Ignore HYPHEN-MINUS (dash) in arg_enum!()

[kbknapp]

From @behnam on November 7, 2017 22:3

Ref: https://docs.rs/clap/2.27.1/clap/macro.arg_enum.html

Would be great to support ignoring ASCII dash character (HYPHEN-MINUS) in the FromStr implementations inside the arg_enum!() macro.

The HYPHEN-MINUS character cannot be part of an enum variant name because of the limits of the language, and therefore, this won't be a breaking change to the current behavior, but only expanding acceptable values for values using an arg_enum-defined type.

Optionally, we would also do so for UNDERSCORE (U+005F LOW LINE), since it's recommended to not use them for Enum variant names. BUT, there are cases where it exists.

Alternate approach

Except alternate strings to match per-variant, so "some-bar" can be defined as an alternate string form for SomeBar variant explicitly.

This approach gives users all the control, but makes it too verbose for simple, common use-cases.

Another alternate approach

Having both solutions above, as they don't overlap: staying agile and flexible.

---

Rust Version

rustc 1.23.0-nightly (e340996ff 2017-11-02)

Affected Version of clap

2.27.1

Expected Behavior Summary

Running command with --foo some-bar should match Foo::SomeBar.

Actual Behavior Summary

Only --foo somebar matches Foo::SomeBar and --foo some-bar results in an error.

Steps to Reproduce the issue

Add the above example to a clap app file.

Sample Code or Link to Sample Code

arg_enum!{
    #[derive(Debug)]
    pub enum Foo {
        SomeBar,
        AnotherBaz
    }
}

Debug output

N/A

Copied from original issue: #1098

[kbknapp]

I would be hesitant to simply ignore - when matching as this could lead to typos, plus would cause us to match Foo::SomeBar to s-omebar, so-mebar, som-ebar, etc.

I would consider Foo::Some_Bar allowing foo_bar or foo-bar and this would be acceptable. Thoughts?

[kbknapp]

From @behnam on November 9, 2017 19:52

Well, I see your point, @kbknapp.

I also like to point out that case-insensitive matching already has a similar issue, where value AbcDef is equal to AbCedf.

Ideally, instead of case-insensitive matching, we would use a case-aware matching, converting the input into CamelCase and perform exact match on the variant names.

to_camel_case() is provided by https://crates.io/crates/case and similar crates.

I'm not sure how much people rely on the current behavior and how far we can go with breaking compatibility. If we don't want to break anything, maybe we can do both in parallel: case-insensitive AND case-aware matching.

This would address your concern, while not expecting people to use underscores in variant names. And is backwards-compatible.

[kbknapp]

I'm going to move this over to the clap-derives]() repo as I intend on this being solved in 3.x via CustomDerive which will replace this macro call.

[pksunkara]

Default is kebab case. You can also rename the case into others. Also, there are aliases if renaming is not enough.