README

A Neos Content Repository search common package used to implement concrete indexing and search functionality.

Related packages

Some of the related packages are:

Flowpack.ElasticSearch.ContentRepositoryAdaptor

To use Elasticsearch for indexing and searching.

Flowpack.SimpleSearch.ContentRepositoryAdaptor

Uses a SQLite database for indexing and search and thus can be used without additional dependencies.

Flowpack.SearchPlugin

A plugin to offer search functionality to users via Fusion rendering.

Realtime Indexing (experimental)

Realtime indexing in Neos.ContentRepository.Search >= 5.0 (compatible with Neos 9) has SEVERE PERFORMANCE PENALTYS when publishing workspaces. We suggest that you do batch indexing in the background instead.

To resolve backreferences (a very common Elasticsearch Use Case for Neos < 9), please now directly use the backReferences() FlowQuery operation available in Neos 9.

If you still want to enable this, add the following to the Settings.yaml (but you have been warned).

# Settings.yaml

Neos:
  ContentRepositoryRegistry:
    presets:
      'default':
        contentGraphProjection:
          catchUpHooks:
            'Neos.ContentRepository.Search:RealtimeIndex':
              factoryObjectName: \Neos\ContentRepository\Search\CatchUpHook\RealTimeIndexCatchUpHookFactory

Inner workings

The NodeIndexingManager listens to signals emitted from Neos Content Repository and the PersistenceManager if realtimeIndexing.enabled is true (which it defaults to).

• nodeAdded, nodeUpdated, afterNodePublishing trigger indexNode()
• nodeRemoved triggers removeNode()
• allObjectsPersisted triggers flushQueues

During a single request the queue with index changes is only flushed once the indexingBatchSize has been reached (see flushQueuesIfNeeded()).

In case the operation queues are flushed, the IndexingManager in turn uses the NodeIndexer to run indexNode() and removeNode() respectively.

If realtimeIndexing.enabled is false, the node index is only updated when built manually.