Add SigBoolean and SBoolean?

[paf31]

Add SigBoolean and SBoolean?

[gbaz]

The current genericBool instance seems fine to me? I'm not sure the advantage of adding another special case constructor here...

The idea was to try to keep the Signature/Spine datatypes as tight as possible to make writing generic functions easier with fewer cases...

[paf31]

Yes, but Signature doesn't give the type constructor name, so I can't deserialize Boolean in a special way in generic-foreign for example.

[paf31]

Also, I think it makes sense to make a special case of Boolean, like we do for every other Prim type. They are built in, after all. Boolean isn't implemented as an ADT.

[gbaz]

When you say, "signature doesn't give the type constructor name" -- isn't this what you're asking for?

toSignature _ = SigProd [{sigConstructor: "true",sigValues: []},
                         {sigConstructor: "false",sigValues: []}]

[paf31]

I'd like

toSignature _ = SigProd "Prim.Boolean" 
                          [{sigConstructor: "True",sigValues: []},
                           {sigConstructor: "False",sigValues: []}]

assuming it were a real ADT.

[gbaz]

gotcha: so really it sounds like fixing #8 is the general solution?

[paf31]

Yes, but this is the easier fix which doesn't need compiler changes. Moreover, I think this is a good change in its own right. Boolean looks like an ADT, but it's not, and why make a special case for every Prim type, but not Boolean? Generic should represent the internal structure of types, and it does, apart from this one case 😄

[gbaz]

Ok, I buy it. 👍

[paf31]

Cool :) I'll test some more and merge my PR then.