gen_degopt_poly(degopt) -> graph_degopt(degopt)

[jarlebring]

The degree optimal polynomial seems so central with the data structures in this package that we can turn it into a constructor for Compgraph instead.

The code:

  (graph,crefs)=gen_degopt_poly(d::Degopt;input=:A)

can be replaced with

g=Compgraph(degopt)
cref=get_degopt_crefs(g);

There are variants of gen_degopt_poly which are very rarely used. I suggest we remove those and replace wherever it is used:

• (graph,crefs)=gen_degopt_poly(x,z;input=:A) => Replace code using this version with create a Degopt
• (graph,crefs)=gen_degopt_poly(k;T=ComplexF64,input=:A) => Introduce a function Degopt(k,T=ComplexF64)

[eringh]

I see the point to emphasize the Degopt more. But will it not be confusing that the interfaces are not similar?
Should we then also change interfaces such that gen_monomial_degopt(a; input=:A), and analogous conversions, returns a Degopt-type rather than (graph,cref)?

[jarlebring]

Okay. My comment comes partially from my feeling now that gen_X_poly is maybe not the best function name, in particular gen_degopt_poly sounds like we generate a Degopt when it is in fact the opposite. It is also easily confused with code generation. Changing all the gen_-functions to graph_ would maybe make more sense. The function graph_degopt_poly() does say more what it is about. Other options?

[eringh]

Sounds good to me!

[jarlebring]

I also changed gen_degopt_poly to graph_degopt. Poly is implicit. 0ef2384