hook_update_N() function to convert <img> tags to Media tokens (Drupal)

convert_img_to_media.php

/**
 * A helper function to convert <img> tags to Media tokens where possible.
 *
 */
function hook_update_N() {
  // A cache to help us when we encounter the same <img> tag multiple times.
  $src_info_cache = array();

  // We'll work directly in the database for performance reasons.
  $tables = array(
    'field_data_body' => 'body_value',
    'field_revision_body' => 'body_value',
    'field_data_comment_body' => 'comment_body_value',
    'field_revision_comment_body' => 'comment_body_value',
  );
  foreach ($tables as $table => $column) {
    if (!db_table_exists($table)) {
      continue;
    }
    $result = db_select($table, 'd')
      ->fields('d')
      ->execute();
    foreach ($result as $row) {
      // Small performance hack to skip field items without any images in them.
      if (strpos($row->$column, '<img') !== FALSE) {
        $dom = new DOMDocument();
        $dom->loadHTML($row->$column);
        $changed = FALSE;

        // Loop through all <img> tags 'src' attributes.
        $xpath = new DOMXPath($dom);
        $list = $xpath->query('//img');
        foreach ($list as $imgTag) {
          $src = $imgTag->getAttribute('src');

          // Attempt to determine if the image is on this site, and find it's
          // underlying FID and image style.
          $src_info = array();
          if (!empty($src_info_cache[$src])) {
            $src_info = $src_info_cache[$src];
          }
          else {
            $src_info = array(
              'fid' => NULL,
              'filename' => NULL,
              'style' => 'panopoly_image_original',
            );
            // NOTE: This only works for 'public' files - we'd need to make a
            // few changes to support private images too.
            if (preg_match('/sites\/[^\/]+\/files\/((?:styles\/)([^\/]+)(?:\/public\/))?(.*)$/', $src, $matches)) {
              if (count($matches) == 4) {
                $src_info['style'] = $matches[2];
                $src_info['filename'] = $matches[3];
              }
              else {
                $src_info['filename'] = $matches[1];
              }

              if (!empty($src_info['filename'])) {
                // Remove any GET arguments from the filename.
                if ($index = strpos($src_info['filename'], '?')) {
                  $src_info['filename'] = substr($src_info['filename'], 0, $index);
                }

                // Finally, do the actually FID lookup based on filename.
                $src_info['fid'] = db_select('file_managed', 'f')
                  ->fields('f', array('fid'))
                  ->condition('f.uri', 'public://' . $src_info['filename'])
                  ->execute()
                  ->fetchField();
              }
            }
            $src_info_cache[$src] = $src_info;
          }

          // If this does refer to an image on this site, then we replace the
          // <img> tag with a Media token.
          if (!empty($src_info['fid'])) {
            $media_info = array(
              'type' => 'media',
              'fid' => $src_info['fid'],
              'fields' => array(),
              'attributes' => array('style' => ''),
            );

            // Translate the style into a File view mode.
            // NOTE: The site I was working with, already used panopoly_images - but your site probably
            //       has it's own set of image styles to convert - feel free to change this!
            $view_mode = '';
            switch ($src_info['style']) {
              case 'panopoly_image_featured':
              case 'panopoly_image_full':
              case 'panopoly_image_original':
                $view_mode = 'default';
                break;

              case 'panopoly_image_half':
              case 'panopoly_image_quarter':
              case 'panopoly_image_square':
              case 'panopoly_image_thumbnail':
                $view_mode = 'teaser';
                break;

              default:
                // If we can't work it out, we just use default.
                $view_mode = 'default';
            }
            $media_info['view_mode'] = $media_info['fields']['format'] = $view_mode;

            // Transfer simple attributes from the 'img' tag.
            foreach (array('width', 'height', 'style') as $attr) {
              if ($imgTag->hasAttribute($attr)) {
                $media_info['attributes'][$attr] = $imgTag->getAttribute($attr);
              }
            }

            // Transfer the special alt/title attributes.
            foreach (array('alt', 'title') as $attr) {
              $value = $imgTag->getAttribute($attr);
              $media_info['attributes'][$attr] = $value;
              $media_info['fields']["field_file_image_{$attr}_text[und][0][value]"] = $value;
            }

            // Process the original 'class' attribute into the media token.
            // NOTE: These classes are specific to the site I originally wrote this code for!
            //       Feel free to remove this section or replace with stuff that makes sense.
            if ($imgTag->hasAttribute('class')) {
              // This builds an array like array('class_name' => 1) for each
              // class on the old <img> tag. Hurray for functional programming!
              $class_index = array_fill_keys(array_filter(array_map('trim', explode(' ', $imgTag->getAttribute('class')))), 1);

              // Map some old classes to new values.
              if (isset($class_index['img-no-border'])) {
                $media_info['attributes']['class'] = 'img-no-border';
              }
              if (isset($class_index['image-center'])) {
                $media_info['attributes']['style'] .= 'display: block; margin-left: auto; margin-right: auto';
              }
              elseif (isset($class_index['image-right'])) {
                $media_info['attributes']['style'] .= 'float: right';
              }
              elseif (isset($class_index['image-left'])) {
                $media_info['attributes']['style'] .= 'float: left';
              }
            }

            $media_token = $dom->createTextNode('[[' . json_encode($media_info) . ']]');
            $imgTag->parentNode->replaceChild($media_token, $imgTag);
            $changed = TRUE;
          }
        }

        // If any changes were made, then we serialize the DOM and update the
        // field value.
        if ($changed) {
          // There is no variant on saveHTML() that will only do the innerHTML,
          // it always includes the outer wrapper, so we have to loop!
          $row->$column = '';
          foreach ($dom->documentElement->firstChild->childNodes as $child) {
            $row->$column .= $dom->saveHTML($child);
          }

          db_update($table)
            ->fields($row)
            ->condition('entity_type', $row->entity_type)
            ->condition('entity_id', $row->entity_id)
            ->condition('revision_id', $row->revision_id)
            ->condition('language', $row->language)
            ->condition('delta', $row->delta)
            ->execute();
        }
      }
    }
  }

  // Since we made the changes directly in the database, we need to clear
  // the filter, field and page caches manually.
  foreach (array('filter', 'field', 'page') as $cache) {
    cache_clear_all('*', "cache_{$cache}", TRUE);
  }
}