refactor(optimizer): make ExprVisitor an object safe trait.

[chenzl25]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

• Related feat: support ExprVisitable for plan nodes #13510
• To implement ExprVisitable, we need to make ExprVisitor an object safe trait, which means we need to remove the fn merge.

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added test labels as necessary. See details.
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

[codecov]

Codecov Report

Attention: 23 lines in your changes are missing coverage. Please review.

Comparison is base (c89c99c) 68.37% compared to head (5be7ef7) 68.42%.
Report is 9 commits behind head on main.

Files	Patch %	Lines
src/frontend/src/expr/mod.rs	78.72%	10 Missing ⚠️
src/frontend/src/expr/pure.rs	66.66%	5 Missing ⚠️
src/frontend/src/expr/expr_visitor.rs	85.71%	3 Missing ⚠️
.../src/optimizer/plan_visitor/input_ref_validator.rs	76.92%	3 Missing ⚠️
...ontend/src/optimizer/plan_node/logical_sys_scan.rs	83.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #13512      +/-   ##
==========================================
+ Coverage   68.37%   68.42%   +0.04%     
==========================================
  Files        1511     1511              
  Lines      259862   259798      -64     
==========================================
+ Hits       177692   177756      +64     
+ Misses      82170    82042     -128     

Flag	Coverage Δ
rust	68.42% <84.76%> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[BugenZhao]

• Is merge and Result really a good practice of a visitor pattern? As all visit methods take &mut self, this seems to imply that one should maintain the states or results in self. In this way, the signature of merge could be like

  fn merge(&mut self, b: Self::Result);

  Or rather, we should never have a merge function?

• Once a method is overridden, users have to duplicate the default implementation there and merge does not help us for boilerplate reduction. Will this work somehow related to optimizer: what's the "correct" form of visitor? #13477?

[chenzl25]

Is merge and Result really a good practice of a visitor pattern?

Although some visitors, such as Has, can easily eliminate the need for merge by using a state, other visitors, like TableScanIoEstimator, might find it challenging to calculate the cost by visiting expressions with the use of only a state, because expressions structure also matters. Therefore, I believe this method still serves a purpose. However, I agree that using &mut self instead of &self for merge would be more appropriate.

[chenzl25]

• Once a method is overridden, users have to duplicate the default implementation there and merge does not help us for boilerplate reduction. Will this work somehow related to optimizer: what's the "correct" form of visitor? #13477?

As mentioned above, I think fn merge is mainly used to capture the expression structure instead of boilerplate reduction, so maybe we can use another PR to resolve that issue.

[BugenZhao]

However, I agree that using &mut self instead of &self for merge would be more appropriate.

Actually I suggest we either... 🥵

• Make visit take &self,
• or either remove merge but keep visit take &mut self

[chenzl25]

However, I agree that using &mut self instead of &self for merge would be more appropriate.

Actually I suggest we either... 🥵

• Make visit take &self,
• or either remove merge but keep visit take &mut self

I prefer to remove merge and return type if we have to.

[chenzl25]

TableScanIoEstimator

But as I mentioned before without a return type and merge function, visitors like TableScanIoEstimator could need to maintain a state by itself which is nontrivial.

[chenzl25]

cc @BugenZhao I removed the merge function right now. PTAL.

[BugenZhao]

LGTM. Thanks for doing this refactoring!

[xxchan]

I revisited #4520 (add return value to visitor) and #5058 (add merge). Still has no strong opinion for either way. So rubber stamp.

[chenzl25]

I revisited #4520 (add return value to visitor) and #5058 (add merge). Still has no strong opinion for either way. So rubber stamp.

Oh, I have already forgotten that both of them were introduced by myself. 🥵 Now I removed them all.

[st1page]

LGTM