Scripting: Ambiguous resolution of `_doc.score`

[jpountz]

Some scripting languages don't make a difference between property access foo.bar and map lookups foo['bar']. This is an issue for the DocLookup class (which is the class behind the _doc variable and implements java.util.Map) since there are two ways that _doc.score can be resolved:

1. DocLookup.getScore()
2. DocLookup.get("score").

The interesting thing is that mvel translates _doc.score to getScore while Groovy seems to translate it to get("score"). So when trying to lookup the score in Groovy, you might see the following error:

GroovyScriptExecutionException[ElasticsearchIllegalArgumentException[No field found for [score] in mapping with types [index_name]].

For reference, if you encounter this issue, you can work around it by using _score instead of _doc.score. So maybe the way to go would be to deprecate _doc.score and make scripting engines expose the score via _score instead. I tend to find it cleaner as well since the score is not really a property of the document (it also depends on the query).

[dakrone]

+1 on using _score instead of _doc.score!

[jpountz]

For reference, here is a confirmation that in case of a conflict Groovy prefers the map access: http://groovy.329449.n5.nabble.com/Property-resolution-clash-for-Map-subclasses-td5717713.html

[dakrone]

@jpountz I looked and we already support _score in Groovy scripts (there's a test for it too). So I think all that needs to be done for this is to add documentation about _score being the preferred method instead of _doc.score. Does that sound good to you?

[jpountz]

+1