It doesn't have a ticket yet, but I would propose that kind annotations were adapted. I have been bitten on a couple of occasions (working with HO type variables) by kind inference putting my type vars into *, and although there is usually a work-around, it is usually mesmerisingly ugly. Stefan
On Fri, Jan 27, 2006 at 01:30:02PM +0000, S.M.Kahrs wrote:
It doesn't have a ticket yet, but I would propose that kind annotations were adapted.
I have been bitten on a couple of occasions (working with HO type variables) by kind inference putting my type vars into *, and although there is usually a work-around, it is usually mesmerisingly ugly.
Do you have an opinion on the alternatives discussed under KindInference?
Do you have an opinion on the alternatives discussed under KindInference?
I probably favour polymorphic kinds, as it is the most elegant solution. (And I agree that in this case kind declarations would not be needed.) What particularly strikes me about the other two proposals is that when I first looked at them I was unsure in what way they'd differ. And that suggests that they are too complicated. In some sense, I'd rather see something toned down, exploiting type definitions only, but being augmented with kind annotations. Stefan
Ross Paterson:
On Fri, Jan 27, 2006 at 01:30:02PM +0000, S.M.Kahrs wrote:
It doesn't have a ticket yet, but I would propose that kind annotations were adapted.
I have been bitten on a couple of occasions (working with HO type variables) by kind inference putting my type vars into *, and although there is usually a work-around, it is usually mesmerisingly ugly.
Do you have an opinion on the alternatives discussed under KindInference?
But we should probably ensure that, if the kind inference proposals don't make it, we revisit the possibility of having kind annotations. (Not sure what the best way is to ensure this in our process. Isaac?) Manuel
Manuel M T Chakravarty
But we should probably ensure that, if the kind inference proposals don't make it, we revisit the possibility of having kind annotations. (Not sure what the best way is to ensure this in our process. Isaac?)
Wiki page and ticket added. (Remember that anyone can add a ticket, if they want to ensure that an idea doesn't get forgotten or overlooked.) Regards, Malcolm
participants (4)
-
Malcolm Wallace -
Manuel M T Chakravarty -
Ross Paterson -
S.M.Kahrs