I know that there are specialization rules for foldl and foldr (among other higher-order functions in the Prelude) with the idea that they will produce usually better generated code.  So, yes, the generated Core will almost certainly be different. 

Whether the code is truly more performant (in time or space) will likely depend on your use case.  Inspecting the Core manually may give you some insights, but unless you're experienced in that domain, you'll get more direct and faster answers by using GHC's profiling tools with some chosen benchmarks.

- David

On Tue, Jul 11, 2017 at 5:20 PM Apoorv Ingle <apoorv.ingle@gmail.com> wrote:
Hi David,

Thanks a lot for the code!

foldr is indeed elegant.
In general is it advisable to use auxiliary functions or foldr/foldl variations.
Does it have any performance benefits or ghc would generate same core language for both the functions?

Regards,
Apoorv

On Jul 11, 2017, at 15:40, David Ringo <davidmringo@gmail.com> wrote:

Hi Apoorv,

There is indeed a left fold:

foldlpart :: [Int] -> [a] -> [[a]]
foldlpart ds ps = result
  where result | null remaining = initial
               | otherwise = initial ++ [remaining]
        (initial, remaining) = foldl aux ([], ps) ds
        aux (l, xs) d = case xs of
          [] -> (l, xs)
          _  -> let (f,s) = splitAt d xs in (l ++ [f], s)

I'm sure someone else can put something better together though.

I much prefer this right fold, since it avoids quadratic behavior incurred with (++) above:

foldrpart :: [Int] -> [a] -> [[a]]
foldrpart ds ps = myFunc ps
  where myFunc = foldr buildMyFunc (: []) ds
        buildMyFunc digit func = \ps ->
          case ps of
            [] -> []
            _  -> let (first, last) = splitAt digit ps
                  in  first : func last

If it's unclear, buildMyFunc is basically composing a bunch of functions which know (from the fold on the list of Ints) how many elements
to take from some list.

Hope this is useful.

- David

On Tue, Jul 11, 2017 at 3:30 PM Apoorv Ingle <apoorv.ingle@gmail.com> wrote:
Hi,

I am trying to write a partition function where we pass group sizes and the list we want to partition into groups 
as arguments and get back a list of groups (or list of lists in this case). My first attempt was by using an auxiliary inner function

{-# LANGUAGE ScopedTypeVariables #-}

module Partition where

partition :: [Int] -> [a] -> [[a]]
partition ds ps = reverse $ paux ds ps []
  where
    paux :: [Int] -> [a] -> [[a]] -> [[a]]
    paux []         []   ps' = ps'
    paux []         ps ps' = [ps] ++ ps’ 
    paux _         []   ps' = ps'
    paux (d:ds') ps ps' = paux ds' (snd (splitAt d ps)) ([fst (splitAt d ps)] ++ ps')

——————

*Partition> partition [2, 3] [1,2,3,4,5]
[[1,2],[3,4,5]]
*Partition> partition [1, 2] [1,2,3,4,5]
[[1],[2,3],[4,5]]
*Partition> partition [1, 2, 5] [1,2,3,4,5]
[[1],[2,3],[4,5]]


I was speculating if we could write the same function using foldl function but haven’t been able to figure it out.
I would really appreciate if you can give me pointers on how we can implement it.

partition' :: [Int] -> [a] -> [[a]]
partition' [] ds = [ds]
partition' ps ds = foldl ??? ???' ???''

contrary to my speculation is it even possible to write such a function using foldl if so why not?

Regards,
Apoorv Ingle
Graduate Student, Computer Science
apoorv.ingle@ku.edu
_______________________________________________
Beginners mailing list
Beginners@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners
_______________________________________________
Beginners mailing list
Beginners@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners

_______________________________________________
Beginners mailing list
Beginners@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners