[BUG] Incorrect serialized aggregation response

[chanon-onman]

What is the bug?

SearchResponse.toJsonString() generates incorrect aggregation response that doesn't match with what is in the documentation. The aggregation keys have extra prefix, e.g. lterms#, sterms#.

How can one reproduce the bug?

Execute the code below to get json serialized of search response.

SearchResponse resp = new SearchResponse.Builder<>()
        .aggregations("aggKey", v -> v.lterms(lv -> lv.buckets(b -> b.array(new ArrayList())).sumOtherDocCount(0)))
        .took(0)
        .timedOut(false)
        .shards(s ->s.failed(0).successful(1).total(1))
        .hits(h -> h.hits(new ArrayList()))
        .build();
System.out.println(resp.toJsonString());

The output is

{
  "took": 0,
  "timed_out": false,
  "_shards": {
    "failed": 0,
    "successful": 1,
    "total": 1
  },
  "hits": {
    "hits": []
  },
  "aggregations": {
    "lterms#aggKey": {    <--------*
      "buckets": [],
      "sum_other_doc_count": 0
    }
  }
}

What is the expected behavior?

Aggregation keys should not have other prefix according to the document: https://opensearch.org/docs/latest/aggregations/

What is your host/environment?

opensearch-java 2.17.0

[Xtansia]

This is not a bug but expected behavior, we (de)serialize aggregations following the behavior of the typed_keys parameter on the search api, this means aggregation keys are prefixed with their type so that there's no ambiguity in deserializing.

There is an existing issue #741 for adding a mechanism to disable the behavior.