rdtr commented on PR #13529: URL: https://github.com/apache/gluten/pull/13529#issuecomment-6764264013
Thanks @philo-he for the review. I added this to the ANSI section of `docs/backends/velox/velox-backend-limitations.md`. It covers: 1. Why an expression can carry a different mode from the session. A view keeps the mode it was created with, and a DataFrame, or a temp view registered from it, keeps the mode it was analyzed with. 2. The list of expressions that fall back on a mismatch. This applies with either value of `spark.gluten.sql.ansiFallback.enabled`. With the default, it matters for a view created with ANSI on and read with ANSI off. 3. `ceil` / `floor` on a double count as ANSI off before Spark 4.3, so they fall back with ANSI on. 4. `arr[i]` with ANSI on falls back, while `get(arr, i)` stays native. I also added a test for the DataFrame case and rebased onto main, which conflicted with the format_string check in `VeloxValidatorApi`. Could you take another look? -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
