X-Git-Url: http://source.bookstackapp.com/bookstack/blobdiff_plain/9e0164f4f45cb68f9dccf96db28c8b05ed493be7..refs/pull/3391/head:/app/Entities/Tools/SearchRunner.php diff --git a/app/Entities/Tools/SearchRunner.php b/app/Entities/Tools/SearchRunner.php index 6296c70e7..a0a44f3a5 100644 --- a/app/Entities/Tools/SearchRunner.php +++ b/app/Entities/Tools/SearchRunner.php @@ -5,14 +5,19 @@ namespace BookStack\Entities\Tools; use BookStack\Auth\Permissions\PermissionService; use BookStack\Auth\User; use BookStack\Entities\EntityProvider; +use BookStack\Entities\Models\BookChild; use BookStack\Entities\Models\Entity; +use BookStack\Entities\Models\Page; +use BookStack\Entities\Models\SearchTerm; +use Illuminate\Database\Connection; use Illuminate\Database\Eloquent\Builder as EloquentBuilder; use Illuminate\Database\Eloquent\Collection as EloquentCollection; +use Illuminate\Database\Eloquent\Relations\BelongsTo; use Illuminate\Database\Query\Builder; -use Illuminate\Database\Query\JoinClause; use Illuminate\Support\Collection; use Illuminate\Support\Facades\DB; use Illuminate\Support\Str; +use SplObjectStorage; class SearchRunner { @@ -33,16 +38,27 @@ class SearchRunner */ protected $queryOperators = ['<=', '>=', '=', '<', '>', 'like', '!=']; + /** + * Retain a cache of score adjusted terms for specific search options. + * From PHP>=8 this can be made into a WeakMap instead. + * + * @var SplObjectStorage + */ + protected $termAdjustmentCache; + public function __construct(EntityProvider $entityProvider, PermissionService $permissionService) { $this->entityProvider = $entityProvider; $this->permissionService = $permissionService; + $this->termAdjustmentCache = new SplObjectStorage(); } /** * Search all entities in the system. * The provided count is for each entity to search, * Total returned could be larger and not guaranteed. + * + * @return array{total: int, count: int, has_more: bool, results: Entity[]} */ public function searchEntities(SearchOptions $searchOpts, string $entityType = 'all', int $page = 1, int $count = 20, string $action = 'view'): array { @@ -64,9 +80,10 @@ class SearchRunner continue; } - $searchQuery = $this->buildQuery($searchOpts, $entityType, $action); + $entityModelInstance = $this->entityProvider->get($entityType); + $searchQuery = $this->buildQuery($searchOpts, $entityModelInstance, $action); $entityTotal = $searchQuery->count(); - $searchResults = $this->getPageOfDataFromQuery($searchQuery, $page, $count); + $searchResults = $this->getPageOfDataFromQuery($searchQuery, $entityModelInstance, $page, $count); if ($entityTotal > ($page * $count)) { $hasMore = true; @@ -98,7 +115,9 @@ class SearchRunner if (!in_array($entityType, $entityTypes)) { continue; } - $search = $this->buildQuery($opts, $entityType)->where('book_id', '=', $bookId)->take(20)->get(); + + $entityModelInstance = $this->entityProvider->get($entityType); + $search = $this->buildQuery($opts, $entityModelInstance)->where('book_id', '=', $bookId)->take(20)->get(); $results = $results->merge($search); } @@ -111,7 +130,8 @@ class SearchRunner public function searchChapter(int $chapterId, string $searchString): Collection { $opts = SearchOptions::fromString($searchString); - $pages = $this->buildQuery($opts, 'page')->where('chapter_id', '=', $chapterId)->take(20)->get(); + $entityModelInstance = $this->entityProvider->get('page'); + $pages = $this->buildQuery($opts, $entityModelInstance)->where('chapter_id', '=', $chapterId)->take(20)->get(); return $pages->sortByDesc('score'); } @@ -119,9 +139,24 @@ class SearchRunner /** * Get a page of result data from the given query based on the provided page parameters. */ - protected function getPageOfDataFromQuery(EloquentBuilder $query, int $page = 1, int $count = 20): EloquentCollection + protected function getPageOfDataFromQuery(EloquentBuilder $query, Entity $entityModelInstance, int $page = 1, int $count = 20): EloquentCollection { + $relations = ['tags']; + + if ($entityModelInstance instanceof BookChild) { + $relations['book'] = function (BelongsTo $query) { + $query->scopes('visible'); + }; + } + + if ($entityModelInstance instanceof Page) { + $relations['chapter'] = function (BelongsTo $query) { + $query->scopes('visible'); + }; + } + return $query->clone() + ->with(array_filter($relations)) ->skip(($page - 1) * $count) ->take($count) ->get(); @@ -130,19 +165,24 @@ class SearchRunner /** * Create a search query for an entity. */ - protected function buildQuery(SearchOptions $searchOpts, string $entityType = 'page', string $action = 'view'): EloquentBuilder + protected function buildQuery(SearchOptions $searchOpts, Entity $entityModelInstance, string $action = 'view'): EloquentBuilder { - $entity = $this->entityProvider->get($entityType); - $entityQuery = $entity->newQuery(); + $entityQuery = $entityModelInstance->newQuery(); + + if ($entityModelInstance instanceof Page) { + $entityQuery->select($entityModelInstance::$listAttributes); + } else { + $entityQuery->select(['*']); + } // Handle normal search terms - $this->applyTermSearch($entityQuery, $searchOpts->searches, $entity); + $this->applyTermSearch($entityQuery, $searchOpts, $entityModelInstance); // Handle exact term matching foreach ($searchOpts->exacts as $inputTerm) { - $entityQuery->where(function (EloquentBuilder $query) use ($inputTerm, $entity) { + $entityQuery->where(function (EloquentBuilder $query) use ($inputTerm, $entityModelInstance) { $query->where('name', 'like', '%' . $inputTerm . '%') - ->orWhere($entity->textField, 'like', '%' . $inputTerm . '%'); + ->orWhere($entityModelInstance->textField, 'like', '%' . $inputTerm . '%'); }); } @@ -155,40 +195,134 @@ class SearchRunner foreach ($searchOpts->filters as $filterTerm => $filterValue) { $functionName = Str::camel('filter_' . $filterTerm); if (method_exists($this, $functionName)) { - $this->$functionName($entityQuery, $entity, $filterValue); + $this->$functionName($entityQuery, $entityModelInstance, $filterValue); } } - return $this->permissionService->enforceEntityRestrictions($entity, $entityQuery, $action); + return $this->permissionService->enforceEntityRestrictions($entityModelInstance, $entityQuery, $action); } /** * For the given search query, apply the queries for handling the regular search terms. */ - protected function applyTermSearch(EloquentBuilder $entityQuery, array $terms, Entity $entity): void + protected function applyTermSearch(EloquentBuilder $entityQuery, SearchOptions $options, Entity $entity): void { + $terms = $options->searches; if (count($terms) === 0) { return; } + $scoredTerms = $this->getTermAdjustments($options); + $scoreSelect = $this->selectForScoredTerms($scoredTerms); + $subQuery = DB::table('search_terms')->select([ 'entity_id', 'entity_type', - DB::raw('SUM(score) as score'), + DB::raw($scoreSelect['statement']), ]); + $subQuery->addBinding($scoreSelect['bindings'], 'select'); + $subQuery->where('entity_type', '=', $entity->getMorphClass()); $subQuery->where(function (Builder $query) use ($terms) { foreach ($terms as $inputTerm) { $query->orWhere('term', 'like', $inputTerm . '%'); } - })->groupBy('entity_type', 'entity_id'); - $entityQuery->join(DB::raw('(' . $subQuery->toSql() . ') as s'), function (JoinClause $join) { - $join->on('id', '=', 'entity_id'); - })->addSelect($entity->getTable() . '.*') - ->selectRaw('s.score') - ->orderBy('score', 'desc'); - $entityQuery->mergeBindings($subQuery); + }); + $subQuery->groupBy('entity_type', 'entity_id'); + + $entityQuery->joinSub($subQuery, 's', 'id', '=', 'entity_id'); + $entityQuery->addSelect('s.score'); + $entityQuery->orderBy('score', 'desc'); + } + + /** + * Create a select statement, with prepared bindings, for the given + * set of scored search terms. + * + * @param array $scoredTerms + * + * @return array{statement: string, bindings: string[]} + */ + protected function selectForScoredTerms(array $scoredTerms): array + { + // Within this we walk backwards to create the chain of 'if' statements + // so that each previous statement is used in the 'else' condition of + // the next (earlier) to be built. We start at '0' to have no score + // on no match (Should never actually get to this case). + $ifChain = '0'; + $bindings = []; + foreach ($scoredTerms as $term => $score) { + $ifChain = 'IF(term like ?, score * ' . (float) $score . ', ' . $ifChain . ')'; + $bindings[] = $term . '%'; + } + + return [ + 'statement' => 'SUM(' . $ifChain . ') as score', + 'bindings' => array_reverse($bindings), + ]; + } + + /** + * For the terms in the given search options, query their popularity across all + * search terms then provide that back as score adjustment multiplier applicable + * for their rarity. Returns an array of float multipliers, keyed by term. + * + * @return array + */ + protected function getTermAdjustments(SearchOptions $options): array + { + if (isset($this->termAdjustmentCache[$options])) { + return $this->termAdjustmentCache[$options]; + } + + $termQuery = SearchTerm::query()->toBase(); + $whenStatements = []; + $whenBindings = []; + + foreach ($options->searches as $term) { + $whenStatements[] = 'WHEN term LIKE ? THEN ?'; + $whenBindings[] = $term . '%'; + $whenBindings[] = $term; + + $termQuery->orWhere('term', 'like', $term . '%'); + } + + $case = 'CASE ' . implode(' ', $whenStatements) . ' END'; + $termQuery->selectRaw($case . ' as term', $whenBindings); + $termQuery->selectRaw('COUNT(*) as count'); + $termQuery->groupByRaw($case, $whenBindings); + + $termCounts = $termQuery->pluck('count', 'term')->toArray(); + $adjusted = $this->rawTermCountsToAdjustments($termCounts); + + $this->termAdjustmentCache[$options] = $adjusted; + + return $this->termAdjustmentCache[$options]; + } + + /** + * Convert counts of terms into a relative-count normalised multiplier. + * + * @param array $termCounts + * + * @return array + */ + protected function rawTermCountsToAdjustments(array $termCounts): array + { + if (empty($termCounts)) { + return []; + } + + $multipliers = []; + $max = max(array_values($termCounts)); + + foreach ($termCounts as $term => $count) { + $percent = round($count / $max, 5); + $multipliers[$term] = 1.3 - $percent; + } + + return $multipliers; } /** @@ -223,7 +357,9 @@ class SearchRunner // We have to do a raw sql query for this since otherwise PDO will quote the value and MySQL will // search the value as a string which prevents being able to do number-based operations // on the tag values. We ensure it has a numeric value and then cast it just to be sure. - $tagValue = (float) trim($query->getConnection()->getPdo()->quote($tagValue), "'"); + /** @var Connection $connection */ + $connection = $query->getConnection(); + $tagValue = (float) trim($connection->getPdo()->quote($tagValue), "'"); $query->whereRaw("value ${tagOperator} ${tagValue}"); } else { $query->where('value', $tagOperator, $tagValue); @@ -244,7 +380,8 @@ class SearchRunner try { $date = date_create($input); $query->where('updated_at', '>=', $date); - } catch (\Exception $e) {} + } catch (\Exception $e) { + } } protected function filterUpdatedBefore(EloquentBuilder $query, Entity $model, $input): void @@ -252,7 +389,8 @@ class SearchRunner try { $date = date_create($input); $query->where('updated_at', '<', $date); - } catch (\Exception $e) {} + } catch (\Exception $e) { + } } protected function filterCreatedAfter(EloquentBuilder $query, Entity $model, $input): void @@ -260,7 +398,8 @@ class SearchRunner try { $date = date_create($input); $query->where('created_at', '>=', $date); - } catch (\Exception $e) {} + } catch (\Exception $e) { + } } protected function filterCreatedBefore(EloquentBuilder $query, Entity $model, $input) @@ -268,7 +407,8 @@ class SearchRunner try { $date = date_create($input); $query->where('created_at', '<', $date); - } catch (\Exception $e) {} + } catch (\Exception $e) { + } } protected function filterCreatedBy(EloquentBuilder $query, Entity $model, $input)