adriangb commented on PR #103: URL: https://github.com/apache/datafusion-site/pull/103#issuecomment-3266449666
> And the downside of that would be you could end up with imbalanced partitions due to skew (e.g. if all your values ended up in a few of the partitions) I think the way to combat that would be to initially set fixed ranges then do a reshuffle where you re-set the ranges (and move the build side data around) to balance the ranges before you start working on the probe side. You can pick arbitrary ranges, they don't have to be evenly spaced. If you knew something about the distributions ahead of time you could probably also optimize the building of the build side. -- 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]
