This PR is way too big. Also, don't regenerate pydoc_data, it's done during the release process.

I think the majority of the diff is from pydoc_data.

I'm on mobile so it's hard to tell (the pydoc data diff doesn't even render). I'll look at the PR substantively over the week though.

It is definitely true that most of the diff here is coming from pydoc_data (and another big chunk from the hard-coded list of dunder methods in pydoc.py).

I had regenerated pydoc_data because part of the change here was adding a couple of topics to it; without it, some of the new topics (__main__, __name__, __instancecheck__, __subclasscheck__) won't work. For now I've reverted all of the changes to pydoc_data other than the ones that are relevant to the additions here; and I can roll those back, too, if that's better (though again, those new topics won't have their help output without regenerating the file locally).

though again, those new topics won't have their help output without regenerating the file locally

I think that's fine, we want to keep the diff down as much as possible. Someone (presumably Hugo) will regenerate it when it's time to do 3.15a1.

OK, sounds fair. Some of my new tests won't pass without the additions to pydoc_data, though; should I remove those tests, too?

Hm, I'm not sure then. We should look into previous PRs to see what they did.

Taking a quick look at git log --patch Lib/pydoc_data/, it looks like there are some (but not many) changes to pydoc_data in between releases. For example, 6266689 and 447a151 seem to be a similar kind of change to what I'm proposing here (adding a reference to the docs and to pydoc) .

I'll leave things here for now (the changes to pydoc_data are only those relevant to the changes in this PR), but I can of course still revert things later if need be.