The TypeClass type class

[Lars Hupel]

Dear lists,

this mail summarises some plans from NE Scala to bring shapeless and
other libraries (scalaz, …) closer together.

*The problem*

You want to derive type class instances for case classes. Imagine you
have a couple of types like:

case class Foo(i: Int, s: String)

and now you want a Monoid for that. Easy enough to define, but you don't
want all that boilerplate.


*Current state of the affairs*

If you were using shapeless, you'd define some isomorphisms, and it'd
derive your stuff automatically, as demonstrated in [1]. However, this
example defines

1) its own Monoid class
2) an implicit to derive a Monoid for HNil
3) an implicit to derive a Monoid for HCons
4) an implicit to derive a Monoid using an Iso


Luckily, there are already two libraries defining Monoids [2,3]. Now, we
just have to find out how to abstract over steps 2 to 4.

Now, Scalaz to the rescue [4]: It contains a type class abstracting over
precisely those things. It has a couple of drawbacks, though:

- it uses Scalaz HLists, which lack all those heavy machinery like
deriving isomorphisms between case classes and HLists
- for some reason, deriving the instances automatically doesn't work
(compiler crashes)

Not a very great rescue, unfortunately.


*How can this be fixed?*

Put this abstraction – a type class called "TypeClass", since it's a
type class abstracting over type classes – into shapeless. I created a
branch in my fork of shapeless, based on the release of 1.2.3 for Scala
2.10 [5,6].

Users who want to derive a Monoid for their case classes now have to do
the following:

1) Define a `TypeClass` instance.
2) Publish an `Iso`.
3) ???
4) PROFIT!

Luckily, we can get rid of step 1 by putting this sort of stuff into a
3rd-party-library, called "shapeless-contrib". It would contain all
sorts of instances for Scalaz and Spire, readily to use for your
convenience.


*What needs to be done?*

I already talked with Miles about the creation of some sort of stable
branch (maybe 1.2.x or even 1.3.x), renaming the macro-paradise efforts
to 2.x [7]. Then, I'll put together a small pull request for my changes
against that branch. Once all that is done, I'll create a new project
which depends on shapeless, scalaz, spire, …

In the end, using your case classes as monoids will require a couple of
library dependencies in your build and some imports in your code. Does
that sound good?

Regards
Lars


[1]
<https://github.com/milessabin/shapeless/blob/master/examples/src/main/scala/shapeless/examples/monoids.scala>

[2] <https://github.com/scalaz/scalaz/tree/scalaz-seven>

[3] <https://github.com/non/spire>

[4]
<https://github.com/scalaz/scalaz/blob/scalaz-seven/typelevel/src/main/scala/scalaz/typelevel/TypeClass.scala>

[5] <https://github.com/larsrh/shapeless/compare/scala-2.10...a500e3f>

[6] Side note: I think for this matter, "2.10 only" is a good choice
since it's stable. It might be possible to support 2.9 too, since my
approach doesn't use any macros (yet :-)).

[7] Since I don't follow the development of shapeless closely, I have no
idea how the branches are organised right now. Miles, can you comment on
that? Maybe 'master' would be the most appropriate branch.

[Runar Bjarnason]

That actually sounds really awesome.

Nicely done!

> --
> You received this message because you are subscribed to the Google Groups "scalaz" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to scalaz+un...@googlegroups.com.
> To post to this group, send email to sca...@googlegroups.com.
> Visit this group at http://groups.google.com/group/scalaz?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Miles Sabin]

On Mon, Feb 18, 2013 at 5:35 AM, Lars Hupel <hu...@in.tum.de> wrote:
> this mail summarises some plans from NE Scala to bring shapeless and
> other libraries (scalaz, …) closer together.

I approve this message :-)

Cheers,


Miles

--
Miles Sabin
tel: +44 7813 944 528
skype: milessabin
gtalk: mi...@milessabin.com
g+: http://www.milessabin.com
http://twitter.com/milessabin

[Alois Cochard]

Sounds like a plan Lars :)

Great stuff!

Alois