Comments on: Looping Syntax in Various Languages

By: Jared

Jared — Thu, 04 Jun 2009 20:52:00 +0000

Hi Liang,

That makes sense as the gcd is the only piece that actually does any work (although obviously it is best to measure to be sure – thanks!). The other part is just iterating around some loops.

By: Liang

Liang — Wed, 03 Jun 2009 14:14:38 +0000

I profile perl version. Three statements in until block in gcd function take most of time.

By: Jared

Jared — Fri, 29 May 2009 21:38:44 +0000

Hi Stuart,

Basically the process was the following:

1. I implemented my solution in emacs lisp using (while …) and (setq …)
I didn’t like it so I had a look at looping in other languages.

2. I liked the look of eager comprehensions so I tried those and I’m possibly interested in learning Ocaml so I tried that. I added the Perl as a reference.

3. I added the benchmarks as I thought other people might be interested (and it looks like I was right – it has had the third most number of visitors in the last 30 days and its only been 2 days). I don’t really care that much.

I guess what came across most strongly was the benchmarking, but what I was most interested in was the various syntaxes for looping.

Hopefully that makes sense.

By: stuart

stuart — Fri, 29 May 2009 10:22:14 +0000

I don’t understand why you are comparing emacs lisp to scheme. I would not have considered emacs lisp a fast implementation of lisp. Wouldn’t it make more sense to compare sbcl (or clisp or clozure or even Clojure) with scheme?or emacs lisp with a common lisp implementation?

By: Jared

Jared — Fri, 29 May 2009 07:13:55 +0000

Wow, byte-compiling made a big difference.

(This is a different machine than my other tests).

before:

Started at: 08:07:17
XXX solutions
Finished at: 08:08:39

After:

Started at: 08:10:54
XXX solutions
Finished at: 08:11:11

And yes, I think that mzscheme has a JIT. However, I’m still surprised that it beats Perl by so much.

By: Jonathan Creekmore

Jonathan Creekmore — Thu, 28 May 2009 21:16:39 +0000

Compiling your solve-it function with compile-defun sped up Emacs-Lisp for me a lot. The uncompiled version of solve-it completed in 56 seconds for me. After compiling it, the time shrank to 9 seconds. That is closer to your expected MzScheme time and I believe that MzScheme will byte-compile things behind the scenes for you.

By: Jared

Jared — Thu, 28 May 2009 19:26:54 +0000

@trimtab – iterate looks nice, but I’m looking for a looping mechanism to improve my emacs coding. I just thought I’d do a quick comparison with looping in other languages while I was messing around.

@Porges – thanks for that – it looks similar in spirit (if a bit different in syntax) to the scheme. How long does it take to run on your machine?

@Jisang – sit-for does redisplay as the manual says. If I don’t add it then the screen does not redraw until the loop has finished executing, which took around 5 minutes in my first effort. I added the sit-for so I could check on the progress.

Thanks for the comments.

By: Jisang Yoo

Jisang Yoo — Thu, 28 May 2009 18:35:35 +0000

What does (sit-for 0) do in the first solve-it defun? Accoding to its describe-function, it redisplays. But why redisplay in the middle of the loop?

By: Porges

Porges — Thu, 28 May 2009 14:41:15 +0000

A direct translation of your algorithm into Haskell yields:

main = print solve

solve::Int
solve = sum [1|
	denom <- [5..10000::Int],
	num <- [1+floor(fromIntegral denom/3)..ceiling(fromIntegral denom/2)-1],
	gcd num denom == 1]

By: trimtab

trimtab — Thu, 28 May 2009 14:24:14 +0000

For Common Lisp, you might be interested in the Iterate package in lieu of Loop.

http://common-lisp.net/project/iterate/

More expressive, extensible, and Lispy! There are others still (e.g. Series), but Iterate seems to be the most powerful.