Piyush Narang

@avibryant - @johnynek mentioned that you might know more so tagging you on this.

Yeah that makes sense. I can keep reducing the resolution (from 2^-16 to 2^-1) and the range gets wider: ``` val qt = longList.map{ QTree(32,-4,1,_,None,None) }.reduce{ qtreeSemigroup.plus(_, _) } scala>...

@johnynek - do you know what graph it fails on? We had to work through a few job failures on the planning phase as the cascading3 planner didn't like the...

I tried out a few variants of the failing job and it seems like the planner seems to trigger due to both `p1` and `p2` performing a cross(hashJoin). If I...

It seems like the behavior matches that - I tried out a run where p1 didn't have the `.cross` which worked and the same for when just p2 didn't have...

+1, changes look good to me. We have ~20 or so failing targets due to these implicits (some fairly simple to fix, some will likely require us to create an...

@ianoc yeah we released 0.17.x at Twitter a few months back. I'll let @fwbrasil / @pankajroark / @ttim chime in if they have any reservations about the fields api breakages.

This sounds pretty cool. Some of the transformations you're proposing are useful and we have occasionally run into them. Our usage of Job vs Execution is split (not sure off...

I think dropping these implicits sounds good to me. I could do a dry run internally to see how many affected customers there are.

Tried running a sandbox with these changes and it looks like we have a decent number of failures (~25 or so targets). A bunch of our ads jobs are on...