class-query-parser.php
720 lines
| 1 | <?php |
| 2 | /** |
| 3 | * Parse a pure text query into WordPress Elasticsearch query. This builds on |
| 4 | * the Query_Builder() to provide search query parsing. |
| 5 | * |
| 6 | * The key part of this parser is taking a user's query string typed into a box |
| 7 | * and converting it into an ES search query. |
| 8 | * |
| 9 | * This varies by application, but roughly it means extracting some parts of the query |
| 10 | * (authors, tags, and phrases) that are treated as a filter. Then taking the |
| 11 | * remaining words and building the correct query (possibly with prefix searching |
| 12 | * if we are doing search as you type) |
| 13 | * |
| 14 | * This class only supports ES 2.x+ |
| 15 | * |
| 16 | * Disables comment chehcks. |
| 17 | * phpcs:disable Squiz.Commenting |
| 18 | * |
| 19 | * This parser builds queries of the form: |
| 20 | * bool: |
| 21 | * must: |
| 22 | * AND match of a single field (ideally an edgengram field) |
| 23 | * filter: |
| 24 | * filter clauses from context (eg @gibrown, #news, etc) |
| 25 | * should: |
| 26 | * boosting of results by various fields |
| 27 | * |
| 28 | * Features supported: |
| 29 | * - search as you type |
| 30 | * - phrases |
| 31 | * - supports querying across multiple languages at once |
| 32 | * |
| 33 | * @package automattic/jetpack-search |
| 34 | */ |
| 35 | |
| 36 | namespace Automattic\Jetpack\Search\WPES; |
| 37 | |
| 38 | if ( ! defined( 'ABSPATH' ) ) { |
| 39 | exit( 0 ); |
| 40 | } |
| 41 | |
| 42 | /** |
| 43 | * Query parser class. |
| 44 | */ |
| 45 | class Query_Parser extends Query_Builder { |
| 46 | protected $orig_query = ''; |
| 47 | protected $current_query = ''; |
| 48 | protected $langs; |
| 49 | protected $avail_langs = array( 'ar', 'bg', 'ca', 'cs', 'da', 'de', 'el', 'en', 'es', 'eu', 'fa', 'fi', 'fr', 'he', 'hi', 'hu', 'hy', 'id', 'it', 'ja', 'ko', 'nl', 'no', 'pt', 'ro', 'ru', 'sv', 'tr', 'zh' ); |
| 50 | |
| 51 | /** |
| 52 | * All fields. |
| 53 | * |
| 54 | * @var array |
| 55 | */ |
| 56 | public $all_fields; |
| 57 | |
| 58 | /** |
| 59 | * Phrase fields. |
| 60 | * |
| 61 | * @var array |
| 62 | */ |
| 63 | public $phrase_fields; |
| 64 | |
| 65 | public function __construct( $user_query, $langs ) { |
| 66 | $this->orig_query = $user_query; |
| 67 | $this->current_query = $this->orig_query; |
| 68 | $this->langs = $this->norm_langs( $langs ); |
| 69 | } |
| 70 | |
| 71 | protected $extracted_phrases = array(); |
| 72 | |
| 73 | public function get_current_query() { |
| 74 | return $this->current_query; |
| 75 | } |
| 76 | |
| 77 | public function set_current_query( $q ) { |
| 78 | $this->current_query = $q; |
| 79 | } |
| 80 | |
| 81 | /////////////////////////////////////////////////////// |
| 82 | // Methods for Building arrays of multilingual fields |
| 83 | |
| 84 | /* |
| 85 | * Normalize language codes |
| 86 | */ |
| 87 | public function norm_langs( $langs ) { |
| 88 | $lst = array(); |
| 89 | foreach ( $langs as $l ) { |
| 90 | $l = strtok( $l, '-_' ); |
| 91 | if ( in_array( $l, $this->avail_langs, true ) ) { |
| 92 | $lst[ $l ] = true; |
| 93 | } else { |
| 94 | $lst['default'] = true; |
| 95 | } |
| 96 | } |
| 97 | return array_keys( $lst ); |
| 98 | } |
| 99 | |
| 100 | public function get_lang_field_suffix() { |
| 101 | if ( ! is_array( $this->langs ) || empty( $this->langs ) ) { |
| 102 | return; |
| 103 | } |
| 104 | |
| 105 | // Returns the first language only |
| 106 | return $this->langs[0]; |
| 107 | } |
| 108 | |
| 109 | /* |
| 110 | * Take a list of field prefixes and expand them for multi-lingual |
| 111 | * with the provided boostings. |
| 112 | */ |
| 113 | public function merge_ml_fields( $fields2boosts, $additional_fields ) { |
| 114 | $flds = array(); |
| 115 | foreach ( $fields2boosts as $f => $b ) { |
| 116 | foreach ( $this->langs as $l ) { |
| 117 | $flds[] = $f . '.' . $l . '^' . $b; |
| 118 | } |
| 119 | } |
| 120 | foreach ( $additional_fields as $f ) { |
| 121 | $flds[] = $f; |
| 122 | } |
| 123 | return $flds; |
| 124 | } |
| 125 | |
| 126 | //////////////////////////////////// |
| 127 | // Extract Fields for Filtering on |
| 128 | |
| 129 | /* |
| 130 | * Extract any @mentions from the user query |
| 131 | * use them as a filter if we can find a wp.com id |
| 132 | * otherwise use them as a |
| 133 | * |
| 134 | * args: |
| 135 | * wpcom_id_field: wp.com id field |
| 136 | * must_query_fields: array of fields to search for matching results (optional) |
| 137 | * boost_query_fields: array of fields to search in for boosting results (optional) |
| 138 | * prefixes: array of prefixes that the user can use to indicate an author |
| 139 | * |
| 140 | * returns true/false of whether any were found |
| 141 | * |
| 142 | * See also: https://github.com/twitter/twitter-text/blob/master/java/src/com/twitter/Regex.java |
| 143 | */ |
| 144 | public function author_field_filter( $args ) { |
| 145 | $defaults = array( |
| 146 | 'wpcom_id_field' => 'author_id', |
| 147 | 'must_query_fields' => null, |
| 148 | 'boost_query_fields' => null, |
| 149 | 'prefixes' => array( '@' ), |
| 150 | ); |
| 151 | $args = wp_parse_args( $args, $defaults ); |
| 152 | |
| 153 | $names = array(); |
| 154 | foreach ( $args['prefixes'] as $p ) { |
| 155 | $found = $this->get_fields( $p ); |
| 156 | if ( $found ) { |
| 157 | foreach ( $found as $f ) { |
| 158 | $names[] = $f; |
| 159 | } |
| 160 | } |
| 161 | } |
| 162 | |
| 163 | if ( empty( $names ) ) { |
| 164 | return false; |
| 165 | } |
| 166 | |
| 167 | foreach ( $args['prefixes'] as $p ) { |
| 168 | $this->remove_fields( $p ); |
| 169 | } |
| 170 | |
| 171 | $user_ids = array(); |
| 172 | |
| 173 | //loop through the matches and separate into filters and queries |
| 174 | foreach ( $names as $n ) { |
| 175 | //check for exact match on login |
| 176 | $userdata = get_user_by( 'login', strtolower( $n ) ); |
| 177 | $filtering = false; |
| 178 | if ( $userdata ) { |
| 179 | $user_ids[ $userdata->ID ] = true; |
| 180 | $filtering = true; |
| 181 | } |
| 182 | |
| 183 | $is_phrase = false; |
| 184 | if ( preg_match( '/"/', $n ) ) { |
| 185 | $is_phrase = true; |
| 186 | $n = preg_replace( '/"/', '', $n ); |
| 187 | } |
| 188 | |
| 189 | if ( ! empty( $args['must_query_fields'] ) && ! $filtering ) { |
| 190 | if ( $is_phrase ) { |
| 191 | $this->add_query( |
| 192 | array( |
| 193 | 'multi_match' => array( |
| 194 | 'fields' => $args['must_query_fields'], |
| 195 | 'query' => $n, |
| 196 | 'type' => 'phrase', |
| 197 | ), |
| 198 | ) |
| 199 | ); |
| 200 | } else { |
| 201 | $this->add_query( |
| 202 | array( |
| 203 | 'multi_match' => array( |
| 204 | 'fields' => $args['must_query_fields'], |
| 205 | 'query' => $n, |
| 206 | ), |
| 207 | ) |
| 208 | ); |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 213 | if ( $is_phrase ) { |
| 214 | $this->add_query( |
| 215 | array( |
| 216 | 'multi_match' => array( |
| 217 | 'fields' => $args['boost_query_fields'], |
| 218 | 'query' => $n, |
| 219 | 'type' => 'phrase', |
| 220 | ), |
| 221 | ), |
| 222 | 'should' |
| 223 | ); |
| 224 | } else { |
| 225 | $this->add_query( |
| 226 | array( |
| 227 | 'multi_match' => array( |
| 228 | 'fields' => $args['boost_query_fields'], |
| 229 | 'query' => $n, |
| 230 | ), |
| 231 | ), |
| 232 | 'should' |
| 233 | ); |
| 234 | } |
| 235 | } |
| 236 | } |
| 237 | |
| 238 | if ( ! empty( $user_ids ) ) { |
| 239 | $user_ids = array_keys( $user_ids ); |
| 240 | $this->add_filter( array( 'terms' => array( $args['wpcom_id_field'] => $user_ids ) ) ); |
| 241 | } |
| 242 | |
| 243 | return true; |
| 244 | } |
| 245 | |
| 246 | /* |
| 247 | * Extract any prefix followed by text use them as a must clause, |
| 248 | * and optionally as a boost to the should query |
| 249 | * This can be used for hashtags. eg #News, or #"current events", |
| 250 | * but also works for any arbitrary field. eg from:Greg |
| 251 | * |
| 252 | * args: |
| 253 | * must_query_fields: array of fields that must match the tag (optional) |
| 254 | * boost_query_fields: array of fields to boost search on (optional) |
| 255 | * prefixes: array of prefixes that the user can use to indicate a tag |
| 256 | * |
| 257 | * returns true/false of whether any were found |
| 258 | * |
| 259 | */ |
| 260 | public function text_field_filter( $args ) { |
| 261 | $defaults = array( |
| 262 | 'must_query_fields' => array( 'tag.name' ), |
| 263 | 'boost_query_fields' => array( 'tag.name' ), |
| 264 | 'prefixes' => array( '#' ), |
| 265 | ); |
| 266 | $args = wp_parse_args( $args, $defaults ); |
| 267 | |
| 268 | $tags = array(); |
| 269 | foreach ( $args['prefixes'] as $p ) { |
| 270 | $found = $this->get_fields( $p ); |
| 271 | if ( $found ) { |
| 272 | foreach ( $found as $f ) { |
| 273 | $tags[] = $f; |
| 274 | } |
| 275 | } |
| 276 | } |
| 277 | |
| 278 | if ( empty( $tags ) ) { |
| 279 | return false; |
| 280 | } |
| 281 | |
| 282 | foreach ( $args['prefixes'] as $p ) { |
| 283 | $this->remove_fields( $p ); |
| 284 | } |
| 285 | |
| 286 | foreach ( $tags as $t ) { |
| 287 | $is_phrase = false; |
| 288 | if ( preg_match( '/"/', $t ) ) { |
| 289 | $is_phrase = true; |
| 290 | $t = preg_replace( '/"/', '', $t ); |
| 291 | } |
| 292 | |
| 293 | if ( ! empty( $args['must_query_fields'] ) ) { |
| 294 | if ( $is_phrase ) { |
| 295 | $this->add_query( |
| 296 | array( |
| 297 | 'multi_match' => array( |
| 298 | 'fields' => $args['must_query_fields'], |
| 299 | 'query' => $t, |
| 300 | 'type' => 'phrase', |
| 301 | ), |
| 302 | ) |
| 303 | ); |
| 304 | } else { |
| 305 | $this->add_query( |
| 306 | array( |
| 307 | 'multi_match' => array( |
| 308 | 'fields' => $args['must_query_fields'], |
| 309 | 'query' => $t, |
| 310 | ), |
| 311 | ) |
| 312 | ); |
| 313 | } |
| 314 | } |
| 315 | |
| 316 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 317 | if ( $is_phrase ) { |
| 318 | $this->add_query( |
| 319 | array( |
| 320 | 'multi_match' => array( |
| 321 | 'fields' => $args['boost_query_fields'], |
| 322 | 'query' => $t, |
| 323 | 'type' => 'phrase', |
| 324 | ), |
| 325 | ), |
| 326 | 'should' |
| 327 | ); |
| 328 | } else { |
| 329 | $this->add_query( |
| 330 | array( |
| 331 | 'multi_match' => array( |
| 332 | 'fields' => $args['boost_query_fields'], |
| 333 | 'query' => $t, |
| 334 | ), |
| 335 | ), |
| 336 | 'should' |
| 337 | ); |
| 338 | } |
| 339 | } |
| 340 | } |
| 341 | |
| 342 | return true; |
| 343 | } |
| 344 | |
| 345 | /* |
| 346 | * Extract anything surrounded by quotes or if there is an opening quote |
| 347 | * that is not complete, and add them to the query as a phrase query. |
| 348 | * Quotes can be either '' or "" |
| 349 | * |
| 350 | * args: |
| 351 | * must_query_fields: array of fields that must match the phrases |
| 352 | * boost_query_fields: array of fields to boost the phrases on (optional) |
| 353 | * |
| 354 | * returns true/false of whether any were found |
| 355 | * |
| 356 | */ |
| 357 | public function phrase_filter( $args ) { |
| 358 | $defaults = array( |
| 359 | 'must_query_fields' => array( 'all_content' ), |
| 360 | 'boost_query_fields' => array( 'title' ), |
| 361 | ); |
| 362 | $args = wp_parse_args( $args, $defaults ); |
| 363 | |
| 364 | $phrases = array(); |
| 365 | if ( preg_match_all( '/"([^"]+)"/', $this->current_query, $matches ) ) { |
| 366 | foreach ( $matches[1] as $match ) { |
| 367 | $phrases[] = $match; |
| 368 | } |
| 369 | $this->current_query = preg_replace( '/"([^"]+)"/', '', $this->current_query ); |
| 370 | } |
| 371 | |
| 372 | if ( preg_match_all( "/'([^']+)'/", $this->current_query, $matches ) ) { |
| 373 | foreach ( $matches[1] as $match ) { |
| 374 | $phrases[] = $match; |
| 375 | } |
| 376 | $this->current_query = preg_replace( "/'([^']+)'/", '', $this->current_query ); |
| 377 | } |
| 378 | |
| 379 | //look for a final, uncompleted phrase |
| 380 | $phrase_prefix = false; |
| 381 | if ( preg_match_all( '/"([^"]+)$/', $this->current_query, $matches ) ) { |
| 382 | $phrase_prefix = $matches[1][0]; |
| 383 | $this->current_query = preg_replace( '/"([^"]+)$/', '', $this->current_query ); |
| 384 | } |
| 385 | if ( preg_match_all( "/(?:'\B|\B')([^']+)$/", $this->current_query, $matches ) ) { |
| 386 | $phrase_prefix = $matches[1][0]; |
| 387 | $this->current_query = preg_replace( "/(?:'\B|\B')([^']+)$/", '', $this->current_query ); |
| 388 | } |
| 389 | |
| 390 | if ( $phrase_prefix ) { |
| 391 | $phrases[] = $phrase_prefix; |
| 392 | } |
| 393 | if ( empty( $phrases ) ) { |
| 394 | return false; |
| 395 | } |
| 396 | |
| 397 | foreach ( $phrases as $p ) { |
| 398 | $this->add_query( |
| 399 | array( |
| 400 | 'multi_match' => array( |
| 401 | 'fields' => $args['must_query_fields'], |
| 402 | 'query' => $p, |
| 403 | 'type' => 'phrase', |
| 404 | ), |
| 405 | ) |
| 406 | ); |
| 407 | |
| 408 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 409 | $this->add_query( |
| 410 | array( |
| 411 | 'multi_match' => array( |
| 412 | 'fields' => $args['boost_query_fields'], |
| 413 | 'query' => $p, |
| 414 | 'operator' => 'and', |
| 415 | ), |
| 416 | ), |
| 417 | 'should' |
| 418 | ); |
| 419 | } |
| 420 | } |
| 421 | |
| 422 | return true; |
| 423 | } |
| 424 | |
| 425 | /* |
| 426 | * Query fields based on the remaining parts of the query |
| 427 | * This could be the final AND part of the query terms to match, or it |
| 428 | * could be boosting certain elements of the query |
| 429 | * |
| 430 | * args: |
| 431 | * must_query_fields: array of fields that must match the remaining terms (optional) |
| 432 | * boost_query_fields: array of fields to boost the remaining terms on (optional) |
| 433 | * |
| 434 | */ |
| 435 | public function remaining_query( $args ) { |
| 436 | $defaults = array( |
| 437 | 'must_query_fields' => null, |
| 438 | 'boost_query_fields' => null, |
| 439 | 'boost_operator' => 'and', |
| 440 | 'boost_query_type' => 'best_fields', |
| 441 | ); |
| 442 | $args = wp_parse_args( $args, $defaults ); |
| 443 | |
| 444 | if ( empty( $this->current_query ) || ctype_space( $this->current_query ) ) { |
| 445 | return; |
| 446 | } |
| 447 | |
| 448 | if ( ! empty( $args['must_query_fields'] ) ) { |
| 449 | $this->add_query( |
| 450 | array( |
| 451 | 'multi_match' => array( |
| 452 | 'fields' => $args['must_query_fields'], |
| 453 | 'query' => $this->current_query, |
| 454 | 'operator' => 'and', |
| 455 | ), |
| 456 | ) |
| 457 | ); |
| 458 | } |
| 459 | |
| 460 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 461 | $this->add_query( |
| 462 | array( |
| 463 | 'multi_match' => array( |
| 464 | 'fields' => $args['boost_query_fields'], |
| 465 | 'query' => $this->current_query, |
| 466 | 'operator' => $args['boost_operator'], |
| 467 | 'type' => $args['boost_query_type'], |
| 468 | ), |
| 469 | ), |
| 470 | 'should' |
| 471 | ); |
| 472 | } |
| 473 | } |
| 474 | |
| 475 | /* |
| 476 | * Query fields using a prefix query (alphabetical expansions on the index). |
| 477 | * This is not recommended. Slower performance and worse relevancy. |
| 478 | * |
| 479 | * (UNTESTED! Copied from old prefix expansion code) |
| 480 | * |
| 481 | * args: |
| 482 | * must_query_fields: array of fields that must match the remaining terms (optional) |
| 483 | * boost_query_fields: array of fields to boost the remaining terms on (optional) |
| 484 | * |
| 485 | */ |
| 486 | public function remaining_prefix_query( $args ) { |
| 487 | $defaults = array( |
| 488 | 'must_query_fields' => array( 'all_content' ), |
| 489 | 'boost_query_fields' => array( 'title' ), |
| 490 | 'boost_operator' => 'and', |
| 491 | 'boost_query_type' => 'best_fields', |
| 492 | ); |
| 493 | $args = wp_parse_args( $args, $defaults ); |
| 494 | |
| 495 | if ( empty( $this->current_query ) || ctype_space( $this->current_query ) ) { |
| 496 | return; |
| 497 | } |
| 498 | |
| 499 | ////////////////////////////////// |
| 500 | // Example cases to think about: |
| 501 | // "elasticse" |
| 502 | // "elasticsearch" |
| 503 | // "elasticsearch " |
| 504 | // "elasticsearch lucen" |
| 505 | // "elasticsearch lucene" |
| 506 | // "the future" - note the stopword which will match nothing! |
| 507 | // "F1" - an exact match that also has tons of expansions |
| 508 | // "こんにちは" ja "hello" |
| 509 | // "こんにちは友人" ja "hello friend" - we just rely on the prefix phrase and ES to split words |
| 510 | // - this could still be better I bet. Maybe we need to analyze with ES first? |
| 511 | // |
| 512 | |
| 513 | ///////////////////////////// |
| 514 | //extract pieces of query |
| 515 | // eg: "PREFIXREMAINDER PREFIXWORD" |
| 516 | // "elasticsearch lucen" |
| 517 | |
| 518 | $prefix_word = false; |
| 519 | $prefix_remainder = false; |
| 520 | if ( preg_match_all( '/([^ ]+)$/', $this->current_query, $matches ) ) { |
| 521 | $prefix_word = $matches[1][0]; |
| 522 | } |
| 523 | |
| 524 | $prefix_remainder = preg_replace( '/([^ ]+)$/', '', $this->current_query ); |
| 525 | if ( ctype_space( $prefix_remainder ) ) { |
| 526 | $prefix_remainder = false; |
| 527 | } |
| 528 | |
| 529 | if ( ! $prefix_word ) { |
| 530 | //Space at the end of the query, so skip using a prefix query |
| 531 | if ( ! empty( $args['must_query_fields'] ) ) { |
| 532 | $this->add_query( |
| 533 | array( |
| 534 | 'multi_match' => array( |
| 535 | 'fields' => $args['must_query_fields'], |
| 536 | 'query' => $this->current_query, |
| 537 | 'operator' => 'and', |
| 538 | ), |
| 539 | ) |
| 540 | ); |
| 541 | } |
| 542 | |
| 543 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 544 | $this->add_query( |
| 545 | array( |
| 546 | 'multi_match' => array( |
| 547 | 'fields' => $args['boost_query_fields'], |
| 548 | 'query' => $this->current_query, |
| 549 | 'operator' => $args['boost_operator'], |
| 550 | 'type' => $args['boost_query_type'], |
| 551 | ), |
| 552 | ), |
| 553 | 'should' |
| 554 | ); |
| 555 | } |
| 556 | } else { |
| 557 | |
| 558 | //must match the prefix word and the prefix remainder |
| 559 | if ( ! empty( $args['must_query_fields'] ) ) { |
| 560 | //need to do an OR across a few fields to handle all cases |
| 561 | $must_q = array( |
| 562 | 'bool' => array( |
| 563 | 'should' => array(), |
| 564 | 'minimum_should_match' => 1, |
| 565 | ), |
| 566 | ); |
| 567 | |
| 568 | //treat all words as an exact search (boosts complete word like "news" |
| 569 | //from prefixes of "newspaper") |
| 570 | $must_q['bool']['should'][] = array( |
| 571 | 'multi_match' => array( |
| 572 | 'fields' => $this->all_fields, |
| 573 | // NOTE: This line has been disabled since $full_text is not available. |
| 574 | // 'query' => $full_text, |
| 575 | 'operator' => 'and', |
| 576 | 'type' => 'cross_fields', |
| 577 | ), |
| 578 | ); |
| 579 | |
| 580 | //always optimistically try and match the full text as a phrase |
| 581 | //prefix "the futu" should try to match "the future" |
| 582 | //otherwise the first stopword kinda breaks |
| 583 | //This also works as the prefix match for a single word "elasticsea" |
| 584 | $must_q['bool']['should'][] = array( |
| 585 | 'multi_match' => array( |
| 586 | 'fields' => $this->phrase_fields, |
| 587 | // NOTE: This line has been disabled since $full_text is not available. |
| 588 | // 'query' => $full_text, |
| 589 | 'operator' => 'and', |
| 590 | 'type' => 'phrase_prefix', |
| 591 | 'max_expansions' => 100, |
| 592 | ), |
| 593 | ); |
| 594 | |
| 595 | if ( $prefix_remainder ) { |
| 596 | //Multiple words found, so treat each word on its own and not just as |
| 597 | //a part of a phrase |
| 598 | //"elasticsearch lucen" => "elasticsearch" exact AND "lucen" prefix |
| 599 | $must_q['bool']['should'][] = array( |
| 600 | 'bool' => array( |
| 601 | 'must' => array( |
| 602 | array( |
| 603 | 'multi_match' => array( |
| 604 | 'fields' => $this->phrase_fields, |
| 605 | 'query' => $prefix_word, |
| 606 | 'operator' => 'and', |
| 607 | 'type' => 'phrase_prefix', |
| 608 | 'max_expansions' => 100, |
| 609 | ), |
| 610 | ), |
| 611 | array( |
| 612 | 'multi_match' => array( |
| 613 | 'fields' => $this->all_fields, |
| 614 | 'query' => $prefix_remainder, |
| 615 | 'operator' => 'and', |
| 616 | 'type' => 'cross_fields', |
| 617 | ), |
| 618 | ), |
| 619 | ), |
| 620 | ), |
| 621 | ); |
| 622 | } |
| 623 | |
| 624 | $this->add_query( $must_q ); |
| 625 | } |
| 626 | |
| 627 | //Now add any boosting of the query |
| 628 | if ( ! empty( $args['boost_query_fields'] ) ) { |
| 629 | //treat all words as an exact search (boosts complete word like "news" |
| 630 | //from prefixes of "newspaper") |
| 631 | $this->add_query( |
| 632 | array( |
| 633 | 'multi_match' => array( |
| 634 | 'fields' => $args['boost_query_fields'], |
| 635 | 'query' => $this->current_query, |
| 636 | 'operator' => $args['boost_query_operator'], |
| 637 | 'type' => $args['boost_query_type'], |
| 638 | ), |
| 639 | ), |
| 640 | 'should' |
| 641 | ); |
| 642 | |
| 643 | //optimistically boost the full phrase prefix match |
| 644 | $this->add_query( |
| 645 | array( |
| 646 | 'multi_match' => array( |
| 647 | 'fields' => $args['boost_query_fields'], |
| 648 | 'query' => $this->current_query, |
| 649 | 'operator' => 'and', |
| 650 | 'type' => 'phrase_prefix', |
| 651 | 'max_expansions' => 100, |
| 652 | ), |
| 653 | ) |
| 654 | ); |
| 655 | } |
| 656 | } |
| 657 | } |
| 658 | |
| 659 | /* |
| 660 | * Boost results based on the lang probability overlaps |
| 661 | * |
| 662 | * args: |
| 663 | * langs2prob: list of languages to search in with associated boosts |
| 664 | */ |
| 665 | public function boost_lang_probs( $langs2prob ) { |
| 666 | foreach ( $langs2prob as $p ) { |
| 667 | $this->add_function( |
| 668 | 'field_value_factor', |
| 669 | array( |
| 670 | 'modifier' => 'none', |
| 671 | 'factor' => $p, |
| 672 | 'missing' => 0.01, //1% chance doc did not have right lang detected |
| 673 | ) |
| 674 | ); |
| 675 | } |
| 676 | } |
| 677 | |
| 678 | //////////////////////////////////// |
| 679 | // Helper Methods |
| 680 | |
| 681 | //Get the text after some prefix. eg @gibrown, or @"Greg Brown" |
| 682 | protected function get_fields( $field_prefix ) { |
| 683 | $regex = '/' . $field_prefix . '(("[^"]+")|([^\\p{Z}]+))/'; |
| 684 | if ( preg_match_all( $regex, $this->current_query, $match ) ) { |
| 685 | return $match[1]; |
| 686 | } |
| 687 | return false; |
| 688 | } |
| 689 | |
| 690 | //Remove the prefix and text from the query |
| 691 | protected function remove_fields( $field_name ) { |
| 692 | $regex = '/' . $field_name . '(("[^"]+")|([^\\p{Z}]+))/'; |
| 693 | $this->current_query = preg_replace( $regex, '', $this->current_query ); |
| 694 | } |
| 695 | |
| 696 | //Best effort string truncation that splits on word breaks |
| 697 | protected function truncate_string( $string, $limit, $break = ' ' ) { |
| 698 | if ( mb_strwidth( $string ) <= $limit ) { |
| 699 | return $string; |
| 700 | } |
| 701 | |
| 702 | // walk backwards from $limit to find first break |
| 703 | $breakpoint = $limit; |
| 704 | $broken = false; |
| 705 | while ( $breakpoint > 0 ) { |
| 706 | if ( mb_strimwidth( $string, $breakpoint, 1 ) === $break ) { |
| 707 | $string = mb_strimwidth( $string, 0, $breakpoint ); |
| 708 | $broken = true; |
| 709 | break; |
| 710 | } |
| 711 | --$breakpoint; |
| 712 | } |
| 713 | // if we weren't able to find a break, need to chop mid-word |
| 714 | if ( ! $broken ) { |
| 715 | $string = mb_strimwidth( $string, 0, $limit ); |
| 716 | } |
| 717 | return $string; |
| 718 | } |
| 719 | } |
| 720 |