Missing bindings for GLib.Variant

[cameronwhite]

A full sweep is probably needed to make sure there are bindings for all the variant types, but the ones I found porting Pinta were:

• bool types: need Variant.Create(bool) and bool GetBool()
• more predefined VariantType constants would be convenient, currently only "s" and "v" have named constants
• dictionary types: the "a{sv}" variant type mapping strings to variants. It's only possible currently to construct an empty one via Variant.CreateEmptyDictionary()

[badcel]

@cameronwhite: I wonder if there should be more VariantType constants or if they should be removed altogether as those are only convenience. If the API is getting generated a user can create every type he wants to via a string. E.g.: VariantType.New("i");

Or if they should be static methods instead to avoid unnecessary memory allocation.

[cameronwhite]

I think at least some sort of constant is useful so that it's more discoverable - otherwise you have to look up the GTK docs to find the right string to use and might make a typo

[badcel]

With constant you mean a string constant?

So you write: VariantType.New(VariantType.String)?

We could have a static method instead:
VariantType.GetString() or VariantType.CreateString()

Or a static property:
VariantType.String

I think I would avoid caching a lot of instances. As new instances are created in all 3 scenarios I would probably not use the static property as it is not obvious that a new instance is created on every call.

What do you think of implementing a custom equality so that two VariantTypes are equal if they represent the same type (e.g. via type_equal)?

This should probably be a separate PR and not be part of #942.

---

Edit: the more i think about it the more I think having helper for the types is not useful as it is possible to create arbitrary complex types and the helpers would never cover all cases. This is probably the reason why there are no helpers in c, too. It is probably best to add extensive documentation to the class.

I think I would prefer a builder class (see #954).

[cameronwhite]

I think a separate build class is probably fine too, that also makes it more clearly separated as a convenience on top of the actual GTK API

[badcel]

A boolean variant can now be created:

gir.core/src/Tests/Libs/GLib-2.0.Tests/Records/VariantTest.cs

Line 12 in 807ef28

public void CanCreateBool(bool value)

In regard to VariantType #954 supersedes this.