coefnames should always return a tuple no mattter the number of terms

[oxinabox]

The number of terms should not change the type of coeffnames,
it should always be a tuple of tuples

1 RHS term

df = (y=1:10, a = 1:10, b=1:10);
f = @formula(y ~ a);
f = apply_schema(f, schema(f, df));

gives:

julia> coefnames(f)
("y", "a")

2 RHS terms

df = (y=1:10, a = 1:10, b=1:10);
f = @formula(y ~ a + b);
f = apply_schema(f, schema(f, df));

gives

julia> coefnames(f)
("y", ["a", "b"])

[matthieugomez]

I agree with this. A related suggestion is that the RHS of the formula could always return a tuple of terms no matter the number of terms. The current situation makes it hard to iterate over the terms on the RHS of the formula for instance.

[kleinschmidt]

I'm also coming around to this. Although I'd suggest it should be a Vector and that it also goes hand-in-hand with the question of whether you shoudl always return an AbstractMatrix from modelcols (then the coefnames correspond to the labels of the second dimension of whatever modelcols gives you.