Const enum variants names

[WaffleLapkin]

I tried to write code like this:

pub const LETTERS: &'static [&'static str] = Letter::variants();

#[derive(EnumVariantNames)]
pub enum Letter { A, B, C }

But I've found that variants fn is not const.

This pr changes the behaviour of EnumVariantNames in that way:

1. Add pub const VARIANTS: &'static [&'static str] = ... to enum
2. make variants fn const
3. deprecate variants fn with a message like "Use `Letter::VARIANTS` instead"

[hobofan]

deprecate variants fn with a message like "Use Letter::VARIANTS instead"

Why deprecate the function if you made the access via the function const too?

[WaffleLapkin]

Because I see no reasons to use the function, when you can use associated constant.

[Peternator7]

Thanks for the PR! Sorry I thought I responded to this last week, but it looks like I forgot to. Based on #74, I don't think we should deprecate the function because it will be extracted into a trait, and those can't be const yet.

Let's definitely get the associated constant added though :)

[WaffleLapkin]

Based on #74, I don't think we should deprecate the function because it will be extracted into a trait, and those can't be const yet.

So I must undeprecate variants and make it non-const again? Ok, I'll make these changes tomorrow :)

[Peternator7]

Thanks! Merging #74 created a conflict so here's the resolution you'll want in the enum variants file.

    quote! {
        impl #impl_generics ::strum::VariantNames for #name #ty_generics #where_clause {
            /// Return a slice containing the names of the variants of this enum
            #[allow(dead_code)]
            fn variants() -> &'static [&'static str] {
                Self::VARIANTS
            }
        }

        impl #impl_generics #name #ty_generics #where_clause {
            /// Names of the variants of this enum
            #[allow(dead_code)]
            pub const VARIANTS: &'static [&'static str] = &[ #(#names),* ];
        }
    }

[WaffleLapkin]

@Peternator7 shouldn't I add VARIANTS const to the trait?

[Peternator7]

@WaffleLapkin, that would be preferred, but if I remember correctly, const in traits isn't supported on stable yet :(

[WaffleLapkin]

@Peternator7 const in traits is supported on stable

[WaffleLapkin]

@Peternator7 I've added VARIANTS const to VariantNames trait and it seems like there is still no need in variants function... Am I wrong?

[WaffleLapkin]

@Peternator7 ping

[Peternator7]

Thanks for the ping; I agree. Let's go ahead and remove the variants function

[Peternator7]

Looks great! Thanks for the feature :)

[WaffleLapkin]

Yay! Btw, don't forget that these changes are breaking now ;)