ES|QL: Change FUSE KEY BY to receive a list of qualifiedName

[ioanatia]

Previously KEY BY was using fields which is actually used by STATS ... BY.
However, what we actually want for the FUSE keys is a list of qualifiedName, whereas fields can also accept expressions like concat(field1, field2).

I checked if something like FUSE KEY BY concat(field1, field2) would actually work previously and it didn't - we still failed with a statement parser error, due to some mishandling ANTLR expression push mode for (.
So even if the syntax looked like it should allow something like concat(field1, field2), in reality this does not actually work even right now on main.

[elasticsearchmachine]

Hi @ioanatia, I've created a changelog YAML for you.

[elasticsearchmachine]

Pinging @elastic/es-search-relevance (Team:Search Relevance)

[ioanatia]

The KEY BY syntax is exercised in many existing tests already, see for example:

elasticsearch/x-pack/plugin/esql/qa/testFixtures/src/main/resources/fuse.csv-spec

Lines 343 to 350 in e8749d9

	FROM books METADATA _id, _score
	| FORK ( WHERE title:"Tolkien" | SORT _score, _id DESC | LIMIT 3
	| EVAL new_id = CONCAT(_id, _id) | DROP _id)
	( WHERE author:"Tolkien" | SORT _score, _id DESC | LIMIT 3 | EVAL new_id = CONCAT(_id, _id) | DROP _id)
	| EVAL new_fork = CASE (_fork == "fork1", "A", "B")
	| EVAL new_score = _score * 2
	| DROP _fork, _score
	| FUSE rrf SCORE BY new_score KEY BY new_id GROUP BY new_fork WITH {"rank_constant": 60, "weights": { "fork1": 0.3, "fork2": 0.7 } }

so I only need to add a statement tests to check that we raise a syntax error when the wrong syntax is used with KEY BY

[afoucret]

This grammar change is quite straightforward.

[carlosdelest]

Makes sense! 👍