search

snoyberg / classy-prelude

A typeclass-based Prelude.
108 stars 15 forks source link

singleton and Set #87

Open gibiansky opened 9 years ago

gibiansky commented 9 years ago

I would like to be able to use singleton with Set a. This is exported by Data.Set, but can't be implemented using classy-prelude because singleton requires IsSequence, which Set isn't. Is this possible some other way?

gregwebs commented 9 years ago

Do you want to use the name singleton or can you use opoint?

gibiansky commented 9 years ago

I'm currently using opoint, but I'm wondering whether there is a design which would allow singleton to be used here. (As someone who isn't fluent in mono-traversable, the names for the typeclass MonoPointed and function opoint are not terribly intuitive; why not just use singleton?)

(Also, I would not have guessed to use opoint if I had not looked at the source of singleton. Though maybe that's documented somewhere, Hackage is just not displaying recent documentation...)

gregwebs commented 9 years ago

The concept here is probably that singleton implies a container of multiple items, so the IsSequence constraint may have been used to make that sensible. But that is limiting.

It would make sense to me to loosen the constraint to (MonoFoldable, MonoPointed), since MonoFoldable contains toList. Although that really means loosening it to MonoPointed and being the same as opoint since MonoPointed requires MonoFoldable for its law.

gibiansky commented 9 years ago

I'm afraid I don't really understand the naming behind opoint and MonoPointed. Why not singleton and CanSingleton? Alternatively, why not just singleton instead of opoint (MonoPointed doesn't really matter so much, it's just the typeclass)

gregwebs commented 9 years ago

CanSingleton is the old naming of ad-hoc typeclasses. MonoPointed follows a law: otoList . mconcat . map opoint == id The pointed concept has a history in Haskell, but I am not opposed re-naming it to singleton (and MonoSingleton.

Would you would be happy with singleton :: MonoPointed seq => Element seq -> seq, or is that not the case?

gibiansky commented 9 years ago

I think the name of the typeclass doesn't matter too much, but having it be named "singleton" is important. Do yes, I think that would be fine.

As an side, could you briefly explain (or give a resource) as to the "pointed" naming? MonoSingleton is more intuitive imi, but didn't really matter for the type class name.

Thanks! On Oct 10, 2014 7:01 PM, "Greg Weber" notifications@github.com wrote:

CanSingleton is the old naming of ad-hoc typeclasses. MonoPointed follows a law: otoList . mconcat . map opoint == id The pointed concept has a history in Haskell, but I am not opposed re-naming it to singleton (and MonoSingleton.

Would you would be happy with singleton :: MonoPointed seq => Element seq -> seq, or is that not the case?

— Reply to this email directly or view it on GitHub https://github.com/snoyberg/classy-prelude/issues/87#issuecomment-58734568 .

gregwebs commented 9 years ago

Haskell tends towards category names rather than intuitive ones :). The current classy-prelude is mostly about using existing abstractions and names. But given that the pointed abstraction has almost no usage in the community we could certainly try to change the name.

https://hackage.haskell.org/package/category-extras-0.53.1/docs/Control-Functor-Pointed.html

snoyberg commented 9 years ago

Currently, mono-traversable and classy-prelude provide a function singletonSet for this use case. This arose from trying to address three different use cases: Sequence/Functor instances like Vector or ByteString, Sets, and Maps. However, it seems like we now have a way (via MonoPointed) to combined the first two cases (Map is still inherently different). I'm not opposed to changing singleton to just being a nicer name for opoint.