Comments on: Lazy Fibonacci

By: Paul Kuliniewicz

Paul Kuliniewicz — Sun, 11 May 2008 02:33:29 +0000

Is there something in that code that’s getting garbled by WordPress?

By: Eric Leons

Eric Leons — Sat, 10 May 2008 10:07:58 +0000

Welcome to the functional party, You are probably not writing in a Haskell-y enough syntax This feels better syntax to me:


main = do argv <- getArgs
          name  hPutStr stderr $ "usage: " ++ name ++ " number\n"
            x:xs -> C.catch (putStr $ "Fibonacci number #" ++ x ++ " is " ++ (show result) ++ "\n")               
                            (\_ -> hPutStr stderr "Must give a nonnegative integer\n")
                     where result = fibonacci !! read x

Ideally I would also remove the catch clause and make better use of the Either datatype

By: Eric Leons

Eric Leons — Sat, 10 May 2008 10:07:18 +0000

Welcome to the functional party,

You are probably not writing in a Haskell-y enough syntax

This feels better syntax to me:

main = do argv <- getArgs
name hPutStr stderr $ “usage: ” ++ name ++ ” number\n”
x:xs -> C.catch (putStr $ “Fibonacci number #” ++ x ++ ” is ” ++ (show result) ++ “\n”)
(\_ -> hPutStr stderr “Must give a nonnegative integer\n”)
where result = fibonacci !! read x

Ideally I would also remove the catch clause and make better use of the Either datatype

By: Paul Kuliniewicz

Paul Kuliniewicz — Sat, 10 May 2008 03:53:38 +0000

Frickin’ monads.

By: Ryan

Ryan — Sat, 10 May 2008 03:32:26 +0000

This is cool! (monads?)

By: Paul Kuliniewicz

Paul Kuliniewicz — Tue, 06 May 2008 23:32:17 +0000

The Haskell code is just using the standard definition of Fibonacci numbers; it’s just written in a way that’s not immediately obvious.

Fun math fact about Haskell: its input/output mechanism is based around monads, which are adapted from category theory. Once I wrap my head a little better around monads in Haskell I’ll probably do a post or two on them, since it is kind of neat once it all clicks together.

By: Jamie Weigandt

Jamie Weigandt — Tue, 06 May 2008 23:01:14 +0000

So I kinda wondered about using CRT and found this little undergrad project:

http://www.math.oregonstate.edu/~math_reu/Proceedings/2003/AM_Final.pdf

Looks like the computations are significantly faster, though I guess they don’t precisely use CRT in the end. I didn’t really read a lot of it, and you could be using that method. I just like that my intuition to use CRT kind of worked out.

By: Paul Kuliniewicz

Paul Kuliniewicz — Tue, 06 May 2008 22:17:56 +0000

Depending on the application, the closed form equation may not be more efficient, especially if (like the Haskell solution) you’re saving a copy of your results for later use. If your list of Fibonacci numbers is in the processor’s cache, it’s possibly faster to look something up in the list than it is to do some floating point operations. Of course, you’d need to actually profile the code to see if that’s truly the case on your target system.

Having said that, I can’t recall ever running into a program that needed to compute Fibonacci numbers where the problem to be solved wasn’t explicitly “compute Fibonacci numbers.”

By: Jamie Weigandt

Jamie Weigandt — Tue, 06 May 2008 05:54:36 +0000

If there’s anything I learned last summer, its that computers and infinity don’t mix, but apparently they can hang out.
This little though experiment may be useful when working with elliptic curves, but its 2am….

Though if you’re computing Fibonacci numbers wouldn’t it be a lot more efficient use the formula involving the golden ratio?

Now I want to try something with chinese remainder and fibonacci numbers, but its still 2am…