I have been working on ARROW-5227 [1] for quite a while now (with help from some other contributors) and the new trait-based physical query plan has now reached feature parity with the previous implementation of query execution which directly executed the logical plan enum and therefore could not be extended by other projects and also didn’t support parallel execution.
I would now like to delete the original query execution code which would reduce the code base by approximately 2,500 lines of code and remove a lot of code duplication. I have a WIP PR for this [2]. I am raising this for discussion here because there is an argument for just deprecating this code for 1.0.0 and then removing in a later release, but I would prefer to delete it prior to the 1.0.0 release to reduce confusion and maintenance overhead. Does anybody have any objections to this? [1] https://issues.apache.org/jira/browse/ARROW-5227 [2] https://github.com/apache/arrow/pull/5583
