[Haskell-cafe] Fastest way to calculate all the ways to interleave two lists

[Arseniy Alekseyev]

I asked David off-list and he explained that the reason his version is
faster is that `head . toList` on Seq is O(1) and the equivalent operation
on difference lists is O(n). Now I want to have a fingertree-backed
[Function a b] type. :)

On 3 April 2016 at 07:05, Arseniy Alekseyev <arseniy.alekseyev at gmail.com>
wrote:

> That (specifically, the benchmark below) shows your thing is faster, but
> I'm not sure why. Maybe it's because Seq is cheaper than a closure, but
> maybe it's something more meaningful than that. Looks like you've guided
> myself roughly to your original solution now so I'm giving up. :)
>
> main = print $ sum $ map head $ take 1000000 $ interleavings
> [[1..100],[5..100]]
>
> On 3 April 2016 at 06:20, David Feuer <david.feuer at gmail.com> wrote:
>
>> Er.. I mean force . map head
>> On Apr 3, 2016 1:14 AM, "David Feuer" <david.feuer at gmail.com> wrote:
>>
>>> I choose the `force (map head)` attack.
>>> On Apr 3, 2016 1:04 AM, "Arseniy Alekseyev" <arseniy.alekseyev at gmail.com>
>>> wrote:
>>>
>>>> I see! At this point I'd say that you probably have the wrong type:
>>>> there are ways to produce n'th interleaving much faster, but let's continue
>>>> optimizing for the hell of it!
>>>>
>>>> i2 :: ([a] -> [b]) -> [a] -> [a] -> [[b]] -> [[b]]
>>>> i2 f [] ys = (f ys :)
>>>> i2 f xs [] = (f xs :)
>>>> i2 f (x : xs) (y : ys) =
>>>>   i2 (f . (x :)) xs (y : ys) . i2 (f . (y :)) (x : xs) ys
>>>>
>>>> interleave2 xs ys = i2 id xs ys []
>>>>
>>>> Seems faster than your original solution on examples I tried it on and
>>>> it has fewer characters. :)
>>>>
>>>> On 3 April 2016 at 05:41, David Feuer <david.feuer at gmail.com> wrote:
>>>>
>>>>> Of course, but something like take k . (!! m)   will cut it down
>>>>> nicely.
>>>>>
>>>>> On Sun, Apr 3, 2016 at 12:39 AM, Arseniy Alekseyev
>>>>> <arseniy.alekseyev at gmail.com> wrote:
>>>>> > Um, the result is exponential in size. A problem will emerge in any
>>>>> > solution. :)
>>>>> >
>>>>> > On 3 April 2016 at 05:38, David Feuer <david.feuer at gmail.com> wrote:
>>>>> >>
>>>>> >> Your lists are very short. Pump them up to thousands of elements
>>>>> each
>>>>> >> and you will see a problem emerge in the naive solution.
>>>>> >>
>>>>> >> On Sun, Apr 3, 2016 at 12:16 AM, Arseniy Alekseyev
>>>>> >> <arseniy.alekseyev at gmail.com> wrote:
>>>>> >> > I measure the following naive solution of interleave2 beating
>>>>> yours in
>>>>> >> > performance:
>>>>> >> >
>>>>> >> > i2 [] ys = [ys]
>>>>> >> > i2 xs [] = [xs]
>>>>> >> > i2 (x : xs) (y : ys) =
>>>>> >> >   fmap (x :) (i2 xs (y : ys)) ++ fmap (y :) (i2 (x : xs) ys)
>>>>> >> >
>>>>> >> > The program I'm benchmarking is:
>>>>> >> >
>>>>> >> > main = print $ sum $ map sum $ interleavings
>>>>> >> > [[1,2,3,4],[5,6,7,8],[9,10,11,12],[1,1,1]]
>>>>> >> >
>>>>> >> >
>>>>> >> >
>>>>> >> > On 3 April 2016 at 04:05, David Feuer <david.feuer at gmail.com>
>>>>> wrote:
>>>>> >> >>
>>>>> >> >> I ran into a fun question today:
>>>>> >> >> http://stackoverflow.com/q/36342967/1477667
>>>>> >> >>
>>>>> >> >> Specifically, it asks how to find all ways to interleave lists
>>>>> so that
>>>>> >> >> the order of elements within each list is preserved. The most
>>>>> >> >> efficient way I found is copied below. It's nicely lazy, and
>>>>> avoids
>>>>> >> >> left-nested appends. Unfortunately, it's pretty seriously ugly.
>>>>> Does
>>>>> >> >> anyone have any idea of a way to do this that's both efficient
>>>>> and
>>>>> >> >> elegant?
>>>>> >> >>
>>>>> >> >> {-# LANGUAGE BangPatterns #-}
>>>>> >> >> import Data.Monoid
>>>>> >> >> import Data.Foldable (toList)
>>>>> >> >> import Data.Sequence (Seq, (|>))
>>>>> >> >>
>>>>> >> >> -- Find all ways to interleave two lists
>>>>> >> >> interleave2 :: [a] -> [a] -> [[a]]
>>>>> >> >> interleave2 xs ys = interleave2' mempty xs ys []
>>>>> >> >>
>>>>> >> >> -- Find all ways to interleave two lists, adding the
>>>>> >> >> -- given prefix to each result and continuing with
>>>>> >> >> -- a given list to append
>>>>> >> >> interleave2' :: Seq a -> [a] -> [a] -> [[a]] -> [[a]]
>>>>> >> >> interleave2' !prefix xs ys rest =
>>>>> >> >>   (toList prefix ++ xs ++ ys)
>>>>> >> >>      : interleave2'' prefix xs ys rest
>>>>> >> >>
>>>>> >> >> -- Find all ways to interleave two lists except for
>>>>> >> >> -- the trivial case of just appending them. Glom
>>>>> >> >> -- the results onto the given list.
>>>>> >> >> interleave2'' :: Seq a -> [a] -> [a] -> [[a]] -> [[a]]
>>>>> >> >> interleave2'' !prefix [] _ = id
>>>>> >> >> interleave2'' !prefix _ [] = id
>>>>> >> >> interleave2'' !prefix xs@(x : xs') ys@(y : ys') =
>>>>> >> >>   interleave2' (prefix |> y) xs ys' .
>>>>> >> >>       interleave2'' (prefix |> x) xs' ys
>>>>> >> >>
>>>>> >> >> -- What the question poser wanted; I don't *think* there's
>>>>> >> >> -- anything terribly interesting to do here.
>>>>> >> >> interleavings :: [[a]] -> [[a]]
>>>>> >> >> interleavings = foldr (concatMap . interleave2) [[]]
>>>>> >> >>
>>>>> >> >>
>>>>> >> >> Thanks,
>>>>> >> >> David
>>>>> >> >> _______________________________________________
>>>>> >> >> Haskell-Cafe mailing list
>>>>> >> >> Haskell-Cafe at haskell.org
>>>>> >> >> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
>>>>> >> >
>>>>> >> >
>>>>> >
>>>>> >
>>>>>
>>>>
>>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20160403/ab705895/attachment.html>