Everything in numpy uses the convention that arrays shapes are called shape... except np.random. This just confused me yet again. Is there any good reason we can't switch to shape= as the canonical name for this argument? Obviously we would continue to support size= as an alias for some long time, but at least I would stop stubbing my toe on this...
I wouldn't have a problem with that as I usually need to think a bit to recall the right keyword.
@njsmith So this can be closed?
Well, I'm on a closing hunt, so that makes me happy. It all sums to zero ;)