Thanks Cancai, Having reviewed the first integration of the adapter I agree it's much more mature now.
Best regards, Alessandro On Thu, Jun 4, 2026, 09:01 Cancai Cai <[email protected]> wrote: > Hi all, > > I would like to share a short update on the Arrow adapter work under > [CALCITE-6304][1]. > > [CALCITE-6304][1] was created as an umbrella issue to improve the Arrow > adapter. With the recent focused issues and PRs, the main functional gaps > in the Arrow adapter have now largely been addressed. > > The Arrow adapter now has a more complete baseline for basic SQL usage: > schema inference, table scans, projections, filters, null handling, and a > broader set of Arrow data types are covered by result-level tests. > > I would describe the adapter as having reached a basic MVP state for common > SQL queries. There are still follow-up areas, especially around more > complex nested types and execution-path cleanup, but the adapter is now in > a much more usable state than when [CALCITE-6304][1] was opened. > > I think this may also be useful as a reference for other adapters. Many > adapters need to go through a similar path: map source-specific types into > Calcite row types, support common SQL projection and filter behavior, > handle null values correctly, and verify result correctness from SQL-level > tests rather than only checking plans or internal helper methods. > > Different adapters have different storage models and execution constraints, > so the implementation details will differ. But the Arrow work provides one > example of how an adapter can reach a basic SQL MVP through focused issues > and tests. > > Thanks, > Cancai > > [1]: https://issues.apache.org/jira/browse/CALCITE-6304 >
