Franklyn Dsouza
Franklyn Dsouza
Submitted the signed CLA to [[email protected]](mailto:[email protected])
> Skewed columns need to be derived from stats, so there should be no requirement for `SKEWED_JOIN_METADATA` From the original issue. @martint suggested getting the feature in with a session...
> @damnMeddlingKid Can you use broadcast join for your skewed queries or build side is too big? We are still working on collecting data from our workload to isolate skewed...
Wanted to bump this, would love to get people's thoughts on the change.
cc: @sopel39 @raunaqmorarka
Hey folks, was wondering how to proceed here.
I've removed the array constructor and fixed all the scope issues. Will look into using the record and fixing getMetadata this week.
> We're gradually trying to get rid of whole-plan optimizers based on PlanOptimizer in favor of incremental optimizers based on Rule. Unfortunately, this specific optimization requires propagating "traits" up the...
I've addressed all remaining comments and have two open comments. @martint could you take a look. I'll also try and run some tests on data this week and pull some...
Any updates on this issue ?