New lint: suggest using the only remaining variant instead of a wildcard

[ebroto]

When matching an enum, if there is only one variant left, suggest using that instead of the wildcard pattern.

For example, for this enum:

enum E {
    First,
    Second,
    Third,
}

When matching it like this:

match e {
    E::First => {}
    E::Second => {}
    _ => {}
}

Suggest the following replacement:

match e {
    E::First => {}
    E::Second => {}
    E::Third => {}
}

The lint should be under the pedantic category.

Note that for #[non_exhaustive] enums the wildcard arm should be kept.

[covaii]

Hi! I think it is covered by the lint "wildcard enum match arm" implemented in matches.rs
It also suggests a fix with an arm consisting of the union of the remaining variants.

declare_clippy_lint! {
    /// **What it does:** Checks for wildcard enum matches using `_`.
    ///
    /// **Why is this bad?** New enum variants added by library updates can be missed.
    ///
    /// **Known problems:** Suggested replacements may be incorrect if guards exhaustively cover some
    /// variants, and also may not use correct path to enum if it's not present in the current scope.
    ///
    /// **Example:**
    /// ```rust
    /// # enum Foo { A(usize), B(usize) }
    /// # let x = Foo::B(1);
    /// match x {
    ///     A => {},
    ///     _ => {},
    /// }
    /// ```
    pub WILDCARD_ENUM_MATCH_ARM,
    restriction,
    "a wildcard enum match arm using `_`"
}

[flip1995]

Not quite. The wildcard_enum_match_arm will trigger for every use of _ => ..., while this lint should only trigger on the once, where only one variant is left to cover.

But thanks for pointing out, that a lint like this is already implemented. This means, that the only thing that has to be implemented here is if variants_left() == 1 { lint_new_lint() }.

[covaii]

I want to try to do that, if you don't mind

[flip1995]

That would be great, if you need help with anything don't hesitate to ask here or open a WIP PR.

[covaii]

I've implemented at least a starting version, but I have some questions.

1. The dogfood_clippy test doesn't pass on 18 matches. Should I fix them or should I turn off the lint? If I should turn it off, then what is the best way to do so? The cargo test output to see the places where the lints are spanned dogfood_clippy.txt
2. I've found a small typo in one of the lint examples. Is it ok to include the fix in the same PR where the lint is implemented?
3. I've named the lint as "redundant_wildcard_enum_match". Is it good enough or is it better to rename it to something else?

[flip1995]

1. Please fix all occurrences. If some cannot be fixed for some reason or another leave an allow there and let the reviewer take a look.
2. Yes, but put it in its own commit please.
3. Maybe match_wildcard_for_single_variants to point out the single variant part of the lint.