Sudoku solver

[Mark Lawrence]

On 29/03/2015 19:03, Marko Rauhamaa wrote:
> BartC <bc at freeuk.com>:
>
>> As Chris mentioned, when I say 'faster than C', I mean X running my
>> algorithm was faster then C running Marko's algoritim (on Ian's data).
>> This was just an illustration of algorithm being more important than
>> language.
>
> Be careful with the benchmark comparisons. Ian's example can be solved
> with the identical algorithm in eight different ways (four corners, left
> or right). I ran the example with my recent Python solver and got these
> times in the eight cases:
>
>      884   s
>        2.5 s
>       13   s
>      499   s
>        5.9 s
>      128   s
>     1360   s
>       36   s
>
> (That suggests a slight modification to the original strategy: solve it
> in each of the eight ways "in parallel" and the worst case should be
> significantly alleviated.)
>
> I'm actually fascinated by the Python modifications that were
> conceptually insignificant but dramatically speeded up the execution
> (7880 seconds to 884 seconds). That's something I'll have to keep in
> mind in the future.
>
>
> Marko
>

One thing I have come to rely on over the years is never, ever trust 
your gut instincts about Python performance, you're almost inevitably 
wrong.  When I first came across the Norvig solver I made a change, 
purely for fun, to replace two calls to len() with a single call and 
save the result.  The run time to solve each puzzle dropped by around 
5%.  I believe this meets your "conceptually insignificant".

-- 
My fellow Pythonistas, ask not what our language can do for you, ask
what you can do for our language.

Mark Lawrence