[Haskell-cafe] Efficient M to N conversion on fusion?

[Felipe Lessa]

On Tue, Feb 16, 2010 at 07:52:51PM +0900, iquiw wrote:
> Is there any efficient way to convert M-length ByteString, Text or
> Vector into N-length one?

This is just a guess, but in vector at least it should be
possible to use sized[1] to tell how much space should be
allocated on the final vector.  For example, something like the
following *untested* code:

import qualified Data.Vector.Stream.Monadic as S
import qualified Data.Vector.Generic as G
import Control.Applicative
import Data.Vector.Stream.Size

-- | The size should be the size of each v b.
concatMapSized :: (G.Vector v a, G.Vector v b) => Size -> (a -> v b) -> v a -> v b
concatMapSized size_each f v =
  let stream = G.stream v
      size   = mult size_each (S.size stream)
  in G.unstream . flip S.sized size . S.concatMap (G.stream . f) $ stream

mult :: Size -> Size -> Size
mult (Exact n) (Exact m) = Exact (n*m)
mult x y = maybe Max Unknown $ (*) <$> upperBound x <*> upperBound y

I really don't know if this would improve or worsen performance,
you should check the core and do benchmarks :).

[1] http://hackage.haskell.org/packages/archive/vector/0.5/doc/html/Data-Vector-Fusion-Stream.html#v%3Asized

--
Felipe.