Can't handle _source-fields with point (".") inside field-name

[chris-searchit]

elasticsearch-carrot2/src/main/java/org/carrot2/elasticsearch/ClusteringAction.java

Lines 1108 to 1124 in a027f04

	String[] fieldNames = spec.field.split("\\.");
	Object value = sourceAsMap;
	// Descend into maps.
	for (String fieldName : fieldNames) {
	if (Map.class.isInstance(value)) {
	value = ((Map<?, ?>) value).get(fieldName);
	if (value == null) {
	// No such key.
	logger.warn(
	"Cannot find field named '{}' from spec: '{}'", fieldName, spec.field);
	break outer;
	}
	} else {
	logger.warn("Field is not a map: {} in spec.: {}", fieldName, spec.field);
	break outer;
	}

_source-fields with point (".") inside field-name are not correct handled.
E.g. following field_mapping will fail, because point inside "text.title"

    "field_mapping": {
      "title"  : ["_source.text.title"]
    }

[dweiss]

I'll take a look. If you can beat me to it and provide a patch, it'd be great.

[dweiss]

Sorry, took a while. Indeed, the path within _source is evaluated with a simple dot-delimited pattern. I wonder how to elegantly solve it though... With some kind of escaping? Or maybe with quoted-path segments?

   "field_mapping": {
      "title"  : ['_source."text.title"']
    }

[chris-searchit]

That seems to be a solution. Unfortunately i can't get unit tests running to provide a patch, sorry.

[dweiss]

The tests pass for me on master (windows and Linux) - strange that you can't get them to work. I do have this in the back of my mind but I admit it is of low priority. Will get there, eventually. Thanks for reporting.

[chris-searchit]

Which JDK version are you using?

[dweiss]

Elasticsearch plugin will enforce the minimum JDK version for each ES release. At the moment master is on 7.9.3 and I'm using JDK 14.0.1. gradlew check passes just fine. What's the error you're getting?