Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
81.03% |
94 / 116 |
|
73.33% |
11 / 15 |
CRAP | |
0.00% |
0 / 1 |
| BaseRegexFeature | |
81.03% |
94 / 116 |
|
73.33% |
11 / 15 |
55.61 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
1 | |||
| getValueDelimiters | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
1 | |||
| parseValue | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
4 | |||
| getFeatureName | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| getCrossSearchStrategy | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| doApplyExtended | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
5 | |||
| getFilterQuery | |
87.50% |
7 / 8 |
|
0.00% |
0 / 1 |
3.02 | |||
| buildHighlightFields | |
87.50% |
7 / 8 |
|
0.00% |
0 / 1 |
3.02 | |||
| getNonRegexFilterQuery | n/a |
0 / 0 |
n/a |
0 / 0 |
0 | |||||
| getRegexHLFlavor | n/a |
0 / 0 |
n/a |
0 / 0 |
0 | |||||
| buildRegexQuery | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
3 | |||
| configureHighlighting | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
2 | |||
| doGetRegexHLFields | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
3 | |||
| buildRegexWithPlugin | |
83.33% |
10 / 12 |
|
0.00% |
0 / 1 |
6.17 | |||
| buildRegexWithGroovy | |
0.00% |
0 / 18 |
|
0.00% |
0 / 1 |
6 | |||
| buildNonRegexHLFields | n/a |
0 / 0 |
n/a |
0 / 0 |
0 | |||||
| isRegexQuery | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
3 | |||
| trimFirstOccurrenceOfSlash | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
3 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace CirrusSearch\Query; |
| 4 | |
| 5 | use CirrusSearch\CrossSearchStrategy; |
| 6 | use CirrusSearch\Extra\Query\SourceRegex; |
| 7 | use CirrusSearch\Parser\AST\KeywordFeatureNode; |
| 8 | use CirrusSearch\Query\Builder\QueryBuildingContext; |
| 9 | use CirrusSearch\Search\Fetch\FetchPhaseConfigBuilder; |
| 10 | use CirrusSearch\Search\Fetch\HighlightedField; |
| 11 | use CirrusSearch\Search\Fetch\HighlightFieldGenerator; |
| 12 | use CirrusSearch\Search\Filters; |
| 13 | use CirrusSearch\Search\SearchContext; |
| 14 | use CirrusSearch\SearchConfig; |
| 15 | use CirrusSearch\WarningCollector; |
| 16 | use Elastica\Query\AbstractQuery; |
| 17 | use Wikimedia\Assert\Assert; |
| 18 | |
| 19 | /** |
| 20 | * Base class supporting regex searches. Works best when combined with the |
| 21 | * wikimedia-extra plugin for elasticsearch, but can also fallback to a groovy |
| 22 | * based implementation. Can be really expensive, but mostly ok if you have the |
| 23 | * extra plugin enabled. |
| 24 | * |
| 25 | * Examples: |
| 26 | * insource:/abc?/ |
| 27 | * |
| 28 | * @see SourceRegex |
| 29 | */ |
| 30 | abstract class BaseRegexFeature extends SimpleKeywordFeature implements FilterQueryFeature, HighlightingFeature { |
| 31 | /** |
| 32 | * @var string[] Elasticsearch field(s) to search against |
| 33 | */ |
| 34 | private $fields; |
| 35 | |
| 36 | /** |
| 37 | * @var bool Is this feature enabled? |
| 38 | */ |
| 39 | private $enabled; |
| 40 | |
| 41 | /** |
| 42 | * @var string Locale used for case conversions. It's important that this |
| 43 | * matches the locale used for lowercasing in the ngram index. |
| 44 | */ |
| 45 | private $languageCode; |
| 46 | |
| 47 | /** |
| 48 | * @var string[] Configuration flags for the regex plugin |
| 49 | */ |
| 50 | private $regexPlugin; |
| 51 | |
| 52 | /** |
| 53 | * @var int The maximum number of automaton states that Lucene's regex |
| 54 | * compilation can expand to (even temporarily). Provides protection |
| 55 | * against overloading the search cluster. Only works when using the |
| 56 | * extra plugin, groovy based execution is unbounded. |
| 57 | */ |
| 58 | private $maxDeterminizedStates; |
| 59 | |
| 60 | /** |
| 61 | * @var string timeout for regex queries |
| 62 | * with the extra plugin |
| 63 | */ |
| 64 | private $shardTimeout; |
| 65 | |
| 66 | /** |
| 67 | * @param SearchConfig $config |
| 68 | * @param string[] $fields |
| 69 | */ |
| 70 | public function __construct( SearchConfig $config, array $fields ) { |
| 71 | $this->enabled = $config->get( 'CirrusSearchEnableRegex' ); |
| 72 | $this->languageCode = $config->get( 'LanguageCode' ); |
| 73 | $this->regexPlugin = $config->getElement( 'CirrusSearchWikimediaExtraPlugin', 'regex' ); |
| 74 | $this->maxDeterminizedStates = $config->get( 'CirrusSearchRegexMaxDeterminizedStates' ); |
| 75 | Assert::precondition( $fields !== [], 'must have at least one field' ); |
| 76 | $this->fields = $fields; |
| 77 | $this->shardTimeout = $config->getElement( 'CirrusSearchSearchShardTimeout', 'regex' ); |
| 78 | } |
| 79 | |
| 80 | /** |
| 81 | * @return string[][] |
| 82 | */ |
| 83 | public function getValueDelimiters() { |
| 84 | return [ |
| 85 | [ |
| 86 | // simple search |
| 87 | 'delimiter' => '"' |
| 88 | ], |
| 89 | [ |
| 90 | // regex searches |
| 91 | 'delimiter' => '/', |
| 92 | // optional case insensitive suffix |
| 93 | 'suffixes' => 'i' |
| 94 | ] |
| 95 | ]; |
| 96 | } |
| 97 | |
| 98 | /** |
| 99 | * @param string $key |
| 100 | * @param string $value |
| 101 | * @param string $quotedValue |
| 102 | * @param string $valueDelimiter |
| 103 | * @param string $suffix |
| 104 | * @param WarningCollector $warningCollector |
| 105 | * @return array|false|null |
| 106 | */ |
| 107 | public function parseValue( $key, $value, $quotedValue, $valueDelimiter, $suffix, WarningCollector $warningCollector ) { |
| 108 | if ( $valueDelimiter === '/' ) { |
| 109 | if ( !$this->enabled ) { |
| 110 | $warningCollector->addWarning( 'cirrussearch-feature-not-available', "$key regex" ); |
| 111 | } |
| 112 | |
| 113 | $pattern = $this->trimFirstOccurrenceOfSlash( $quotedValue ); |
| 114 | |
| 115 | if ( $pattern === '' ) { |
| 116 | $warningCollector->addWarning( 'cirrussearch-regex-empty-expression', $key ); |
| 117 | } |
| 118 | |
| 119 | return [ |
| 120 | 'type' => 'regex', |
| 121 | 'pattern' => $pattern, |
| 122 | 'insensitive' => $suffix === 'i', |
| 123 | ]; |
| 124 | } |
| 125 | return parent::parseValue( $key, $value, $quotedValue, $valueDelimiter, $suffix, $warningCollector ); |
| 126 | } |
| 127 | |
| 128 | /** |
| 129 | * @param string $key |
| 130 | * @param string $valueDelimiter |
| 131 | * @return string |
| 132 | */ |
| 133 | public function getFeatureName( $key, $valueDelimiter ) { |
| 134 | if ( $valueDelimiter === '/' ) { |
| 135 | return 'regex'; |
| 136 | } |
| 137 | return parent::getFeatureName( $key, $valueDelimiter ); |
| 138 | } |
| 139 | |
| 140 | /** |
| 141 | * @param KeywordFeatureNode $node |
| 142 | * @return CrossSearchStrategy |
| 143 | */ |
| 144 | public function getCrossSearchStrategy( KeywordFeatureNode $node ) { |
| 145 | if ( $node->getDelimiter() === '/' ) { |
| 146 | return CrossSearchStrategy::hostWikiOnlyStrategy(); |
| 147 | } else { |
| 148 | return CrossSearchStrategy::allWikisStrategy(); |
| 149 | } |
| 150 | } |
| 151 | |
| 152 | /** |
| 153 | * @param SearchContext $context |
| 154 | * @param string $key |
| 155 | * @param string $value |
| 156 | * @param string $quotedValue |
| 157 | * @param bool $negated |
| 158 | * @param string $delimiter |
| 159 | * @param string $suffix |
| 160 | * @return array |
| 161 | */ |
| 162 | public function doApplyExtended( SearchContext $context, $key, $value, $quotedValue, $negated, $delimiter, $suffix ) { |
| 163 | $parsedValue = $this->parseValue( $key, $value, $quotedValue, $delimiter, $suffix, $context ); |
| 164 | if ( $this->isRegexQuery( $parsedValue ) ) { |
| 165 | if ( !$this->enabled ) { |
| 166 | return [ null, false ]; |
| 167 | } |
| 168 | '@phan-var array $parsedValue'; |
| 169 | $pattern = $parsedValue['pattern']; |
| 170 | $insensitive = $parsedValue['insensitive']; |
| 171 | |
| 172 | if ( $pattern === '' ) { |
| 173 | $context->setResultsPossible( false ); |
| 174 | |
| 175 | return [ null, false ]; |
| 176 | } |
| 177 | |
| 178 | $filter = $this->buildRegexQuery( $pattern, $insensitive ); |
| 179 | if ( !$negated ) { |
| 180 | $this->configureHighlighting( $pattern, $insensitive, $context->getFetchPhaseBuilder() ); |
| 181 | } |
| 182 | return [ $filter, false ]; |
| 183 | } else { |
| 184 | return $this->doApply( $context, $key, $value, $quotedValue, $negated ); |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | /** |
| 189 | * @inheritDoc |
| 190 | */ |
| 191 | public function getFilterQuery( KeywordFeatureNode $node, QueryBuildingContext $context ) { |
| 192 | $parsedValue = $node->getParsedValue(); |
| 193 | if ( $this->isRegexQuery( $parsedValue ) ) { |
| 194 | if ( !$this->enabled ) { |
| 195 | return null; |
| 196 | } |
| 197 | '@phan-var array $parsedValue'; |
| 198 | $pattern = $parsedValue['pattern']; |
| 199 | $insensitive = $parsedValue['insensitive']; |
| 200 | return $this->buildRegexQuery( $pattern, $insensitive ); |
| 201 | } else { |
| 202 | return $this->getNonRegexFilterQuery( $node, $context ); |
| 203 | } |
| 204 | } |
| 205 | |
| 206 | /** |
| 207 | * @inheritDoc |
| 208 | */ |
| 209 | public function buildHighlightFields( KeywordFeatureNode $node, QueryBuildingContext $context ) { |
| 210 | $parsedValue = $node->getParsedValue(); |
| 211 | if ( $this->isRegexQuery( $parsedValue ) ) { |
| 212 | if ( !$this->enabled ) { |
| 213 | return []; |
| 214 | } |
| 215 | '@phan-var array $parsedValue'; |
| 216 | $pattern = $parsedValue['pattern']; |
| 217 | $insensitive = $parsedValue['insensitive']; |
| 218 | return $this->doGetRegexHLFields( $context->getHighlightFieldGenerator(), $pattern, $insensitive ); |
| 219 | } |
| 220 | return $this->buildNonRegexHLFields( $node, $context ); |
| 221 | } |
| 222 | |
| 223 | /** |
| 224 | * Obtain the filter when the keyword is used in non regex mode. |
| 225 | * This method will be called on syntax like keyword:word or keyword:"phrase" |
| 226 | * @param KeywordFeatureNode $node |
| 227 | * @param QueryBuildingContext $context |
| 228 | * @return AbstractQuery|null |
| 229 | */ |
| 230 | abstract protected function getNonRegexFilterQuery( KeywordFeatureNode $node, QueryBuildingContext $context ); |
| 231 | |
| 232 | /** |
| 233 | * Determine the flavor of regex highlighting to apply. |
| 234 | * @return string one of: java, lucene, lucene_extended, lucene_anchored |
| 235 | */ |
| 236 | abstract protected function getRegexHLFlavor(): string; |
| 237 | |
| 238 | /** |
| 239 | * @param string $pattern |
| 240 | * @param bool $insensitive |
| 241 | * @return AbstractQuery |
| 242 | */ |
| 243 | private function buildRegexQuery( $pattern, $insensitive ) { |
| 244 | return $this->regexPlugin && in_array( 'use', $this->regexPlugin ) |
| 245 | ? $this->buildRegexWithPlugin( $pattern, $insensitive ) |
| 246 | : $this->buildRegexWithGroovy( $pattern, $insensitive ); |
| 247 | } |
| 248 | |
| 249 | /** |
| 250 | * @param string $pattern |
| 251 | * @param bool $insensitive |
| 252 | * @param FetchPhaseConfigBuilder $fetchPhaseConfigBuilder |
| 253 | */ |
| 254 | private function configureHighlighting( $pattern, $insensitive, FetchPhaseConfigBuilder $fetchPhaseConfigBuilder ) { |
| 255 | foreach ( $this->doGetRegexHLFields( $fetchPhaseConfigBuilder, $pattern, $insensitive ) as $f ) { |
| 256 | $fetchPhaseConfigBuilder->addHLField( $f ); |
| 257 | } |
| 258 | } |
| 259 | |
| 260 | /** |
| 261 | * @param HighlightFieldGenerator $generator |
| 262 | * @param string $pattern |
| 263 | * @param bool $insensitive |
| 264 | * @return HighlightedField[] |
| 265 | */ |
| 266 | private function doGetRegexHLFields( HighlightFieldGenerator $generator, $pattern, $insensitive ) { |
| 267 | $fields = []; |
| 268 | if ( !$generator->supportsRegexFields() ) { |
| 269 | return $fields; |
| 270 | } |
| 271 | $regexFlavor = $this->getRegexHLFlavor(); |
| 272 | foreach ( $this->fields as $field => $hlTarget ) { |
| 273 | $fields[] = $generator->newRegexField( "$field.plain", $hlTarget, |
| 274 | $pattern, $insensitive, HighlightedField::COSTLY_EXPERT_SYNTAX_PRIORITY, |
| 275 | $regexFlavor ); |
| 276 | } |
| 277 | return $fields; |
| 278 | } |
| 279 | |
| 280 | /** |
| 281 | * Builds a regular expression query using the wikimedia-extra plugin. |
| 282 | * |
| 283 | * @param string $pattern The regular expression to match |
| 284 | * @param bool $insensitive Should the match be case insensitive? |
| 285 | * @return AbstractQuery Regular expression query |
| 286 | */ |
| 287 | private function buildRegexWithPlugin( $pattern, $insensitive ) { |
| 288 | $filters = []; |
| 289 | // TODO: Update plugin to accept multiple values for the field property |
| 290 | // so that at index time we can create a single trigram index with |
| 291 | // copy_to instead of creating multiple queries. |
| 292 | foreach ( $this->fields as $field => $hlTarget ) { |
| 293 | $filter = new SourceRegex( $pattern, $field, $field . '.trigram' ); |
| 294 | // set some defaults |
| 295 | $filter->setMaxDeterminizedStates( $this->maxDeterminizedStates ); |
| 296 | if ( isset( $this->regexPlugin['max_ngrams_extracted'] ) && is_numeric( $this->regexPlugin['max_ngrams_extracted'] ) ) { |
| 297 | $filter->setMaxNgramsExtracted( (int)$this->regexPlugin['max_ngrams_extracted'] ); |
| 298 | } |
| 299 | if ( isset( $this->regexPlugin['max_ngram_clauses'] ) && is_numeric( $this->regexPlugin['max_ngram_clauses'] ) ) { |
| 300 | $filter->setMaxNgramClauses( (int)$this->regexPlugin['max_ngram_clauses'] ); |
| 301 | } |
| 302 | $filter->setCaseSensitive( !$insensitive ); |
| 303 | $filter->setLocale( $this->languageCode ); |
| 304 | |
| 305 | $filters[] = $filter; |
| 306 | } |
| 307 | |
| 308 | return Filters::booleanOr( $filters ); |
| 309 | } |
| 310 | |
| 311 | /** |
| 312 | * Builds a regular expression query using groovy. It's significantly less |
| 313 | * good than the wikimedia-extra plugin, but it's something. |
| 314 | * |
| 315 | * @param string $pattern The regular expression to match |
| 316 | * @param bool $insensitive Should the match be case insensitive? |
| 317 | * @return AbstractQuery Regular expression query |
| 318 | */ |
| 319 | private function buildRegexWithGroovy( $pattern, $insensitive ) { |
| 320 | $filters = []; |
| 321 | foreach ( $this->fields as $field ) { |
| 322 | $script = <<<GROOVY |
| 323 | import org.apache.lucene.util.automaton.*; |
| 324 | sourceText = _source.get("{$field}"); |
| 325 | if (sourceText == null) { |
| 326 | false; |
| 327 | } else { |
| 328 | if (automaton == null) { |
| 329 | if (insensitive) { |
| 330 | locale = new Locale(language); |
| 331 | pattern = pattern.toLowerCase(locale); |
| 332 | } |
| 333 | regexp = new RegExp(pattern, RegExp.ALL ^ RegExp.AUTOMATON); |
| 334 | automaton = new CharacterRunAutomaton(regexp.toAutomaton()); |
| 335 | } |
| 336 | if (insensitive) { |
| 337 | sourceText = sourceText.toLowerCase(locale); |
| 338 | } |
| 339 | automaton.run(sourceText); |
| 340 | } |
| 341 | |
| 342 | GROOVY; |
| 343 | |
| 344 | $filters[] = new \Elastica\Query\Script( new \Elastica\Script\Script( |
| 345 | $script, |
| 346 | [ |
| 347 | 'pattern' => '.*(' . $pattern . ').*', |
| 348 | 'insensitive' => $insensitive, |
| 349 | 'language' => $this->languageCode, |
| 350 | // The null here creates a slot in which the script will shove |
| 351 | // an automaton while executing. |
| 352 | 'automaton' => null, |
| 353 | 'locale' => null, |
| 354 | ], |
| 355 | 'groovy' |
| 356 | ) ); |
| 357 | } |
| 358 | |
| 359 | return Filters::booleanOr( $filters ); |
| 360 | } |
| 361 | |
| 362 | /** |
| 363 | * @param KeywordFeatureNode $node |
| 364 | * @param QueryBuildingContext $context |
| 365 | * @return HighlightedField[] |
| 366 | */ |
| 367 | abstract public function buildNonRegexHLFields( KeywordFeatureNode $node, QueryBuildingContext $context ); |
| 368 | |
| 369 | /** |
| 370 | * @param array|null $parsedValue |
| 371 | * @return bool |
| 372 | */ |
| 373 | private function isRegexQuery( ?array $parsedValue = null ) { |
| 374 | return is_array( $parsedValue ) && isset( $parsedValue['type'] ) && |
| 375 | $parsedValue['type'] === 'regex'; |
| 376 | } |
| 377 | |
| 378 | /** |
| 379 | * @param string $quotedValue |
| 380 | * @return false|string |
| 381 | */ |
| 382 | private function trimFirstOccurrenceOfSlash( string $quotedValue ) { |
| 383 | $pattern = $quotedValue; |
| 384 | if ( str_starts_with( $pattern, '/' ) ) { |
| 385 | $pattern = substr( $pattern, 1 ); |
| 386 | } |
| 387 | if ( str_ends_with( $pattern, '/' ) ) { |
| 388 | $pattern = substr( $pattern, 0, -1 ); |
| 389 | } |
| 390 | |
| 391 | return $pattern; |
| 392 | } |
| 393 | } |