Variant Refactor

[PadraigK]

Here is some refactoring in advance of adding back the Signal macro.

The core idea is to leverage the type system a bit more, which lets us reduce the amount of code we're generating and make it possible to use protocols as generic constraints in the Signal macro stuff (coming soon).

By marking all types that can be stored directly in a native Godot Variant as VariantRepresentable we can add a single init(_ variant: some VariantRepresentable) method to Variant, instead of needing a separate initializer for each type. This gives us a contract where we (and the type-checker) know that any VariantRepresentable type can travel in and out of a Variant.

Previously this two-way travel worked, but it was not expressed in a way that the type system could reason about it, so we needed to add things like toVariant() and an extra protocol to promise that a given type would implement toVariant().

I'll call out some of the good things here:

• public required init? (_ from: Variant) is no longer generated on each object, instead the VariantStorable protocol has a default implementation that can initialize any VariantStorable from a Variant.
• public func toVariant () -> Variant no longer needed and is removed from generation — instead you can just do Variant(vector)
• Many initializers are removed from Variant as they're now handled by init(_ variant: some VariantRepresentable)
• Conformances to VariantRepresentable and VariantStorable are not code-generated, they're just implemented as extensions. I think this is an improvement because the compiler can tell us about any problems instead of needing to generate the code first.

And here are some compromises that enable the good things above:

• init (content: ContentType) and init() are public and required.
• Various ContentType internals are public
  * VariantCollection < String > is no longer possible because String is not a direct 1:1 VariantRepresentable type, so these now generate as VariantCollection < GString >. There are still conveniences around String ←→ GString, so I am not sure if this is a problem or not. Update: I did some more work so that any VariantStorable types (including String) can be used anywhere a VariantRepresentable can, so this change is no longer necessary.

[PadraigK]

API Diffs here: https://gist.github.com/PadraigK/750bfccd51101109689fbe77edaa0a58

[PadraigK]

Hold off on reviewing/merging this please, I have a couple more improvements that address the compromises above.

[PadraigK]

Hold off on reviewing/merging this please, I have a couple more improvements that address the compromises above.

Ok, this is in better shape now, it no longer messes up VariantCollection<String>

[PadraigK]

Updated Diffs: https://gist.github.com/PadraigK/750bfccd51101109689fbe77edaa0a58

[migueldeicaza]

My initial take is that I like this patch very much.

I would love if you could review the existing documentation in SwiftGodot/Sources/SwiftGodot/SwiftGodot.docc/Variants.md as part of this PR, to make sure they do not contradict each other.

[PadraigK]

My initial take is that I like this patch very much.

I would love if you could review the existing documentation in SwiftGodot/Sources/SwiftGodot/SwiftGodot.docc/Variants.md as part of this PR, to make sure they do not contradict each other.

Thanks — I reviewed these and the other two spots you mentioned. Please take another look when you have a chance.

[migueldeicaza]

Thank you for your contribution!

[migueldeicaza]

One question: how does this pass a subclass of Wrapper/Object to a Variant?

[migueldeicaza]

Ok, figured that out - but it does not seem like it is possible to create Nil values right now:

Variant (Nil())

This produces:

Initializer 'init(_:)' requires that 'Nil' conform to 'VariantStorable'