drill
drill copied to clipboard
DRILL-4566: TDigest, median, and quantile functions
Adds tdigest, tdigest_merge, median, and quantile UDFS Adds reduce aggregates rule to reduce: median(x) -> tdigest_median(tdigest(x)) quantile(q, x) -> tdigest_quantile(q, tdigest(x)) Adds two-phase aggregate rule for tdigest
Also adds ability to specify tolerances in the TestBuilder
Any idea when this pull request will be implemented in mainstream release 1.10?
There was a brief discussion on the drill-dev mailing list a few days after this PR was posted. Unfortunately the discussion did not culminate in any decision. The discussion was mainly around what syntax should we use for these functions, since they are actually approximate functions.
If you want to revive the discussion, or propose a resolution, feel free to pick up this PR and make sure it can rebase on latest drill master. I probably won't be getting to it for at least a few weeks. But if someone else takes it up and gets it into a mergeable state, and others in the community are in agreement, I think we can merge it.
@StevenMPhillips Is anyone reviewing this? Can you set the original JIRA to be in a Reviewable state?