[CALCITE-7479] Remove redundant aggregate group keys with FD

[xiedeyantu]

Jira Link

CALCITE-7479

Changes Proposed

[xiedeyantu]

Just draft.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
93.3% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[mihaibudiu]

This looks fine.
How come it is never applicable?
Are all our tests using optimal group by keys?
Or we just don't have many tables with keys?

[xiedeyantu]

This looks fine. How come it is never applicable? Are all our tests using optimal group by keys? Or we just don't have many tables with keys?

"How come it is never applicable?" I didn't quite understand the meaning of this sentence. Because the functional dependencies we currently support are limited, I used the primary key as an example.

[mihaibudiu]

No other tests were modified

[xiedeyantu]

I'm concerned about whether this rewrite is meaningful, as it requires the introduction of
"single_value".

[mihaibudiu]

I think there are other uses of SINGLE_VALUE in the compiler.

[xiedeyantu]

No other tests were modified

Because this new rule is not set as a "default" rule, it doesn't have any other impact.

[mihaibudiu]

Can you make it default and see what happens?

[xiedeyantu]

I think there are other uses of SINGLE_VALUE in the compiler.

Are you suggesting that I don't need to worry too much about "single_value"? Can the compiler do more?

[xiedeyantu]

Can you make it default and see what happens?

Of course, I can create a new branch to add it as a default rule and observe the outcome later.

[mihaibudiu]

Doesn't even have to be a branch, you can run the experiment and report the results.

[mihaibudiu]

There are lots of uses of SINGLE_VALUE, I think that one is fine.
I am more worried about whether the functional dependency is sufficiently conservative.