[SPARK-27280][SQL]infer more filter from join or condition

[windpiger]

What changes were proposed in this pull request?

current spark can handle this query to infer fitler from join condition:

select * from t1 join t2 where ((t1.a = t2.d and t1.b = 1) or (t1.a = t2.d and t1.b = 5))

(t1.b =1 or t1.b =5) can be infered, and pushed to table t1.

but it can not handle following query:

select * from t1 join t2 where (t1.a = t2.d and t1.b = 1 and t2.e=3 ) or (t1.a = t2.d and t1.b = 5 and t2.e=7)

(t1.b=1 or t1.b=5) should be infered and push to t1
(t2.e=3 or t2.e=7) should be infered and push to t2

How was this patch tested?

unit test added

[beliefer]

Is this SQL grammar supported by Hive or standard SQL?

[beliefer]

Is this SQL grammar supported by Hive or standard SQL?

[AngersZhuuuu]

hive can do this

[SparkQA]

Test build #104008 has finished for PR 24228 at commit 879c98e.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AngersZhuuuu]

#24973 I want to do the same thing for partition puring

[maropu]

I also think we need not a band-aid fix, but a general solution for this issue as @cloud-fan suggested in #24973. Have you checked the previous activity for the CNF normalization? https://issues.apache.org/jira/browse/SPARK-6624

[AngersZhuuuu]

I also think we need not a band-aid fix, but a general solution for this issue as @cloud-fan suggested in #24973. Have you checked the previous activity for the CNF normalization? https://issues.apache.org/jira/browse/SPARK-6624

thanks for Issue link. I will work on this in partition predicate part.

[github-actions]

We're closing this PR because it hasn't been updated in a while.
This isn't a judgement on the merit of the PR in any way. It's just
a way of keeping the PR queue manageable.

If you'd like to revive this PR, please reopen it!