WP Engine Search

Integrate Elasticsearch with WordPress for users of the WP Engine platform.

Background

Let's face it, WordPress search is rudimentary at best. Lackluster performance, inflexible and rigid matching algorithms (which means no comprehension of 'close' queries), the inability to search metadata and taxonomy information, no way to determine categories of your results and most importantly the overall relevancy of results is poor.

Elasticsearch is a search server based on Lucene. It provides a distributed, multitenant-capable full-text search engine with a RESTful web interface and schema-free JSON documents.

Coupling WordPress with Elasticsearch allows us to do amazing things with search including:

• Relevant results
• Autosuggest
• Fuzzy matching (catch misspellings as well as 'close' queries)
• Proximity and geographic queries
• Search metadata
• Search taxonomies
• Facets
• Search all sites on a multisite install
• The list goes on...

Purpose

Elasticsearch is a complex topic and integration results in complex problems. The goal of WP Engine Search is to fully marry WordPress with Elasticsearch. This plugin integrates with the WP_Query object returning results from Elasticsearch instead of MySQL.

There are other Elasticsearch integration plugins available for WordPress. WP Engine Search, unlike most others, offers multi-site search out of the box. Also, rather than providing a limited, clunky UI, we elected to instead provide full control via WP-CLI.

Installation

1. Download the current development release and install it using the WordPress plugin uploader GUI.

Configuration

Important! Members of the WP Engine Search alpha do not need to follow these steps! The plugin will be configured for you by our Labs team. However, if you have an Elasticsearch instance you would like to use, feel free to follow the steps below.

1. Activate the plugin with wp-cli. Remember to use the --network flag for Multisite installs!

wp plugin activate wpe-search [--network]

2. Using wp-cli, configure the host of your Elasticsearch server.

wp ep4wpe set-host 192.168.50.4

3. (Optional) If your instance of Elasticsearch is running on a different port, use wp-cli to set the port number:

wp ep4wpe set-port 9200

Note: Port 9200 is used by default if not manually configured.

Index Initialization

1. Using wp-cli, do an initial sync (with mapping) with your Elasticsearch server.

wp ep4wpe index --setup [--network-wide]

Note: Use the --network-wide flag only if the install has Multisite configured.

Index names are automatically generated based on site URL. Once your index (or, in the case of Multisite, indices) has been initialized, WP_Query will be integrated with Elasticsearch and support a few special parameters.

Usage

After running an index, WP Engine Search integrates with WP_Query if and only if the query is a search or the ep_integrate parameter is passed (see below).

The end goal is to support all the parameters available to WP_Query so the transition is completely transparent. Right now, our WP_Query integration supports many of the relevant WP_Query parameters. It also adds a couple special ones.

Supported WP_Query Parameters

• s (string)

  Search keyword. By default used to search against post_title, post_content, and post_excerpt.

• posts_per_page (int)

  Number of posts to show per page. Use -1 to show all posts (the offset parameter is ignored with a -1 value). Set the paged parameter to paginate based on posts_per_page.

• tax_query (array)

  Filter posts by terms in taxonomies. Takes an array of form:

  new WP_Query( array(
      's'         => 'search phrase',
      'tax_query' => array(
          array(
              'taxonomy' => 'taxonomy-name',
              'terms'    => array( ... ),
          ),
      ),
  ) );

  tax_query accepts an array of arrays where each inner array only supports taxonomy (string) and terms (string|array) parameters. terms is a slug, either in string or array form.

• meta_query (array)

  Filter posts by post meta conditions. Takes an array of form:

  new WP_Query( array(
      's'          => 'search phrase',
      'meta_query' => array(
          array(
              'key'   => 'key_name',
              'value' => 'meta value',
              'compare' => '=',
          ),
      ),
  ) );

  meta_query accepts an array of arrays where each inner array only supports key (string), value (string|array|int), and compare (string) parameters. compare supports the following:

  * `=` - Posts will be returned that have a post meta key corresponding to `key` and a value that equals the value passed to `value`.
  * `!=` - Posts will be returned that have a post meta key corresponding to `key` and a value that does NOT equal the value passed to `value`.
  * `EXISTS` - Posts will be returned that have a post meta key corresponding to `key`.
  * `NOT EXISTS` - Posts will be returned that do not have a post meta key corresponding to `key`.
  

  The outer array also supports a relation (string) parameter. By default relation is set to AND:

  new WP_Query( array(
      's'          => 'search phrase',
      'meta_query' => array(
          array(
              'key'   => 'key_name',
              'value' => 'meta value',
              'compare' => '=',
          ),
          array(
              'key'   => 'key_name2',
              'value' => 'meta value',
              'compare' => '!=',
          ),
          'relation' => 'AND',
      ),
  ) );

  Possible values for relation are OR and AND. If relation is set to AND, all inner queries must be true for a post to be returned. If relation is set to OR, only one of the inner meta queries must be true for the post to be returned.

• post_type (string/array)

  Filter posts by post type. any wil search all public post types. WP_Query defaults to either post or any if no post_type is provided depending on the context of the query. This is confusing. WP Engine Search will ALWAYS default to any if no post_type is provided. If you want to search for post posts, you MUST specify post as the post_type.

• offset (int)

  Number of posts to skip in ascending order.

• paged (int)

  Page number of posts to be used with posts_per_page.

• author (int)

  Show posts associated with certain author ID.

• author_name (string)

  Show posts associated with certain author. Use user_nicename (NOT name).

• orderby (string)

  Order results by field name instead of relevance. Currently only supports: title, name, date, and relevance (default).

• order (string)

  Which direction to order results in. Accepts ASC and DESC. Default is DESC.

The following are special parameters that are only supported by WP Engine Search.

• search_fields (array)

  If not specified, defaults to array( 'post_title', 'post_excerpt', 'post_content' ).

  • post_title (string)

    Applies current search to post titles.

  • post_content (string)

    Applies current search to post content.

  • post_excerpt (string)

    Applies current search to post excerpts.

  • taxonomies (string => array/string)

    Applies the current search to terms within a taxonomy or taxonomies. The following will fuzzy search across post_title, post_excerpt, post_content, and terms within taxonomies category and post_tag:

    new WP_Query( array(
        's'             => 'term search phrase',
        'search_fields' => array(
            'post_title',
            'post_content',
            'post_excerpt',
            'taxonomies' => array( 'category', 'post_tag' ),
        ),
    ) );

  • meta (string => array/string)

    Applies the current search to post meta. The following will fuzzy search across post_title, post_excerpt, post_content, and post meta keys meta_key_1 and meta_key_2:

    new WP_Query( array(
        's'             => 'meta search phrase',
        'search_fields' => array(
            'post_title',
            'post_content',
            'post_excerpt',
            'meta' => array( 'meta_key_1', 'meta_key_2' ),
        ),
    ) );

  • author_name (string)

    Applies the current search to author login names. The following will fuzzy search across post_title, post_excerpt, post_content and author user_login:

    new WP_Query( array(
        's'             => 'username',
        'search_fields' => array(
            'post_title',
            'post_content',
            'post_excerpt',
            'author_name',
        ),
    ) );

• aggs (array)

  Add aggregation results to your search result. For example:

  new WP_Query( array(
      's'    => 'search phrase',
      'aggs' => array(
          'name'       => 'name-of-aggregation', // (can be whatever you'd like)
          'use-filter' => true // (*bool*) used if you'd like to apply the other filters (i.e. post type, tax_query, author), to the aggregation
          'aggs'       => array(
              'name'  => 'name-of-sub-aggregation',
              'terms' => array(
                  'field' => 'terms.name-of-taxonomy.name-of-term',
              ),
          ),
      ),
  ) );

• sites (int/string/array)

  This parameter only applies in a multi-site environment. It lets you search for posts on specific sites or across the network.

  By default, sites defaults to current which searches the current site on the network:

  new WP_Query( array(
      's'     => 'search phrase',
      'sites' => 'current',
  ) );

  You can search on all sites across the network:

  new WP_Query( array(
      's'     => 'search phrase',
      'sites' => 'all',
  ) );

  You can also specify specific sites by id on the network:

  new WP_Query( array(
      's'     => 'search phrase',
      'sites' => 3,
  ) );

  You can even specify a group of specific sites on the network:

  new WP_Query( array(
      's'     => 'search phrase',
      'sites' => array( 2, 3 ),
  ) );

  Note: Nesting cross-site WP_Query loops can result in unexpected behavior.

• ep_integrate (bool)

  Allows you to perform queries without passing a search parameter. This is pretty powerful as you can leverage Elasticsearch to retrieve queries that are too complex for MySQL (such as a 5-dimensional taxonomy query). For example:

  Get 20 of the lastest posts

  new WP_Query( array(
      'ep_integrate'   => true,
      'post_type'      => 'post',
      'posts_per_page' => 20,
  ) );

  Get all posts with a specific category

  new WP_Query( array(
      'ep_integrate'   => true,
      'post_type'      => 'post',
      'posts_per_page' => -1,
      'category'       => 5,
  ) );

Related Posts Widget

The plugin makes available a widget called WP Engine Related Posts, with one configurable option: the number of posts to display in the widget, up to a maximum of 10.

Content is only generated if the widget is on a single post page, and uses the post ID to call Elasticsearch using its "more like this API".

The widget generates relatively generic HTML that may be styled to match any theme. This is an example of the output with the option set to only one post.

<ul class="ep4wpe-related-posts">
<li>
  <a href="http://dhap.wpengine.com/parent-child-relationship/" rel="bookmark" title="Parent / Child Relationship">Parent / Child Relationship</a>
  <span class="post-date">June 30, 2013</span>
</li>
</ul>

Supported WP-CLI Commands

The following commands are supported by WP Engine Search:

• wp ep4wpe index [--setup] [--network-wide] [--posts-per-page] [--no-bulk]

  Index all posts in the current blog. --network-wide will force indexing on all the blogs in the network. --setup will clear the index first and re-send the put mapping. --posts-per-page let's you determine the amount of posts to be indexed per bulk index (or cycle). --no-bulk let's you disable bulk indexing.

• wp ep4wpe activate

  Turns on WP Engine Search integration. Integration is automatically deactivated during indexing.

• wp ep4wpe deactivate

  Turns off WP Engine Search integration. Integration is automatically deactivated during indexing.

• wp ep4wpe delete-index [--network-wide]

  Deletes the current blog index. --network-wide will force every index on the network to be deleted.

• wp ep4wpe is-active

  Checks whether WP Engine Search is currently integration active. This is different than whether the plugin is WordPress active or not. During indexing, integration will be deactivated automatically.

• wp ep4wpe put-mapping [--network-wide]

  Sends plugin put mapping to the current blog index. --network-wide will force mappings to be sent for every index in the network.

• wp ep4wpe recreate-network-alias

  Recreates the alias index which points to every index in the network.

• wp ep4wpe stats

  Returns basic stats on Elasticsearch instance i.e. number of documents in current index as well as disk space used.

• wp ep4wpe status

Development

Setup

Follow the configuration instructions above to setup the plugin.

Testing

Within the terminal, change directories to the plugin folder. Initialize your testing environment by running the following command:

For HGV users:

bash bin/install-wp-tests.sh wordpress_test root '' localhost latest

For VVV users:

bash bin/install-wp-tests.sh wordpress_test root root localhost latest

For VIP Quickstart users:

bash bin/install-wp-tests.sh wordpress_test root '' localhost latest

where:

• wordpress_test is the name of the test database (all data will be deleted!)
• root is the MySQL user name
• root is the MySQL user password (if you're running VVV). Blank if you're running VIP Quickstart.
• localhost is the MySQL server host
• latest is the WordPress version; could also be 3.7, 3.6.2 etc.

Our test suite depends on a running Elasticsearch server. You can supply a host to PHPUnit as an environmental variable like so:

EP_HOST="http://192.168.50.4:9200" phpunit

Dockunit

WP Engine Search contains a valid Dockunit file for running unit tests across a variety of environments locally (PHP 5.2 and 5.5). It assumes the address of your Elasticsearch server is http://192.168.50.4:9200. You can use Dockunit by running:

dockunit

Issues

If you identify any errors or have an idea for improving the plugin, please open an issue. We're excited to see what the community thinks of this project, and we would love your input!

License

WP Engine Searchis free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.