impl rename_all for #[derive(Display)] (#216)

[ModProg]

Resolves #216

Checklist

• Documentation is updated (if required)
• Tests are added/updated (if required)
• CHANGELOG entry is added (if required)

[tyranron]

@ModProg hmmm... I don't see why rename_all argument should allow implicit naming for other traits rather than Display only? Could you elaborate on that?

For me, the situation assert_eq!(format!("{:o}", Enum::VariantOne), "variant_one"); seems quite awkward.

[ModProg]

My idea was you could use different display traits for different ways of rename_all if you want to.

[tyranron]

@JelteF what do you think on this? I don't particularly like this idea of abusing different fmt traits for different naming, but maybe it's just me?

[ModProg]

I have seen crates use different format traits for stuff like this in the past, but I don't mind removing the functionality. i.e. make rename_all only available for Display.

[JelteF]

I agree that this is weird: assert_eq!(format!("{:o}", Enum::VariantOne), "variant_one");

But I also think assert_eq!(format!("{:o}", Enum::VariantOne), "VariantOne"); is weird. And as I understand that's what we do currently. I would be happier to see the Octal derive simply fail for unit enum variants without an #[octal(...)] attribute. Instead of defaulting to stringifying them in the same way as we do by default for Display, but that would be a breaking change. And doing another major release to fix that doesn't seem worth it.

So my vote would be, let's do whatever is easiest code wise. And let's create an issue that we add to the 3.0 milestone, to remove this weird behaviour completely then.

[tyranron]

@ModProg instead of modifying the ContainerAttributes with an additional slot (which is useless for the Debug), I think it's better to leave it "as is", and:

1. Move RenameAllAttribute into display module.
2. Inside display module do:

   struct ContainerAttributes {
       rename_all: RenameAllAttribute,
       common: super::ContainerAttributes,
   }

   and parse it as Either<RenameAllAttribute, super::ContainerAttributes> to omit repeating parsing of the super::ContainerAttributes twice.

[ModProg]

I thought It'd make sense to allow to have rename_all and bounds in the same attribute. But that will only at most save one line anyway.