More efficient code, but slower program

[2QdxY4RzWzUUiLuE at potatochowder.com]

On 2022-07-27 at 17:48:47 +0200,
Regarding "Re: More efficient code, but slower program,"
Cecil Westerhof via Python-list <python-list at python.org> wrote:

> ram at zedat.fu-berlin.de (Stefan Ram) writes:
> 
> > Cecil Westerhof <Cecil at decebal.nl> writes:
> >>values = [*range(100)]
> >
> >   In many cases, any iterable is just fine and a list is not
> >   required, just as peudo-random numbers often are just fine and
> >   real-world entropy is not required.
> 
> In this case both are. I must select (several times) a random element
> from the list. So I need the list.
> I also want the randomness to be as good as possible to make the
> 'simulation' as good as possible.

"[A]s good as possible" for simulations and tests may not require the
cryptographic quality numbers from SystemRandom.  Many/most pseudo
random number generators are optimized for statistically normalized
outputs, and are repeatable as a bonus (again, often a requirement for
certain types of simulations; YMMV).

Also, what if you shuffled the list first (e.g., with random.shuffle)
and then iterated through it directly?  Repeatedly removing arbitrary
elements from the middle of a list is potentially expensive.