Github user cloud-fan commented on the pull request:
https://github.com/apache/spark/pull/9844#issuecomment-158831881
This is really a good catch, thanks @dilipbiswal
The problems is that, normal operator should be resolved based on its
child, but `Sort` operator can be resolved based on its grandchild. So we have
3 rules that can resolve `Sort`: `ResolveReferences`,
`ResolveSortReferences`(if grandchild is `Project`) and
`ResolveAggregateFunctions`(if grandchild is `Aggregate`).
For your example, `select c1 as a , c2 as b from tab group by c1, c2 order
by a, c2`, we need to resolve `a` and `c2` for `Sort`. Firstly `a` will be
resolved in `ResolveReferences` based on its child, and when we reach
`ResolveAggregateFunctions`, we will try to resolve both `a` and `c2` based on
its grandchild, but failed because `a` is not a legal aggregate expression.
I think we can just fix the problem directly, i.e. only pick up unresolved
`SortOrder`s and try to resolve it based on grandchild in
`ResolveAggregateFunctions`. @dilipbiswal what do you think?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
