Using the Java API for scripting in Elasticsearch 5.6

29 Nov 2017, Fabian Steeg | 🏷 lobid-resources 

To create a custom aggregation in lobid-resources for NWBib, I used the Java API for scripting in Elasticsearch 5.6. It was surprisingly hard to find details on this, since there was no full, current documentation available. So here is the gist of it.

First, we specify the scripting language to use, and a script ID:

String lang = "painless";
String id = "topic-aggregation";

Then, our actual script content (as an example, we just pick the document’s type field here):

String script = "doc['type']";

Next, we have to create a JSON string to pass to the Elasticsearch API. Our goal is something like this:

{ "script": { "lang": "painless", "source": "doc['type']" } }

To avoid escaping issues, we use a JSON lib here, like this:

ObjectNode scriptObject = Json.newObject();
scriptObject.putObject("script")
	.put("lang", lang).put("source", script);

We then wrap the JSON string into a BytesArray object:

BytesArray bytes = new BytesArray(scriptObject.toString());

And finally use that object to store the script in our cluster:

PutStoredScriptResponse response = client.admin().cluster()
	.preparePutStoredScript().setId(id)
	.setContent(bytes, XContentType.JSON).get();

Later, we can create a Script object based in the script ID:

Script script = new Script(
	ScriptType.STORED, lang, id, Collections.emptyMap());

Which we can then use, e.g. in our Aggregation:

searchRequest.addAggregation(
	AggregationBuilders.terms("topic").script(script));

For details on other Elasticsearch APIs and the painless scripting language, see the Elasticsearch documentation.

Comments? Feedback? Just add an annotation with hypothes.is.