Download the PHP package mikemadisonweb/yii2-elasticsearch without Composer

Elasticsearch Extension for Yii2

Yii2 extension for integration with Elasticsearch version 5.0 and above, based on official elasticsearch-php low-level client.

Compared to elasticsearch-php this extension has more intuitive way of doing things, like index documents, search, percolate (reverse search), building complex filter conditions using simple SQL-like language. Also, it's highly configurable and extensible, it's not tightly tied to ActiveRecord models, but this can be easily implemented.

The documentation is relevant for the latest stable version of the extension.

Installation

The preferred way to install this extension is through composer.

Either run

or add

to the require section of your composer.json file.

Configuration

Let's suppose that you have a huge database of blog posts and you want to use full-text search on title and body fields, also search for keywords as is and maybe filter by categories and tags:

You can configure multiple elasticsearch clients if you need. This could be different search clusters used by your application. Each client can be defined as multiple hosts, this parameter contains list of hosts or ip addresses of your elasticsearch servers, optionally followed by a port(default is 9200). You can find more options in elasticsearch-php documentation. By default connection to either one of these servers is calculated using round-robin strategy.

Consider your needs on performance and durability when selecting the number of shards and replicas. The right amount of these parameters is a compromise between those characteristics. Above example uses the default number of shards and replicas, so if that is what you want, you can omit these parameters in configuration.

After you specify the name of the index and the client using this index, you need to specify the fields that are required for the search. These fields can be grouped into mappings the way you find it logical and convenient to search for(e.g. 'posts'). In terms of full-text search, any single piece of data that you add to an index called a document. An easy way to understand the meaning of the mapping is to think of it as a document type. By default, Elasticsearch doesn't force you to provide that kind of schema for your data, so if your document has more fields than listed in mapping, those new fields would be stored as well without any problems. To change that behavior you can set dynamic to strict, unspecified fields upon insert will raise an error then.

When you insert new document each field is processed based on it's datatype. For example, text datatype is needed to index full-text values (these values are analyzed) and keyword is searchable by their exact value, which is useful for filtering or aggregations. Please note that if you want to store an array, there is no need to explicitly define array datatype, you can store an array of integers in integer datatype.

By default Elasticsearch store incoming documents upon indexing, when you retrieve search results for it, you will find the original content in _source field. You can disable that behavior by setting parameter in mapping config:

From the other hand you can set 'index' => false in particular mapping field to prevent that field to be indexed.

Elasticsearch also has a special _all field, which concatenates the values of all of the other fields. It's used when you don't specify the list of fields to search in, but it does not get stored.

Extension comes with console command to create and delete indices depending on configuration:

Usage

Create index

After you configure your indices there is the command to create them on Elasticsearch:

If you don't provide an index name it will create all defined indices.

Indexing Documents

After you have created an index you can insert new records like so:

Insert method has a second optional parameter - an unique id. You may use any string you want as an id, every document in the index has that id and if you don't pass it to the method id would be automatically generated by Elasticsearch.

Other useful methods in the Indexer class:

Batch insert is a way to speed up indexing process and it is an only method returning an array instead of ElasticResponse object. An array of documents in the batch can be numerical(without an id) and associative with ids as keys.

Searching

Extension provide a builder to simplify the process of query composition. To execute a full-text search on 'title' field you should use match() method:

There are also ways to filter the results, sort them or select specific portion of it:

Please note that if you use analyzed filed(full-text field) in where() method you will probably get wrong results because that method should be used only for non-analyzed datatypes(keyword, integer, boolean etc). There is optional parameters in match() method that you most certainly should keep eye on:

Most of the times you would need to match query string to particular fields, but if you want to search different strings in different fields (e.g. 'some text' on 'title' and/or 'another data' on 'body') then you can choose logical operator in condition parameter for your needs. If you searching on multiple fields you can define operator and type parameters to specify the logic for searching on these fields.

More complex filter causes can be specified in a SQL-like fashion:

If by any chance you feel limited to Finder methods you can pass raw JSON and interact directly with Elasticsearch API:

Analysis

There is a huge amount of options Elasticearch provides for analyzing text. Analyzers, normalizers, token filters are out of the scope of this documentation, you can find loads of information about them in official docs.

This is an example analysis config for the Russian language with stop-words filter and snowball stemmer:

Highlighting

Elasticsearch provides a way to highlight search terms in the source text. If you search for 'New possibilities' the word 'possibility' found in some document would be surrounded with tags like so <em>possibility</em>. These highlighted results would be in the 'highlight' field of the response, not in the '_source' field. There is a Finder method highlight() to turn it on for a single query. You can also pass various parameters as a second argument:

To turn on this functionality globally you should enable it in index configuration. Default sort order and limit options can be configured there as well:

Asterisk sign means it would highlight in all configured analyzed fields, but there is a catch. If you want to receive highlighted results in all text fields, then you shouldn't pass _all as a search field to the match() function.