Rename field of Complements

[blegat]

The field of the Complements set should be renamed as it does not satisfy the convention that it is equal to MOI.dimension.
See #947

[matbesancon]

is there a possible name for the replacement? output_dimension / half_dimension?

[odow]

I guess another options to define

struct Complements <: AbstractVectorSet
    dimension::Int
    Complements(d::Int) = new(2 * d)
end

[matbesancon]

Can't that be tricky for users (thinking that the argument passed as a dimension is the same as accessed as c.dimension)

[odow]

Currently people write

@constraint(model, [x, y] in Complements(1))

so I don't mind what happens so long as that doesn't change.

To be honest, I'm not really sure that this change is critical. Sure it's a convention, but Complements is a weird case that doesn't have to follow convention. I would be in favor of closing this issue with no change.

(If anything, we should rename the dimension(::AbstractSet) function to be output_dimension, but that boat has sailed.)

[odow]

Closing because I don't think this needs to change. The usage is documented, and complements is a weird set. Users shouldn't be accessing .dimension anyway, they should access MOI.dimension(set).

[odow]

This inconsistency annoyed me here: #1332. I changed my mind.