On Thu, 9 Dec 2021 at 06:49, Christopher Barker <pythonchb@gmail.com> wrote:
(I started writing this this morning, and there’s been quite a bit more discussion since, but I don’t think it’s completely too late)
Your email client is doing "white on white" again. You should try to get that fixed :-( I agree, without an actual proposal for how deferred expressions work, it's impossible to be specific about what might be affected. And holding one proposal hostage to something that might never happen isn't reasonable. So I'm happy to class a lot of the discussion about deferred expressions as off-topic. My point was more specific, though - that whenever deferred expressions have been discussed in the past (and they do come up every so often) it's fairly clear that one case they would be expected to handle would be calculating function defaults at call time (at least, that's my impression). So if they ever do get implemented, they would render this PEP obsolete. I'm not aware of any case where we've added a language feature knowing that people were interested in something strictly more general. Indeed, switch statements were rejected for years because they "weren't general enough", without any specific more general feature being on the table. So my objection is that I don't see why this particular PEP warrants an exception to that precedent, and why in this specific case, "never is often better than *right* now" doesn't apply. What's changed to make late bound defaults worth fixing right now? That's *not* a rhetorical question - I'm happy if someone can tell me, if now is the right time to do this, then what's different about getting this in 3.11, as opposed to 3.8, or 3.14? And I'm fine with an imprecise answer here - assignment expressions got in mostly because "they've been asked about for years, we have the momentum right now, and Guido wants to do it". If the argument here is similarly imprecise, that's fine - a precise argument would be *stronger*, but a lot of the concerns being raised are imprecise, I'm not expecting to hold the responses to a standard the concerns don't achieve. But I would like a response. Paul PS While I'm posting, a *huge* thanks to Rob Cliffe for preparing and posting that summary of the concerns. That was extremely helpful.