1 package Koha::SearchEngine::Elasticsearch;
3 # Copyright 2015 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.
20 use base qw(Class::Accessor);
25 use Koha::SearchFields;
26 use Koha::SearchMarcMaps;
34 use Data::Dumper; # TODO remove
36 __PACKAGE__->mk_ro_accessors(qw( index ));
37 __PACKAGE__->mk_accessors(qw( sort_fields ));
39 # Constants to refer to the standard index names
40 Readonly our $BIBLIOS_INDEX => 'biblios';
41 Readonly our $AUTHORITIES_INDEX => 'authorities';
45 Koha::SearchEngine::Elasticsearch - Base module for things using elasticsearch
53 The name of the index to use, generally 'biblios' or 'authorities'.
63 my $self = $class->SUPER::new(@_);
64 # Check for a valid index
65 croak('No index name provided') unless $self->index;
69 =head2 get_elasticsearch_params
71 my $params = $self->get_elasticsearch_params();
73 This provides a hashref that contains the parameters for connecting to the
74 ElasicSearch servers, in the form:
77 'nodes' => ['127.0.0.1:9200', 'anotherserver:9200'],
78 'index_name' => 'koha_instance_index',
81 This is configured by the following in the C<config> block in koha-conf.xml:
84 <server>127.0.0.1:9200</server>
85 <server>anotherserver:9200</server>
86 <index_name>koha_instance</index_name>
91 sub get_elasticsearch_params {
94 # Copy the hash so that we're not modifying the original
95 my $conf = C4::Context->config('elasticsearch');
96 die "No 'elasticsearch' block is defined in koha-conf.xml.\n" if ( !$conf );
97 my $es = { %{ $conf } };
99 # Helpfully, the multiple server lines end up in an array for us anyway
100 # if there are multiple ones, but not if there's only one.
101 my $server = $es->{server};
102 delete $es->{server};
103 if ( ref($server) eq 'ARRAY' ) {
105 # store it called 'nodes' (which is used by newer Search::Elasticsearch)
106 $es->{nodes} = $server;
109 $es->{nodes} = [$server];
112 die "No elasticsearch servers were specified in koha-conf.xml.\n";
114 die "No elasticserver index_name was specified in koha-conf.xml.\n"
115 if ( !$es->{index_name} );
116 # Append the name of this particular index to our namespace
117 $es->{index_name} .= '_' . $self->index;
119 $es->{key_prefix} = 'es_';
123 =head2 get_elasticsearch_settings
125 my $settings = $self->get_elasticsearch_settings();
127 This provides the settings provided to elasticsearch when an index is created.
128 These can do things like define tokenisation methods.
130 A hashref containing the settings is returned.
134 sub get_elasticsearch_settings {
137 # Ultimately this should come from a file or something, and not be
144 tokenizer => 'icu_tokenizer',
145 filter => ['icu_folding'],
147 analyser_standard => {
148 tokenizer => 'icu_tokenizer',
149 filter => ['icu_folding'],
158 =head2 get_elasticsearch_mappings
160 my $mappings = $self->get_elasticsearch_mappings();
162 This provides the mappings that get passed to elasticsearch when an index is
167 sub get_elasticsearch_mappings {
170 # TODO cache in the object?
173 _all => {type => "string", analyzer => "analyser_standard"},
177 include_in_all => JSON::false,
184 my $marcflavour = lc C4::Context->preference('marcflavour');
185 $self->_foreach_mapping(
187 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
188 return if $marc_type ne $marcflavour;
189 # TODO if this gets any sort of complexity to it, it should
190 # be broken out into its own function.
192 # TODO be aware of date formats, but this requires pre-parsing
193 # as ES will simply reject anything with an invalid date.
199 if ($es_type eq 'boolean') {
200 $mappings->{data}{properties}{$name} = _elasticsearch_mapping_for_boolean( $name, $es_type, $facet, $suggestible, $sort, $marc_type );
201 return; #Boolean cannot have facets nor sorting nor suggestions
203 $mappings->{data}{properties}{$name} = _elasticsearch_mapping_for_default( $name, $es_type, $facet, $suggestible, $sort, $marc_type );
207 $mappings->{data}{properties}{ $name . '__facet' } = {
212 $mappings->{data}{properties}{ $name . '__suggestion' } = {
213 type => 'completion',
214 analyzer => 'simple',
215 search_analyzer => 'simple',
218 # Sort may be true, false, or undef. Here we care if it's
219 # anything other than undef.
221 $mappings->{data}{properties}{ $name . '__sort' } = {
222 search_analyzer => "analyser_phrase",
223 analyzer => "analyser_phrase",
225 include_in_all => JSON::false,
232 $sort_fields{$name} = 1;
236 $self->sort_fields(\%sort_fields);
240 =head2 _elasticsearch_mapping_for_*
242 Get the ES mappings for the given data type or a special mapping case
244 Receives the same parameters from the $self->_foreach_mapping() dispatcher
248 sub _elasticsearch_mapping_for_boolean {
249 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
257 sub _elasticsearch_mapping_for_default {
258 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
261 search_analyzer => "analyser_standard",
262 analyzer => "analyser_standard",
266 search_analyzer => "analyser_phrase",
267 analyzer => "analyser_phrase",
277 sub reset_elasticsearch_mappings {
278 my $mappings_yaml = C4::Context->config('intranetdir') . '/admin/searchengine/elasticsearch/mappings.yaml';
279 my $indexes = LoadFile( $mappings_yaml );
281 while ( my ( $index_name, $fields ) = each %$indexes ) {
282 while ( my ( $field_name, $data ) = each %$fields ) {
283 my $field_type = $data->{type};
284 my $field_label = $data->{label};
285 my $mappings = $data->{mappings};
286 my $search_field = Koha::SearchFields->find_or_create({ name => $field_name, label => $field_label, type => $field_type }, { key => 'name' });
287 for my $mapping ( @$mappings ) {
288 my $marc_field = Koha::SearchMarcMaps->find_or_create({ index_name => $index_name, marc_type => $mapping->{marc_type}, marc_field => $mapping->{marc_field} });
289 $search_field->add_to_search_marc_maps($marc_field, { facet => $mapping->{facet} || 0, suggestible => $mapping->{suggestible} || 0, sort => $mapping->{sort} } );
295 # This overrides the accessor provided by Class::Accessor so that if
296 # sort_fields isn't set, then it'll generate it.
301 $self->_sort_fields_accessor(@_);
304 my $val = $self->_sort_fields_accessor();
307 # This will populate the accessor as a side effect
308 $self->get_elasticsearch_mappings();
309 return $self->_sort_fields_accessor();
312 # Provides the rules for data conversion.
313 sub get_fixer_rules {
316 my $marcflavour = lc C4::Context->preference('marcflavour');
319 $self->_foreach_mapping(
321 my ( $name, $type, $facet, $suggestible, $sort, $marc_type, $marc_field ) = @_;
322 return if $marc_type ne $marcflavour;
325 # There's a bug when using 'split' with something that
327 # The split makes everything into nested arrays, but that's not
328 # really a big deal, ES doesn't mind.
329 $options = '-split => 1' unless $marc_field =~ m|_/| || $type eq 'sum';
330 push @rules, "marc_map('$marc_field','${name}.\$append', $options)";
332 push @rules, "marc_map('$marc_field','${name}__facet.\$append', $options)";
336 #"marc_map('$marc_field','${name}__suggestion.input.\$append', $options)"; #must not have nested data structures in .input
337 "marc_map('$marc_field','${name}__suggestion.input.\$append')";
339 if ( $type eq 'boolean' ) {
341 # boolean gets special handling, basically if it doesn't exist,
342 # it's added and set to false. Otherwise we can't query it.
344 "unless exists('$name') add_field('$name', 0) end";
346 if ($type eq 'sum' ) {
347 push @rules, "sum('$name')";
349 # Sort is a bit special as it can be true, false, undef. For
350 # fixer rules, we care about "true", or "undef" if there is
351 # special handling of this field from other one. "undef" means
352 # to do the default thing, which is make it sortable.
353 if ($self->sort_fields()->{$name}) {
354 if ($sort || !defined $sort) {
355 push @rules, "marc_map('$marc_field','${name}__sort.\$append', $options)";
361 push @rules, "move_field(_id,es_id)"; #Also you must set the Catmandu::Store::ElasticSearch->new(key_prefix: 'es_');
365 =head2 _foreach_mapping
367 $self->_foreach_mapping(
369 my ( $name, $type, $facet, $suggestible, $sort, $marc_type,
372 return unless $marc_type eq 'marc21';
373 print "Data comes from: " . $marc_field . "\n";
377 This allows you to apply a function to each entry in the elasticsearch mappings
378 table, in order to build the mappings for whatever is needed.
380 In the provided function, the files are:
386 The field name for elasticsearch (corresponds to the 'mapping' column in the
391 The type for this value, e.g. 'string'.
395 True if this value should be facetised. This only really makes sense if the
396 field is understood by the facet processing code anyway.
400 True if this is a field that a) needs special sort handling, and b) if it
401 should be sorted on. False if a) but not b). Undef if not a). This allows,
402 for example, author to be sorted on but not everything marked with "author"
403 to be included in that sort.
407 A string that indicates the MARC type that this mapping is for, e.g. 'marc21',
408 'unimarc', 'normarc'.
412 A string that describes the MARC field that contains the data to extract.
413 These are of a form suited to Catmandu's MARC fixers.
419 sub _foreach_mapping {
420 my ( $self, $sub ) = @_;
422 # TODO use a caching framework here
423 my $search_fields = Koha::Database->schema->resultset('SearchField')->search(
425 'search_marc_map.index_name' => $self->index,
427 { join => { search_marc_to_fields => 'search_marc_map' },
429 'search_marc_to_fields.facet',
430 'search_marc_to_fields.suggestible',
431 'search_marc_to_fields.sort',
432 'search_marc_map.marc_type',
433 'search_marc_map.marc_field',
445 while ( my $search_field = $search_fields->next ) {
449 $search_field->get_column('facet'),
450 $search_field->get_column('suggestible'),
451 $search_field->get_column('sort'),
452 $search_field->get_column('marc_type'),
453 $search_field->get_column('marc_field'),
460 die process_error($@);
462 This parses an Elasticsearch error message and produces a human-readable
463 result from it. This result is probably missing all the useful information
464 that you might want in diagnosing an issue, so the warning is also logged.
466 Note that currently the resulting message is not internationalised. This
467 will happen eventually by some method or other.
472 my ($self, $msg) = @_;
474 warn $msg; # simple logging
476 # This is super-primitive
477 return "Unable to understand your search query, please rephrase and try again.\n" if $msg =~ /ParseException/;
479 return "Unable to perform your search. Please try again.\n";
490 =item Chris Cormack C<< <chrisc@catalyst.net.nz> >>
492 =item Robin Sheat C<< <robin@catalyst.net.nz> >>
494 =item Jonathan Druart C<< <jonathan.druart@bugs.koha-community.org> >>