[darcs-devel] darcs patch: make Hopefully have type witnesses

[Jason Dagit]

On Fri, Feb 15, 2008 at 10:58 AM, David Roundy <droundy at darcs.net> wrote:
> On Fri, Feb 15, 2008 at 10:18:04AM -0800, Jason Dagit wrote:
>  > Sun Feb 10 01:00:34 PST 2008  Jason Dagit <dagit at codersbase.com>
>  >   * make Hopefully have type witnesses
>
>  Thanks! This looks good, but I've got one question...
>
>  >  \begin{code}
>  > -get_common_and_uncommon :: RepoPatch p => (PatchSet p C(x),PatchSet p C(y)) ->
>  > -                           ([PatchInfo],(RL (RL (PatchInfoAnd p)) :\/: RL (RL (PatchInfoAnd p))) C(x y))
>  > -get_common_and_uncommon_or_missing :: RepoPatch p => (PatchSet p C(x),PatchSet p C(y)) ->
>  > -                                      Either PatchInfo ([PatchInfo],(RL (RL (PatchInfoAnd p)) :\/: RL (RL (PatchInfoAnd p))) C(x y))
>  > +get_common_and_uncommon :: (RepoPatch p, (Invert (PatchInfoAnd p)))
>  > +                        => (PatchSet p C(x),PatchSet p C(y))
>  > +                        -> ([PatchInfo],(RL (RL (PatchInfoAnd p)) :\/: RL (RL (PatchInfoAnd p))) C(x y))
>  > +get_common_and_uncommon_or_missing :: (RepoPatch p, (Invert (PatchInfoAnd p)))
>  > +                                   => (PatchSet p C(x),PatchSet p C(y))
>  > +                                   ->  Either PatchInfo ([PatchInfo],(RL (RL (PatchInfoAnd p)) :\/: RL (RL (PatchInfoAnd p))) C(x y))
>
>  Why is this Invert instance required here? It seems to me like the compiler
>  ought to be able to infer from RepoPatch p that (Invert (PatchInfoAnd
>  p))...
>
>  I haven't actually looked into why this instance isn't inferred, since I
>  thought I'd run it by you first.

I actually don't know either.  I did have to change some instances in
HopefullyPrivate.lhs.  Maybe doing so in there requires that I change
the instance of RepoPatch.

>From the sounds of it, RepoPatch is intended to take care of this
Invert instance?  I'll look at fixing that.

Jason