Github user cloud-fan commented on the issue:
https://github.com/apache/spark/pull/21123
This PR needs to be rebased once #22009 gets in.
> I think that PR should remove SaveMode and should always use the new
logical plans for v2 operations.
I don't agree according to the fact that these new logical plans are not
ready. As you said, we can interpret `SaveMode` and generate these new logical
plans. However, this mapping is not mentioned in the logical plan
standardization design doc and I doubt if it's doable. As you said, the
behavior of `SaveMode` is ambiguous, it's up to the data source to interpret
the `SaveMode`, so we need to pass `SaveMode` to the data source in the API.
BTW from a project manager's point of view, it's risky to make data source
v2 unusable until a big feature is finished. We should do things incrementally
and always prepare for the worst case. Although very unlike, if we can't finish
the logical plan standardization before 3.0, we should still have a usable data
source v2 at that time.
cc @rxin @marmbrus
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
