Do a query on the full-text search index for a word or words.

This function is normally only called by each module that supports the indexed search (and thus, implements hook_update_index()).

Results are retrieved in two logical passes. However, the two passes are joined together into a single query. And in the case of most simple queries the second pass is not even used.

The first pass selects a set of all possible matches, which has the benefit of also providing the exact result set for simple "AND" or "OR" searches.

The second portion of the query further refines this set by verifying advanced text conditions (such as negative or phrase matches).

The used query object has the tag 'search_$module' and can be further extended with hook_query_alter().

Hierarchy

Expanded class hierarchy of SearchQuery

1 string reference to 'SearchQuery'
SearchMatchTestCase::_testQueries in modules/search/search.test
Run predefine queries looking for indexed terms.

File

modules/search/search.extender.inc, line 27
Search query extender and helper functions.

View source
class SearchQuery extends SelectQueryExtender {

  /**
   * The search query that is used for searching.
   *
   * @var string
   */
  protected $searchExpression;

  /**
   * Type of search (search module).
   *
   * This maps to the value of the type column in search_index, and is equal
   * to the machine-readable name of the module that implements
   * hook_search_info().
   *
   * @var string
   */
  protected $type;

  /**
   * Positive and negative search keys.
   *
   * @var array
   */
  protected $keys = array(
    'positive' => array(),
    'negative' => array(),
  );

  /**
   * Indicates whether the first pass query requires complex conditions (LIKE).
   *
   * @var boolean.
   */
  protected $simple = TRUE;

  /**
   * Conditions that are used for exact searches.
   *
   * This is always used for the second pass query but not for the first pass,
   * unless $this->simple is FALSE.
   *
   * @var DatabaseCondition
   */
  protected $conditions;

  /**
   * Indicates how many matches for a search query are necessary.
   *
   * @var int
   */
  protected $matches = 0;

  /**
   * Array of search words.
   *
   * These words have to match against {search_index}.word.
   *
   * @var array
   */
  protected $words = array();

  /**
   * Multiplier for the normalized search score.
   *
   * This value is calculated by the first pass query and multiplied with the
   * actual score of a specific word to make sure that the resulting calculated
   * score is between 0 and 1.
   *
   * @var float
   */
  protected $normalize;

  /**
   * Indicates whether the first pass query has been executed.
   *
   * @var boolean
   */
  protected $executedFirstPass = FALSE;

  /**
   * Stores score expressions.
   *
   * @var array
   *
   * @see addScore()
   */
  protected $scores = array();

  /**
   * Stores arguments for score expressions.
   *
   * @var array
   */
  protected $scoresArguments = array();

  /**
   * Stores multipliers for score expressions.
   *
   * @var array
   */
  protected $multiply = array();

  /**
   * Whether or not search expressions were ignored.
   *
   * The maximum number of AND/OR combinations exceeded can be configured to
   * avoid Denial-of-Service attacks. Expressions beyond the limit are ignored.
   *
   * @var boolean
   */
  protected $expressionsIgnored = FALSE;

  /**
   * Sets up the search query expression.
   *
   * @param $query
   *   A search query string, which can contain options.
   * @param $module
   *   The search module. This maps to {search_index}.type in the database.
   *
   * @return
   *   The SearchQuery object.
   */
  public function searchExpression($expression, $module) {
    $this->searchExpression = $expression;
    $this->type = $module;

    // Add a search_* tag. This needs to be added before any preExecute methods
    // for decorated queries are called, as $this->prepared will be set to TRUE
    // and tags added in the execute method will never get used. For example,
    // if $query is extended by 'SearchQuery' then 'PagerDefault', the
    // search-specific tag will be added too late (when preExecute() has
    // already been called from the PagerDefault extender), and as a
    // consequence will not be available to hook_query_alter() implementations,
    // nor will the correct hook_query_TAG_alter() implementations get invoked.
    // See node_search_execute().
    $this
      ->addTag('search_' . $module);
    return $this;
  }

  /**
   * Applies a search option and removes it from the search query string.
   *
   * These options are in the form option:value,value2,value3.
   *
   * @param $option
   *   Name of the option.
   * @param $column
   *   Name of the database column to which the value should be applied.
   *
   * @return
   *   TRUE if a value for that option was found, FALSE if not.
   */
  public function setOption($option, $column) {
    if ($values = search_expression_extract($this->searchExpression, $option)) {
      $or = db_or();
      foreach (explode(',', $values) as $value) {
        $or
          ->condition($column, $value);
      }
      $this
        ->condition($or);
      $this->searchExpression = search_expression_insert($this->searchExpression, $option);
      return TRUE;
    }
    return FALSE;
  }

  /**
   * Parses the search query into SQL conditions.
   *
   * We build two queries that match the dataset bodies.
   */
  protected function parseSearchExpression() {

    // Matchs words optionally prefixed by a dash. A word in this case is
    // something between two spaces, optionally quoted.
    preg_match_all('/ (-?)("[^"]+"|[^" ]+)/i', ' ' . $this->searchExpression, $keywords, PREG_SET_ORDER);
    if (count($keywords) == 0) {
      return;
    }

    // Classify tokens.
    $or = FALSE;
    $warning = '';
    $limit_combinations = variable_get('search_and_or_limit', 7);

    // The first search expression does not count as AND.
    $and_count = -1;
    $or_count = 0;
    foreach ($keywords as $match) {
      if ($or_count && $and_count + $or_count >= $limit_combinations) {

        // Ignore all further search expressions to prevent Denial-of-Service
        // attacks using a high number of AND/OR combinations.
        $this->expressionsIgnored = TRUE;
        break;
      }
      $phrase = FALSE;

      // Strip off phrase quotes.
      if ($match[2][0] == '"') {
        $match[2] = substr($match[2], 1, -1);
        $phrase = TRUE;
        $this->simple = FALSE;
      }

      // Simplify keyword according to indexing rules and external
      // preprocessors. Use same process as during search indexing, so it
      // will match search index.
      $words = search_simplify($match[2]);

      // Re-explode in case simplification added more words, except when
      // matching a phrase.
      $words = $phrase ? array(
        $words,
      ) : preg_split('/ /', $words, -1, PREG_SPLIT_NO_EMPTY);

      // Negative matches.
      if ($match[1] == '-') {
        $this->keys['negative'] = array_merge($this->keys['negative'], $words);
      }
      elseif ($match[2] == 'OR' && count($this->keys['positive'])) {
        $last = array_pop($this->keys['positive']);

        // Starting a new OR?
        if (!is_array($last)) {
          $last = array(
            $last,
          );
        }
        $this->keys['positive'][] = $last;
        $or = TRUE;
        $or_count++;
        continue;
      }
      elseif ($match[2] == 'AND' || $match[2] == 'and') {
        $warning = $match[2];
        continue;
      }
      else {
        if ($match[2] == 'or') {
          $warning = $match[2];
        }
        if ($or) {

          // Add to last element (which is an array).
          $this->keys['positive'][count($this->keys['positive']) - 1] = array_merge($this->keys['positive'][count($this->keys['positive']) - 1], $words);
        }
        else {
          $this->keys['positive'] = array_merge($this->keys['positive'], $words);
          $and_count++;
        }
      }
      $or = FALSE;
    }

    // Convert keywords into SQL statements.
    $this->conditions = db_and();
    $simple_and = FALSE;
    $simple_or = FALSE;

    // Positive matches.
    foreach ($this->keys['positive'] as $key) {

      // Group of ORed terms.
      if (is_array($key) && count($key)) {
        $simple_or = TRUE;
        $any = FALSE;
        $queryor = db_or();
        foreach ($key as $or) {
          list($num_new_scores) = $this
            ->parseWord($or);
          $any |= $num_new_scores;
          $queryor
            ->condition('d.data', "% {$or} %", 'LIKE');
        }
        if (count($queryor)) {
          $this->conditions
            ->condition($queryor);

          // A group of OR keywords only needs to match once.
          $this->matches += $any > 0;
        }
      }
      else {
        $simple_and = TRUE;
        list($num_new_scores, $num_valid_words) = $this
          ->parseWord($key);
        $this->conditions
          ->condition('d.data', "% {$key} %", 'LIKE');
        if (!$num_valid_words) {
          $this->simple = FALSE;
        }

        // Each AND keyword needs to match at least once.
        $this->matches += $num_new_scores;
      }
    }
    if ($simple_and && $simple_or) {
      $this->simple = FALSE;
    }

    // Negative matches.
    foreach ($this->keys['negative'] as $key) {
      $this->conditions
        ->condition('d.data', "% {$key} %", 'NOT LIKE');
      $this->simple = FALSE;
    }
    if ($warning == 'or') {
      drupal_set_message(t('Search for either of the two terms with uppercase <strong>OR</strong>. For example, <strong>cats OR dogs</strong>.'));
    }
  }

  /**
   * Helper function for parseQuery().
   */
  protected function parseWord($word) {
    $num_new_scores = 0;
    $num_valid_words = 0;

    // Determine the scorewords of this word/phrase.
    $split = explode(' ', $word);
    foreach ($split as $s) {
      $num = is_numeric($s);
      if ($num || drupal_strlen($s) >= variable_get('minimum_word_size', 3)) {
        if (!isset($this->words[$s])) {
          $this->words[$s] = $s;
          $num_new_scores++;
        }
        $num_valid_words++;
      }
    }

    // Return matching snippet and number of added words.
    return array(
      $num_new_scores,
      $num_valid_words,
    );
  }

  /**
   * Executes the first pass query.
   *
   * This can either be done explicitly, so that additional scores and
   * conditions can be applied to the second pass query, or implicitly by
   * addScore() or execute().
   *
   * @return
   *   TRUE if search items exist, FALSE if not.
   */
  public function executeFirstPass() {
    $this
      ->parseSearchExpression();
    if (count($this->words) == 0) {
      form_set_error('keys', format_plural(variable_get('minimum_word_size', 3), 'You must include at least one positive keyword with 1 character or more.', 'You must include at least one positive keyword with @count characters or more.'));
      return FALSE;
    }
    if ($this->expressionsIgnored) {
      drupal_set_message(t('Your search used too many AND/OR expressions. Only the first @count terms were included in this search.', array(
        '@count' => variable_get('search_and_or_limit', 7),
      )), 'warning');
    }
    $this->executedFirstPass = TRUE;
    if (!empty($this->words)) {
      $or = db_or();
      foreach ($this->words as $word) {
        $or
          ->condition('i.word', $word);
      }
      $this
        ->condition($or);
    }

    // Build query for keyword normalization.
    $this
      ->join('search_total', 't', 'i.word = t.word');
    $this
      ->condition('i.type', $this->type)
      ->groupBy('i.type')
      ->groupBy('i.sid')
      ->having('COUNT(*) >= :matches', array(
      ':matches' => $this->matches,
    ));

    // Clone the query object to do the firstPass query;
    $first = clone $this->query;

    // For complex search queries, add the LIKE conditions to the first pass query.
    if (!$this->simple) {
      $first
        ->join('search_dataset', 'd', 'i.sid = d.sid AND i.type = d.type');
      $first
        ->condition($this->conditions);
    }

    // Calculate maximum keyword relevance, to normalize it.
    $first
      ->addExpression('SUM(i.score * t.count)', 'calculated_score');
    $this->normalize = $first
      ->range(0, 1)
      ->orderBy('calculated_score', 'DESC')
      ->execute()
      ->fetchField();
    if ($this->normalize) {
      return TRUE;
    }
    return FALSE;
  }

  /**
   * Adds a custom score expression to the search query.
   *
   * Score expressions are used to order search results. If no calls to
   * addScore() have taken place, a default keyword relevance score will be
   * used. However, if at least one call to addScore() has taken place, the
   * keyword relevance score is not automatically added.
   *
   * Note that you must use this method to add ordering to your searches, and
   * not call orderBy() directly, when using the SearchQuery extender. This is
   * because of the two-pass system the SearchQuery class uses to normalize
   * scores.
   *
   * @param $score
   *   The score expression, which should evaluate to a number between 0 and 1.
   *   The string 'i.relevance' in a score expression will be replaced by a
   *   measure of keyword relevance between 0 and 1.
   * @param $arguments
   *   Query arguments needed to provide values to the score expression.
   * @param $multiply
   *   If set, the score is multiplied with this value. However, all scores
   *   with multipliers are then divided by the total of all multipliers, so
   *   that overall, the normalization is maintained.
   *
   * @return object
   *   The updated query object.
   */
  public function addScore($score, $arguments = array(), $multiply = FALSE) {
    if ($multiply) {
      $i = count($this->multiply);

      // Modify the score expression so it is multiplied by the multiplier,
      // with a divisor to renormalize.
      $score = "CAST(:multiply_{$i} AS DECIMAL) * COALESCE(( " . $score . "), 0) / CAST(:total_{$i} AS DECIMAL)";

      // Add an argument for the multiplier. The :total_$i argument is taken
      // care of in the execute() method, which is when the total divisor is
      // calculated.
      $arguments[':multiply_' . $i] = $multiply;
      $this->multiply[] = $multiply;
    }
    $this->scores[] = $score;
    $this->scoresArguments += $arguments;
    return $this;
  }

  /**
   * Executes the search.
   *
   * If not already done, this executes the first pass query. Then the complex
   * conditions are applied to the query including score expressions and
   * ordering.
   *
   * @return
   *   FALSE if the first pass query returned no results, and a database result
   *   set if there were results.
   */
  public function execute() {
    if (!$this->executedFirstPass) {
      $this
        ->executeFirstPass();
    }
    if (!$this->normalize) {
      return new DatabaseStatementEmpty();
    }

    // Add conditions to query.
    $this
      ->join('search_dataset', 'd', 'i.sid = d.sid AND i.type = d.type');
    $this
      ->condition($this->conditions);
    if (empty($this->scores)) {

      // Add default score.
      $this
        ->addScore('i.relevance');
    }
    if (count($this->multiply)) {

      // Re-normalize scores with multipliers by dividing by the total of all
      // multipliers. The expressions were altered in addScore(), so here just
      // add the arguments for the total.
      $i = 0;
      $sum = array_sum($this->multiply);
      foreach ($this->multiply as $total) {
        $this->scoresArguments[':total_' . $i] = $sum;
        $i++;
      }
    }

    // Replace the pseudo-expression 'i.relevance' with a measure of keyword
    // relevance in all score expressions, using string replacement. Careful
    // though! If you just print out a float, some locales use ',' as the
    // decimal separator in PHP, while SQL always uses '.'. So, make sure to
    // set the number format correctly.
    $relevance = number_format(1.0 / $this->normalize, 10, '.', '');
    $this->scores = str_replace('i.relevance', '(' . $relevance . ' * i.score * t.count)', $this->scores);

    // Add all scores together to form a query field.
    $this
      ->addExpression('SUM(' . implode(' + ', $this->scores) . ')', 'calculated_score', $this->scoresArguments);

    // If an order has not yet been set for this query, add a default order
    // that sorts by the calculated sum of scores.
    if (count($this
      ->getOrderBy()) == 0) {
      $this
        ->orderBy('calculated_score', 'DESC');
    }

    // Add useful metadata.
    $this
      ->addMetaData('normalize', $this->normalize)
      ->fields('i', array(
      'type',
      'sid',
    ));
    return $this->query
      ->execute();
  }

  /**
   * Builds the default count query for SearchQuery.
   *
   * Since SearchQuery always uses GROUP BY, we can default to a subquery. We
   * also add the same conditions as execute() because countQuery() is called
   * first.
   */
  public function countQuery() {

    // Clone the inner query.
    $inner = clone $this->query;

    // Add conditions to query.
    $inner
      ->join('search_dataset', 'd', 'i.sid = d.sid AND i.type = d.type');
    $inner
      ->condition($this->conditions);

    // Remove existing fields and expressions, they are not needed for a count
    // query.
    $fields =& $inner
      ->getFields();
    $fields = array();
    $expressions =& $inner
      ->getExpressions();
    $expressions = array();

    // Add the sid as the only field and count them as a subquery.
    $count = db_select($inner
      ->fields('i', array(
      'sid',
    )), NULL, array(
      'target' => 'slave',
    ));

    // Add the COUNT() expression.
    $count
      ->addExpression('COUNT(*)');
    return $count;
  }

}

Members

Name Modifiers Type Description Overridessort ascending
SelectQueryExtender::__construct public function 2
SearchQuery::searchExpression public function Sets up the search query expression.
SearchQuery::setOption public function Applies a search option and removes it from the search query string.
SearchQuery::parseSearchExpression protected function Parses the search query into SQL conditions.
SearchQuery::parseWord protected function Helper function for parseQuery().
SearchQuery::executeFirstPass public function Executes the first pass query.
SearchQuery::addScore public function Adds a custom score expression to the search query.
SearchQuery::execute public function Executes the search. Overrides SelectQueryExtender::execute
SearchQuery::countQuery public function Builds the default count query for SearchQuery. Overrides SelectQueryExtender::countQuery
SelectQueryExtender::nextPlaceholder public function Implements QueryPlaceholderInterface::nextPlaceholder(). Overrides QueryPlaceholderInterface::nextPlaceholder
SelectQueryExtender::addTag public function Adds a tag to a query. Overrides QueryAlterableInterface::addTag
SelectQueryExtender::hasTag public function Determines if a given query has a given tag. Overrides QueryAlterableInterface::hasTag
SelectQueryExtender::hasAllTags public function Determines if a given query has all specified tags. Overrides QueryAlterableInterface::hasAllTags
SelectQueryExtender::hasAnyTag public function Determines if a given query has any specified tag. Overrides QueryAlterableInterface::hasAnyTag
SelectQueryExtender::addMetaData public function Adds additional metadata to the query. Overrides QueryAlterableInterface::addMetaData
SelectQueryExtender::getMetaData public function Retrieves a given piece of metadata. Overrides QueryAlterableInterface::getMetaData
SelectQueryExtender::condition public function Helper function: builds the most common conditional clauses. Overrides QueryConditionInterface::condition
SelectQueryExtender::conditions public function Gets a complete list of all conditions in this conditional clause. Overrides QueryConditionInterface::conditions
SelectQueryExtender::arguments public function Gets a complete list of all values to insert into the prepared statement. Overrides QueryConditionInterface::arguments
SelectQueryExtender::where public function Adds an arbitrary WHERE clause to the query. Overrides QueryConditionInterface::where
SelectQueryExtender::compile public function Compiles the saved conditions for later retrieval. Overrides QueryConditionInterface::compile
SelectQueryExtender::havingCondition public function Helper function to build most common HAVING conditional clauses. Overrides SelectQueryInterface::havingCondition
SelectQueryExtender::havingConditions public function
SelectQueryExtender::havingArguments public function
SelectQueryExtender::having public function
SelectQueryExtender::havingCompile public function
SelectQueryExtender::extend public function Enhance this object by wrapping it in an extender object. Overrides QueryExtendableInterface::extend
SelectQueryExtender::getFields public function Returns a reference to the fields array for this query. Overrides SelectQueryInterface::getFields
SelectQueryExtender::getExpressions public function Returns a reference to the expressions array for this query. Overrides SelectQueryInterface::getExpressions
SelectQueryExtender::getOrderBy public function Returns a reference to the order by array for this query. Overrides SelectQueryInterface::getOrderBy
SelectQueryExtender::getGroupBy public function Returns a reference to the group-by array for this query. Overrides SelectQueryInterface::getGroupBy
SelectQueryExtender::getTables public function Returns a reference to the tables array for this query. Overrides SelectQueryInterface::getTables
SelectQueryExtender::getUnion public function Returns a reference to the union queries for this query. This include queries for UNION, UNION ALL, and UNION DISTINCT. Overrides SelectQueryInterface::getUnion
SelectQueryExtender::getArguments public function Compiles and returns an associative array of the arguments for this prepared statement. Overrides SelectQueryInterface::getArguments
SelectQueryExtender::isPrepared public function Indicates if preExecute() has already been called on that object. Overrides SelectQueryInterface::isPrepared
SelectQueryExtender::preExecute public function Generic preparation and validation for a SELECT query. Overrides SelectQueryInterface::preExecute
SelectQueryExtender::distinct public function Sets this query to be DISTINCT. Overrides SelectQueryInterface::distinct
SelectQueryExtender::addField public function Adds a field to the list to be SELECTed. Overrides SelectQueryInterface::addField
SelectQueryExtender::fields public function Add multiple fields from the same table to be SELECTed. Overrides SelectQueryInterface::fields
SelectQueryExtender::addExpression public function Adds an expression to the list of "fields" to be SELECTed. Overrides SelectQueryInterface::addExpression
SelectQueryExtender::join public function Default Join against another table in the database. Overrides SelectQueryInterface::join
SelectQueryExtender::innerJoin public function Inner Join against another table in the database. Overrides SelectQueryInterface::innerJoin
SelectQueryExtender::leftJoin public function Left Outer Join against another table in the database. Overrides SelectQueryInterface::leftJoin
SelectQueryExtender::rightJoin public function Right Outer Join against another table in the database. Overrides SelectQueryInterface::rightJoin
SelectQueryExtender::addJoin public function Join against another table in the database. Overrides SelectQueryInterface::addJoin
SelectQueryExtender::orderBy public function Orders the result set by a given field. Overrides SelectQueryInterface::orderBy
SelectQueryExtender::orderRandom public function Orders the result set by a random value. Overrides SelectQueryInterface::orderRandom
SelectQueryExtender::range public function Restricts a query to a given range in the result set. Overrides SelectQueryInterface::range
SelectQueryExtender::union public function Add another Select query to UNION to this one. Overrides SelectQueryInterface::union
SelectQueryExtender::groupBy public function Groups the result set by the specified field. Overrides SelectQueryInterface::groupBy
SelectQueryExtender::forUpdate public function Add FOR UPDATE to the query. Overrides SelectQueryInterface::forUpdate
SelectQueryExtender::isNull function Sets a condition that the specified field be NULL. Overrides QueryConditionInterface::isNull
SelectQueryExtender::isNotNull function Sets a condition that the specified field be NOT NULL. Overrides QueryConditionInterface::isNotNull
SelectQueryExtender::__toString public function
SelectQueryExtender::__clone public function Clone magic method. Overrides SelectQueryInterface::__clone
SelectQueryExtender::__call public function Magic override for undefined methods.
SelectQueryExtender::exists public function Sets a condition that the specified subquery returns values. Overrides QueryConditionInterface::exists
SelectQueryExtender::notExists public function Sets a condition that the specified subquery returns no values. Overrides QueryConditionInterface::notExists
SelectQueryExtender::uniqueIdentifier public function Implements QueryPlaceholderInterface::uniqueIdentifier(). Overrides QueryPlaceholderInterface::uniqueIdentifier
SelectQueryExtender::compiled public function Check whether a condition has been previously compiled. Overrides QueryConditionInterface::compiled
SearchQuery::$searchExpression protected property The search query that is used for searching.
SearchQuery::$type protected property Type of search (search module).
SearchQuery::$keys protected property Positive and negative search keys.
SearchQuery::$simple protected property Indicates whether the first pass query requires complex conditions (LIKE).
SearchQuery::$conditions protected property Conditions that are used for exact searches.
SearchQuery::$matches protected property Indicates how many matches for a search query are necessary.
SearchQuery::$words protected property Array of search words.
SearchQuery::$normalize protected property Multiplier for the normalized search score.
SearchQuery::$executedFirstPass protected property Indicates whether the first pass query has been executed.
SearchQuery::$scores protected property Stores score expressions.
SearchQuery::$scoresArguments protected property Stores arguments for score expressions.
SearchQuery::$multiply protected property Stores multipliers for score expressions.
SearchQuery::$expressionsIgnored protected property Whether or not search expressions were ignored.
SelectQueryExtender::$query protected property The SelectQuery object we are extending/decorating.
SelectQueryExtender::$connection protected property The connection object on which to run this query.
SelectQueryExtender::$placeholder protected property The placeholder counter.
SelectQueryExtender::$uniqueIdentifier protected property A unique identifier for this query object.