alamb commented on issue #2064: URL: https://github.com/apache/arrow-datafusion/issues/2064#issuecomment-1076149642
@yjshen I think this is a common problem with vectorized evaluation engines, where the overhead of checking for overflow somewhat reduces the effectiveness of the vectorization I think your suggestion of using `take_all_valid` is a reasonable one The other approach I have heard of (though never seen implemented) is to implement some sort of 'poisoning' / deferred exception tracking that leaves the value of `a / 0` as `nulll` but doesn't throw the error until later My personal opinion is go for correct (aka the `take` approach you outline) and then we can optimize for speed if/when evaluating CASE becomes a bottleneck -- 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]
