Is the solving a real problem that you've had or is this just a theoretical issue?

In general, the module shies away from such checks unless we find that they are really necessary. For example, gauss doesn't check to see if the sigma argument is negative. Mostly, these calls need to be short and fast without anything extraneous.

I happened to check the coverage of random module today and found some uncovered path which led to this PR.

I agree that the function should be short and fast, but should we keep the same standard? For example, the check for p prevents the function from a infinite loop which is considered necessary, but why do we check for n < 0? That'll just be a case of "invalud input, invalid output". Should we remove that check?

In gammavariate, we check if beta <= 0, that'll be another example of "invalid input, invalid output', as it was just used as a multiplication factor.

However, I understand if you don't want to add the extra 0 and integer check on n. I do think the comparison with 0 makes more sense than not c. Or we can only keep the new test cases.

Or we can only keep the new test cases.

I think that would be best.

I think that would be best.

Could you share your thoughts about if not c vs if c == 0? That's actually the part I have a rather strong feeling. I think the latter statement is significantly better in this case.

#81620 has been closed a long time ago. Could you please open a new issue for discussing the feasibility of such changes?

Could you share your thoughts about if not c vs if c == 0?

It's a bit of a personal preference. This module tends to use the former such as the if y in betavariate. The former is also faster than if c == 0 and is consistent with the practice for sequences of writing if not seq instead of if len(seq) == 0. Using the truthy value of an object is common in the Python (and some other languages). It is something you get used to.

@serhiy-storchaka I don't think we need to open another issue for a minor code edit. I'll just remove the issue reference from this PR.

It's a bit of a personal preference. This module tends to use the former such as the if y in betavariate. The former is also faster than if c == 0 and is consistent with the practice for sequences of writing if not seq instead of if len(seq) == 0. Using the truthy value of an object is common in the Python (and some other languages). It is something you get used to.

I'm actually perfectly fine with truthy value check for sequences. I only have hesitations about numerical values. It is faster than c == 0.

Anyway, I removed the code change in random.py and I agree we don't need a separate issue for this minor change.

Thanks for the extra tests and for the polite discussion.