geoffreyclaude commented on PR #27181: URL: https://github.com/apache/datafusion/pull/27181#issuecomment-6703921224
Thanks for the ping @alamb! I already had [#25186](https://github.com/apache/datafusion/pull/25186) open for this, but thought it was waiting on the parent epic. I’ve now rebased it onto main and updated it to use the shared float-semantics API. For constant float `IN` lists, it adds the missing zero sign once when constructing the filter. This avoids scanning each input batch to normalize its zeros. The benchmarks look promising: about [3 µs for a three-entry Float32 list](https://github.com/apache/datafusion/pull/25186#issuecomment-6703634181), compared with [6 µs here](https://github.com/apache/datafusion/pull/27181#issuecomment-6698452098). I also found a nested-list regression here. This returns `false`, although the first entry matches exactly: ```sql SELECT [-0.0] IN ([-0.0], [1.0], [2.0], [3.0]); ``` Arrays are normalized recursively, but scalars containing lists aren’t. #25186 handles both consistently. Would you mind having a look at #25186 as an alternative? -- 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]
