Github user KurtYoung commented on the issue:
https://github.com/apache/flink/pull/3511
@heytitle @ggevay Great work!
We are working on a project to fully code generate the algorithm Flink
runtime used, and borrowed lots of idea of this PR, thanks! IMHO, in addition
to these changes, there are still some potential improvements we can do about
the sorter, like deserialization when comparing the real records. To achieve
this, we need more type information control and flexible code generating
supports, so we choose to do it through the Table API & SQL. How do you see
this approach?
---
