andygrove opened a new pull request #32195: URL: https://github.com/apache/spark/pull/32195
### What changes were proposed in this pull request? AQE has an optimization where it attempts to reuse compatible exchanges but it does not take into account whether the exchanges are columnar or not, resulting in incorrect reuse under some circumstances. ### Why are the changes needed? When using the [RAPIDS Accelerator for Apache Spark](https://github.com/NVIDIA/spark-rapids) we sometimes see a new query stage correctly create a row-based exchange and then Spark replaces it with a cached columnar exchange, which is not compatible, and this causes queries to fail. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? The patch has been tested with the query that highlighted this issue. I looked at writing unit tests for this but it would involve implementing a mock columnar exchange in the tests so would be quite a bit of work. If anyone has ideas on other ways to test this I am happy to hear them. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
