Significant_terms aggregation Java API returns unformatted strings for Dates and IPs

[markharwood]

The date and ip field types return unformatted strings for Bucket.getKeyAsString() in the signficant_terms aggregation responses. These values, like the REST api should return formatted values.

[markharwood]

This issue may have become irrelevant.
As a result of the #17746 change it is not possible to to run significant_terms aggs on numeric-based fields and users will need to define a special string-based mapping for significant_terms analysis e.g:

        "myIpField": {
           "type": "ip",
           "fields":{
               "asKeyword":{
                   "type":"keyword"
               }
           }
        }

..and then run sig terms analysis on the myIpField.asKeyword version of the field.

We may return to this formatting issue if we figure out a way to reintroduce support for significant_terms on numeric types with their new scheme for Lucene indexing.

[markharwood]

For the record, the formatting solution was simple - it was copying the format.format() call found in SignificantLongTerms.Bucket.toXContent method for the key_as_string JSON property and applying it to the corresponding getKeyAsString() method used by Java clients.