The java 8 group by collector's default implementation uses underlying hash maps. Any iteration of the results ends up iterating over hash map entries in a way that is not conducive to spliterators. Specifically, things like sizing and subsizing don't work on the spliterators. Since many use cases for the collectors result in an immutable stream, and collect the entire stream in the process, it should be possible to have a result which can be part of a divide and conquer strategy based on streams and spliterators.
There are some places in the code that hard materialize to a list in order to attain these properties. They should be fixed once a suitable group by collector implementation is available. [ Full content available at: https://github.com/apache/incubator-druid/issues/6421 ] This message was relayed via gitbox.apache.org for [email protected]
