[Haskell-cafe] Any precedent or plan for guaranteed-safe Eq and Ord instances?

[Stijn van Drongelen]

On Wed, Oct 2, 2013 at 4:17 PM, Tom Ellis <
tom-lists-haskell-cafe-2013 at jaguarpaw.co.uk> wrote:

> What's the benefit of this requirement, as opposed to, for example
>
>    False <= _ = True
>

I was trying to cover for void types, where the only sensible definitions
are

    instance Eq Void where
        _ == _ = error "void (==)"

    instance Ord Void where
        _ <= _ = error "void (<=)"

This is because reflexivity must be guaranteed, so undefined == undefined
may not yield False, but I doubt error "foo" == (let x = x in x) should
yield True either. But perhaps this exception deserves its own rule.

On Wed, Oct 2, 2013 at 5:36 PM, Roman Cheplyaka <roma at ro-che.info> wrote:

> * Stijn van Drongelen <rhymoid at gmail.com> [2013-10-02 15:46:42+0200]
> > I do think something has to be done to have an Eq and Ord with more
> strict
> > laws.
> >
> > * Operators in Eq and Ord diverge iff any of their parameters are bottom.
>
> This outlaws the Eq instances of lists, trees, and other (co)recursive
> types.
>
> Furthermore, in this formulation, even Eq for tuples is illegal, because
>
>   (undefined, something) == somethingElse
>
> is going to diverge.
>
> Roman
>

I knew this was going to bite me in the ass. Let me try again:

* Operators in Eq and Ord may only diverge when any of their parameters are
bottom.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20131002/7c35cb71/attachment.htm>