[clean-list] Re: Clean versus Haskell

[Philippos Apolinarius]

Hi, Henrique.
I couldn't resist adding a few comments. Clean is much better than Haskell, in the following senses:

(i) It generates faster code, smaller exec files, and uses less memory. The compiler is also much smaller than GHC. Clean has even more features than Haskell, but occupies less space in your hard disk.
    (a) Distribution: GHC 6.10.4 --- 52M versus Clean 2.2 --- 14 M. 
    (b) Fibonacci compiled code (below): GHC  --- 1M versus Clean 2.2 --- 57 K
    (c) Fibonacci time (fib 40): GHC --- 7.422 s versus Clean 2.2 --- 1.7s
    (d) Schmidtt's Neural net: GHC --- 12.8 s versus Clean 2.2 --- 0.37 s

(ii) Clean has efficient array processing, as you can see from the example of Neural Network. Clean takes less than 0.1s to train and run the neural network in my machine.  I won't tell you how long Haskell takes, because you will not believe me. A hint: Haskell takes much, much longer.

(ii) Clean is safer, since it does not allow things like trying to write into a closed file.

Well, Henrique, I would be very happy if I had an option  in functional programming. Competition is good. However, it seems that Clean is alone in the arena. I wrote many letters to the Haskell community, requesting smaller exec files, and better array processing. This is necessary to make Haskell a realy good tool.

By the way, the Neural Network example comes from a paper by Peter Schmidtt et al. 

//Fibonacci in Clean
module fibo
import StdEnv, ArgEnv

fib n | n  < 2    = 1
      | otherwise = fib (n-1) + fib (n-2)

Start= fib (toInt argv.[1])
where
   argv= getCommandLine

-- Fibonacci in Haskell
module Main( main ) where
 import System( getArgs )


 fib n | n  < 2    = 1
       | otherwise = fib (n-1) + fib (n-2)

 -- Function f_print  prints the n'th Fibonacci number
 f_print n = print(show n ++ "th Fibonacci number is " ++ show (fib n))

 -- Function main is the entry point of the program
 main = do
    args <- getArgs
    if (length args /= 1)
        then putStr "Usage: f1a <n>"
        else (f_print (read (head args)))
    putStr "\n"


// Neural network in Clean
module arrays
import StdEnv

sig x = 1.0 / (1.0 + (exp (~ x)))

error vt = loop vt 0 0.0
where
    loop vt i acc | i>= size vt = (acc, vt)
    loop vt i acc
      # (e, ww) = vt![i] 
      = loop ww (i+1) (acc+e)

na vt=:{[4]=x, [5]=y, [6]=z} [i1, i2]=
  (sig (x+y*i1+z*i2), vt)

ns vt [i1, i2, i3]
  # (v0, vt)= vt![0]
  # (v1, vt)= vt![1]
  # (v2, vt)= vt![2]
  # (v3, vt)= vt![3]
  = (sig(v0+i1*v1+i2*v2+i3*v3), vt)
  
gate vt [i1, i2]
    # (n1, vt)= na vt [i1, i2]
    # (n2, vt)= ns vt [i1, n1, i2]
    = (n2, vt)
 
errSum v2 [] acc= (acc, v2)
errSum v2 [ex:s] acc
   # v= hd ex
   # (vc, v3)= gate v2 (tl ex)
   = errSum v3 s (acc+(vc-v)*(vc-v))
   
updweight vt nvt err0 ns exs = loop 0 vt nvt 
where
   dx = 0.01
   mu = 0.5
   loop i vt vs | i > ns = (vs, vt)
   loop i vt vs
     # (v, v1) = vt![i]
     # v2 = {v1 & [i]= v+dx}
     # (nerr, v3)= errSum v2 exs 0.0
     # nv= v + mu*(err0 - nerr)/dx
     = loop (i+1) {v3& [i]=v} {vs & [i]=nv}

train exs = loop v1 v2
where
    v1 :: .{#Real}
    v1 = {0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0}
    v2 :: .{#Real}
    v2 = {0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0}
    loop vt nvt
      # (nerr, vt) = errSum vt exs 0.0
      | nerr < 0.01 =vt
      # (vt, nvt)= updweight vt nvt nerr 6 exs
      = loop vt nvt
 
Start = xor 0.0 0.0 // (xor 0.0 0.0, xor 0.0 1.0, xor 1.0 0.0, xor 1.0 1.0)
where
    vt :: .{#Real}
    vt = train exs
    exs = [eg e\\ e <- [0,1,2,3,3,2,1,0]]
    xor i1 i2= fst(gate vt [i1,i2])
    
eg 0 =  [0.0, 1.0, 1.0]
eg 1 = [1.0, 1.0, 0.0]
eg 2 = [1.0, 0.0, 1.0]
eg 3 = [0.0, 0.0, 0.0]


--- On Mon, 10/12/09, rinus plasmeijer <rinus at cs.ru.nl> wrote:

From: rinus plasmeijer <rinus at cs.ru.nl>
Subject: [clean-list] Re: Clean
To: "Henrique" <henrique_gusmao_ at hotmail.com>
Cc: clean-list at science.ru.nl
Received: Monday, October 12, 2009, 1:22 AM



 
 


Hi Henrique

 
> Is this e-mail still working? http://clean.cs.ru.nl/
 
 
Yes.

 
>I want to know more about Clean: news? 
updates?
> What is the future of Clean?
 
See: http://wiki.clean.cs.ru.nl/Latest_developments
We are working on a new version which allows you to 
mix Clean and Haskell 98 code.
It is a lot of work, it will still take some time. 
 
> Is Haskell killing Clean? 
Haskell is certainly much more used, which is also 
the raison for adding a Haskell front end.

> I would like to 
learn and test Clean, and maybe then use it commercially. 
> Where do I 
can download it? I send e-mail to clean at cs.ru.nl , asked for 
Clean at the site above, but I got no answer.

We did not got your 
email.
To download browse to: http://clean.cs.ru.nl/Download/main/main.htm
 
> Obs: I am using windows platform. 
Thats fine, it should work on any windows platform.
 

Greetings,
 
Rinus
 
-----Inline Attachment Follows-----

_______________________________________________
clean-list mailing list
clean-list at science.ru.nl
http://mailman.science.ru.nl/mailman/listinfo/clean-list



      __________________________________________________________________
Looking for the perfect gift? Give the gift of Flickr! 

http://www.flickr.com/gift/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.science.ru.nl/pipermail/clean-list/attachments/20091012/5c30c928/attachment.html