Fix bug in check_method_implementation

[schillic]

• fix bug in check_method_implementation
• implement vertices_list for HPolytope >> see Vertices_list for HPolytope #189

[schillic]

@mforets: How should we implement vertices_list?
One idea: Implement the intersection for two Hyperplanes (result: a Hyperplane of lower dimension). Then intersect all defining hyperplanes of the polytope as long as possible until you arrive at 1D. Is this necessary for every (unordered) pair of constraints?

[mforets]

does the vertices list functionality depend on the "check_method_implementation"? how does that work?

[schillic]

It just checks if the method exists. When #189 is merged, it should work (we should still try).

[mforets]

do you refer to this test?

@test check_method_implementation(AbstractPolytope, vertices_list,
                                  Function[S -> (S{Float64},)])

[schillic]

Yes. I just rebased with master. If it gets green, you can merge.

[mforets]

the access of the method vertices_list requires Polyhedra (it's hidden under a @require macro, since we don't want it to be loaded at startup). Maybe we needs using Polyhedra before the test of this file?

[schillic]

Indeed:

julia> method_exists(vertices_list, (HPolytope{Float64},))
false

julia> using Polyhedra

julia> method_exists(vertices_list, (HPolytope{Float64},))
true

[schillic]

Unfortunately, when adding using Polyhedra, I get an error with Halfspace, see #194.

LoadError: UndefVarError: HalfSpace not defined

[schillic]

For now my fix is to load this test set last. We should still investigate #194 another time.

Another solution would be to provide a default implementation outside the @require which just throws an error. But then there might be a "redefinition" warning every time.