Are we really going to accept an explicit argument None for the default base? I'm not sure what I think of that.

Edit: to be clear, I see that it is convenient to be able to pass an explicit argument to mean 'give me the default', but log(x, base=None) reads wrong to me. We're doing a logarithm without a base?

Are we really going to accept an explicit argument None for the default base?

Not yet (see the issue and mentioned discussion). That's why this is a draft.

it is convenient to be able to pass an explicit argument to mean 'give me the default'

Unfortunately, we can't pass other default in this case (no such entity as the Euler's number in the stdlib). Before it was base=math.e, but that's wrong: with the default value this functions returns the natural logarithm of x, not log(x)/log(math.e).

We're doing a logarithm without a base?

Yes, without this argument it's assumed to be the natural logarithm of x. The difference is just in the log function signature, it was before log(x[, base]) - not that currently the inspect.Signature does support.

We don't have core dev consensus on this: @rhettinger in particular is strongly opposed to the introduction of None here. Let's close on the "status quo wins a stalemate" basis.

@mdickinson, I appreciate your judgment here, but unless I miss that, the perm() function case was just ignored by @rhettinger (both in the issue and in the discussion thread). Same module, the default value was introduced by... Is that was an error, or?

I see point to make the inspect module/AC more expressive to describe de-facto existing syntax (multiple signatures), that has no pure-Python counterpart yet. That clearly does make sense for more complex cases (e.g. range). But does it make sense in the given example? I'm not sure that it was shown in the discussion. Unfortunately, there are no other options suggested (c.f. multiple signatures) to solve the issue, e.g. I doubt it's possible to use base=math.e default & provide correct inspect.Signature's.

And there is another, yet related, issue: log() is slow (see pr description) just to keep the "helpful and correct" "signature". Maybe it does make sense to revert the AC stuff here to workaround this?

@skirpichev I understand (and share) your frustration, but I don't think anything's going to be achieved by further discussion here. It's not that big a deal, and it's time to let it go.

@mdickinson, ok. But you closed the issue as well. That sounds (at least for me): "we won't any progress here at all" (that includes Raymond's suggestion). If this somehow was related to my personal involvement in the discussion - please reconsider: I can promise to be outside of any further discussions and/or any touching of math/cmath code, whatever you feel better.

What I actually want - some public interface, that will allow me to add introspection capabilities to external extensions. Personally I'm satisfied with the current "state of art" (except for #101872 maybe), but now I see why all this stuff (e.g. __signature__ or __text_signature__ attributes) can't be made public yet. This bug is an example. Don't hide problems.