Daniel McCloy

> when/how often should we be running the script to update our kitchen sink pages? historically it's been ad-hoc. I guess it makes sense to make that part of the...

> I don't understand why the installer can't add this file in the first place (I thought that you were also discussing non-installer MNE). It will. See https://github.com/mne-tools/mne-python/pull/12900#issuecomment-2420341049. The problem...

> users without the installer will not have the `MNE-Python` class. They will have `python3`, so this workaround will not really work. IIUC the wmclass is set at runtime by...

In theory it can be done directly with pytest using a `-k test_name_regex` flag, assuming you're in a suitable environment. The `nox` setup is newish and I'm not very familiar...

This comment has the relevant backstory links that @cbrnr forgot to include: https://github.com/mne-tools/mne-python/issues/12388#issuecomment-1910876635 > we started with https://github.com/mne-tools/mne-python/pull/11838 (but did it a dumb way, without .pyi files). This caused problems...

> It's not the entire backstory, just some aspect. Sure. I wasn't trying to imply that your description wasn't relevant. Just that there are some other things that are also...

> Of course I'm not saying I omitted something intentionally, come on! I guess I misunderstood. Saying "I think I summarized all important points in my initial post" implied that...

Well, the ship may have already sailed, but I'd like to take the time to enumerate some arguments on *both* sides of this issue. #### in favor of keeping lazy...

> I think the ideal solution would be to not import everything into the global `mne` namespace, so having to `import mne.decoding` or even `import mne.io` would be perfectly fine...

> you can still interactively explore if you first `import mne.viz` which is *even slower* than importing everything eagerly in the first place. > (to stick with your example, it...