Elastic Extensible Search

An extensible search implementation for Elastic Search.

Installation

composer require nyeholt/silverstripe-extensible-elastic

NOTE: Until heyday/silverstripe-elastica#10 is merged in, you will need to add the patch file at

https://gist.github.com/nyeholt/e3f4b1745b9af2cc3265df0e548d2b7c

Configuration

Add the following to your project's config

---
Name: elastic_config
---
nglasl\extensible\ExtensibleSearchPage:
  custom_search_engines:
    Symbiote\ElasticSearch\ElasticaSearchEngine: 'Elastic'

PageController:
  extensions:
    - 'nglasl\extensible\ExtensibleSearchExtension'
    - 'Symbiote\ElasticSearch\ElasticaSearchController'

Page:
  extensions:
    - 'Symbiote\ElasticSearch\ElasticaSearchable'

SilverStripe\Core\Injector\Injector:
  ElasticClient:
    class: Elastica\Client
    constructor:
      host_details: 
        host: elastic
        port: 9200
        # transport: AwsAuthV4 - this is needed for AWS search service compatibility; it adds credentials support
  Symbiote\ElasticSearch\ElasticaSearch:
    properties:
      searchService: %$Heyday\Elastica\ElasticaService
  Heyday\Elastica\ElasticaService:
    class: Symbiote\ElasticSearch\ExtensibleElasticService
    constructor:
      client: %$ElasticClient
      index: my-index

To add additional types for selection in an extensible search page config; note namespaces are supported.

---
Name: search_page_config
---
Symbiote\ElasticSearch\ElasticaSearch:
  additional_search_types:
    My\Namespaced\Class: Friendly Label

Run /dev/tasks/Symbiote-ElasticSearch-VersionedReindexTask

Note: Reindex will ONLY reindex items that have the Searchable extension applied. There's also
a DataDiscovery extension that will grab taxonomy terms if available.

---
Name: elastic_data_config
---
SilverStripe\CMS\Model\SiteTree:
  extensions:
    - Symbiote\ElasticSearch\ElasticaSearchable
    # for extra boosting options - Symbiote\ElasticSearch\DataDiscovery

API

To define your own custom field structures in the elastic index, you need to

• define your field mappings for the 'rebuild' phase
• add data for those fields during the indexing phase

public function updateElasticMappings($mappings = []) {
    $mappings['Identifier'] = ['type' => 'keyword'];
    $mappings['ContentType'] = ['type' => 'keyword'];
}

public function updateElasticDoc(Document $document)
{
    $document->set('Identifier', $this->Identifier);
    $document->set('ContentType', $this->ContentType);
}

Details

How do I use the BoostTerms field?

BoostTerms are used for subsequent querying, either direct through the builder or by the "Boost values" and
"Boost fields with field/value matches" options on the Extensible Search Page.

The field hint states to use the word "important" in this field to boost the record super high in result sets. This
requires you to set the "Boost fields with field/value matches" to have an entry of

BoostTerms:important : 10

in the search page to boost records with that set. Additionally, set the "Boost values" for BoostTerms to be higher
than all other fields for any match to contribute highly.

Why the separate ElasticaSearchable extension?

The base Heyday Elastic module doesn't handle indexing of Versioned content directly;
ElasticaSearchable provides a few overrides that take into account versioned content.

Can I get rid of stale results?

You can prune old results by creating the PruneStaleResultsJob ; this
takes as parameters

• The field:value filter to use; typically something like ClassName:MyDataClass.
  If you don't want a filter applied, pass the string 'null'
• How old something should be until it's considered 'old' in strtotime format
• How frequently to run the job in seconds, ie 86400 for every day
• How many to delete in each batch, typically around 1000

ClassName:My\Data\Class
-1 month
86400
1000