I don't think it's that hard. We just never got round to it. Can you log a JIRA case, and submit a pull request when you're done. Be sure to add a test similar to UdfTest.testUserDefinedAggregateFunction.
Also, there's no code currently that would call merge, but don't let that stop you. EnumerableAggregate generates code that calls EnumerableDefaults.groupBy, which groups all values of the same key together. We'd use merge if we used a parallel or distributed algorithm. Another piece of code that could call merge is roll-up: see SubstitutionVisitor.getRollup. Julian On Tue, May 9, 2017 at 8:03 AM, James <[email protected]> wrote: > I was implementing UDAF based on calcite, but found that merge method is > not implemented: > > https://github.com/apache/calcite/blob/0938c7b6d767e3242874d87a30d9112512d9243a/core/src/main/java/org/apache/calcite/schema/impl/AggregateFunctionImpl.java#L86 > > I just wonder why it is not implemented? As a newbie to calcite, it seems > not so hard to implement the merge. Are there any issue blocking the > implemention of merge, or just a matter of time? If it is just a matter of > time, I'd like to spend some time implementing it. > > Thanks in advance. > James
