Cognitive Complexity PR, re-gitted

[felix91gr]

A continuation of the work done in #3963 due to git conflicts.

References:

• A significant portion of the conversation surrounding this lint has happened at Suggestion for new lint: Cognitive Complexity #3793 as well.
• Previous PR (Cognitive Complexity (step 2 (and maybe 3) out of 3+): implementation! #3963)
• The original paper

Overall todo:

• Remove the ICE described by Manish.

• Decide if and how to score the constructs that are not being scored right now at all (most of them were not considered in the paper and so I thought that maybe there could be a default, low value, for non-specified kinds of expressions. Either do that, or just skip them for now and expand upon them in later versions).

• Roll back the implementation of flexible scores (done in one of the last commits of the original PR). This is because we want to keep this PR as simple as possible, and steering too far out of where the paper scores things will make future expansions harder to do in the right direction. For now, the minimal, the better.

• Write the complete specification for what's been implemented (the Draft should be a good base, and the original paper should be used as a reference). This should end up in the docs directory of Clippy, I think.

• Finish writing up test cases and tidy up the test folder for this lint.

Overall already done:

• Implemented the entire spec described in the Draft.
• Changed Lint from Late to Early pass.
• Downgraded Lint to Nursery (in preparation for next step) (thanks Manish!)

[felix91gr]

@vityafx I think this is much better now as a base. You should be able to base your work on it.

It doesn't pass the dogfood test (which is basically Clippy linting itself) but that's okay, I should be able to fix those soon. The code is, or should be, pretty much sound already.

[felix91gr]

@Manishearth there's a new ExprKind, Let. You also left a comment for me about the if let statements:

XXXManishearth don't count complexity of if let
(For the IF-Case)

I'm guessing they're now being treated differently in the AST? What were the changes made to the If nodes, and what do the Let nodes cover?

(iirc there was work being done in the syntax so that one could chain if-lets together or something, right? Do you happen to know which RFC it was, and if it is that one that caused the change?)

[Manishearth]

Yeah so now if let is represented as an If block with a Let condition

[felix91gr]

Ohh. And this is the documentation for ExprKind::Let as well:

A let pat = expr expression that is only semantically allowed in the condition of if / while expressions. (e.g., if let 0 = x { .. }).

So, that ExprKind is only generated in the context of an if let or while let. Right?

[Manishearth]

yep

[felix91gr]

Awesome, Manish, thank you.

[felix91gr]

Btw I talked it over with Oliver and we've decided to keep this as close to the original paper as possible. Therefore syntax nodes that aren't sufficiently described there will not score at all. This is so as to make this PR as simple as possible, and to facilitate future expansions upon the lint.

[bors]

☔ The latest upstream changes (presumably #4523) made this pull request unmergeable. Please resolve the merge conflicts.

[felix91gr]

Please go here to the lint's discussion, to see why I'm closing the PR 🙂