On Wed, 2009-01-14 at 12:45 -0800, Sukit Tretriluxana wrote:
> Hi all,
> 
> I was looking around Stroustrup's website and found a simple program
> that he showed how standard library can be used to make the program
> succinct and safe. See
> http://www.research.att.com/~bs/bs_faq2.html#simple-program. I
> wondered how a Haskell program equivalent to it looks like and I came
> up with the code below.
> 
> import qualified Control.Exception as E
> 
> main = E.catch (interact reverseDouble) (\_ -> print "format error")
> 
> reverseDouble = unlines . doIt . words
>    where doIt = intro . toStrings . reverse . toDoubles . input
>          toDoubles = map (read::String->Double)
>          toStrings = map show
>          input = takeWhile (/= "end")
>          intro l = ("read " ++ (show $ length l) ++ " elements") :
>                    "elements in reversed order" :

My only criticism is that I find code written with lots of secondary
definitions like this confusing; so I would inline most of the
definitions:

  reverseDouble =
        unlines
      . intro
      . map show
      . reverse
      . map (read :: String -> Double)
      . takeWhile (/= "end")
      . words
    where
      intro l =
        ("read " ++ show (length l) ++ " elements") :
        "elements in reversed order" :
        l

I observe also in passing that the cast on read is somewhat inelegant;
in a real application, the consumer of map read's output would specify
its type sufficiently that the cast would be un-necessary.

For example, the program could be specified to compute sines instead:

main = E.catch (interact unlines . intro . map (show . sin . read) .
words) $ \ _ -> print "format error"
  where
    intro l =
      ("read " ++ show (length l) ++ " arguments") :
      "computed sins" :
      l

(Others will no doubt object to the use of lazy I/O.  I disagree in
principle with those objections.)

jcc

PS Stroustrup's comments about vectors are at best half right; push_back
may extend the vector's length correctly, but operator[] on a vector
certainly does not do bounds checking.


_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe

Reply via email to