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::Exceptions::Config;
26 use Koha::SearchFields;
27 use Koha::SearchMarcMaps;
33 use Search::Elasticsearch;
37 use Data::Dumper; # TODO remove
39 __PACKAGE__->mk_ro_accessors(qw( index ));
40 __PACKAGE__->mk_accessors(qw( sort_fields ));
42 # Constants to refer to the standard index names
43 Readonly our $BIBLIOS_INDEX => 'biblios';
44 Readonly our $AUTHORITIES_INDEX => 'authorities';
48 Koha::SearchEngine::Elasticsearch - Base module for things using elasticsearch
56 The name of the index to use, generally 'biblios' or 'authorities'.
66 my $self = $class->SUPER::new(@_);
67 # Check for a valid index
68 croak('No index name provided') unless $self->index;
72 =head2 get_elasticsearch_params
74 my $params = $self->get_elasticsearch_params();
76 This provides a hashref that contains the parameters for connecting to the
77 ElasicSearch servers, in the form:
80 'nodes' => ['127.0.0.1:9200', 'anotherserver:9200'],
81 'index_name' => 'koha_instance_index',
84 This is configured by the following in the C<config> block in koha-conf.xml:
87 <server>127.0.0.1:9200</server>
88 <server>anotherserver:9200</server>
89 <index_name>koha_instance</index_name>
94 sub get_elasticsearch_params {
97 # Copy the hash so that we're not modifying the original
98 my $conf = C4::Context->config('elasticsearch');
99 die "No 'elasticsearch' block is defined in koha-conf.xml.\n" if ( !$conf );
100 my $es = { %{ $conf } };
102 # Helpfully, the multiple server lines end up in an array for us anyway
103 # if there are multiple ones, but not if there's only one.
104 my $server = $es->{server};
105 delete $es->{server};
106 if ( ref($server) eq 'ARRAY' ) {
108 # store it called 'nodes' (which is used by newer Search::Elasticsearch)
109 $es->{nodes} = $server;
112 $es->{nodes} = [$server];
115 die "No elasticsearch servers were specified in koha-conf.xml.\n";
117 die "No elasticserver index_name was specified in koha-conf.xml.\n"
118 if ( !$es->{index_name} );
119 # Append the name of this particular index to our namespace
120 $es->{index_name} .= '_' . $self->index;
122 $es->{key_prefix} = 'es_';
126 =head2 get_elasticsearch_settings
128 my $settings = $self->get_elasticsearch_settings();
130 This provides the settings provided to elasticsearch when an index is created.
131 These can do things like define tokenisation methods.
133 A hashref containing the settings is returned.
137 sub get_elasticsearch_settings {
140 # Ultimately this should come from a file or something, and not be
148 char_filter => ['icu_normalizer'],
153 tokenizer => 'icu_tokenizer',
154 filter => ['icu_folding'],
156 analyser_standard => {
157 tokenizer => 'icu_tokenizer',
158 filter => ['icu_folding'],
167 =head2 get_elasticsearch_mappings
169 my $mappings = $self->get_elasticsearch_mappings();
171 This provides the mappings that get passed to elasticsearch when an index is
176 sub get_elasticsearch_mappings {
179 # TODO cache in the object?
182 _all => {type => "string", analyzer => "analyser_standard"},
186 include_in_all => JSON::false,
193 my $marcflavour = lc C4::Context->preference('marcflavour');
194 $self->_foreach_mapping(
196 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
197 return if $marc_type ne $marcflavour;
198 # TODO if this gets any sort of complexity to it, it should
199 # be broken out into its own function.
201 # TODO be aware of date formats, but this requires pre-parsing
202 # as ES will simply reject anything with an invalid date.
208 if ($es_type eq 'boolean') {
209 $mappings->{data}{properties}{$name} = _elasticsearch_mapping_for_boolean( $name, $es_type, $facet, $suggestible, $sort, $marc_type );
210 return; #Boolean cannot have facets nor sorting nor suggestions
212 $mappings->{data}{properties}{$name} = _elasticsearch_mapping_for_default( $name, $es_type, $facet, $suggestible, $sort, $marc_type );
216 $mappings->{data}{properties}{ $name . '__facet' } = {
221 $mappings->{data}{properties}{ $name . '__suggestion' } = {
222 type => 'completion',
223 analyzer => 'simple',
224 search_analyzer => 'simple',
227 # Sort is a bit special as it can be true, false, undef.
228 # We care about "true" or "undef",
229 # "undef" means to do the default thing, which is make it sortable.
230 if ($sort || !defined $sort) {
231 $mappings->{data}{properties}{ $name . '__sort' } = {
232 search_analyzer => "analyser_phrase",
233 analyzer => "analyser_phrase",
235 include_in_all => JSON::false,
242 $sort_fields{$name} = 1;
246 $self->sort_fields(\%sort_fields);
250 =head2 _elasticsearch_mapping_for_*
252 Get the ES mappings for the given data type or a special mapping case
254 Receives the same parameters from the $self->_foreach_mapping() dispatcher
258 sub _elasticsearch_mapping_for_boolean {
259 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
267 sub _elasticsearch_mapping_for_default {
268 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
271 search_analyzer => "analyser_standard",
272 analyzer => "analyser_standard",
276 search_analyzer => "analyser_phrase",
277 analyzer => "analyser_phrase",
285 normalizer => "my_normalizer",
291 sub reset_elasticsearch_mappings {
292 my $mappings_yaml = C4::Context->config('intranetdir') . '/admin/searchengine/elasticsearch/mappings.yaml';
293 my $indexes = LoadFile( $mappings_yaml );
295 while ( my ( $index_name, $fields ) = each %$indexes ) {
296 while ( my ( $field_name, $data ) = each %$fields ) {
297 my $field_type = $data->{type};
298 my $field_label = $data->{label};
299 my $mappings = $data->{mappings};
300 my $search_field = Koha::SearchFields->find_or_create({ name => $field_name, label => $field_label, type => $field_type }, { key => 'name' });
301 for my $mapping ( @$mappings ) {
302 my $marc_field = Koha::SearchMarcMaps->find_or_create({ index_name => $index_name, marc_type => $mapping->{marc_type}, marc_field => $mapping->{marc_field} });
303 $search_field->add_to_search_marc_maps($marc_field, { facet => $mapping->{facet} || 0, suggestible => $mapping->{suggestible} || 0, sort => $mapping->{sort} } );
309 # This overrides the accessor provided by Class::Accessor so that if
310 # sort_fields isn't set, then it'll generate it.
314 $self->_sort_fields_accessor(@_);
317 my $val = $self->_sort_fields_accessor();
320 # This will populate the accessor as a side effect
321 $self->get_elasticsearch_mappings();
322 return $self->_sort_fields_accessor();
325 # Provides the rules for data conversion.
326 sub get_fixer_rules {
329 my $marcflavour = lc C4::Context->preference('marcflavour');
332 $self->_foreach_mapping(
334 my ( $name, $type, $facet, $suggestible, $sort, $marc_type, $marc_field ) = @_;
335 return if $marc_type ne $marcflavour;
338 push @rules, "marc_map('$marc_field','${name}.\$append', $options)";
340 push @rules, "marc_map('$marc_field','${name}__facet.\$append', $options)";
344 #"marc_map('$marc_field','${name}__suggestion.input.\$append', '')"; #must not have nested data structures in .input
345 "marc_map('$marc_field','${name}__suggestion.input.\$append')";
347 if ( $type eq 'boolean' ) {
349 # boolean gets special handling, basically if it doesn't exist,
350 # it's added and set to false. Otherwise we can't query it.
352 "unless exists('$name') add_field('$name', 0) end";
354 if ($type eq 'sum' ) {
355 push @rules, "sum('$name')";
357 if ($self->sort_fields()->{$name}) {
358 if ($sort || !defined $sort) {
359 push @rules, "marc_map('$marc_field','${name}__sort.\$append', $options)";
365 push @rules, "move_field(_id,es_id)"; #Also you must set the Catmandu::Store::ElasticSearch->new(key_prefix: 'es_');
369 =head2 _foreach_mapping
371 $self->_foreach_mapping(
373 my ( $name, $type, $facet, $suggestible, $sort, $marc_type,
376 return unless $marc_type eq 'marc21';
377 print "Data comes from: " . $marc_field . "\n";
381 This allows you to apply a function to each entry in the elasticsearch mappings
382 table, in order to build the mappings for whatever is needed.
384 In the provided function, the files are:
390 The field name for elasticsearch (corresponds to the 'mapping' column in the
395 The type for this value, e.g. 'string'.
399 True if this value should be facetised. This only really makes sense if the
400 field is understood by the facet processing code anyway.
404 True if this is a field that a) needs special sort handling, and b) if it
405 should be sorted on. False if a) but not b). Undef if not a). This allows,
406 for example, author to be sorted on but not everything marked with "author"
407 to be included in that sort.
411 A string that indicates the MARC type that this mapping is for, e.g. 'marc21',
412 'unimarc', 'normarc'.
416 A string that describes the MARC field that contains the data to extract.
417 These are of a form suited to Catmandu's MARC fixers.
423 sub _foreach_mapping {
424 my ( $self, $sub ) = @_;
426 # TODO use a caching framework here
427 my $search_fields = Koha::Database->schema->resultset('SearchField')->search(
429 'search_marc_map.index_name' => $self->index,
431 { join => { search_marc_to_fields => 'search_marc_map' },
433 'search_marc_to_fields.facet',
434 'search_marc_to_fields.suggestible',
435 'search_marc_to_fields.sort',
436 'search_marc_map.marc_type',
437 'search_marc_map.marc_field',
449 while ( my $search_field = $search_fields->next ) {
453 $search_field->get_column('facet'),
454 $search_field->get_column('suggestible'),
455 $search_field->get_column('sort'),
456 $search_field->get_column('marc_type'),
457 $search_field->get_column('marc_field'),
464 die process_error($@);
466 This parses an Elasticsearch error message and produces a human-readable
467 result from it. This result is probably missing all the useful information
468 that you might want in diagnosing an issue, so the warning is also logged.
470 Note that currently the resulting message is not internationalised. This
471 will happen eventually by some method or other.
476 my ($self, $msg) = @_;
478 warn $msg; # simple logging
480 # This is super-primitive
481 return "Unable to understand your search query, please rephrase and try again.\n" if $msg =~ /ParseException/;
483 return "Unable to perform your search. Please try again.\n";
486 =head2 _read_configuration
488 my $conf = _read_configuration();
490 Reads the I<configuration file> and returns a hash structure with the
491 configuration information. It raises an exception if mandatory entries
494 The hashref structure has the following form:
497 'nodes' => ['127.0.0.1:9200', 'anotherserver:9200'],
498 'index_name' => 'koha_instance',
501 This is configured by the following in the C<config> block in koha-conf.xml:
504 <server>127.0.0.1:9200</server>
505 <server>anotherserver:9200</server>
506 <index_name>koha_instance</index_name>
511 sub _read_configuration {
515 my $conf = C4::Context->config('elasticsearch');
516 Koha::Exceptions::Config::MissingEntry->throw(
517 "Missing 'elasticsearch' block in config file")
518 unless defined $conf;
520 if ( $conf && $conf->{server} ) {
521 my $nodes = $conf->{server};
522 if ( ref($nodes) eq 'ARRAY' ) {
523 $configuration->{nodes} = $nodes;
526 $configuration->{nodes} = [$nodes];
530 Koha::Exceptions::Config::MissingEntry->throw(
531 "Missing 'server' entry in config file for elasticsearch");
534 if ( defined $conf->{index_name} ) {
535 $configuration->{index_name} = $conf->{index_name};
538 Koha::Exceptions::Config::MissingEntry->throw(
539 "Missing 'index_name' entry in config file for elasticsearch");
542 return $configuration;
553 =item Chris Cormack C<< <chrisc@catalyst.net.nz> >>
555 =item Robin Sheat C<< <robin@catalyst.net.nz> >>
557 =item Jonathan Druart C<< <jonathan.druart@bugs.koha-community.org> >>