[Haskell-cafe] type-class and subclasses

[lloyd allison]

thank you for the suggestions.
I obviously have some more reading to do(!) and experiments to try
regards
Lloyd
--

On Thu, 10 Apr 2025 at 18:42, Andreas Källberg <anka.213 at gmail.com> wrote:

> GADTs does indeed resolve the issue in this specific case, as in that case
> the constructor carries the evidence of the equality constraint. However if
> we for example wanted to add a constructor method to your pair class:
>
> class Pair pt where
>    -- …
>    mkPair :: u -> v -> pt u v
>
> then it would no longer work, as you no longer have a pt in the arguments
> that carries the equality proof between u and v.
>
> If you want it to work in that case, then one option is that you could add
> a constraint-kinded type parameter to your class, which lets instances
> specify custom constraints to the type variables:
>
> class Pair c pt where
>   mkPair :: c u v => u -> v -> pt u v
>
> instance Pair (~) Ptype where
>   -- …
>
> this can however lead to issues with ambiguous types, since users of the
> type class methods needs to know the constraint to decide which instance to
> use. We can resolve this either using functional dependencies or type
> families:
>
> class PairFunDep c pt | pt -> c where
>   mkPair :: c u v => u -> v -> pt u v
>
> instance PairFunDep (~) Ptype where
>   -- …
>
> class PairTypeFam pt where
>   type PairConstraint pt :: u -> v -> Constraint
>   mkPair :: PairConstraint pt u v => u -> v -> pt u v
>
> instance PairTypeFam Ptype where
>   type PairConstraint Ptype u v = u ~ v
>   -- …
>
> Regards,
> Anka
>
> > On 10 Apr 2025, at 04:33, lloyd allison <proflandy at gmail.com> wrote:
> > 
> > I have a general question:
> > I would like to have a type-class C in which types that are instances of
> C have some type parameters and then a subclass S of C where some of the
> type parameters of an instance of S are more constrained in some way --
> perhaps by being equal and/or being instances of some other class.
> >
> > I have an entirely artificial example (att.) that illustrates this
> situation and struggled (it's a while since I wrote much Haskell) to get it
> past the ghc type checker until stumbling across GADTs which do seem to do
> the trick, so I could leave it at that but...
> >  ...I am very far from sure that this is the correct way to look at the
> problem and am hoping for some illumination.
> >
> > regards
> > Lloyd.
> > --
> > No AI was used in composing this message.
> >
> >
> > <eg.hs>
> > _______________________________________________
> > Haskell-Cafe mailing list
> > To (un)subscribe, modify options or view archives go to:
> > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> > Only members subscribed via the mailman list are allowed to post.
>


-- 
No AI was used in composing this message.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20250416/fc2848e7/attachment.html>