adriangb commented on issue #20324: URL: https://github.com/apache/datafusion/issues/20324#issuecomment-3904434055
> > Maybe I'm just being biased here but I personally think the expensive to evaluate but not helpful join dynamic filters are just a pathological case of "expensive low selectivity filter". So while disabling them specifically (via existing feature flag or code change) may be an immediate win that lets us turn on filter pushdown I don't think they're actually "special", other filters will have the same issues / this is why we do see other slowdowns and have been fighting this for a while. > > I agree > > That being said, if we need to have an intermediate state where we disable dynamic filters that might be ok. Totally agree. My understanding is it’s only the hash join ones that are an issue that we don’t have an immediate solution to and probably require restructuring the filters themselves or having the more adaptive system in place. So if possible, let’s disable this those and not for example, the top K ones. -- 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]
