2.8 and Partially Applying Scala type variables

On Sun, Feb 21, 2010 at 12:46:47PM -0800, Vlad Patryshev wrote:
> I've been using this trick:
> http://dibblego.wordpress.com/2008/09/22/partially-applying-scala-type-v...
> for
> a while.
>
> Seems like it does not work in 2.8; I wonder if it is intentional, and
> whether there's a solution.

It's usually helpful to define "does not work", but I see you did on the
web page. That it doesn't work as given is intentional, but you
shouldn't do it that way anyway. Give the wrapper a name and it will
work, something like:

class Joiner[M[_]] {
def apply[A](m: M[M[A]]): M[A] = error("todo")
}

def join[M[_]] = new Joiner[M]

As Paul points out, the trick still works with a named class. I have a
feeling the change was related to changes in the way structural types
are handled, but I'm not sure.

If you're interested in this trick, you're probably in the target
market for one of the tastier features of Scala 2.8: type inference
for type constructors:

def foo[M[_], A](m: M[M[A]]) = 0
// M and A are both inferred!
foo(List(List(1)))

-jason

On Sun, Feb 21, 2010 at 9:46 PM, Vlad Patryshev wrote:
> Hi,
> I've been using this
> trick: http://dibblego.wordpress.com/2008/09/22/partially-applying-scala-type-variables/ for
> a while.
> Seems like it does not work in 2.8; I wonder if it is intentional, and
> whether there's a solution.
>
> --
> Thanks,
> -Vlad
>

Yay!! Under what circumstances will this work?

Matt

On Sun, Feb 21, 2010 at 3:09 PM, Jason Zaugg wrote:
> As Paul points out, the trick still works with a named class. I have a
> feeling the change was related to changes in the way structural types
> are handled, but I'm not sure.
>
> If you're interested in this trick, you're probably in the target
> market for one of the tastier features of Scala 2.8: type inference
> for type constructors:
>
>  def foo[M[_], A](m: M[M[A]]) = 0
>  // M and A are both inferred!
>  foo(List(List(1)))
>
> -jason
>
> On Sun, Feb 21, 2010 at 9:46 PM, Vlad Patryshev wrote:
>> Hi,
>> I've been using this
>> trick: http://dibblego.wordpress.com/2008/09/22/partially-applying-scala-type-variables/ for
>> a while.
>> Seems like it does not work in 2.8; I wonder if it is intentional, and
>> whether there's a solution.
>>
>> --
>> Thanks,
>> -Vlad
>>
>

Vlad Patryshev wrote:
> Hi,
>
> I've been using this trick:
> http://dibblego.wordpress.com/2008/09/22/partially-applying-scala-type-v...
> for
> a while.
>
> Seems like it does not work in 2.8; I wonder if it is intentional, and
> whether there's a solution.
>
>
G'day Vlad, name the class instead of using structural types.

This is far less important now that higher kinds get inferred. Thanks
Adriaan!