This plugin is a Search Plugin, to be used together with

• Flowpack.ElasticSearch.ContentRepositoryAdaptor or
• Flowpack.SimpleSearch.ContentRepositoryAdaptor.

Install via composer with your favorite adaptor:

ElasticSearch

composer require flowpack/searchplugin flowpack/elasticsearch-contentrepositoryadaptor

SimpleSearch

composer require flowpack/searchplugin flowpack/simplesearch-contentrepositoryadaptor

Inclusion of the routes from this package into your main Configuration/Routes.yaml is no longer needed as of Flow 4.0.

It is usually a good idea to specify a custom index name for a project, instead of the default typo3cr. That way no conflicts can arise when multiple projects use the same Elasticsearch server.

To specify a custom index name, the following is needed:

Neos:
  ContentRepository:
    Search:
      elasticSearch:
        indexName: acmecom

The pagination search results can be configured via Fusion. The following shows the defaults:

prototype(Flowpack.SearchPlugin:Search).configuration {
    itemsPerPage = 25
    insertAbove = false
    insertBelow = true
    maximumNumberOfLinks = 10
}

The result list is rendered using a Fusion object of type nodeType 'SearchResult' for each hit. Thus you can easily adjust the rendering per type like this for an imaginary Acme.AcmeCom:Product nodetype:

prototype(Acme.AcmeCom:ProductSearchResult) < prototype(Neos.Neos:DocumentSearchResult) {
    templatePath = 'resource://Acme.AcmeCom/Private/Templates/SearchResult/ProductSearchResult.html'
}

Feel free to use the DocumentSearchResult.html in the Flowpack.SearchPlugin as an example.

The default search form template comes with a data-autocomplete-source attribute pointing to the SuggestController of this package.

To use this term suggester, you need to configure the indexing like this, to define a custom analyzer to be used:

Flowpack:
  ElasticSearch:
    indexes:
      default:      # client name used to connect (see Flowpack.ElasticSearch.clients)
        acmecom:    # your (custom) index name
          settings:
            analysis:
              filter:
                autocompleteFilter:
                  max_shingle_size: 5
                  min_shingle_size: 2
                  type: 'shingle'
              analyzer:
                autocomplete:
                  filter: [ 'lowercase', 'autocompleteFilter' ]
                  char_filter: [ 'html_strip' ]
                  type: 'custom'
                  tokenizer: 'standard'

Then you need to configure the node types to be be included in the suggestion building, this can be done like this:

'Neos.Neos:Document':
  superTypes:
    'Flowpack.SearchPlugin:SuggestableMixin': true
    'Flowpack.SearchPlugin:AutocompletableMixin': true

'Neos.Neos:Shortcut':
  superTypes:
    'Flowpack.SearchPlugin:SuggestableMixin': false
    'Flowpack.SearchPlugin:AutocompletableMixin': false

'Neos.NodeTypes:TitleMixin':
  superTypes:
    'Flowpack.SearchPlugin:SuggestableMixin': true
    'Flowpack.SearchPlugin:AutocompletableMixin': true

When fed with a term parameter via a GET request, the SuggestController will return a JSON-encoded array of suggestions from Elasticsearch. They are fetched with a term suggester from the _all field, i.e. "the fulltext index".

These can be used to provide autocompletion on the search input using a JS library of your choice. In case you need to build the URI to the suggest controller yourself, this is what the form uses:

{f:uri.action(action: 'index', controller: 'Suggest', package: 'Flowpack.SearchPlugin', format: 'json', absolute: 1, arguments: {contextNodeIdentifier: node.identifier, dimensionCombination: dimensionCombination})}

The suggestionContext determines, if a result should be displayed in suggetions. By default, nodes that are hidden are excluded. In many projects, the search als takes the Neos.Seo metaRobotsNoindex property into account or excludes certain nodeTypes. In order to adjust the suggestion context to your search logic, you can write your custom logic and switch the implementation of the Flowpack\SearchPlugin\Suggestion\SuggestionContextInterface via Objects.yaml

To create valid links for suggestions the following package can be used:

• Punktde.OutOfBandRendering

Connect the given example of this package with a custom fusion object:

pathToProtoType = Your.Vendor:Content.NodeLink

And finally add the new property to the plugin settings:

Flowpack:
  SearchPlugin:
    searchAsYouType:
      suggestions:
        sourceFields:
          - title
          - __myProperty

The plugin comes with a controller that can be reached like this per default, using GET:

{f:uri.action(action: 'search', controller: 'AjaxSearch', package: 'Flowpack.SearchPlugin', arguments: {node: node, q: ''}, absolute: 1)}

It expects the search term as a parameter named q (as defined in AjaxSearch.fusion). This controller renders the search results and returns them as HTML without any of the page template. It can therefore be used to request search results via AJAX and display the result by adding it to the DOM as needed.

It is recommended to remove characters, which are reserved in Elasticsearch from the search term to prevent errors. There is an eel helper to replace them before submitting the search like this:

prototype(Flowpack.SearchPlugin:Search) {
    searchTerm = ${Flowpack.SearchPlugin.SearchTerm.sanitize(request.arguments.search)}
}

Keep in mind, that this blocks the explicit use of wildcards (*) and phrase search ("search exactly this") for your users, in case you want to support that.