[Haskell-cafe] Alternative: you can fool many people some time, and some people many time, but...

David Menendez dave at zednenem.com
Thu Sep 29 21:14:53 UTC 2016

On Thu, Sep 29, 2016 at 4:28 PM, Jake <jake.waksbaum at gmail.com> wrote:

> I think this may have something to do with the default definition of many in the definition of Alternative <http://hackage.haskell.org/package/base->:
> many :: f <http://hackage.haskell.org/package/base-> a <http://hackage.haskell.org/package/base-> -> f <http://hackage.haskell.org/package/base-> [a <http://hackage.haskell.org/package/base->]many <http://hackage.haskell.org/package/base-> v <http://hackage.haskell.org/package/base-> = many_v <http://hackage.haskell.org/package/base->  where    many_v <http://hackage.haskell.org/package/base-> = some_v <http://hackage.haskell.org/package/base-> <|> <http://hackage.haskell.org/package/base-> pure <http://hackage.haskell.org/package/base-> []    some_v <http://hackage.haskell.org/package/base-> = (fmap <http://hackage.haskell.org/package/base-> (:) v <http://hackage.haskell.org/package/base->) <*> <http://hackage.haskell.org/package/base-> many_v <http://hackage.haskell.org/package/base->
> many_v and some_v are mutually recursive functions, and it may be that
> this prevents the thunks from being made available to take in some way. I'm
> really not sure though, this is just an idea about why this is not quite
> the same as (take $ repeat 1)

The problem is that many is creating an infinite sum that’s nested to the
left. So you’re trying to compute

(((… <|> Just [1,1,1]) <|> Just [1,1]) <|> Just [1]) <|> Just 1

which will never terminate because Maybe is strict in the first argument to

As a practical matter, the Alternative instance for Maybe should probably
be changed to either call error or return Just (repeat v).

Similarly, we should probably flip the order for many in the instance for

Dave Menendez <dave at zednenem.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20160929/a564456d/attachment.html>

More information about the Haskell-Cafe mailing list