31 KiB
Elastica integration in Symfony2
Installation
Bundle and Dependencies
For Symfony 2.0.x projects, you must use a 1.x release of this bundle. Please check the bundle tags or the Packagist page for information on Symfony and Elastica compatibility.
Add FOSElasticaBundle to your application's composer.json file:
{
"require": {
"friendsofsymfony/elastica-bundle": "3.0.*@dev"
}
}
Install the bundle and its dependencies with the following command:
$ php composer.phar update friendsofsymfony/elastica-bundle
You may rely on Composer to fetch the appropriate version of Elastica. Lastly, enable the bundle in your application kernel:
// app/AppKernel.php
public function registerBundles()
{
$bundles = array(
// ...
new FOS\ElasticaBundle\FOSElasticaBundle(),
);
}
Elasticsearch
Instructions for installing and deploying Elasticsearch may be found here.
Basic configuration
Declare a client
Elasticsearch client is comparable to a database connection. Most of the time, you will need only one.
#app/config/config.yml
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
If your client requires Basic HTTP Authentication, you can specify an Authorization Header to
include in HTTP requests. The Authorization Header value is a base64 encoded string that
includes the authentication username and password, and can be obtained by running the following
command in your terminal:
php -r "Print 'Basic ' . base64_encode('your_auth_username' . ':' . 'your_auth_password');"
A sample configuration with Basic HTTP Authentication is:
#app/config/config.yml
fos_elastica:
clients:
default:
host: example.com
port: 80
headers:
Authorization: "Basic jdumrGK7rY9TMuQOPng7GZycmxyMHNoir=="
A client configuration can also override the Elastica logger to change the used class logger: <your logger class> or to simply disable it logger: false. Disabling the logger should be done on production because it can cause a memory leak.
Declare a serializer
Elastica can handle objects instead of data arrays if a serializer callable is configured
#app/config/config.yml
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: callback_class
serializer: serializer
callback_class is the name of a class having a public method serialize($object) and should
extends from FOS\ElasticaBundle\Serializer\Callback.
serializer is the service id for the actual serializer, e.g. serializer if you're using
JMSSerializerBundle. If this is configured you can use \Elastica\Type::addObject instead of
\Elastica\Type::addDocument to add data to the index. The bundle provides a default implementation
with a serializer service id 'serializer' that can be turned on by adding the following line to your config.
#app/config/config.yml
fos_elastica:
serializer: ~
Declare an index
Elasticsearch index is comparable to Doctrine entity manager. Most of the time, you will need only one.
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
Here we created a "website" index, that uses our "default" client.
Our index is now available as a service: fos_elastica.index.website. It is an instance of \Elastica\Index.
If you need to have different index name from the service name, for example,
in order to have different indexes for different environments then you can
use the index_name key to change the index name. The service name will
remain the same across the environments:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
website:
client: default
index_name: website_qa
The service id will be fos_elastica.index.website but the underlying index name is website_qa.
Declare a type
Elasticsearch type is comparable to Doctrine entity repository.
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
types:
user:
mappings:
username: { boost: 5 }
firstName: { boost: 3 }
lastName: { boost: 3 }
aboutMe: ~
Our type is now available as a service: fos_elastica.index.website.user. It is an instance of \Elastica\Type.
Declaring serializer groups
If you are using the JMSSerializerBundle for serializing objects passed to elastica you can define serializer groups per type.
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: %classname%
serializer: serializer
indexes:
website:
client: default
types:
user:
mappings:
username: { boost: 5 }
firstName: { boost: 3 }
lastName: { boost: 3 }
aboutMe:
serializer:
groups: [elastica, Default]
Declaring parent field
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
types:
comment:
mappings:
date: { boost: 5 }
content: ~
_parent: { type: "post", property: "post", identifier: "id" }
The parent field declaration has the following values:
type: The parent type.property: The property in the child entity where to look for the parent entity. It may be ignored if is equal to the parent type.identifier: The property in the parent entity which has the parent identifier. Defaults toid.
Note that to create a document with a parent, you need to call setParent on the document rather than setting a _parent field.
If you do this wrong, you will see a RoutingMissingException as elasticsearch does not know where to store a document that should have a parent but does not specify it.
Declaring nested or object
Note that object can autodetect properties
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
types:
post:
mappings:
date: { boost: 5 }
title: { boost: 3 }
content: ~
comments:
type: "nested"
properties:
date: { boost: 5 }
content: ~
user:
type: "object"
approver:
type: "object"
properties:
date: { boost: 5 }
Doctrine ORM and object mappings
Objects operate in the same way as the nested results but they need to have associations set up in Doctrine ORM so that they can be referenced correctly when indexing.
If an "Entity was not found" error occurs while indexing, a null association has been discovered in the database. A custom Doctrine query must be used to utilize left joins instead of the default inner join.
Populate the types
php app/console fos:elastica:populate
This command deletes and creates the declared indexes and types. It applies the configured mappings to the types.
This command needs providers to insert new documents in the elasticsearch types. There are 2 ways to create providers. If your elasticsearch type matches a Doctrine repository or a Propel query, go for the persistence automatic provider. Or, for complete flexibility, go for a manual provider.
Persistence automatic provider
If we want to index the entities from a Doctrine repository or a Propel query, some configuration will let ElasticaBundle do it for us.
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
types:
user:
mappings:
username: { boost: 5 }
firstName: { boost: 3 }
# more mappings...
persistence:
driver: orm # orm, mongodb, propel are available
model: Application\UserBundle\Entity\User
provider: ~
Three drivers are actually supported: orm, mongodb, and propel.
Use a custom Doctrine query builder
You can control which entities will be indexed by specifying a custom query builder method.
persistence:
driver: orm
model: Application\UserBundle\Entity\User
provider:
query_builder_method: createIsActiveQueryBuilder
Your repository must implement this method and return a Doctrine query builder.
Propel doesn't support this feature yet.
Change the batch size
By default, ElasticaBundle will index documents by packets of 100. You can change this value in the provider configuration.
persistence:
driver: orm
model: Application\UserBundle\Entity\User
provider:
batch_size: 100
Change the document identifier field
By default, ElasticaBundle will use the id field of your entities as the elasticsearch document identifier.
You can change this value in the persistence configuration.
persistence:
driver: orm
model: Application\UserBundle\Entity\User
identifier: id
Manual provider
Create a service with the tag "fos_elastica.provider" and attributes for the index and type for which the service will provide.
<service id="acme.search_provider.user" class="Acme\UserBundle\Search\UserProvider">
<tag name="fos_elastica.provider" index="website" type="user" />
<argument type="service" id="fos_elastica.index.website.user" />
</service>
Its class must implement FOS\ElasticaBundle\Provider\ProviderInterface.
<?php
namespace Acme\UserBundle\Provider;
use FOS\ElasticaBundle\Provider\ProviderInterface;
use Elastica\Type;
use Elastica\Document;
class UserProvider implements ProviderInterface
{
protected $userType;
public function __construct(Type $userType)
{
$this->userType = $userType;
}
/**
* Insert the repository objects in the type index
*
* @param \Closure $loggerClosure
* @param array $options
*/
public function populate(\Closure $loggerClosure = null, array $options = array())
{
if ($loggerClosure) {
$loggerClosure('Indexing users');
}
$document = new Document();
$document->setData(array('username' => 'Bob'));
$this->userType->addDocuments(array($document));
}
}
You will find a more complete implementation example in src/FOS/ElasticaBundle/Doctrine/AbstractProvider.php.
Search
You can just use the index and type Elastica objects, provided as services, to perform searches.
/** var Elastica\Type */
$userType = $this->container->get('fos_elastica.index.website.user');
/** var Elastica\ResultSet */
$resultSet = $userType->search('bob');
Doctrine/Propel finder
If your elasticsearch type is bound to a Doctrine entity repository or a Propel query, you can get your entities instead of Elastica results when you perform a search. Declare that you want a Doctrine/Propel finder in your configuration:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
serializer:
callback_class: FOS\ElasticaBundle\Serializer\Callback
serializer: serializer
indexes:
website:
client: default
types:
user:
mappings:
# your mappings
persistence:
driver: orm
model: Application\UserBundle\Entity\User
provider: ~
finder: ~
You can now use the fos_elastica.finder.website.user service:
/** var FOS\ElasticaBundle\Finder\TransformedFinder */
$finder = $container->get('fos_elastica.finder.website.user');
/** var array of Acme\UserBundle\Entity\User */
$users = $finder->find('bob');
/** var array of Acme\UserBundle\Entity\User limited to 10 results */
$users = $finder->find('bob', 10);
You can even get paginated results!
Pagerfanta:
/** var Pagerfanta\Pagerfanta */
$userPaginator = $finder->findPaginated('bob');
/** Number of results to be used for paging the results */
$countOfResults = $userPaginator->getNbResults();
Knp paginator:
$paginator = $this->get('knp_paginator');
$userPaginator = $paginator->paginate($finder->createPaginatorAdapter('bob'));
You can also get both the Elastica results and the entities together from the finder. You can then access the score, highlights etc. from the Elastica\Result whilst still also getting the entity.
/** var array of FOS\ElasticaBundle\HybridResult */
$hybridResults = $finder->findHybrid('bob');
foreach ($hybridResults as $hybridResult) {
/** var Acme\UserBundle\Entity\User */
$user = $hybridResult->getTransformed();
/** var Elastica\Result */
$result = $hybridResult->getResult();
}
If you would like to access facets while using Pagerfanta they can be accessed through the Adapter seen in the example below.
$query = new \Elastica\Query();
$facet = new \Elastica\Facet\Terms('tags');
$facet->setField('companyGroup');
$query->addFacet($facet);
$companies = $finder->findPaginated($query);
$companies->setMaxPerPage($params['limit']);
$companies->setCurrentPage($params['page']);
$facets = $companies->getAdapter()->getFacets());
Index wide finder
You can also define a finder that will work on the entire index. Adjust your index configuration as per below:
fos_elastica:
indexes:
website:
client: default
finder: ~
You can now use the index wide finder service fos_elastica.finder.website:
/** var FOS\ElasticaBundle\Finder\MappedFinder */
$finder = $container->get('fos_elastica.finder.website');
// Returns a mixed array of any objects mapped
$results = $finder->find('bob');
Repositories
As well as using the finder service for a particular Doctrine/Propel entity you can use a manager service for each driver and get a repository for an entity to search against. This allows you to use the same service rather than the particular finder. For example:
/** var FOS\ElasticaBundle\Manager\RepositoryManager */
$repositoryManager = $container->get('fos_elastica.manager.orm');
/** var FOS\ElasticaBundle\Repository */
$repository = $repositoryManager->getRepository('UserBundle:User');
/** var array of Acme\UserBundle\Entity\User */
$users = $repository->find('bob');
You can also specify the full name of the entity instead of the shortcut syntax:
/** var FOS\ElasticaBundle\Repository */
$repository = $repositoryManager->getRepository('Application\UserBundle\Entity\User');
The 2.0 branch doesn't support using
UserBundle:Userstyle syntax and you must use the full name of the entity. .
Default Manager
If you are only using one driver then its manager service is automatically aliased
to fos_elastica.manager. So the above example could be simplified to:
/** var FOS\ElasticaBundle\Manager\RepositoryManager */
$repositoryManager = $container->get('fos_elastica.manager');
/** var FOS\ElasticaBundle\Repository */
$repository = $repositoryManager->getRepository('UserBundle:User');
/** var array of Acme\UserBundle\Entity\User */
$users = $repository->find('bob');
If you use multiple drivers then you can choose which one is aliased to fos_elastica.manager
using the default_manager parameter:
fos_elastica:
default_manager: mongodb #defaults to orm
clients:
default: { host: localhost, port: 9200 }
#--
Custom Repositories
As well as the default repository you can create a custom repository for an entity and add
methods for particular searches. These need to extend FOS\ElasticaBundle\Repository to have
access to the finder:
<?php
namespace Acme\ElasticaBundle\SearchRepository;
use FOS\ElasticaBundle\Repository;
class UserRepository extends Repository
{
public function findWithCustomQuery($searchText)
{
// build $query with Elastica objects
return $this->find($query);
}
}
To use the custom repository specify it in the mapping for the entity:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
website:
client: default
types:
user:
mappings:
# your mappings
persistence:
driver: orm
model: Application\UserBundle\Entity\User
provider: ~
finder: ~
repository: Acme\ElasticaBundle\SearchRepository\UserRepository
Then the custom queries will be available when using the repository returned from the manager:
/** var FOS\ElasticaBundle\Manager\RepositoryManager */
$repositoryManager = $container->get('fos_elastica.manager');
/** var FOS\ElasticaBundle\Repository */
$repository = $repositoryManager->getRepository('UserBundle:User');
/** var array of Acme\UserBundle\Entity\User */
$users = $repository->findWithCustomQuery('bob');
Alternatively you can specify the custom repository using an annotation in the entity:
<?php
namespace Application\UserBundle\Entity;
use FOS\ElasticaBundle\Configuration\Search;
/**
* @Search(repositoryClass="Acme\ElasticaBundle\SearchRepository\UserRepository")
*/
class User
{
//---
}
Realtime, selective index update
If you use the Doctrine integration, you can let ElasticaBundle update the indexes automatically when an object is added, updated or removed. It uses Doctrine lifecycle events to schedule updates and then synchronizes changes either before or after flush.
Propel doesn't support this feature yet.
Declare that you want to update the index in real time:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
website:
client: default
types:
user:
mappings:
# your mappings
persistence:
driver: orm
model: Application\UserBundle\Entity\User
listener: ~ # by default, listens to "insert", "update" and "delete" and updates `postFlush`
Now the index is automatically updated each time the state of the bound Doctrine repository changes.
No need to repopulate the whole "user" index when a new User is created.
You can also choose to only listen for some of the events:
persistence:
listener:
insert: true
update: false
delete: true
By default, the ElasticSearch index will be updated after flush. To update before flushing, set immediate
to true:
persistence:
listener:
insert: true
update: false
delete: true
immediate: true
Using
immediateto update ElasticSearch before flush completes may cause the ElasticSearch index to fall out of sync with the source database in the event of a crash during the flush itself, such as in the case of a bad query.
Checking an entity method for listener
If you use listeners to update your index, you may need to validate your
entities before you index them (e.g. only index "public" entities). Typically,
you'll want the listener to be consistent with the provider's query criteria.
This may be achieved by using the is_indexable_callback config parameter:
persistence:
listener:
is_indexable_callback: "isPublic"
If is_indexable_callback is a string and the entity has a method with the
specified name, the listener will only index entities for which the method
returns true. Additionally, you may provide a service and method name pair:
persistence:
listener:
is_indexable_callback: [ "%custom_service_id%", "isIndexable" ]
In this case, the callback_class will be the isIndexable() method on the specified
service and the object being considered for indexing will be passed as the only
argument. This allows you to do more complex validation (e.g. ACL checks).
If you have the Symfony ExpressionLanguage component installed, you can use expressions to evaluate the callback:
persistence:
listener:
is_indexable_callback: "user.isActive() && user.hasRole('ROLE_USER')"
As you might expect, new entities will only be indexed if the callback_class returns
true. Additionally, modified entities will be updated or removed from the
index depending on whether the callback_class returns true or false, respectively.
The delete listener disregards the callback_class.
Propel doesn't support this feature yet.
Ignoring missing index results
By default, FOSElasticaBundle will throw an exception if the results returned from Elasticsearch are different from the results it finds from the chosen persistence provider. This may pose problems for a large index where updates do not occur instantly or another process has removed the results from your persistence provider without updating Elasticsearch.
The error you're likely to see is something like: 'Cannot find corresponding Doctrine objects for all Elastica results.'
To solve this issue, each mapped object can be configured to ignore the missing results:
persistence:
elastica_to_model_transformer:
ignore_missing: true
Advanced elasticsearch configuration
Any setting can be specified when declaring a type. For example, to enable a custom analyzer, you could write:
fos_elastica:
indexes:
doc:
settings:
index:
analysis:
analyzer:
my_analyzer:
type: custom
tokenizer: lowercase
filter : [my_ngram]
filter:
my_ngram:
type: "nGram"
min_gram: 3
max_gram: 5
types:
blog:
mappings:
title: { boost: 8, analyzer: my_analyzer }
Overriding the Client class to suppress exceptions
By default, exceptions from the Elastica client library will propagate through
the bundle's Client class. For instance, if the elasticsearch server is offline,
issuing a request will result in an Elastica\Exception\Connection being thrown.
Depending on your needs, it may be desirable to suppress these exceptions and
allow searches to fail silently.
One way to achieve this is to override the fos_elastica.client.class service
container parameter with a custom class. In the following example, we override
the Client::request() method and return the equivalent of an empty search
response if an exception occurred.
<?php
namespace Acme\ElasticaBundle;
use FOS\ElasticaBundle\Client as BaseClient;
use Elastica\Exception\ExceptionInterface;
use Elastica\Response;
class Client extends BaseClient
{
public function request($path, $method, $data = array())
{
try {
return parent::request($path, $method, $data);
} catch (ExceptionInterface $e) {
return new Response('{"took":0,"timed_out":false,"hits":{"total":0,"max_score":0,"hits":[]}}');
}
}
}
Clients as Tagged Services
Clients will be tagged as fos_elastica.client, which makes it possible to
retrieve all clients from the service container and interact with them via a
compiler pass. See
Working with Tagged Services
for more information.
Example of Advanced Query
If you would like to perform more advanced queries, here is one example using the snowball stemming algorithm.
It searches for Article entities using title, tags, and categoryIds.
Results must match at least one specified categoryIds, and should match the
title or tags criteria. Additionally, we define a snowball analyzer to
apply to queries against the title field.
$finder = $this->container->get('fos_elastica.finder.website.article');
$boolQuery = new \Elastica\Query\Bool();
$fieldQuery = new \Elastica\Query\Text();
$fieldQuery->setFieldQuery('title', 'I am a title string');
$fieldQuery->setFieldParam('title', 'analyzer', 'my_analyzer');
$boolQuery->addShould($fieldQuery);
$tagsQuery = new \Elastica\Query\Terms();
$tagsQuery->setTerms('tags', array('tag1', 'tag2'));
$boolQuery->addShould($tagsQuery);
$categoryQuery = new \Elastica\Query\Terms();
$categoryQuery->setTerms('categoryIds', array('1', '2', '3'));
$boolQuery->addMust($categoryQuery);
$data = $finder->find($boolQuery);
Configuration:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
site:
settings:
index:
analysis:
analyzer:
my_analyzer:
type: snowball
language: English
types:
article:
mappings:
title: { boost: 10, analyzer: my_analyzer }
tags:
categoryIds:
persistence:
driver: orm
model: Acme\DemoBundle\Entity\Article
provider:
finder:
Filtering Results and Executing a Default Query
If may want to omit certain results from a query, filtering can be more performant than a basic query because the filter results can be cached. In turn, the query is run against only a subset of the results. A common use case for filtering would be if your data has fields that indicate whether records are "active" or "inactive". The following example illustrates how to issue such a query with Elastica:
$query = new \Elastica\Query\QueryString($queryString);
$term = new \Elastica\Filter\Term(array('active' => true));
$filteredQuery = new \Elastica\Query\Filtered($query, $term);
$results = $this->container->get('fos_elastica.finder.index.type')->find($filteredQuery);
Date format example
If you want to specify a date format:
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
site:
types:
user:
mappings:
username: { type: string }
lastlogin: { type: date, format: basic_date_time }
birthday: { type: date, format: "yyyy-MM-dd" }
Dynamic templates
Dynamic templates allow to define mapping templates that will be applied when dynamic introduction of fields / objects happens.
fos_elastica:
clients:
default: { host: localhost, port: 9200 }
indexes:
site:
types:
user:
dynamic_templates:
my_template_1:
match: apples_*
mapping:
type: float
my_template_2:
match: *
match_mapping_type: string
mapping:
type: string
index: not_analyzed
mappings:
username: { type: string }