1 package Koha::SearchEngine::Elasticsearch::Search;
3 # Copyright 2014 Catalyst IT
5 # This file is part of Koha.
7 # Koha is free software; you can redistribute it and/or modify it under the
8 # terms of the GNU General Public License as published by the Free Software
9 # Foundation; either version 3 of the License, or (at your option) any later
12 # Koha is distributed in the hope that it will be useful, but WITHOUT ANY
13 # WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
14 # A PARTICULAR PURPOSE. See the GNU General Public License for more details.
16 # You should have received a copy of the GNU General Public License along
17 # with Koha; if not, write to the Free Software Foundation, Inc.,
18 # 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
22 Koha::SearchEngine::Elasticsearch::Search - search functions for Elasticsearch
27 Koha::SearchEngine::Elasticsearch::Search->new( { index => $index } );
28 my $builder = Koha::SearchEngine::Elasticsearch::QueryBuilder->new(
29 { index => $index } );
30 my $query = $builder->build_query('perl');
31 my $results = $searcher->search($query);
32 print "There were " . $results->total . " results.\n";
43 use base qw(Koha::SearchEngine::Elasticsearch);
45 use C4::AuthoritiesMarc;
47 use Koha::AuthorisedValues;
48 use Koha::SearchEngine::QueryBuilder;
49 use Koha::SearchEngine::Search;
50 use Koha::Exceptions::Elasticsearch;
52 use Catmandu::Store::ElasticSearch;
54 use Data::Dumper; #TODO remove
58 Koha::SearchEngine::Elasticsearch::Search->mk_accessors(qw( store ));
62 my $results = $searcher->search($query, $page, $count, %options);
64 Run a search using the query. It'll return C<$count> results, starting at page
65 C<$page> (C<$page> counts from 1, anything less that, or C<undef> becomes 1.)
66 C<$count> is also the number of entries on a page.
68 C<%options> is a hash containing extra options:
74 If provided, this overrides the C<$page> value, and specifies the record as
75 an offset (i.e. the number of the record to start with), rather than a page.
84 my ($self, $query, $page, $count, %options) = @_;
86 my $params = $self->get_elasticsearch_params();
87 # 20 is the default number of results per page
88 $query->{size} = $count // 20;
89 # ES doesn't want pages, it wants a record to start from.
90 if (exists $options{offset}) {
91 $query->{from} = $options{offset};
93 $page = (!defined($page) || ($page <= 0)) ? 0 : $page - 1;
94 $query->{from} = $page * $query->{size};
96 my $elasticsearch = $self->get_elasticsearch();
98 $elasticsearch->search(
99 index => $params->{index_name},
104 die $self->process_error($@);
111 my $count = $searcher->count($query);
113 This mimics a search request, but just gets the result count instead. That's
114 faster than pulling all the data in, usually.
119 my ( $self, $query ) = @_;
121 my $params = $self->get_elasticsearch_params();
123 Catmandu::Store::ElasticSearch->new( %$params, trace_calls => 0, ) )
126 my $search = $self->store->bag->search( %$query);
127 my $count = $search->total() || 0;
133 my ( $error, $results, $facets ) = $search->search_compat(
134 $query, $simple_query, \@sort_by, \@servers,
135 $results_per_page, $offset, undef, $item_types,
139 A search interface somewhat compatible with L<C4::Search->getRecords>. Anything
140 that is returned in the query created by build_query_compat will probably
141 get ignored here, along with some other things (like C<@servers>.)
147 $self, $query, $simple_query, $sort_by,
148 $servers, $results_per_page, $offset, $branches,
149 $item_types, $query_type, $scan
153 return $self->_aggregation_scan( $query, $results_per_page, $offset );
157 if ( !defined $offset or $offset < 0 ) {
160 $options{offset} = $offset;
161 my $results = $self->search($query, undef, $results_per_page, %options);
163 # Convert each result into a MARC::Record
165 # opac-search expects results to be put in the
166 # right place in the array, according to $offset
168 my $hits = $results->{'hits'};
169 foreach my $es_record (@{$hits->{'hits'}}) {
170 $records[$index++] = $self->decode_record_from_result($es_record->{'_source'});
173 # consumers of this expect a name-spaced result, we provide the default
176 $result{biblioserver}{hits} = $hits->{'total'};
177 $result{biblioserver}{RECORDS} = \@records;
178 return (undef, \%result, $self->_convert_facets($results->{aggregations}));
181 =head2 search_auth_compat
183 my ( $results, $total ) =
184 $searcher->search_auth_compat( $query, $offset, $count, $skipmetadata, %options );
186 This has a similar calling convention to L<search>, however it returns its
187 results in a form the same as L<C4::AuthoritiesMarc::SearchAuthorities>.
191 sub search_auth_compat {
192 my ($self, $query, $offset, $count, $skipmetadata, %options) = @_;
194 if ( !defined $offset or $offset <= 0 ) {
197 # Uh, authority search uses 1-based offset..
198 $options{offset} = $offset - 1;
199 my $database = Koha::Database->new();
200 my $schema = $database->schema();
201 my $res = $self->search($query, undef, $count, %options);
203 my $bib_searcher = Koha::SearchEngine::Elasticsearch::Search->new({index => 'biblios'});
205 my $hits = $res->{'hits'};
206 foreach my $es_record (@{$hits->{'hits'}}) {
207 my $record = $es_record->{'_source'};
210 # We are using the authid to create links, we should honor the authid as stored in the db, not
211 # the 001 which, in some circumstances, can contain other data
212 my $authid = $es_record->{_id};
215 $result{authid} = $authid;
217 if (!defined $skipmetadata || !$skipmetadata) {
218 # TODO put all this info into the record at index time so we
219 # don't have to go and sort it all out now.
220 my $authtypecode = $record->{authtype};
221 my $rs = $schema->resultset('AuthType')
222 ->search( { authtypecode => $authtypecode } );
224 # FIXME there's an assumption here that we will get a result.
225 # the original code also makes an assumption that some provided
226 # authtypecode may sometimes be used instead of the one stored
227 # with the record. It's not documented why this is the case, so
228 # it's not reproduced here yet.
229 my $authtype = $rs->single;
230 my $auth_tag_to_report = $authtype ? $authtype->auth_tag_to_report : "";
231 my $marc = $self->decode_record_from_result($record);
232 my $mainentry = $marc->field($auth_tag_to_report);
235 foreach ( $mainentry->subfields() ) {
236 $reported_tag .= '$' . $_->[0] . $_->[1];
239 # Turn the resultset into a hash
240 $result{authtype} = $authtype ? $authtype->authtypetext : $authtypecode;
241 $result{reported_tag} = $reported_tag;
243 # Reimplementing BuildSummary is out of scope because it'll be hard
245 C4::AuthoritiesMarc::BuildSummary( $marc, $result{authid},
247 $result{used} = $self->count_auth_use($bib_searcher, $authid);
249 push @records, \%result;
251 return ( \@records, $hits->{'total'} );
254 =head2 count_auth_use
256 my $count = $auth_searcher->count_auth_use($bib_searcher, $authid);
258 This runs a search to determine the number of records that reference the
259 specified authid. C<$bib_searcher> must be something compatible with
260 elasticsearch, as the query is built in this function.
265 my ($self, $bib_searcher, $authid) = @_;
270 # query => { match_all => {} },
271 filter => { term => { 'koha-auth-number' => $authid } }
275 $bib_searcher->count($query);
278 =head2 simple_search_compat
280 my ( $error, $marcresults, $total_hits ) =
281 $searcher->simple_search( $query, $offset, $max_results, %options );
283 This is a simpler interface to the searching, intended to be similar enough to
284 L<C4::Search::SimpleSearch>.
292 A thing to search for. It could be a simple string, or something constructed
293 with the appropriate QueryBuilder module.
297 How many results to skip from the start of the results.
299 =item C<$max_results>
301 The max number of results to return. The default is 100 (because unlimited
302 is a pretty terrible thing to do.)
306 These options are unused by Elasticsearch
316 if something went wrong, this'll contain some kind of error
319 =item C<$marcresults>
321 an arrayref of MARC::Records (note that this is different from the
322 L<C4::Search> version which will return plain XML, but too bad.)
326 the total number of results that this search could have returned.
332 sub simple_search_compat {
333 my ($self, $query, $offset, $max_results) = @_;
335 return ('No query entered', undef, undef) unless $query;
338 $offset = 0 if not defined $offset or $offset < 0;
339 $options{offset} = $offset;
340 $max_results //= 100;
342 unless (ref $query) {
343 # We'll push it through the query builder to sanitise everything.
344 my $qb = Koha::SearchEngine::QueryBuilder->new({index => $self->index});
345 (undef,$query) = $qb->build_query_compat(undef, [$query]);
347 my $results = $self->search($query, undef, $max_results, %options);
349 my $hits = $results->{'hits'};
350 foreach my $es_record (@{$hits->{'hits'}}) {
351 push @records, $self->decode_record_from_result($es_record->{'_source'});
353 return (undef, \@records, $hits->{'total'});
356 =head2 extract_biblionumber
358 my $biblionumber = $searcher->extract_biblionumber( $searchresult );
360 $searchresult comes from simple_search_compat.
362 Returns the biblionumber from the search result record.
366 sub extract_biblionumber {
367 my ( $self, $searchresultrecord ) = @_;
368 return Koha::SearchEngine::Search::extract_biblionumber( $searchresultrecord );
371 =head2 decode_record_from_result
372 my $marc_record = $self->decode_record_from_result(@result);
374 Extracts marc data from Elasticsearch result and decodes to MARC::Record object
378 sub decode_record_from_result {
379 # Result is passed in as array, will get flattened
380 # and first element will be $result
381 my ( $self, $result ) = @_;
382 if ($result->{marc_format} eq 'base64ISO2709') {
383 return MARC::Record->new_from_usmarc(decode_base64($result->{marc_data}));
385 elsif ($result->{marc_format} eq 'MARCXML') {
386 return MARC::Record->new_from_xml($result->{marc_data}, 'UTF-8', uc C4::Context->preference('marcflavour'));
388 elsif ($result->{marc_format} eq 'ARRAY') {
389 return $self->_array_to_marc($result->{marc_data_array});
392 Koha::Exceptions::Elasticsearch->throw("Missing marc_format field in Elasticsearch result");
396 =head2 max_result_window
398 Returns the maximum number of results that can be fetched
400 This directly requests Elasticsearch for the setting index.max_result_window (or
401 the default value for this setting in case it is not set)
405 sub max_result_window {
409 Catmandu::Store::ElasticSearch->new(%{ $self->get_elasticsearch_params })
410 ) unless $self->store;
412 my $index_name = $self->store->index_name;
413 my $settings = $self->store->es->indices->get_settings(
414 index => $index_name,
415 params => { include_defaults => 'true', flat_settings => 'true' },
418 my $max_result_window = $settings->{$index_name}->{settings}->{'index.max_result_window'};
419 $max_result_window //= $settings->{$index_name}->{defaults}->{'index.max_result_window'};
421 return $max_result_window;
424 =head2 _convert_facets
426 my $koha_facets = _convert_facets($es_facets);
428 Converts elasticsearch facets types to the form that Koha expects.
429 It expects the ES facet name to match the Koha type, for example C<itype>,
430 C<au>, C<su-to>, etc.
434 sub _convert_facets {
435 my ( $self, $es, $exp_facet ) = @_;
439 # These should correspond to the ES field names, as opposed to the CCL
440 # things that zebra uses.
444 itype => 'ItemTypes',
445 location => 'Location',
446 'su-geo' => 'Places',
447 'title-series' => 'Series',
449 ccode => 'CollectionCodes',
450 holdingbranch => 'HoldingLibrary',
451 homebranch => 'HomeLibrary',
454 my @facetable_fields =
455 Koha::SearchEngine::Elasticsearch->get_facetable_fields;
456 for my $f (@facetable_fields) {
457 next unless defined $f->facet_order;
458 $type_to_label{ $f->name } =
459 { order => $f->facet_order, label => $label{ $f->name } };
462 # We also have some special cases, e.g. itypes that need to show the
463 # value rather than the code.
464 my @itypes = Koha::ItemTypes->search;
465 my @libraries = Koha::Libraries->search;
466 my $library_names = { map { $_->branchcode => $_->branchname } @libraries };
467 my @locations = Koha::AuthorisedValues->search( { category => 'LOC' } );
468 my $opac = C4::Context->interface eq 'opac' ;
470 itype => { map { $_->itemtype => $_->description } @itypes },
471 location => { map { $_->authorised_value => ( $opac ? ( $_->lib_opac || $_->lib ) : $_->lib ) } @locations },
472 holdingbranch => $library_names,
473 homebranch => $library_names
477 while ( my ( $type, $data ) = each %$es ) {
478 next if !exists( $type_to_label{$type} );
480 # We restrict to the most popular $limit !results
481 my $limit = C4::Context->preference('FacetMaxCount');
483 type_id => $type . '_id',
484 "type_label_$type_to_label{$type}{label}" => 1,
485 type_link_value => $type,
486 order => $type_to_label{$type}{order},
488 $limit = @{ $data->{buckets} } if ( $limit > @{ $data->{buckets} } );
489 foreach my $term ( @{ $data->{buckets} }[ 0 .. $limit - 1 ] ) {
490 my $t = $term->{key};
491 my $c = $term->{doc_count};
493 if ( exists( $special{$type} ) ) {
494 $label = $special{$type}->{$t} // $t;
499 push @{ $facet->{facets} }, {
501 facet_link_value => $t,
502 facet_title_value => $t . " ($c)",
503 facet_label_value => $label, # TODO either truncate this,
504 # or make the template do it like it should anyway
505 type_link_value => $type,
508 push @facets, $facet if exists $facet->{facets};
511 @facets = sort { $a->{order} <=> $b->{order} } @facets;
515 =head2 _aggregation_scan
517 my $result = $self->_aggregration_scan($query, 10, 0);
519 Perform an aggregation request for scan purposes.
523 sub _aggregation_scan {
524 my ($self, $query, $results_per_page, $offset) = @_;
526 if (!scalar(keys %{$query->{aggregations}})) {
533 return (undef, \%result, undef);
535 my ($field) = keys %{$query->{aggregations}};
536 $query->{aggregations}{$field}{terms}{size} = 1000;
537 my $results = $self->search($query, 1, 0);
539 # Convert each result into a MARC::Record
540 my (@records, $index);
541 # opac-search expects results to be put in the
542 # right place in the array, according to $offset
543 $index = $offset - 1;
545 my $count = scalar(@{$results->{aggregations}{$field}{buckets}});
546 for (my $index = $offset; $index - $offset < $results_per_page && $index < $count; $index++) {
547 my $bucket = $results->{aggregations}{$field}{buckets}->[$index];
548 # Scan values are expressed as:
549 # - MARC21: 100a (count) and 245a (term)
550 # - UNIMARC: 200f (count) and 200a (term)
551 my $marc = MARC::Record->new;
552 $marc->encoding('UTF-8');
553 if (C4::Context->preference('marcflavour') eq 'UNIMARC') {
554 $marc->append_fields(
555 MARC::Field->new((200, ' ', ' ', 'f' => $bucket->{doc_count}))
557 $marc->append_fields(
558 MARC::Field->new((200, ' ', ' ', 'a' => $bucket->{key}))
561 $marc->append_fields(
562 MARC::Field->new((100, ' ', ' ', 'a' => $bucket->{doc_count}))
564 $marc->append_fields(
565 MARC::Field->new((245, ' ', ' ', 'a' => $bucket->{key}))
568 $records[$index] = $marc->as_usmarc();
570 # consumers of this expect a namespaced result, we provide the default
573 $result{biblioserver}{hits} = $count;
574 $result{biblioserver}{RECORDS} = \@records;
575 return (undef, \%result, undef);