opt: add rule to reduce IS DISTINCT FROM NULL in join filters

[DrewKimball]

This PR introduces a normalization rule that removes an IS NOT NULL
condition on a not-null column from the ON condition of a left or full
join.

Ex:

SELECT * FROM abc FULL JOIN abc AS abc2 ON abc.a IS NOT NULL;
=>
SELECT * FROM abc FULL JOIN abc AS abc2 ON True;

Fixes #48173

Release note: None

[cockroach-teamcity]

This change is 

[mgartner]

Reviewed 1 of 3 files at r1.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @DrewKimball)

---

pkg/sql/opt/norm/testdata/rules/join, line 2157 at r1 (raw file):

      └── a.k:1 = a2.f:9 [outer=(1,9), constraints=(/1: (/NULL - ]; /9: (/NULL - ]), fd=(1)==(9), (9)==(1)]

# Can't simplify: one equality condition has columns from same side of join.

It sounds like this is a test case we still want. This test happens to fail now because a.f=a.f is simplified to a IS DISTINCT FROM NULL, but the comment here claims that the point of the test is to ensure full-join aren't simplified when columns from the same side of the join are referenced in one of the ON filters.

I think you could test this same case by changing a.f=a.f to a.f=a.i, and it would remain a full-join even with your normalization rule.

(As an aside, this is a good reminder why comments explaining each test are extremely helpful.)

[DrewKimball]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @mgartner)

---

pkg/sql/opt/norm/testdata/rules/join, line 2157 at r1 (raw file):

Previously, mgartner (Marcus Gartner) wrote…

It sounds like this is a test case we still want. This test happens to fail now because a.f=a.f is simplified to a IS DISTINCT FROM NULL, but the comment here claims that the point of the test is to ensure full-join aren't simplified when columns from the same side of the join are referenced in one of the ON filters.

I think you could test this same case by changing a.f=a.f to a.f=a.i, and it would remain a full-join even with your normalization rule.

(As an aside, this is a good reminder why comments explaining each test are extremely helpful.)

Ah, nice catch. Done.

[mgartner]

Reviewed 1 of 3 files at r1, 1 of 1 files at r2.
Reviewable status: complete! 1 of 0 LGTMs obtained

[DrewKimball]

TFTR!

[DrewKimball]

bors r+

[craig]

This PR was included in a batch that was canceled, it will be automatically retried

[craig]

Build failed (retrying...):

• GitHub CI (Cockroach)

[craig]

Build succeeded:

• GitHub CI (Cockroach)