304 my ( $wh, $op, $val ) = @{$s}{qw(where operator value)};
305 $wh = '_all' if $wh eq '';
306 if ( $op eq 'is' || $op eq '=' || $op eq 'exact' ) {
308 # look for something that matches a term completely
309 # note, '=' is about numerical vals. May need special handling.
310 # Also, we lowercase our search because the ES
311 # index lowercases its values, and term searches don't get the
312 # search analyzer applied to them.
313 push @query_parts, { match_phrase => {"$wh.phrase" => lc $val} };
315 elsif ( $op eq 'start' ) {
316 # startswith search, uses lowercase untokenized version of heading
317 push @query_parts, { match_phrase_prefix => {"$wh.phrase" => lc $val} };
320 # regular wordlist stuff
321 my @tokens = $self->_split_query( $val );
322 foreach my $token ( @tokens ) {
323 $token = $self->_truncate_terms(
324 $self->_clean_search_term( $token )
327 my $query = $self->_join_queries( @tokens );
328 push @query_parts, { query_string => { default_field => $wh, query => $query } };
332 # Merge the query parts appropriately
333 # 'should' behaves like 'or'
334 # 'must' behaves like 'and'
335 # Zebra results seem to match must so using that here
336 my $query = { query =>
338 { must => \@query_parts }
343 if ( exists $search->{sort} ) {
344 foreach my $k ( keys %{ $search->{sort} } ) {
345 my $f = $self->_sort_field($k);
346 $s{$f} = $search->{sort}{$k};
348 $search->{sort} = \%s;
352 $query->{sort} = [ $search->{sort} ] if exists $search->{sort};
358 =head2 build_authorities_query_compat
361 $builder->build_authorities_query_compat( \@marclist, \@and_or,
362 \@excluding, \@operator, \@value, $authtypecode, $orderby );
364 This builds a query for searching for authorities, in the style of
365 L<C4::AuthoritiesMarc::SearchAuthorities>.
373 An arrayref containing where the particular term should be searched for.
374 Options are: mainmainentry, mainentry, match, match-heading, see-from, and
375 thesaurus. If left blank, any field is used.
379 Totally ignored. It is never used in L<C4::AuthoritiesMarc::SearchAuthorities>.
387 What form of search to do. Options are: is (phrase, no truncation, whole field
388 must match), = (number exact match), exact (phrase, no truncation, whole field
389 must match). If left blank, then word list, right truncated, anywhere is used.
393 The actual user-provided string value to search for.
397 The authority type code to search within. If blank, then all will be searched.
401 The order to sort the results by. Options are Relevance, HeadingAsc,
402 HeadingDsc, AuthidAsc, AuthidDsc.
406 marclist, operator, and value must be the same length, and the values at
407 index /i/ all relate to each other.
409 This returns a query, which is a black box object that can be passed to the
410 appropriate search object.
414 our $koha_to_index_name = {
415 mainmainentry => 'heading-main',
416 mainentry => 'heading',
418 'match-heading' => 'match-heading',
419 'see-from' => 'match-heading-see-from',
420 thesaurus => 'subject-heading-thesaurus',
425 sub build_authorities_query_compat {
426 my ( $self, $marclist, $and_or, $excluding, $operator, $value,
427 $authtypecode, $orderby )
430 # This turns the old-style many-options argument form into a more
431 # extensible hash form that is understood by L<build_authorities_query>.
434 # Convert to lower case
435 $marclist = [map(lc, @{$marclist})];
436 $orderby = lc $orderby;
438 # Make sure everything exists
439 foreach my $m (@$marclist) {
440 Koha::Exceptions::WrongParameter->throw("Invalid marclist field provided: $m")
441 unless exists $koha_to_index_name->{$m};
443 for ( my $i = 0 ; $i < @$value ; $i++ ) {
444 next unless $value->[$i]; #clean empty form values, ES doesn't like undefined searches
447 where => $koha_to_index_name->{$marclist->[$i]},
448 operator => $operator->[$i],
449 value => $value->[$i],
455 ( $orderby =~ /^heading/ ) ? 'heading'
456 : ( $orderby =~ /^auth/ ) ? 'local-number'
459 my $sort_order = ( $orderby =~ /asc$/ ) ? 'asc' : 'desc';
460 %sort = ( $sort_field => $sort_order, );
463 searches => \@searches,
464 authtypecode => $authtypecode,
466 $search{sort} = \%sort if %sort;
467 my $query = $self->build_authorities_query( \%search );
471 =head2 _convert_sort_fields
473 my @sort_params = _convert_sort_fields(@sort_by)
475 Converts the zebra-style sort index information into elasticsearch-style.
477 C<@sort_by> is the same as presented to L<build_query_compat>, and it returns
478 something that can be sent to L<build_query>.
482 sub _convert_sort_fields {
483 my ( $self, @sort_by ) = @_;
485 # Turn the sorting into something we care about.
486 my %sort_field_convert = (
487 acqdate => 'date-of-acquisition',
489 call_number => 'local-classification',
490 popularity => 'issues',
491 relevance => undef, # default
493 pubdate => 'date-of-publication',
495 my %sort_order_convert =
496 ( qw( desc desc ), qw( dsc desc ), qw( asc asc ), qw( az asc ), qw( za desc ) );
498 # Convert the fields and orders, drop anything we don't know about.
499 grep { $_->{field} } map {
500 my ( $f, $d ) = /(.+)_(.+)/;
502 field => $sort_field_convert{$f},
503 direction => $sort_order_convert{$d}
508 =head2 _convert_index_fields
510 my @index_params = $self->_convert_index_fields(@indexes);
512 Converts zebra-style search index notation into elasticsearch-style.
514 C<@indexes> is an array of index names, as presented to L<build_query_compat>,
515 and it returns something that can be sent to L<build_query>.
517 B<TODO>: this will pull from the elasticsearch mappings table to figure out
522 our %index_field_convert = (
526 'lcn' => 'local-classification',
527 'callnum' => 'local-classification',
528 'record-type' => 'rtype',
529 'mc-rtype' => 'rtype',
531 'lc-card' => 'lc-card-number',
532 'sn' => 'local-number',
533 'yr' => 'date-of-publication',
534 'pubdate' => 'date-of-publication',
535 'acqdate' => 'date-of-acquisition',
536 'date/time-last-modified' => 'date-time-last-modified',
537 'dtlm' => 'date/time-last-modified',
538 'diss' => 'dissertation-information',
541 'music-number' => 'identifier-publisher-for-music',
542 'number-music-publisher' => 'identifier-publisher-for-music',
543 'music' => 'identifier-publisher-for-music',
544 'ident' => 'identifier-standard',
545 'cpn' => 'corporate-name',
546 'cfn' => 'conference-name',
547 'pn' => 'personal-name',
552 'rcn' => 'record-control-number',
554 'su-to' => 'subject',
555 #'su-geo' => 'subject',
556 'su-ut' => 'subject',
558 'se' => 'title-series',
559 'ut' => 'title-uniform',
560 'an' => 'koha-auth-number',
561 'authority-number' => 'koha-auth-number',
564 'rank' => 'relevance',
565 'phr' => 'st-phrase',
566 'wrdl' => 'st-word-list',
567 'rt' => 'right-truncation',
568 'rtrn' => 'right-truncation',
569 'ltrn' => 'left-truncation',
570 'rltrn' => 'left-and-right',
571 'mc-itemtype' => 'itemtype',
572 'mc-ccode' => 'ccode',
573 'branch' => 'homebranch',
574 'mc-loc' => 'location',
575 'stocknumber' => 'number-local-acquisition',
576 'inv' => 'number-local-acquisition',
578 'mc-itype' => 'itype',
579 'aub' => 'author-personal-bibliography',
580 'auo' => 'author-in-order',
584 'frequency-code' => 'ff8-18',
585 'illustration-code' => 'ff8-18-21',
586 'regularity-code' => 'ff8-19',
587 'type-of-serial' => 'ff8-21',
588 'format' => 'ff8-23',
589 'conference-code' => 'ff8-29',
590 'festschrift-indicator' => 'ff8-30',
591 'index-indicator' => 'ff8-31',
594 'literature-code' => 'lf',
595 'biography' => 'bio',
597 'biography-code' => 'bio',
598 'l-format' => 'ff7-01-02',
599 'lex' => 'lexile-number',
600 'hi' => 'host-item-number',
601 'itu' => 'index-term-uncontrolled',
602 'itg' => 'index-term-genre',
604 my $field_name_pattern = '[\w\-]+';
605 my $multi_field_pattern = "(?:\\.$field_name_pattern)*";
607 sub _convert_index_fields {
608 my ( $self, @indexes ) = @_;
610 my %index_type_convert =
611 ( __default => undef, phr => 'phrase', rtrn => 'right-truncate' );
613 # Convert according to our table, drop anything that doesn't convert.
614 # If a field starts with mc- we save it as it's used (and removed) later
615 # when joining things, to indicate we make it an 'OR' join.
616 # (Sorry, this got a bit ugly after special cases were found.)
617 grep { $_->{field} } map {
618 # Lower case all field names
619 my ( $f, $t ) = map(lc, split /,/);
626 field => exists $index_field_convert{$f} ? $index_field_convert{$f} : $f,
627 type => $index_type_convert{ $t // '__default' }
629 $r->{field} = ($mc . $r->{field}) if $mc && $r->{field};
634 =head2 _convert_index_strings
636 my @searches = $self->_convert_index_strings(@searches);
638 Similar to L<_convert_index_fields>, this takes strings of the form
639 B<field:search term> and rewrites the field from zebra-style to
640 elasticsearch-style. Anything it doesn't understand is returned verbatim.
644 sub _convert_index_strings {
645 my ( $self, @searches ) = @_;
647 foreach my $s (@searches) {
649 my ( $field, $term ) = $s =~ /^\s*([\w,-]*?):(.*)/;
650 unless ( defined($field) && defined($term) ) {
654 my ($conv) = $self->_convert_index_fields($field);
655 unless ( defined($conv) ) {
659 push @res, $conv->{field} . ":"
660 . $self->_modify_string_by_type( %$conv, operand => $term );
665 =head2 _convert_index_strings_freeform
667 my $search = $self->_convert_index_strings_freeform($search);
669 This is similar to L<_convert_index_strings>, however it'll search out the
670 things to change within the string. So it can handle strings such as
671 C<(su:foo) AND (su:bar)>, converting the C<su> appropriately.
673 If there is something of the form "su,complete-subfield" or something, the
674 second part is stripped off as we can't yet handle that. Making it work
675 will have to wait for a real query parser.
679 sub _convert_index_strings_freeform {
680 my ( $self, $search ) = @_;
681 # @TODO: Currenty will alter also fields contained within quotes:
682 # `searching for "stuff cn:123"` for example will become
683 # `searching for "stuff local-number:123"
685 # Fixing this is tricky, one possibility:
686 # https://stackoverflow.com/questions/19193876/perl-regex-to-match-a-string-that-is-not-enclosed-in-quotes
687 # Still not perfect, and will not handle escaped quotes within quotes and assumes balanced quotes.
689 # Another, not so elegant, solution could be to replace all quoted content with placeholders, and put
690 # them back when processing is done.
692 # Lower case field names
693 $search =~ s/($field_name_pattern)(?:,[\w-]*)?($multi_field_pattern):/\L$1\E$2:/og;
694 # Resolve possible field aliases
695 $search =~ s/($field_name_pattern)($multi_field_pattern):/(exists $index_field_convert{$1} ? $index_field_convert{$1} : $1)."$2:"/oge;
699 =head2 _modify_string_by_type
701 my $str = $self->_modify_string_by_type(%index_field);
703 If you have a search term (operand) and a type (phrase, right-truncated), this
704 will convert the string to have the function in lucene search terms, e.g.
705 wrapping quotes around it.
709 sub _modify_string_by_type {
710 my ( $self, %idx ) = @_;
712 my $type = $idx{type} || '';
713 my $str = $idx{operand};
714 return $str unless $str; # Empty or undef, we can't use it.
716 $str .= '*' if $type eq 'right-truncate';
717 $str = '"' . $str . '"' if $type eq 'phrase';
723 my $query_str = $self->_join_queries(@query_parts);
725 This takes a list of query parts, that might be search terms on their own, or
726 booleaned together, or specifying fields, or whatever, wraps them in
727 parentheses, and ANDs them all together. Suitable for feeding to the ES
730 Note: doesn't AND them together if they specify an index that starts with "mc"
731 as that was a special case in the original code for dealing with multiple
732 choice options (you can't search for something that has an itype of A and
733 and itype of B otherwise.)
738 my ( $self, @parts ) = @_;
740 my @norm_parts = grep { defined($_) && $_ ne '' && $_ !~ /^mc-/ } @parts;
742 map { s/^mc-//r } grep { defined($_) && $_ ne '' && $_ =~ /^mc-/ } @parts;
743 return () unless @norm_parts + @mc_parts;
744 return ( @norm_parts, @mc_parts )[0] if @norm_parts + @mc_parts == 1;
746 @mc_parts ? '(' . ( join ' OR ', map { "($_)" } @mc_parts ) . ')' : ();
748 # Handy trick: $x || () inside a join means that if $x ends up as an
749 # empty string, it gets replaced with (), which makes join ignore it.
750 # (bad effect: this'll also happen to '0', this hopefully doesn't matter
753 join( ' AND ', map { "($_)" } @norm_parts ) || (),
759 my @phrased_queries = $self->_make_phrases(@query_parts);
761 This takes the supplied queries and forces them to be phrases by wrapping
762 quotes around them. It understands field prefixes, e.g. 'subject:' and puts
763 the quotes outside of them if they're there.
768 my ( $self, @parts ) = @_;
769 map { s/^\s*(\w*?:)(.*)$/$1"$2"/r } @parts;
772 =head2 _create_query_string
774 my @query_strings = $self->_create_query_string(@queries);
776 Given a list of hashrefs, it will turn them into a lucene-style query string.
777 The hash should contain field, type (both for the indexes), operator, and
782 sub _create_query_string {
783 my ( $self, @queries ) = @_;
786 my $otor = $_->{operator} ? $_->{operator} . ' ' : '';
787 my $field = $_->{field} ? $_->{field} . ':' : '';
789 my $oand = $self->_modify_string_by_type(%$_);
790 "$otor($field$oand)";
794 =head2 _clean_search_term
796 my $term = $self->_clean_search_term($term);
798 This cleans a search term by removing any funny characters that may upset
799 ES and give us an error. It also calls L<_convert_index_strings_freeform>
800 to ensure those parts are correct.
804 sub _clean_search_term {
805 my ( $self, $term ) = @_;
807 # Lookahead for checking if we are inside quotes
808 my $lookahead = '(?=(?:[^\"]*+\"[^\"]*+\")*+[^\"]*+$)';
810 # Some hardcoded searches (like with authorities) produce things like
811 # 'an=123', when it ought to be 'an:123' for our purposes.
814 $term = $self->_convert_index_strings_freeform($term);
817 # Remove unbalanced quotes
818 my $unquoted = $term;
819 my $count = ($unquoted =~ tr/"/ /);
820 if ($count % 2 == 1) {
824 # Remove unquoted colons that have whitespace on either side of them
825 $term =~ s/(\:[:\s]+|[:\s]+:)$lookahead//g;
830 =head2 _fix_limit_special_cases
832 my $limits = $self->_fix_limit_special_cases($limits);
834 This converts any special cases that the limit specifications have into things
835 that are more readily processable by the rest of the code.
837 The argument should be an arrayref, and it'll return an arrayref.
841 sub _fix_limit_special_cases {
842 my ( $self, $limits ) = @_;
845 foreach my $l (@$limits) {
847 # This is set up by opac-search.pl
848 if ( $l =~ /^yr,st-numeric,ge=/ ) {
849 my ( $start, $end ) =
850 ( $l =~ /^yr,st-numeric,ge=(.*) and yr,st-numeric,le=(.*)$/ );
851 next unless defined($start) && defined($end);
852 push @new_lim, "copydate:[$start TO $end]";
854 elsif ( $l =~ /^yr,st-numeric=/ ) {
855 my ($date) = ( $l =~ /^yr,st-numeric=(.*)$/ );
856 next unless defined($date);
857 push @new_lim, "copydate:$date";
859 elsif ( $l =~ /^available$/ ) {
860 push @new_lim, 'onloan:0';
871 my $field = $self->_sort_field($field);
873 Given a field name, this works out what the actual name of the field to sort
874 on should be. A '__sort' suffix is added for fields with a sort version, and
875 for text fields either '.phrase' (for sortable versions) or '.raw' is appended
876 to avoid sorting on a tokenized value.
883 my $mappings = $self->get_elasticsearch_mappings();
884 my $textField = defined $mappings->{data}{properties}{$f}{type} && $mappings->{data}{properties}{$f}{type} eq 'text';
885 if (!defined $self->sort_fields()->{$f} || $self->sort_fields()->{$f}) {
887 # We need to add '.phrase' to text fields, otherwise it'll sort
888 # based on the tokenised form.
889 $f .= '.phrase' if $textField;
891 # We need to add '.raw' to text fields without a sort field,
892 # otherwise it'll sort based on the tokenised form.
893 $f .= '.raw' if $textField;
898 =head2 _truncate_terms
900 my $query = $self->_truncate_terms($query);
902 Given a string query this function appends '*' wildcard to all terms except
903 operands and double quoted strings.
907 sub _truncate_terms {
908 my ( $self, $query ) = @_;
910 my @tokens = $self->_split_query( $query );
912 # Filter out empty tokens
913 my @words = grep { $_ !~ /^\s*$/ } @tokens;
915 # Append '*' to words if needed, ie. if it ends in a word character and is not a keyword
918 (/\W$/ or grep {lc($w) eq $_} qw/and or not/) ? $_ : "$_*";
921 return join ' ', @terms;
926 my @token = $self->_split_query($query_str);
928 Given a string query this function splits it to tokens taking into account
929 any field prefixes and quoted strings.
933 my $tokenize_split_re = qr/((?:${field_name_pattern}${multi_field_pattern}:)?"[^"]+"|\s+)/;
936 my ( $self, $query ) = @_;
938 # '"donald duck" title:"the mouse" and peter" get split into
939 # ['', '"donald duck"', '', ' ', '', 'title:"the mouse"', '', ' ', 'and', ' ', 'pete']
940 my @tokens = split $tokenize_split_re, $query;
942 # Filter out empty values
943 @tokens = grep( /\S/, @tokens );