/** * Optimize query in the context of specified index * * @param \Zend\Search\Lucene\SearchIndex $index * @return \Zend\Search\Lucene\Search\Query\AbstractQuery */ public function optimize(Lucene\SearchIndex $index) { $subqueries = array(); $signs = array(); // Optimize all subqueries foreach ($this->_subqueries as $id => $subquery) { $subqueries[] = $subquery->optimize($index); $signs[] = ($this->_signs === null)? true : $this->_signs[$id]; } // Remove insignificant subqueries foreach ($subqueries as $id => $subquery) { if ($subquery instanceof Insignificant) { // Insignificant subquery has to be removed anyway unset($subqueries[$id]); unset($signs[$id]); } } if (count($subqueries) == 0) { // Boolean query doesn't has non-insignificant subqueries return new Insignificant(); } // Check if all non-insignificant subqueries are prohibited $allProhibited = true; foreach ($signs as $sign) { if ($sign !== false) { $allProhibited = false; break; } } if ($allProhibited) { return new Insignificant(); } // Check for empty subqueries foreach ($subqueries as $id => $subquery) { if ($subquery instanceof EmptyResult) { if ($signs[$id] === true) { // Matching is required, but is actually empty return new EmptyResult(); } else { // Matching is optional or prohibited, but is empty // Remove it from subqueries and signs list unset($subqueries[$id]); unset($signs[$id]); } } } // Check, if reduced subqueries list is empty if (count($subqueries) == 0) { return new EmptyResult(); } // Check if all non-empty subqueries are prohibited $allProhibited = true; foreach ($signs as $sign) { if ($sign !== false) { $allProhibited = false; break; } } if ($allProhibited) { return new EmptyResult(); } // Check, if reduced subqueries list has only one entry if (count($subqueries) == 1) { // It's a query with only one required or optional clause // (it's already checked, that it's not a prohibited clause) if ($this->getBoost() == 1) { return reset($subqueries); } $optimizedQuery = clone reset($subqueries); $optimizedQuery->setBoost($optimizedQuery->getBoost()*$this->getBoost()); return $optimizedQuery; } // Prepare first candidate for optimized query $optimizedQuery = new self($subqueries, $signs); $optimizedQuery->setBoost($this->getBoost()); $terms = array(); $tsigns = array(); $boostFactors = array(); // Try to decompose term and multi-term subqueries foreach ($subqueries as $id => $subquery) { if ($subquery instanceof Term) { $terms[] = $subquery->getTerm(); $tsigns[] = $signs[$id]; $boostFactors[] = $subquery->getBoost(); // remove subquery from a subqueries list unset($subqueries[$id]); unset($signs[$id]); } else if ($subquery instanceof MultiTerm) { $subTerms = $subquery->getTerms(); $subSigns = $subquery->getSigns(); if ($signs[$id] === true) { // It's a required multi-term subquery. // Something like '... +(+term1 -term2 term3 ...) ...' // Multi-term required subquery can be decomposed only if it contains // required terms and doesn't contain prohibited terms: // ... +(+term1 term2 ...) ... => ... +term1 term2 ... // // Check this $hasRequired = false; $hasProhibited = false; if ($subSigns === null) { // All subterms are required $hasRequired = true; } else { foreach ($subSigns as $sign) { if ($sign === true) { $hasRequired = true; } else if ($sign === false) { $hasProhibited = true; break; } } } // Continue if subquery has prohibited terms or doesn't have required terms if ($hasProhibited || !$hasRequired) { continue; } foreach ($subTerms as $termId => $term) { $terms[] = $term; $tsigns[] = ($subSigns === null)? true : $subSigns[$termId]; $boostFactors[] = $subquery->getBoost(); } // remove subquery from a subqueries list unset($subqueries[$id]); unset($signs[$id]); } else { // $signs[$id] === null || $signs[$id] === false // It's an optional or prohibited multi-term subquery. // Something like '... (+term1 -term2 term3 ...) ...' // or // something like '... -(+term1 -term2 term3 ...) ...' // Multi-term optional and required subqueries can be decomposed // only if all terms are optional. // // Check if all terms are optional. $onlyOptional = true; if ($subSigns === null) { // All subterms are required $onlyOptional = false; } else { foreach ($subSigns as $sign) { if ($sign !== null) { $onlyOptional = false; break; } } } // Continue if non-optional terms are presented in this multi-term subquery if (!$onlyOptional) { continue; } foreach ($subTerms as $termId => $term) { $terms[] = $term; $tsigns[] = ($signs[$id] === null)? null /* optional */ : false /* prohibited */; $boostFactors[] = $subquery->getBoost(); } // remove subquery from a subqueries list unset($subqueries[$id]); unset($signs[$id]); } } } // Check, if there are no decomposed subqueries if (count($terms) == 0 ) { // return prepared candidate return $optimizedQuery; } // Check, if all subqueries have been decomposed and all terms has the same boost factor if (count($subqueries) == 0 && count(array_unique($boostFactors)) == 1) { $optimizedQuery = new MultiTerm($terms, $tsigns); $optimizedQuery->setBoost(reset($boostFactors)*$this->getBoost()); return $optimizedQuery; } // This boolean query can't be transformed to Term/MultiTerm query and still contains // several subqueries // Separate prohibited terms $prohibitedTerms = array(); foreach ($terms as $id => $term) { if ($tsigns[$id] === false) { $prohibitedTerms[] = $term; unset($terms[$id]); unset($tsigns[$id]); unset($boostFactors[$id]); } } if (count($terms) == 1) { $clause = new Term(reset($terms)); $clause->setBoost(reset($boostFactors)); $subqueries[] = $clause; $signs[] = reset($tsigns); // Clear terms list $terms = array(); } else if (count($terms) > 1 && count(array_unique($boostFactors)) == 1) { $clause = new MultiTerm($terms, $tsigns); $clause->setBoost(reset($boostFactors)); $subqueries[] = $clause; // Clause sign is 'required' if clause contains required terms. 'Optional' otherwise. $signs[] = (in_array(true, $tsigns))? true : null; // Clear terms list $terms = array(); } if (count($prohibitedTerms) == 1) { // (boost factors are not significant for prohibited clauses) $subqueries[] = new Term(reset($prohibitedTerms)); $signs[] = false; // Clear prohibited terms list $prohibitedTerms = array(); } else if (count($prohibitedTerms) > 1) { // prepare signs array $prohibitedSigns = array(); foreach ($prohibitedTerms as $id => $term) { // all prohibited term are grouped as optional into multi-term query $prohibitedSigns[$id] = null; } // (boost factors are not significant for prohibited clauses) $subqueries[] = new MultiTerm($prohibitedTerms, $prohibitedSigns); // Clause sign is 'prohibited' $signs[] = false; // Clear terms list $prohibitedTerms = array(); } /** @todo Group terms with the same boost factors together */ // Check, that all terms are processed // Replace candidate for optimized query if (count($terms) == 0 && count($prohibitedTerms) == 0) { $optimizedQuery = new self($subqueries, $signs); $optimizedQuery->setBoost($this->getBoost()); } return $optimizedQuery; }
/** * Optimize query in the context of specified index * * @param \Zend\Search\Lucene\SearchIndex $index * @return \Zend\Search\Lucene\Search\Query\AbstractQuery */ public function optimize(Lucene\SearchIndex $index) { // Check, that index contains all phrase terms foreach ($this->_terms as $term) { if (!$index->hasTerm($term)) { return new EmptyResult(); } } if (count($this->_terms) == 1) { // It's one term query $optimizedQuery = new Term(reset($this->_terms)); $optimizedQuery->setBoost($this->getBoost()); return $optimizedQuery; } if (count($this->_terms) == 0) { return new EmptyResult(); } return $this; }
/** * Re-write query into primitive queries in the context of specified index * * @param \Zend\Search\Lucene\SearchIndexInterface $index * @return \Zend\Search\Lucene\Search\Query\AbstractQuery */ public function rewrite(Lucene\SearchIndexInterface $index) { // Allow to use wildcards within phrases // They are either removed by text analyzer or used as a part of keyword for keyword fields // // if (strpos($this->_phrase, '?') !== false || strpos($this->_phrase, '*') !== false) { // require_once 'Zend/Search/Lucene/Search/QueryParserException.php'; // throw new Zend_Search_Lucene_Search_QueryParserException('Wildcards are only allowed in a single terms.'); // } // Split query into subqueries if field name is not specified if ($this->_field === null) { $query = new Query\Boolean(); $query->setBoost($this->getBoost()); if (Lucene\Lucene::getDefaultSearchField() === null) { $searchFields = $index->getFieldNames(true); } else { $searchFields = array(Lucene\Lucene::getDefaultSearchField()); } foreach ($searchFields as $fieldName) { $subquery = new Phrase($this->_phrase, $this->_phraseEncoding, $fieldName); $subquery->setSlop($this->getSlop()); $query->addSubquery($subquery->rewrite($index)); } $this->_matches = $query->getQueryTerms(); return $query; } // Recognize exact term matching (it corresponds to Keyword fields stored in the index) // encoding is not used since we expect binary matching $term = new Index\Term($this->_phrase, $this->_field); if ($index->hasTerm($term)) { $query = new Query\Term($term); $query->setBoost($this->getBoost()); $this->_matches = $query->getQueryTerms(); return $query; } // tokenize phrase using current analyzer and process it as a phrase query $tokens = Analyzer::getDefault()->tokenize($this->_phrase, $this->_phraseEncoding); if (count($tokens) == 0) { $this->_matches = array(); return new Query\Insignificant(); } if (count($tokens) == 1) { $term = new Index\Term($tokens[0]->getTermText(), $this->_field); $query = new Query\Term($term); $query->setBoost($this->getBoost()); $this->_matches = $query->getQueryTerms(); return $query; } //It's non-trivial phrase query $position = -1; $query = new Query\Phrase(); foreach ($tokens as $token) { $position += $token->getPositionIncrement(); $term = new Index\Term($token->getTermText(), $this->_field); $query->addTerm($term, $position); $query->setSlop($this->getSlop()); } $this->_matches = $query->getQueryTerms(); return $query; }
/** * Optimize query in the context of specified index * * @param \Zend\Search\Lucene\SearchIndex $index * @return \Zend\Search\Lucene\Search\Query\AbstractQuery */ public function optimize(Lucene\SearchIndex $index) { $terms = $this->_terms; $signs = $this->_signs; foreach ($terms as $id => $term) { if (!$index->hasTerm($term)) { if ($signs === null || $signs[$id] === true) { // Term is required return new EmptyResult(); } else { // Term is optional or prohibited // Remove it from terms and signs list unset($terms[$id]); unset($signs[$id]); } } } // Check if all presented terms are prohibited $allProhibited = true; if ($signs === null) { $allProhibited = false; } else { foreach ($signs as $sign) { if ($sign !== false) { $allProhibited = false; break; } } } if ($allProhibited) { return new EmptyResult(); } /** * @todo make an optimization for repeated terms * (they may have different signs) */ if (count($terms) == 1) { // It's already checked, that it's not a prohibited term // It's one term query with one required or optional element $optimizedQuery = new Term(reset($terms)); $optimizedQuery->setBoost($this->getBoost()); return $optimizedQuery; } if (count($terms) == 0) { return new EmptyResult(); } $optimizedQuery = new MultiTerm($terms, $signs); $optimizedQuery->setBoost($this->getBoost()); return $optimizedQuery; }
/** * Re-write query into primitive queries in the context of specified index * * @param \Zend\Search\Lucene\SearchIndex $index * @throws \Zend\Search\Lucence\Search\Exception\QueryParserException * @return \Zend\Search\Lucene\Search\Query\AbstractQuery */ public function rewrite(Lucene\SearchIndex $index) { if ($this->_field === null) { $query = new Query\MultiTerm(); $query->setBoost($this->getBoost()); $hasInsignificantSubqueries = false; if (Lucene\Lucene::getDefaultSearchField() === null) { $searchFields = $index->getFieldNames(true); } else { $searchFields = array(Lucene\Lucene::getDefaultSearchField()); } foreach ($searchFields as $fieldName) { $subquery = new Term($this->_word, $this->_encoding, $fieldName); $rewrittenSubquery = $subquery->rewrite($index); foreach ($rewrittenSubquery->getQueryTerms() as $term) { $query->addTerm($term); } if ($rewrittenSubquery instanceof Query\Insignificant) { $hasInsignificantSubqueries = true; } } if (count($query->getTerms()) == 0) { $this->_matches = array(); if ($hasInsignificantSubqueries) { return new Query\Insignificant(); } else { return new Query\EmptyResult(); } } $this->_matches = $query->getQueryTerms(); return $query; } // ------------------------------------- // Recognize exact term matching (it corresponds to Keyword fields stored in the index) // encoding is not used since we expect binary matching $term = new Index\Term($this->_word, $this->_field); if ($index->hasTerm($term)) { $query = new Query\Term($term); $query->setBoost($this->getBoost()); $this->_matches = $query->getQueryTerms(); return $query; } // ------------------------------------- // Recognize wildcard queries /** @todo check for PCRE unicode support may be performed through Zend_Environment in some future */ if (@preg_match('/\\pL/u', 'a') == 1) { $word = iconv($this->_encoding, 'UTF-8', $this->_word); $wildcardsPattern = '/[*?]/u'; $subPatternsEncoding = 'UTF-8'; } else { $word = $this->_word; $wildcardsPattern = '/[*?]/'; $subPatternsEncoding = $this->_encoding; } $subPatterns = preg_split($wildcardsPattern, $word, -1, PREG_SPLIT_OFFSET_CAPTURE); if (count($subPatterns) > 1) { // Wildcard query is recognized $pattern = ''; foreach ($subPatterns as $id => $subPattern) { // Append corresponding wildcard character to the pattern before each sub-pattern (except first) if ($id != 0) { $pattern .= $word[$subPattern[1] - 1]; } // Check if each subputtern is a single word in terms of current analyzer $tokens = Analyzer\Analyzer::getDefault()->tokenize($subPattern[0], $subPatternsEncoding); if (count($tokens) > 1) { throw new QueryParserException('Wildcard search is supported only for non-multiple word terms'); } foreach ($tokens as $token) { $pattern .= $token->getTermText(); } } $term = new Index\Term($pattern, $this->_field); $query = new Query\Wildcard($term); $query->setBoost($this->getBoost()); // Get rewritten query. Important! It also fills terms matching container. $rewrittenQuery = $query->rewrite($index); $this->_matches = $query->getQueryTerms(); return $rewrittenQuery; } // ------------------------------------- // Recognize one-term multi-term and "insignificant" queries $tokens = Analyzer\Analyzer::getDefault()->tokenize($this->_word, $this->_encoding); if (count($tokens) == 0) { $this->_matches = array(); return new Query\Insignificant(); } if (count($tokens) == 1) { $term = new Index\Term($tokens[0]->getTermText(), $this->_field); $query = new Query\Term($term); $query->setBoost($this->getBoost()); $this->_matches = $query->getQueryTerms(); return $query; } //It's not insignificant or one term query $query = new Query\MultiTerm(); /** * @todo Process $token->getPositionIncrement() to support stemming, synonyms and other * analizer design features */ foreach ($tokens as $token) { $term = new Index\Term($token->getTermText(), $this->_field); $query->addTerm($term, true); // all subterms are required } $query->setBoost($this->getBoost()); $this->_matches = $query->getQueryTerms(); return $query; }
/** * Re-write query into primitive queries in the context of specified index * * @param \Zend\Search\Lucene\IndexInterface $index * @return \Zend\Search\Lucene\Search\Query\AbstractQuery * @throws \Zend\Search\Lucene\Exception */ public function rewrite(Lucene\IndexInterface $index) { $this->_matches = array(); $this->_scores = array(); $this->_termKeys = array(); if ($this->_term->field === null) { // Search through all fields $fields = $index->getFieldNames(true); } else { $fields = array($this->_term->field); } $prefix = Index\Term::getPrefix($this->_term->text, $this->_prefixLength); $prefixByteLength = strlen($prefix); $prefixUtf8Length = Index\Term::getLength($prefix); $termLength = Index\Term::getLength($this->_term->text); $termRest = substr($this->_term->text, $prefixByteLength); // we calculate length of the rest in bytes since levenshtein() is not UTF-8 compatible $termRestLength = strlen($termRest); $scaleFactor = 1 / (1 - $this->_minimumSimilarity); $maxTerms = Lucene\Lucene::getTermsPerQueryLimit(); foreach ($fields as $field) { $index->resetTermsStream(); if ($prefix != '') { $index->skipTo(new Index\Term($prefix, $field)); while ($index->currentTerm() !== null && $index->currentTerm()->field == $field && substr($index->currentTerm()->text, 0, $prefixByteLength) == $prefix) { // Calculate similarity $target = substr($index->currentTerm()->text, $prefixByteLength); $maxDistance = isset($this->_maxDistances[strlen($target)]) ? $this->_maxDistances[strlen($target)] : $this->_calculateMaxDistance($prefixUtf8Length, $termRestLength, strlen($target)); if ($termRestLength == 0) { // we don't have anything to compare. That means if we just add // the letters for current term we get the new word $similarity = $prefixUtf8Length == 0 ? 0 : 1 - strlen($target) / $prefixUtf8Length; } else { if (strlen($target) == 0) { $similarity = $prefixUtf8Length == 0 ? 0 : 1 - $termRestLength / $prefixUtf8Length; } else { if ($maxDistance < abs($termRestLength - strlen($target))) { //just adding the characters of term to target or vice-versa results in too many edits //for example "pre" length is 3 and "prefixes" length is 8. We can see that //given this optimal circumstance, the edit distance cannot be less than 5. //which is 8-3 or more precisesly abs(3-8). //if our maximum edit distance is 4, then we can discard this word //without looking at it. $similarity = 0; } else { $similarity = 1 - levenshtein($termRest, $target) / ($prefixUtf8Length + min($termRestLength, strlen($target))); } } } if ($similarity > $this->_minimumSimilarity) { $this->_matches[] = $index->currentTerm(); $this->_termKeys[] = $index->currentTerm()->key(); $this->_scores[] = ($similarity - $this->_minimumSimilarity) * $scaleFactor; if ($maxTerms != 0 && count($this->_matches) > $maxTerms) { throw new Lucene\Exception('Terms per query limit is reached.'); } } $index->nextTerm(); } } else { $index->skipTo(new Index\Term('', $field)); while ($index->currentTerm() !== null && $index->currentTerm()->field == $field) { // Calculate similarity $target = $index->currentTerm()->text; $maxDistance = isset($this->_maxDistances[strlen($target)]) ? $this->_maxDistances[strlen($target)] : $this->_calculateMaxDistance(0, $termRestLength, strlen($target)); if ($maxDistance < abs($termRestLength - strlen($target))) { //just adding the characters of term to target or vice-versa results in too many edits //for example "pre" length is 3 and "prefixes" length is 8. We can see that //given this optimal circumstance, the edit distance cannot be less than 5. //which is 8-3 or more precisesly abs(3-8). //if our maximum edit distance is 4, then we can discard this word //without looking at it. $similarity = 0; } else { $similarity = 1 - levenshtein($termRest, $target) / min($termRestLength, strlen($target)); } if ($similarity > $this->_minimumSimilarity) { $this->_matches[] = $index->currentTerm(); $this->_termKeys[] = $index->currentTerm()->key(); $this->_scores[] = ($similarity - $this->_minimumSimilarity) * $scaleFactor; if ($maxTerms != 0 && count($this->_matches) > $maxTerms) { throw new Lucene\Exception('Terms per query limit is reached.'); } } $index->nextTerm(); } } $index->closeTermsStream(); } if (count($this->_matches) == 0) { return new EmptyResult(); } else { if (count($this->_matches) == 1) { return new Term(reset($this->_matches)); } else { $rewrittenQuery = new Boolean(); array_multisort($this->_scores, SORT_DESC, SORT_NUMERIC, $this->_termKeys, SORT_ASC, SORT_STRING, $this->_matches); $termCount = 0; foreach ($this->_matches as $id => $matchedTerm) { $subquery = new Term($matchedTerm); $subquery->setBoost($this->_scores[$id]); $rewrittenQuery->addSubquery($subquery); $termCount++; if ($termCount >= self::MAX_CLAUSE_COUNT) { break; } } return $rewrittenQuery; } } }