Add a real type

[lilizoey]

Currently a lot of struct and functions in rust are hard-coded to use f32s. Such as Vector2, Vector3 etc. But in godot these structs and functions may be either float or double depending on whether godot is compiled with precision=double or not.

In godot the type used is called real_t and will be either a float or double. We could do the same in our bindings. As it is, our bindings will likely break if we try to use it with a godot that is compiled with precision=double enabled.

The float type in godot, and thus the one used by variant, is always a double, see: https://docs.godotengine.org/en/stable/classes/class_float.html.

[Bromeon]

godot-cpp could also serve as an inspiration here.

[ttencate]

This is what godot-cpp does:

https://github.com/godotengine/godot-cpp/blob/9d1c396c54fc3bdfcc7da4f3abcb52b14f6cce8f/include/godot_cpp/core/defs.hpp#L102-L106

This REAL_T_IS_DOUBLE is set manually through scons when compiling godot-cpp, using a precision=double option just like the engine itself:

https://github.com/godotengine/godot-cpp/blob/2b7094f342beb11166f86a6309bba86fe4b21838/SConstruct#L207-L208

In our case, maybe we could even detect it from the Godot binary and generate the appropriate code?

[lilizoey]

In our case, maybe we could even detect it from the Godot binary and generate the appropriate code?

i couldn't really find a way to reliably detect it from the godot binary, but if you can find something that'd be useful! currently i've just added a feature called precision-double in #149.

[Bromeon]

Regarding detection, I got this answer from Godot devs:

There are feature tags called single and double that you can add to your *.gdextension file to make sure Godot loads the correct library

If you want to check it at runtime, then something like OS::get_singleton()->has_feature("double")
should probably work
Although I guess you probably wouldn't get to that point, if you try to do it from a library expecting a single-precision ABI while using a double-precision engine build or vice versa, so using the feature tags and separate libraries are probably your best bet