Senhaji Rhazi hamza writes:
> I think there is a way to map x <- [1; + inf] to y <- [0;1] by
> putting y = 1/x
I don't think that's the point. I'll put money on a poor choice of
value for the random parameter in the example: the OP chose the Pareto
variate because that's what he mentioned earlier. But there are
plenty of distributions besides uniform that can be parametrized to
have support included in [0,1].
To me, it also seems a little odd that random.choices supports only
discrete populations (possibly with weights) but doesn't support float
populations. I think
r.choices(random=lambda: random.paretovariate(1.75), k=k)
would be a nice way to spell "sample with replacement from Pareto
distribution of size k". But as Steven d'Aprano points out, you can
also spell it
[ random.paretovariate(1.75) for _ in range(k) ]
so there's no real need to generalize Random.choices.
On the other hand,
r.sample(random=lambda: random.paretovariate(1.75), k=k)
doesn't make immediate sense to me. What does "sampling without
replacement" mean if you don't have an explicit population? (Note
that random.sample doesn't support weights for a similar reason,
although it could support counts.) So this suggests to me that the
proposed ``random`` parameter to sequence methods is actually a
spurious generalization, and has different interpretations for
different methods.
So I'm not against this proposal at this point, but I think it needs
to be fleshed out more, both in terms of the principle we're trying to
implement, and some more detailed examples.
Steve
_______________________________________________
Python-ideas mailing list -- python-ideas@python.org
To unsubscribe send an email to python-ideas-le...@python.org
https://mail.python.org/mailman3/lists/python-ideas.python.org/
Message archived at
https://mail.python.org/archives/list/python-ideas@python.org/message/JR3G22W4H3JTO4KC5DJGCTEZKI26KHUP/
Code of Conduct: http://python.org/psf/codeofconduct/
