From: Kristof Bastiaensen <kristof@...>
Date: 2006-08-03T23:25:06+09:00
Subject: Re: For performance, write it in OCaml

On Thu, 03 Aug 2006 21:59:40 +0900, Peter Hickman wrote:

> Jon Harrop wrote:
>> Peter Hickman wrote:
>>   
>>> I will take the revised Ocaml version and time it and check the output
>>> tonight.
>>>     
>>>     
>> Yeah. I didn't bother doing that. ;-)
>>
>>
> Not wanting to name and shame but there was one submission (off list)
> that was faster for completely the wrong reasons. Besides I get a warm
> fuzzy feeling when the results match because the pain of checking which
> program is actually wrong is too great. As my Perl, Java and C versions
> all use the same Perl to pre compute their values they all produced the
> same results, consistent yes but are they correct? I'm pretty confident
> now, especially now that I have independently written programs produce
> the same results.
> 
> <snip>

I used the following tests:

"ruby latin.rb | wc -l" to test if the number of squares matches the
number that is written in the wikipedia article.
Then "ruby latin.rb | sort | uniq | wc -l" shows if all the squares that
are output by the program are different.
 All that remains to be done is to check that each latin square is
actually a latin square. That wouldn't be so much work, but I just looked
at some of the output, and trusted that the rest of the squares are
correct too :)

Regards,
Kristof