drtomc:
> Well, maybe I shoud be asking a higher level question then.
>
> I have a function
>
> tidy = reverse . dropWhile punk . reverse . dropWhile punk
> where
> punk = isPunctuation . chr . fromIntegral
>
> which is leading to a significant amount of allocation, and you can see why.
>
> The way I'd like to write it is
>
> tidy = dropWhile punk . dropWhileEnd punk
> where ....
>
> which has the obvious advantage of avoiding quite a bit of
> intermediate allocation.
>
> Is there a another way?
>
> I note that since I'm using a nice declarative language, the compiler
> CLEARLY should be transforming the first form into the second. :-)
I'd just manually write a 'tidy' loop (in the Data.ByteString style) (which
would avoid all allocations), since it seems pretty useful.
Something in this style:
findIndexOrEnd :: (Word8 -> Bool) -> ByteString -> Int
findIndexOrEnd k (PS x s l) =
inlinePerformIO $ withForeignPtr x $ \f -> go (f `plusPtr` s) 0
where
go !ptr !n | n >= l = return l
| otherwise = do w <- peek ptr
if k w
then return n
else go (ptr `plusPtr` 1) (n+1)
If its costly, since that'll make it non-costly.
-- Don
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe
