Is there anything that an author can do to prevent a well-formed Interval Join from being 'optimised' into a regular, non-interval join by later queries? Does the Flink runtime make any guarantees that when it alters an interval join into a regular join, that the safety of the original interval join is retained?
Many thanks in advance for any insight that anyone can offer. Regards, Mark On Wed, 15 Mar 2023 at 16:02, Leonard Xu <xbjt...@gmail.com> wrote: > > > > CREATE TEMPORARY VIEW filteredResults AS > > SELECT * from suspiciousOrders WHERE small_ts > large_ts; > > Looks like after added the condition, the final expanded query should not > match the condition[1] of an interval join that leads to the planner > recognize it as an interval join. It’s not a bug, interval join is a > special case of regular join, thus the result would be still correct. > > [1] > https://nightlies.apache.org/flink/flink-docs-release-1.14/zh/docs/dev/datastream/operators/joining/#interval-join
