[Haskell] An Alternative Data.List.Zipper

[David Menendez]

On Sat, Jan 17, 2009 at 9:13 PM,  <roconnor at theorem.ca> wrote:
> On Sat, 17 Jan 2009, David Menendez wrote:
>
>> instance Applicative f => Applicative (Backwards f) where
>>   pure = B . pure
>>   f <*> x = B (unB f <**> unB x)
>
> probably should be f <*> x = B (unB x <**> unB f)

I always get that backwards.

> anyhow, this should be part of Control.Applicative.

I agree. But until it is, we'll have to make do.

>> This may be terminological confusion. I would have said that a
>> PointedList *is* a zipper.
>
> This is what I thought, and what how I initally advised Jeff.  However after
> carefully reading both <http://en.wikibooks.org/wiki/Haskell/Zippers> and
> <http://www.cs.nott.ac.uk/~ctm/Dissect.pdf> (Clowns to the left of me,
> jokers to the right), it is clear that formally a zipper focuses on a
> sub-list / sub-tree rather than focusing on a particular node of a list or a
> tree.  Hence zippers are not comonads.

Funny, I was just reading that paper earlier today.

You're right. I wasn't making enough of a distinction between zippers
and structures with one selected element. It's the latter which are
comonads.

-- 
Dave Menendez <dave at zednenem.com>
<http://www.eyrie.org/~zednenem/>