Bug 15451: Koha::CsvProfiles - Remove GetCsvProfile
[koha.git] / Koha / ElasticSearch.pm
blobd56c344c8a389ac9afbae6e3434c54b18f0aa3ef
1 package Koha::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
10 # version.
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);
22 use C4::Context;
24 use Koha::Database;
26 use Carp;
27 use JSON;
28 use Modern::Perl;
29 use Readonly;
31 use Data::Dumper; # TODO remove
33 __PACKAGE__->mk_ro_accessors(qw( index ));
34 __PACKAGE__->mk_accessors(qw( sort_fields ));
36 # Constants to refer to the standard index names
37 Readonly our $BIBLIOS_INDEX => 'biblios';
38 Readonly our $AUTHORITIES_INDEX => 'authorities';
40 =head1 NAME
42 Koha::ElasticSearch - Base module for things using elasticsearch
44 =head1 ACCESSORS
46 =over 4
48 =item index
50 The name of the index to use, generally 'biblios' or 'authorities'.
52 =back
54 =head1 FUNCTIONS
56 =cut
58 sub new {
59 my $class = shift @_;
60 my $self = $class->SUPER::new(@_);
61 # Check for a valid index
62 croak('No index name provided') unless $self->index;
63 return $self;
66 =head2 get_elasticsearch_params
68 my $params = $self->get_elasticsearch_params();
70 This provides a hashref that contains the parameters for connecting to the
71 ElasicSearch servers, in the form:
74 'nodes' => ['127.0.0.1:9200', 'anotherserver:9200'],
75 'index_name' => 'koha_instance_index',
78 This is configured by the following in the C<config> block in koha-conf.xml:
80 <elasticsearch>
81 <server>127.0.0.1:9200</server>
82 <server>anotherserver:9200</server>
83 <index_name>koha_instance</index_name>
84 </elasticsearch>
86 =cut
88 sub get_elasticsearch_params {
89 my ($self) = @_;
91 # Copy the hash so that we're not modifying the original
92 my $conf = C4::Context->config('elasticsearch');
93 die "No 'elasticsearch' block is defined in koha-conf.xml.\n" if ( !$conf );
94 my $es = { %{ $conf } };
96 # Helpfully, the multiple server lines end up in an array for us anyway
97 # if there are multiple ones, but not if there's only one.
98 my $server = $es->{server};
99 delete $es->{server};
100 if ( ref($server) eq 'ARRAY' ) {
102 # store it called 'nodes' (which is used by newer Search::Elasticsearch)
103 $es->{nodes} = $server;
105 elsif ($server) {
106 $es->{nodes} = [$server];
108 else {
109 die "No elasticsearch servers were specified in koha-conf.xml.\n";
111 die "No elasticserver index_name was specified in koha-conf.xml.\n"
112 if ( !$es->{index_name} );
113 # Append the name of this particular index to our namespace
114 $es->{index_name} .= '_' . $self->index;
115 return $es;
118 =head2 get_elasticsearch_settings
120 my $settings = $self->get_elasticsearch_settings();
122 This provides the settings provided to elasticsearch when an index is created.
123 These can do things like define tokenisation methods.
125 A hashref containing the settings is returned.
127 =cut
129 sub get_elasticsearch_settings {
130 my ($self) = @_;
132 # Ultimately this should come from a file or something, and not be
133 # hardcoded.
134 my $settings = {
135 index => {
136 analysis => {
137 analyzer => {
138 analyser_phrase => {
139 tokenizer => 'keyword',
140 filter => ['lowercase'],
142 analyser_standard => {
143 tokenizer => 'standard',
144 filter => ['lowercase'],
150 return $settings;
153 =head2 get_elasticsearch_mappings
155 my $mappings = $self->get_elasticsearch_mappings();
157 This provides the mappings that get passed to elasticsearch when an index is
158 created.
160 =cut
162 sub get_elasticsearch_mappings {
163 my ($self) = @_;
165 # TODO cache in the object?
166 my $mappings = {
167 data => {
168 properties => {
169 record => {
170 store => "yes",
171 include_in_all => JSON::false,
172 type => "string",
174 '_all.phrase' => {
175 search_analyzer => "analyser_phrase",
176 index_analyzer => "analyser_phrase",
177 type => "string",
182 my %sort_fields;
183 my $marcflavour = lc C4::Context->preference('marcflavour');
184 $self->_foreach_mapping(
185 sub {
186 my ( $name, $type, $facet, $suggestible, $sort, $marc_type ) = @_;
187 return if $marc_type ne $marcflavour;
188 # TODO if this gets any sort of complexity to it, it should
189 # be broken out into its own function.
191 # TODO be aware of date formats, but this requires pre-parsing
192 # as ES will simply reject anything with an invalid date.
193 my $es_type =
194 $type eq 'boolean'
195 ? 'boolean'
196 : 'string';
197 $mappings->{data}{properties}{$name} = {
198 search_analyzer => "analyser_standard",
199 index_analyzer => "analyser_standard",
200 type => $es_type,
201 fields => {
202 phrase => {
203 search_analyzer => "analyser_phrase",
204 index_analyzer => "analyser_phrase",
205 type => "string",
206 copy_to => "_all.phrase",
208 raw => {
209 "type" => "string",
210 "index" => "not_analyzed",
214 $mappings->{data}{properties}{$name}{null_value} = 0
215 if $type eq 'boolean';
216 if ($facet) {
217 $mappings->{data}{properties}{ $name . '__facet' } = {
218 type => "string",
219 index => "not_analyzed",
222 if ($suggestible) {
223 $mappings->{data}{properties}{ $name . '__suggestion' } = {
224 type => 'completion',
225 index_analyzer => 'simple',
226 search_analyzer => 'simple',
229 # Sort may be true, false, or undef. Here we care if it's
230 # anything other than undef.
231 if (defined $sort) {
232 $mappings->{data}{properties}{ $name . '__sort' } = {
233 search_analyzer => "analyser_phrase",
234 index_analyzer => "analyser_phrase",
235 type => "string",
236 include_in_all => JSON::false,
237 fields => {
238 phrase => {
239 search_analyzer => "analyser_phrase",
240 index_analyzer => "analyser_phrase",
241 type => "string",
245 $sort_fields{$name} = 1;
249 $self->sort_fields(\%sort_fields);
250 return $mappings;
253 # This overrides the accessor provided by Class::Accessor so that if
254 # sort_fields isn't set, then it'll generate it.
255 sub sort_fields {
256 my $self = shift;
258 if (@_) {
259 $self->_sort_fields_accessor(@_);
260 return;
262 my $val = $self->_sort_fields_accessor();
263 return $val if $val;
265 # This will populate the accessor as a side effect
266 $self->get_elasticsearch_mappings();
267 return $self->_sort_fields_accessor();
270 # Provides the rules for data conversion.
271 sub get_fixer_rules {
272 my ($self) = @_;
274 my $marcflavour = lc C4::Context->preference('marcflavour');
275 my @rules;
276 $self->_foreach_mapping(
277 sub {
278 my ( $name, $type, $facet, $suggestible, $sort, $marc_type, $marc_field ) = @_;
279 return if $marc_type ne $marcflavour;
280 my $options = '';
282 # There's a bug when using 'split' with something that
283 # selects a range
284 # The split makes everything into nested arrays, but that's not
285 # really a big deal, ES doesn't mind.
286 $options = '-split => 1' unless $marc_field =~ m|_/| || $type eq 'sum';
287 push @rules, "marc_map('$marc_field','${name}', $options)";
288 if ($facet) {
289 push @rules, "marc_map('$marc_field','${name}__facet', $options)";
291 if ($suggestible) {
292 push @rules,
293 "marc_map('$marc_field','${name}__suggestion.input.\$append', $options)";
295 if ( $type eq 'boolean' ) {
297 # boolean gets special handling, basically if it doesn't exist,
298 # it's added and set to false. Otherwise we can't query it.
299 push @rules,
300 "unless exists('$name') add_field('$name', 0) end";
302 if ($type eq 'sum' ) {
303 push @rules, "sum('$name')";
305 # Sort is a bit special as it can be true, false, undef. For
306 # fixer rules, we care about "true", or "undef" if there is
307 # special handling of this field from other one. "undef" means
308 # to do the default thing, which is make it sortable.
309 if ($self->sort_fields()->{$name}) {
310 if ($sort || !defined $sort) {
311 push @rules, "marc_map('$marc_field','${name}__sort', $options)";
316 return \@rules;
319 =head2 _foreach_mapping
321 $self->_foreach_mapping(
322 sub {
323 my ( $name, $type, $facet, $suggestible, $sort, $marc_type,
324 $marc_field )
325 = @_;
326 return unless $marc_type eq 'marc21';
327 print "Data comes from: " . $marc_field . "\n";
331 This allows you to apply a function to each entry in the elasticsearch mappings
332 table, in order to build the mappings for whatever is needed.
334 In the provided function, the files are:
336 =over 4
338 =item C<$name>
340 The field name for elasticsearch (corresponds to the 'mapping' column in the
341 database.
343 =item C<$type>
345 The type for this value, e.g. 'string'.
347 =item C<$facet>
349 True if this value should be facetised. This only really makes sense if the
350 field is understood by the facet processing code anyway.
352 =item C<$sort>
354 True if this is a field that a) needs special sort handling, and b) if it
355 should be sorted on. False if a) but not b). Undef if not a). This allows,
356 for example, author to be sorted on but not everything marked with "author"
357 to be included in that sort.
359 =item C<$marc_type>
361 A string that indicates the MARC type that this mapping is for, e.g. 'marc21',
362 'unimarc', 'normarc'.
364 =item C<$marc_field>
366 A string that describes the MARC field that contains the data to extract.
367 These are of a form suited to Catmandu's MARC fixers.
369 =back
371 =cut
373 sub _foreach_mapping {
374 my ( $self, $sub ) = @_;
376 # TODO use a caching framework here
377 my $search_fields = Koha::Database->schema->resultset('SearchField')->search(
379 'search_marc_map.index_name' => $self->index,
381 { join => { search_marc_to_fields => 'search_marc_map' },
382 '+select' => [
383 'search_marc_to_fields.facet',
384 'search_marc_to_fields.suggestible',
385 'search_marc_to_fields.sort',
386 'search_marc_map.marc_type',
387 'search_marc_map.marc_field',
389 '+as' => [
390 'facet',
391 'suggestible',
392 'sort',
393 'marc_type',
394 'marc_field',
399 while ( my $search_field = $search_fields->next ) {
400 $sub->(
401 $search_field->name,
402 $search_field->type,
403 $search_field->get_column('facet'),
404 $search_field->get_column('suggestible'),
405 $search_field->get_column('sort'),
406 $search_field->get_column('marc_type'),
407 $search_field->get_column('marc_field'),
412 =head2 process_error
414 die process_error($@);
416 This parses an Elasticsearch error message and produces a human-readable
417 result from it. This result is probably missing all the useful information
418 that you might want in diagnosing an issue, so the warning is also logged.
420 Note that currently the resulting message is not internationalised. This
421 will happen eventually by some method or other.
423 =cut
425 sub process_error {
426 my ($self, $msg) = @_;
428 warn $msg; # simple logging
430 # This is super-primitive
431 return "Unable to understand your search query, please rephrase and try again.\n" if $msg =~ /ParseException/;
433 return "Unable to perform your search. Please try again.\n";
438 __END__
440 =head1 AUTHOR
442 =over 4
444 =item Chris Cormack C<< <chrisc@catalyst.net.nz> >>
446 =item Robin Sheat C<< <robin@catalyst.net.nz> >>
448 =back
450 =cut