choosing random numbers with weights/probability?

[Al Christians]

It looks like generating the random number takes much longer than
searching a short list, so that the method I described, which takes
two or more random numbers per selection, is a big loser on speed 
unless one is searching a list of 100 elements or more.  I  think
it might be better for statistical properties of the results, 
depending on lots of things (the magnitudes of the weights and
the quality of the rng), but in simple applications, that probably 
doesn't matter much, and IDK for sure.

Al 


Darrell wrote:
> 
> Orig:
>  Time: 10.28 sec
> {'one': 25167, 'two': 25144, 'three': 49689}
> 
> Terry's:
>  Time: 9.94 sec
> {'one': 24983, 'three': 50020, 'two': 24997}
> 
> --
> --Darrell