Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
55.32% |
52 / 94 |
|
28.57% |
2 / 7 |
CRAP | |
0.00% |
0 / 1 |
| LexemeTermResult | |
55.32% |
52 / 94 |
|
28.57% |
2 / 7 |
35.07 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getSourceFiltering | |
0.00% |
0 / 5 |
|
0.00% |
0 / 1 |
2 | |||
| getFields | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| getHighlightingConfiguration | |
0.00% |
0 / 32 |
|
0.00% |
0 / 1 |
2 | |||
| transformElasticsearchResult | |
93.88% |
46 / 49 |
|
0.00% |
0 / 1 |
8.01 | |||
| extractLanguageCode | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| createEmptyResult | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| 1 | <?php |
| 2 | namespace Wikibase\Lexeme\Search\Elastic; |
| 3 | |
| 4 | use CirrusSearch\Search\BaseResultsType; |
| 5 | use Elastica\ResultSet; |
| 6 | use Language; |
| 7 | use Wikibase\DataModel\Entity\EntityIdParser; |
| 8 | use Wikibase\DataModel\Term\Term; |
| 9 | use Wikibase\Lexeme\DataAccess\LexemeDescription; |
| 10 | use Wikibase\Lib\Interactors\TermSearchResult; |
| 11 | use Wikibase\Lib\Store\FallbackLabelDescriptionLookupFactory; |
| 12 | use Wikibase\Search\Elastic\EntitySearchUtils; |
| 13 | |
| 14 | /** |
| 15 | * This result type implements the result for searching a Wikibase Lexeme. |
| 16 | * |
| 17 | * @license GPL-2.0-or-later |
| 18 | * @author Stas Malyshev |
| 19 | */ |
| 20 | class LexemeTermResult extends BaseResultsType { |
| 21 | |
| 22 | /** |
| 23 | * @var EntityIdParser |
| 24 | */ |
| 25 | private $idParser; |
| 26 | |
| 27 | /** |
| 28 | * Display language |
| 29 | * @var Language |
| 30 | */ |
| 31 | private $displayLanguage; |
| 32 | |
| 33 | /** |
| 34 | * @var FallbackLabelDescriptionLookupFactory |
| 35 | */ |
| 36 | private $termLookupFactory; |
| 37 | |
| 38 | /** |
| 39 | * @param EntityIdParser $idParser |
| 40 | * @param Language $displayLanguage User display language |
| 41 | * @param FallbackLabelDescriptionLookupFactory $termLookupFactory |
| 42 | * Lookup factory for assembling descriptions |
| 43 | */ |
| 44 | public function __construct( |
| 45 | EntityIdParser $idParser, |
| 46 | Language $displayLanguage, |
| 47 | FallbackLabelDescriptionLookupFactory $termLookupFactory |
| 48 | ) { |
| 49 | $this->idParser = $idParser; |
| 50 | $this->termLookupFactory = $termLookupFactory; |
| 51 | $this->displayLanguage = $displayLanguage; |
| 52 | } |
| 53 | |
| 54 | /** |
| 55 | * Get the source filtering to be used loading the result. |
| 56 | * |
| 57 | * @return string[] |
| 58 | */ |
| 59 | public function getSourceFiltering() { |
| 60 | return array_merge( parent::getSourceFiltering(), [ |
| 61 | LemmaField::NAME, |
| 62 | LexemeLanguageField::NAME, |
| 63 | LexemeCategoryField::NAME, |
| 64 | ] ); |
| 65 | } |
| 66 | |
| 67 | /** |
| 68 | * Get the fields to load. Most of the time we'll use source filtering instead but |
| 69 | * some fields aren't part of the source. |
| 70 | * |
| 71 | * @return string[] |
| 72 | */ |
| 73 | public function getFields() { |
| 74 | return []; |
| 75 | } |
| 76 | |
| 77 | /** |
| 78 | * Get the highlighting configuration. |
| 79 | * |
| 80 | * @param array $highlightSource configuration for how to highlight the source. |
| 81 | * Empty if source should be ignored. |
| 82 | * @return array|null highlighting configuration for elasticsearch |
| 83 | */ |
| 84 | public function getHighlightingConfiguration( array $highlightSource ) { |
| 85 | $config = [ |
| 86 | 'pre_tags' => [ '' ], |
| 87 | 'post_tags' => [ '' ], |
| 88 | 'fields' => [], |
| 89 | ]; |
| 90 | $config['fields']['title'] = [ |
| 91 | 'type' => 'experimental', |
| 92 | 'fragmenter' => "none", |
| 93 | 'number_of_fragments' => 0, |
| 94 | 'matched_fields' => [ 'title.keyword' ] |
| 95 | ]; |
| 96 | $config['fields']["lemma"] = [ |
| 97 | 'type' => 'experimental', |
| 98 | 'fragmenter' => "none", |
| 99 | 'number_of_fragments' => 0, |
| 100 | 'options' => [ |
| 101 | 'skip_if_last_matched' => true, |
| 102 | ], |
| 103 | 'matched_fields' => [ 'lemma.prefix' ] |
| 104 | ]; |
| 105 | $config['fields']["lexeme_forms.representation"] = [ |
| 106 | 'type' => 'experimental', |
| 107 | 'fragmenter' => "none", |
| 108 | 'number_of_fragments' => 0, |
| 109 | "matched_fields" => [ |
| 110 | "lexeme_forms.representation.prefix", |
| 111 | ], |
| 112 | 'options' => [ |
| 113 | 'skip_if_last_matched' => true, |
| 114 | ], |
| 115 | ]; |
| 116 | |
| 117 | return $config; |
| 118 | } |
| 119 | |
| 120 | /** |
| 121 | * Convert search result from ElasticSearch result set to TermSearchResult. |
| 122 | * @param ResultSet $result |
| 123 | * @return TermSearchResult[] Set of search results, the types of which vary by implementation. |
| 124 | */ |
| 125 | public function transformElasticsearchResult( ResultSet $result ) { |
| 126 | $rawResults = $entityIds = []; |
| 127 | foreach ( $result->getResults() as $r ) { |
| 128 | $sourceData = $r->getSource(); |
| 129 | $entityId = EntitySearchUtils::parseOrNull( $sourceData['title'], $this->idParser ); |
| 130 | if ( !$entityId ) { |
| 131 | // Can not parse entity ID - skip it |
| 132 | continue; |
| 133 | } |
| 134 | |
| 135 | $lemmaCode = self::extractLanguageCode( $sourceData ); |
| 136 | |
| 137 | // Highlight part contains information about what has actually been matched. |
| 138 | $highlight = $r->getHighlights(); |
| 139 | |
| 140 | if ( !empty( $highlight['title'] ) ) { |
| 141 | // If we matched title, this means it's a match by ID |
| 142 | $matchedTermType = 'entityId'; |
| 143 | $matchedTerm = new Term( 'qid', $sourceData['title'] ); |
| 144 | } elseif ( empty( $highlight['lemma'] ) && empty( $highlight['lexeme_forms.representation'] ) ) { |
| 145 | // Something went wrong, we don't have any highlighting data |
| 146 | continue; |
| 147 | } elseif ( !empty( $highlight['lemma'] ) ) { |
| 148 | // We matched lemma |
| 149 | $matchedTermType = 'label'; |
| 150 | $matchedTerm = new Term( $lemmaCode, $highlight['lemma'][0] ); |
| 151 | } else { |
| 152 | // matched one of the forms |
| 153 | $matchedTermType = 'alias'; |
| 154 | // @phan-suppress-next-line PhanTypePossiblyInvalidDimOffset |
| 155 | $matchedTerm = new Term( $lemmaCode, $highlight['lexeme_forms.representation'][0] ); |
| 156 | } |
| 157 | |
| 158 | $lang = $sourceData['lexeme_language']['entity']; |
| 159 | $category = $sourceData['lexical_category']; |
| 160 | |
| 161 | $entityIds[$lang] = EntitySearchUtils::parseOrNull( $lang, $this->idParser ); |
| 162 | $entityIds[$category] = EntitySearchUtils::parseOrNull( $category, $this->idParser ); |
| 163 | |
| 164 | // Doing two-stage resolution here since we want to prefetch all labels for |
| 165 | // auxiliary entities before using them to construct descriptions. |
| 166 | $rawResults[$entityId->getSerialization()] = [ |
| 167 | 'id' => $entityId, |
| 168 | // TODO: this assumes we always take the first lemma. Maybe we should use |
| 169 | // the shortest language code or something. That would require us to index |
| 170 | // lemma language codes though. |
| 171 | 'lemma' => $sourceData['lemma'][0], |
| 172 | 'term' => $matchedTerm, |
| 173 | 'type' => $matchedTermType, |
| 174 | 'lang' => $lang, |
| 175 | 'langcode' => $lemmaCode, |
| 176 | 'category' => $category |
| 177 | ]; |
| 178 | } |
| 179 | |
| 180 | $langCode = $this->displayLanguage->getCode(); |
| 181 | if ( $entityIds ) { |
| 182 | // Create prefetched lookup |
| 183 | $termLookup = $this->termLookupFactory->newLabelDescriptionLookup( $this->displayLanguage, |
| 184 | array_filter( $entityIds ) ); |
| 185 | $descriptionMaker = new LexemeDescription( $termLookup, $this->idParser, |
| 186 | $this->displayLanguage ); |
| 187 | // Create full descriptons and instantiate TermSearchResult objects |
| 188 | return array_map( static function ( $raw ) use ( $descriptionMaker, $langCode ) { |
| 189 | return new TermSearchResult( |
| 190 | $raw['term'], |
| 191 | $raw['type'], |
| 192 | $raw['id'], |
| 193 | new Term( $raw['langcode'], $raw['lemma'] ), |
| 194 | // We are lying somewhat here, as description might be from fallback languages, |
| 195 | // but I am not sure there's any better way here. |
| 196 | new Term( $langCode, |
| 197 | $descriptionMaker->createDescription( $raw['id'], $raw['lang'], |
| 198 | $raw['category'] ) ) |
| 199 | ); |
| 200 | }, $rawResults ); |
| 201 | } else { |
| 202 | return []; |
| 203 | } |
| 204 | } |
| 205 | |
| 206 | /** |
| 207 | * @param array $sourceData the source data returned by elastic |
| 208 | * @return string the lexeme_language code if set, 'und' otherwise. |
| 209 | */ |
| 210 | public static function extractLanguageCode( array $sourceData ) { |
| 211 | if ( empty( $sourceData['lexeme_language']['code'] ) ) { |
| 212 | return 'und'; |
| 213 | } else { |
| 214 | return $sourceData['lexeme_language']['code']; |
| 215 | } |
| 216 | } |
| 217 | |
| 218 | /** |
| 219 | * @return TermSearchResult[] Empty set of search results |
| 220 | */ |
| 221 | public function createEmptyResult() { |
| 222 | return []; |
| 223 | } |
| 224 | |
| 225 | } |