Basis inconsistencies between GDScript and C++

[Jummit]

godot::Godot::print(Basis(Vector3(1, 2, 3), Vector3(4, 5, 6), Vector3(7, 8, 9)))
// prints 1, 2, 3, 4, 5, 6, 7, 8, 9
godot::Godot::print(Basis(Vector3(1, 2, 3), Vector3(4, 5, 6), Vector3(7, 8, 9)).y);
// prints 2, 5, 8

print(Basis(Vector3(1, 2, 3), Vector3(4, 5, 6), Vector3(7, 8, 9)))
# prints ((1, 4, 7), (2, 5, 8), (3, 6, 9))
print(Basis(Vector3(1, 2, 3), Vector3(4, 5, 6), Vector3(7, 8, 9)).y)
# prints (4, 5, 6)

This could be related to how basis printing works, or with how they are stored, but it affects calculations (GDScript code ported directly to C++ behaves differently, so it should definitely made more consistent.

If I'm missing something here, I'd be happy if someone could clear it up.

Related: godotengine/godot#14553

[Zylann]

Printing might be different, but behaviour should at least be the same. You could check that with some test cases to see if the functionality works. If it does, then we can fix the printing to be transposed. If it doesn't work, in doubt, we could re-port the whole class by copying Godot's over to make it up to date.
Although, it's also possible that GDScript exposes Godot's Basis differently than the way it's implemented internally, which could explain the difference.

[sheepandshepherd]

I think printing uses the same convention as GDScript already, even though operator String in Basis formats it without parentheses. The numbers are the same if you let Godot format it:

godot::Godot::print("{0}", Basis(Vector3(1, 2, 3), Vector3(4, 5, 6), Vector3(7, 8, 9)));

The linked issue and PR should have fixed x/y/z already too.

I assume the Basis(Vector3, Vector3, Vector3) constructor is what's causing the incorrect values in both C++ lines. The Basis is just being constructed with the wrong row/column convention, the one used in the core engine.

[BastiaanOlij]

I assume the Basis(Vector3, Vector3, Vector3) constructor is what's causing the incorrect values in both C++ lines. The Basis is just being constructed with the wrong row/column convention, the one used in the core engine.

If that is true (and from the above print statements I would conclude the same), that is a pretty serious bug

[DmitriySalnikov]

Basis(row0, row1, row2) is still broken. #533 seems to be working correctly.

[aaronfranke]

PR #859 updated Basis to match the engine core, so if there is an inconsistency then it also exists in the engine.

[akien-mga]

This issue is pretty old by now and things have changed in 4.0+, so closing.