andygrove commented on PR #4948: URL: https://github.com/apache/datafusion-comet/pull/4948#issuecomment-5004453770
The collation guard looks right. One question on routing. Since #4728, an `Unsupported` result is offered to the JVM codegen dispatcher, but only when the serde mixes in `CodegenDispatchFallback`. These don't, so a single collated operand falls the whole projection or filter back to Spark. This same PR opts `CometStrToMap` into that trait for exactly this reason. Could `CollationAwareBinaryPredicate` mix in `CodegenDispatchFallback` too, the way `CometFromUnixTime` does? That would run Spark's own collation-aware `doGenCode` inside the Comet pipeline and keep the rest of the operator native. If there's a reason the dispatcher can't handle these, a comment capturing that would help. -- 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]
