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
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::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;
53 use MIME::Base64 qw( decode_base64 );
55 Koha::SearchEngine::Elasticsearch::Search->mk_accessors(qw( store ));
59 my $results = $searcher->search($query, $page, $count, %options);
61 Run a search using the query. It'll return C<$count> results, starting at page
62 C<$page> (C<$page> counts from 1, anything less that, or C<undef> becomes 1.)
63 C<$count> is also the number of entries on a page.
65 C<%options> is a hash containing extra options:
71 If provided, this overrides the C<$page> value, and specifies the record as
72 an offset (i.e. the number of the record to start with), rather than a page.
81 my ($self, $query, $page, $count, %options) = @_;
83 # 20 is the default number of results per page
84 $query->{size} = $count // 20;
85 # ES doesn't want pages, it wants a record to start from.
86 if (exists $options{offset}) {
87 $query->{from} = $options{offset};
89 $page = (!defined($page) || ($page <= 0)) ? 0 : $page - 1;
90 $query->{from} = $page * $query->{size};
92 my $elasticsearch = $self->get_elasticsearch();
94 $elasticsearch->search(
95 index => $self->index_name,
100 die $self->process_error($@);
107 my $count = $searcher->count($query);
109 This mimics a search request, but just gets the result count instead. That's
110 faster than pulling all the data in, usually.
115 my ( $self, $query ) = @_;
116 my $elasticsearch = $self->get_elasticsearch();
118 # TODO: Probably possible to exclude results
119 # and just return number of hits
120 my $result = $elasticsearch->search(
121 index => $self->index_name,
125 return $result->{hits}->{total};
130 my ( $error, $results, $facets ) = $search->search_compat(
131 $query, $simple_query, \@sort_by, \@servers,
132 $results_per_page, $offset, undef, $item_types,
136 A search interface somewhat compatible with L<C4::Search->getRecords>. Anything
137 that is returned in the query created by build_query_compat will probably
138 get ignored here, along with some other things (like C<@servers>.)
144 $self, $query, $simple_query, $sort_by,
145 $servers, $results_per_page, $offset, $branches,
146 $item_types, $query_type, $scan
150 return $self->_aggregation_scan( $query, $results_per_page, $offset );
154 if ( !defined $offset or $offset < 0 ) {
157 $options{offset} = $offset;
158 my $results = $self->search($query, undef, $results_per_page, %options);
160 # Convert each result into a MARC::Record
162 # opac-search expects results to be put in the
163 # right place in the array, according to $offset
165 my $hits = $results->{'hits'};
166 foreach my $es_record (@{$hits->{'hits'}}) {
167 $records[$index++] = $self->decode_record_from_result($es_record->{'_source'});
170 # consumers of this expect a name-spaced result, we provide the default
173 $result{biblioserver}{hits} = $hits->{'total'};
174 $result{biblioserver}{RECORDS} = \@records;
175 return (undef, \%result, $self->_convert_facets($results->{aggregations}));
178 =head2 search_auth_compat
180 my ( $results, $total ) =
181 $searcher->search_auth_compat( $query, $offset, $count, $skipmetadata, %options );
183 This has a similar calling convention to L<search>, however it returns its
184 results in a form the same as L<C4::AuthoritiesMarc::SearchAuthorities>.
188 sub search_auth_compat {
189 my ($self, $query, $offset, $count, $skipmetadata, %options) = @_;
191 if ( !defined $offset or $offset <= 0 ) {
194 # Uh, authority search uses 1-based offset..
195 $options{offset} = $offset - 1;
196 my $database = Koha::Database->new();
197 my $schema = $database->schema();
198 my $res = $self->search($query, undef, $count, %options);
200 my $bib_searcher = Koha::SearchEngine::Elasticsearch::Search->new({index => 'biblios'});
202 my $hits = $res->{'hits'};
203 foreach my $es_record (@{$hits->{'hits'}}) {
204 my $record = $es_record->{'_source'};
207 # We are using the authid to create links, we should honor the authid as stored in the db, not
208 # the 001 which, in some circumstances, can contain other data
209 my $authid = $es_record->{_id};
212 $result{authid} = $authid;
214 if (!defined $skipmetadata || !$skipmetadata) {
215 # TODO put all this info into the record at index time so we
216 # don't have to go and sort it all out now.
217 my $authtypecode = $record->{authtype};
218 my $rs = $schema->resultset('AuthType')
219 ->search( { authtypecode => $authtypecode } );
221 # FIXME there's an assumption here that we will get a result.
222 # the original code also makes an assumption that some provided
223 # authtypecode may sometimes be used instead of the one stored
224 # with the record. It's not documented why this is the case, so
225 # it's not reproduced here yet.
226 my $authtype = $rs->single;
227 my $auth_tag_to_report = $authtype ? $authtype->auth_tag_to_report : "";
228 my $marc = $self->decode_record_from_result($record);
229 my $mainentry = $marc->field($auth_tag_to_report);
232 foreach ( $mainentry->subfields() ) {
233 $reported_tag .= '$' . $_->[0] . $_->[1];
236 # Turn the resultset into a hash
237 $result{authtype} = $authtype ? $authtype->authtypetext : $authtypecode;
238 $result{reported_tag} = $reported_tag;
240 # Reimplementing BuildSummary is out of scope because it'll be hard
242 C4::AuthoritiesMarc::BuildSummary( $marc, $result{authid},
244 $result{used} = $self->count_auth_use($bib_searcher, $authid);
246 push @records, \%result;
248 return ( \@records, $hits->{'total'} );
251 =head2 count_auth_use
253 my $count = $auth_searcher->count_auth_use($bib_searcher, $authid);
255 This runs a search to determine the number of records that reference the
256 specified authid. C<$bib_searcher> must be something compatible with
257 elasticsearch, as the query is built in this function.
262 my ($self, $bib_searcher, $authid) = @_;
267 # query => { match_all => {} },
268 filter => { term => { 'koha-auth-number' => $authid } }
272 $bib_searcher->count($query);
275 =head2 simple_search_compat
277 my ( $error, $marcresults, $total_hits ) =
278 $searcher->simple_search( $query, $offset, $max_results, %options );
280 This is a simpler interface to the searching, intended to be similar enough to
281 L<C4::Search::SimpleSearch>.
289 A thing to search for. It could be a simple string, or something constructed
290 with the appropriate QueryBuilder module.
294 How many results to skip from the start of the results.
296 =item C<$max_results>
298 The max number of results to return. The default is 100 (because unlimited
299 is a pretty terrible thing to do.)
303 These options are unused by Elasticsearch
313 if something went wrong, this'll contain some kind of error
316 =item C<$marcresults>
318 an arrayref of MARC::Records (note that this is different from the
319 L<C4::Search> version which will return plain XML, but too bad.)
323 the total number of results that this search could have returned.
329 sub simple_search_compat {
330 my ($self, $query, $offset, $max_results) = @_;
332 return ('No query entered', undef, undef) unless $query;
335 $offset = 0 if not defined $offset or $offset < 0;
336 $options{offset} = $offset;
337 $max_results //= 100;
339 unless (ref $query) {
340 # We'll push it through the query builder to sanitise everything.
341 my $qb = Koha::SearchEngine::QueryBuilder->new({index => $self->index});
342 (undef,$query) = $qb->build_query_compat(undef, [$query]);
344 my $results = $self->search($query, undef, $max_results, %options);
346 my $hits = $results->{'hits'};
347 foreach my $es_record (@{$hits->{'hits'}}) {
348 push @records, $self->decode_record_from_result($es_record->{'_source'});
350 return (undef, \@records, $hits->{'total'});
353 =head2 extract_biblionumber
355 my $biblionumber = $searcher->extract_biblionumber( $searchresult );
357 $searchresult comes from simple_search_compat.
359 Returns the biblionumber from the search result record.
363 sub extract_biblionumber {
364 my ( $self, $searchresultrecord ) = @_;
365 return Koha::SearchEngine::Search::extract_biblionumber( $searchresultrecord );
368 =head2 decode_record_from_result
369 my $marc_record = $self->decode_record_from_result(@result);
371 Extracts marc data from Elasticsearch result and decodes to MARC::Record object
375 sub decode_record_from_result {
376 # Result is passed in as array, will get flattened
377 # and first element will be $result
378 my ( $self, $result ) = @_;
379 if ($result->{marc_format} eq 'base64ISO2709') {
380 return MARC::Record->new_from_usmarc(decode_base64($result->{marc_data}));
382 elsif ($result->{marc_format} eq 'MARCXML') {
383 return MARC::Record->new_from_xml($result->{marc_data}, 'UTF-8', uc C4::Context->preference('marcflavour'));
385 elsif ($result->{marc_format} eq 'ARRAY') {
386 return $self->_array_to_marc($result->{marc_data_array});
389 Koha::Exceptions::Elasticsearch->throw("Missing marc_format field in Elasticsearch result");
393 =head2 max_result_window
395 Returns the maximum number of results that can be fetched
397 This directly requests Elasticsearch for the setting index.max_result_window (or
398 the default value for this setting in case it is not set)
402 sub max_result_window {
405 my $elasticsearch = $self->get_elasticsearch();
407 my $response = $elasticsearch->indices->get_settings(
408 index => $self->index_name,
409 flat_settings => 'true',
410 include_defaults => 'true'
413 my $max_result_window = $response->{$self->index_name}->{settings}->{'index.max_result_window'};
414 $max_result_window //= $response->{$self->index_name}->{defaults}->{'index.max_result_window'};
416 return $max_result_window;
419 =head2 _convert_facets
421 my $koha_facets = _convert_facets($es_facets);
423 Converts elasticsearch facets types to the form that Koha expects.
424 It expects the ES facet name to match the Koha type, for example C<itype>,
425 C<au>, C<su-to>, etc.
429 sub _convert_facets {
430 my ( $self, $es, $exp_facet ) = @_;
434 # These should correspond to the ES field names, as opposed to the CCL
435 # things that zebra uses.
439 itype => 'ItemTypes',
440 location => 'Location',
441 'su-geo' => 'Places',
442 'title-series' => 'Series',
444 ccode => 'CollectionCodes',
445 holdingbranch => 'HoldingLibrary',
446 homebranch => 'HomeLibrary',
449 my @facetable_fields =
450 Koha::SearchEngine::Elasticsearch->get_facetable_fields;
451 for my $f (@facetable_fields) {
452 next unless defined $f->facet_order;
453 $type_to_label{ $f->name } =
454 { order => $f->facet_order, label => $label{ $f->name } };
457 # We also have some special cases, e.g. itypes that need to show the
458 # value rather than the code.
459 my @itypes = Koha::ItemTypes->search;
460 my @libraries = Koha::Libraries->search;
461 my $library_names = { map { $_->branchcode => $_->branchname } @libraries };
462 my @locations = Koha::AuthorisedValues->search( { category => 'LOC' } );
463 my $opac = C4::Context->interface eq 'opac' ;
465 itype => { map { $_->itemtype => $_->description } @itypes },
466 location => { map { $_->authorised_value => ( $opac ? ( $_->lib_opac || $_->lib ) : $_->lib ) } @locations },
467 holdingbranch => $library_names,
468 homebranch => $library_names
472 while ( my ( $type, $data ) = each %$es ) {
473 next if !exists( $type_to_label{$type} );
475 # We restrict to the most popular $limit !results
476 my $limit = C4::Context->preference('FacetMaxCount');
478 type_id => $type . '_id',
479 "type_label_$type_to_label{$type}{label}" => 1,
480 type_link_value => $type,
481 order => $type_to_label{$type}{order},
483 $limit = @{ $data->{buckets} } if ( $limit > @{ $data->{buckets} } );
484 foreach my $term ( @{ $data->{buckets} }[ 0 .. $limit - 1 ] ) {
485 my $t = $term->{key};
486 my $c = $term->{doc_count};
488 if ( exists( $special{$type} ) ) {
489 $label = $special{$type}->{$t} // $t;
494 push @{ $facet->{facets} }, {
496 facet_link_value => $t,
497 facet_title_value => $t . " ($c)",
498 facet_label_value => $label, # TODO either truncate this,
499 # or make the template do it like it should anyway
500 type_link_value => $type,
503 push @facets, $facet if exists $facet->{facets};
506 @facets = sort { $a->{order} <=> $b->{order} } @facets;
510 =head2 _aggregation_scan
512 my $result = $self->_aggregration_scan($query, 10, 0);
514 Perform an aggregation request for scan purposes.
518 sub _aggregation_scan {
519 my ($self, $query, $results_per_page, $offset) = @_;
521 if (!scalar(keys %{$query->{aggregations}})) {
528 return (undef, \%result, undef);
530 my ($field) = keys %{$query->{aggregations}};
531 $query->{aggregations}{$field}{terms}{size} = 1000;
532 my $results = $self->search($query, 1, 0);
534 # Convert each result into a MARC::Record
535 my (@records, $index);
536 # opac-search expects results to be put in the
537 # right place in the array, according to $offset
538 $index = $offset - 1;
540 my $count = scalar(@{$results->{aggregations}{$field}{buckets}});
541 for (my $index = $offset; $index - $offset < $results_per_page && $index < $count; $index++) {
542 my $bucket = $results->{aggregations}{$field}{buckets}->[$index];
543 # Scan values are expressed as:
544 # - MARC21: 100a (count) and 245a (term)
545 # - UNIMARC: 200f (count) and 200a (term)
546 my $marc = MARC::Record->new;
547 $marc->encoding('UTF-8');
548 if (C4::Context->preference('marcflavour') eq 'UNIMARC') {
549 $marc->append_fields(
550 MARC::Field->new((200, ' ', ' ', 'f' => $bucket->{doc_count}))
552 $marc->append_fields(
553 MARC::Field->new((200, ' ', ' ', 'a' => $bucket->{key}))
556 $marc->append_fields(
557 MARC::Field->new((100, ' ', ' ', 'a' => $bucket->{doc_count}))
559 $marc->append_fields(
560 MARC::Field->new((245, ' ', ' ', 'a' => $bucket->{key}))
563 $records[$index] = $marc->as_usmarc();
565 # consumers of this expect a namespaced result, we provide the default
568 $result{biblioserver}{hits} = $count;
569 $result{biblioserver}{RECORDS} = \@records;
570 return (undef, \%result, undef);