Full-text-search query execution in Amazon Neptune

In a query that includes full-text-search, Neptune tries to put the full-text-search calls first, before other parts of the query. This reduces the number of calls to OpenSearch and in most cases significantly improves performance. However, this is by no means a hard-and-fast rule. There are situations, for example, where a PatternNode or UnionNode may precede a full-text search call.

Consider the following Gremlin query to a database that contains 100,000 instances of Person:

g.withSideEffect('Neptune#fts.endpoint', 'your-es-endpoint-URL')
 .hasLabel('Person')
 .has('name', 'Neptune#fts marcello~');

If this query were executed in the order in which the steps appear then 100,000 solutions would flow into OpenSearch, causing hundreds of OpenSearch calls. In fact, Neptune calls OpenSearch first and then joins results with the Neptune results. In most cases, this is much faster than executing the query in the original order.

You can prevent this re-ordering of query-step execution using the noReordering query hint:

g.withSideEffect('Neptune#fts.endpoint', 'your-es-endpoint-URL')
 .withSideEffect('Neptune#noReordering', true)
 .hasLabel('Person')
 .has('name', 'Neptune#fts marcello~');

In this second case, the .hasLabel step is executed first and the .has('name', 'Neptune#fts marcello~') step second.

For another example, consider a SPARQL query against the same kind of data:

PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX neptune-fts: <http://aws.amazon.com/neptune/vocab/v01/services/fts#>
SELECT ?person WHERE {
  ?person rdf:type foaf:Person .
  SERVICE neptune-fts:search {
    neptune-fts:config neptune-fts:endpoint 'http://your-es-endpoint.com' .
    neptune-fts:config neptune-fts:field foaf:name .
    neptune-fts:config neptune-fts:query 'mike' .
    neptune-fts:config neptune-fts:return ?person .
  }
}

Here again, Neptune executes the SERVICE part of the query first, and then joins the results with the Person data. You can suppress this behavior using the joinOrder query hint:

PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX neptune-fts: <http://aws.amazon.com/neptune/vocab/v01/services/fts#>
PREFIX hint: <http://aws.amazon.com/neptune/vocab/v01/QueryHints#>
SELECT ?person WHERE {
  hint:Query hint:joinOrder "Ordered" .
  ?person rdf:type foaf:Person .
  SERVICE neptune-fts:search {
    neptune-fts:config neptune-fts:endpoint 'http://your-es-endpoint.com' .
    neptune-fts:config neptune-fts:field foaf:name .
    neptune-fts:config neptune-fts:query 'mike' .
    neptune-fts:config neptune-fts:return ?person .
  }
}

Again, in the second query the parts are executed in the order they appear in the query.