You are here

function _search_find_match_with_simplify in Drupal 9

Same name and namespace in other branches
  1. 8 core/modules/search/search.module \_search_find_match_with_simplify()

Finds an appropriate keyword in text.

Parameters

string $key: The keyword to find.

string $text: The text to search for the keyword.

string $boundary: Regular expression for the boundary character class (characters that indicate spaces between words).

string|null $langcode: Language code for the language of $text, if known.

Return value

string|null A segment of $text that is between word boundary characters that either matches $key directly, or matches $key when both this text segment and $key are processed by \Drupal\search\SearchTextProcessorInterface::analyze(). If a matching text segment is not located, NULL is returned.

1 call to _search_find_match_with_simplify()
search_excerpt in core/modules/search/search.module
Returns snippets from a piece of text, with search keywords highlighted.

File

core/modules/search/search.module, line 472
Enables site-wide keyword searching.

Code

function _search_find_match_with_simplify($key, $text, $boundary, $langcode = NULL) {
  $preceded_by_boundary = '(?<=' . $boundary . ')';
  $followed_by_boundary = '(?=' . $boundary . ')';

  // See if $key appears as-is. When testing, make sure $text starts/ends with
  // a space, because we require $key to be surrounded by word boundary
  // characters.
  $temp = trim($key);
  if ($temp == '') {
    return NULL;
  }
  if (preg_match('/' . $preceded_by_boundary . preg_quote($temp, '/') . $followed_by_boundary . '/iu', ' ' . $text . ' ')) {
    return $temp;
  }

  // See if there is a match after lower-casing and removing diacritics in
  // both, which should preserve the string length.
  $new_text = mb_strtolower($text);
  $new_text = \Drupal::service('transliteration')
    ->removeDiacritics($new_text);
  $new_key = mb_strtolower($temp);
  $new_key = \Drupal::service('transliteration')
    ->removeDiacritics($new_key);
  if (preg_match('/' . $preceded_by_boundary . preg_quote($new_key, '/') . $followed_by_boundary . '/u', ' ' . $new_text . ' ')) {
    $position = mb_strpos($new_text, $new_key);
    return mb_substr($text, $position, mb_strlen($new_key));
  }

  // Run both text and key through text processor.

  /** @var \Drupal\search\SearchTextProcessorInterface $text_processor */
  $text_processor = \Drupal::service('search.text_processor');
  $simplified_key = trim($text_processor
    ->analyze($key, $langcode));
  $simplified_text = trim($text_processor
    ->analyze($text, $langcode));
  if ($simplified_key == '' || $simplified_text == '' || strpos($simplified_text, $simplified_key) === FALSE) {

    // The simplified keyword and text do not match at all, or are empty.
    return NULL;
  }

  // Split $text into words, keeping track of where the word boundaries are.
  $words = preg_split('/' . $boundary . '+/u', $text, NULL, PREG_SPLIT_OFFSET_CAPTURE);

  // Add an entry pointing to the end of the string, for the loop below.
  $words[] = [
    '',
    strlen($text),
  ];

  // Using a binary search, find the earliest possible ending position in
  // $text where it will still match the keyword after applying
  // \Drupal\search\SearchTextProcessorInterface::analyze().
  $start_index = 0;
  $start_pos = $words[$start_index][1];
  $min_end_index = 1;
  $max_end_index = count($words) - 1;
  while ($max_end_index > $min_end_index) {

    // Check the index half way between min and max. See if we ended there,
    // if we would still have a match.
    $proposed_end_index = floor(($max_end_index + $min_end_index) / 2);
    $proposed_end_pos = $words[$proposed_end_index][1];

    // Since the split was done with preg_split(), the positions are byte counts
    // not character counts, so use substr() not mb_substr() here.
    $trial_text = trim($text_processor
      ->analyze(substr($text, $start_pos, $proposed_end_pos - $start_pos), $langcode));
    if (strpos($trial_text, $simplified_key) !== FALSE) {

      // The proposed endpoint is fine, text still matches.
      $max_end_index = $proposed_end_index;
    }
    else {

      // The proposed endpoint index is too early, so the earliest possible
      // OK ending point would be the next index.
      $min_end_index = $proposed_end_index + 1;
    }
  }

  // Now do the same for the starting position: using a binary search, find the
  // latest possible starting position in $text where it will still match the
  // keyword after applying
  // \Drupal\search\SearchTextProcessorInterface::analyze().
  $end_index = $min_end_index;
  $end_pos = $words[$end_index][1];
  $min_start_index = 0;
  $max_start_index = $end_index - 1;
  while ($max_start_index > $min_start_index) {

    // Check the index half way between min and max. See if we started there,
    // if we would still have a match.
    $proposed_start_index = ceil(($max_start_index + $min_start_index) / 2);
    $proposed_start_pos = $words[$proposed_start_index][1];

    // Since the split was done with preg_split(), the positions are byte counts
    // not character counts, so use substr() not mb_substr() here.
    $trial_text = trim($text_processor
      ->analyze(substr($text, $proposed_start_pos, $end_pos - $proposed_start_pos), $langcode));
    if (strpos($trial_text, $simplified_key) !== FALSE) {

      // The proposed start point is fine, text still matches.
      $min_start_index = $proposed_start_index;
    }
    else {

      // The proposed start point index is too late, so the latest possible
      // OK starting point would be the previous index.
      $max_start_index = $proposed_start_index - 1;
    }
  }
  $start_index = $max_start_index;

  // Return the matching text. We need to use substr() here and not the
  // mb_substr() function, because the indices in $words came from preg_split(),
  // so they are Unicode-safe byte positions, not character positions.
  return trim(substr($text, $words[$start_index][1], $words[$end_index][1] - $words[$start_index][1]));
}