I think that would be great. Plus, we could use Calcite as a reference implementation for streaming SQL, to compare with other implementations from Flink, Beam etc.
I am encouraging the developers of MATCH_RECOGNIZE to contribute a reference implementation to Calcite. With that, Calcite would have complex event processing, albeit not scalable or particularly efficient. Julian > On Sep 27, 2017, at 1:52 AM, Simeon Zverinski <[email protected]> > wrote: > > Hi all, > > I'm pretty new to Calcite and understood that it's mainly supposed to be used > for building data processing engines. I've also found in examples/unit tests > that Calcite has its own engine as well. It looks like this engine doesn't > implement all the features though. For example, Streaming Group By is not yet > supported (https://issues.apache.org/jira/browse/CALCITE-1345) and also not > all time grouping functions are fully implemented, such as TUMBLE. Anyway I > found this small engine to be very useful for implementing small embedded > solutions when you don't want to introduce dependencies on something big such > as Flink. > > What do you think about using pure Calcite for embedded solutions? Any hints > on how to implement missing functions? > > Thanks, > > Simeon >
