Why does Option not extend the Iterable trait?

On Fri, Apr 09, 2010 at 04:00:06PM +0000, christopher marshall wrote:
> I asked this on StackOverflow but there have been no answers. Option
> in 2.8 even declares an iterator method but it does not implement
> Iterable, instead relying on an implicit conversion. Is this for
> reasons of conceptual purity, or is there some practical issue with
> it?

Option is not a sufficiently general container to fully interoperate
with the other collections.

scala> Seq(1) flatMap (_ => Iterable(1, 2))
res0: Seq[Int] = List(1, 2)

scala> Some(1) flatMap (_ => Iterable(1, 2))

On 04/09/10 14:04, Paul Phillips wrote:
> On Fri, Apr 09, 2010 at 04:00:06PM +0000, christopher marshall wrote:
>> I asked this on StackOverflow but there have been no answers. Option
>> in 2.8 even declares an iterator method but it does not implement
>> Iterable, instead relying on an implicit conversion. Is this for
>> reasons of conceptual purity, or is there some practical issue with
>> it?
>
> Option is not a sufficiently general container to fully interoperate
> with the other collections.
>
> scala> Seq(1) flatMap (_ => Iterable(1, 2))
> res0: Seq[Int] = List(1, 2)
>
> scala> Some(1) flatMap (_ => Iterable(1, 2))
>
= res0 :) // same blocks passed
Seriously though,

Some(x) : Option[T] flatMap ~> Seq[T] = List(x)
None : Option[T] flatMap ~> Seq[T] = List()

But as we've learned before with matz & POLS, the S is both subjective and
usually refers to the powers that be :-)

-Martin

On Fri, Apr 09, 2010 at 02:14:01PM -0400, Martin S. Weber wrote:
> Some(x) : Option[T] flatMap ~> Seq[T] = List(x)
> None : Option[T] flatMap ~> Seq[T] = List()

Now Option has ceased to be a monad.

> But as we've learned before with matz & POLS, the S is both subjective
> and usually refers to the powers that be :-)

Yes, I would find it rather surprising if Option's monadic methods
starting returning Lists. POLS may be subjective but programming isn't
all about avoiding surprises: there are more important things, like
consistency of theoretical underpinnings.

This in itself I don't see the problem with. For instance:   scala> Map(1 -> 2, 3 -> 4) flatMap ( _ => Map('a -> 'b))
res44: scala.collection.immutable.Map[Symbol,Symbol] = Map(('a,'b))   scala> Map(1 -> 2, 3 -> 4) flatMap ( _ => Iterable('a, 'b'))
res45: scala.collection.immutable.Iterable[Any] = List('a, b, 'a, b)

The problem I do see is that CanBuildFrom can't help here. It would require overload, but one can't overload based a passed function's type parameters. I take your point, though.
On Fri, Apr 9, 2010 at 3:04 PM, Paul Phillips <paulp@improving.org> wrote:

On Fri, Apr 09, 2010 at 04:00:06PM +0000, christopher marshall wrote:
> I asked this on StackOverflow but there have been no answers. Option
> in 2.8 even declares an iterator method but it does not implement
> Iterable, instead relying on an implicit conversion. Is this for
> reasons of conceptual purity, or is there some practical issue with
> it?

Option is not a sufficiently general container to fully interoperate
with the other collections.

scala> Seq(1) flatMap (_ => Iterable(1, 2))
res0: Seq[Int] = List(1, 2)

scala> Some(1) flatMap (_ => Iterable(1, 2))
<your answer here>

--
Paul Phillips      | Atheists dig the best foxholes.
In Theory          |
Empiricist         |
up hill, pi pals!  |----------* http://www.improving.org/paulp/ *----------

--
Daniel C. Sobral

I travel to the future all the time.

Thanks - great answer. It seems to re-inforce Tony Morris' point about subclassing

> Date: Fri, 9 Apr 2010 11:18:32 -0700
> From: paulp@improving.org
> Subject: Re: [scala-user] Why does Option not extend the Iterable trait?
>
> On Fri, Apr 09, 2010 at 02:14:01PM -0400, Martin S. Weber wrote:
> > Some(x) : Option[T] flatMap ~> Seq[T] = List(x)
> > None : Option[T] flatMap ~> Seq[T] = List()
>
> Now Option has ceased to be a monad.


Get a free e-mail account with Hotmail. Sign-up now.