Add pip.main as func raising clean error

[sirosen]

This is follow-up/an alternative to #5149

This has upsides and downsides. It improves the errors and guidance for those who invoke pip.main, but it changes the time of failure for some modules attempting to import and use pip.main. Instead of failing at import time, they'll fail when they invoke pip.main().
On the plus side, it means that modules which import pip.main but only use it in some of their functions/classes may remain otherwise functional, so the scope of breakage is reduced.

Modules using pip.main() are going to be broken one way or another, and although ImportErrors will be superior for a number of users, it forbids us from sending good and clear messaging about what to do to remediate.

Advertising pip<10 as "maybe working" is probably a bit controversial, but it's practical/useful info. Whether or not pip takes a hardline stance on this is a matter for discussion/debate.

[pfmoore]

I remain -1 on this. There's no evidence that it'll result in any change in user behaviour - the people encountering this error are typically doing things like sudo pip install -U pip which breaks their system. Giving an alternative message that explains how to call pip without using pip.main is no help to them, as it's the distro that need to make that change, and there's no reason to think they aren't doing that, in conjunction with updating their system packages to pip 10.

[sirosen]

The ImportError they'll see without this doesn't help those users either.

This helps package maintainers who have been calling pip.main and need to fix their usage.
Instead of a blunt ImportError, they'll get information on how to fix their packages.

[pfmoore]

Well, they could always read the release notes, and the docs, and the various release emails and announcements, and look at the issue tracker for discussions of similar issues...

[sirosen]

Yes, the information is already out there. Why does that make it bad to present it in a more cogent and direct manner?

[pfmoore]

I didn't say it was bad, just that I'm against doing it. I don't think there's any real benefit, and it's something we'd have to maintain and potentially debate (I already object to the "you may find pip.main works on pip<10" comment, people will read that as "pin to pip<10"). But that's mostly "I'm not going to do anything on this" not "I'm going to block this".

[sirosen]

Something is better than nothing if it can be done without making things worse.
If you're not on the listhosts, there's no warning that packages you're using are about to break, or that your careless use of sudo pip install -U is about to land you in hot water. (Not that there necessarily needs to be. You can't save everyone.)

I already object to the "you may find pip.main works on pip<10" comment, people will read that as "pin to pip<10"

If you're on a distro that doesn't support pip10 yet, and you sudo pip install -U package-with-pip-dependency, that's exactly the information which you need and the correct thing to do until your distro updates.

shrugs I don't have a super-strong commitment to this. I just think having a clear message will save less clue-ful users a lot of headache.

[absognety]

pip.main() for pip 10 doesn't seem to work seems like the function main in the latest version has been dropped. However we can use subprocess.check_call([]) with similar arguments ex: subprocess.check_call(['install','package_name']) to install a package and also to upgrade a package

[BrownTruck]

Hello!

I am an automated bot and I have noticed that this pull request is not currently able to be merged. If you are able to either merge the master branch into this pull request or rebase this pull request against master then it will eligible for code review and hopefully merging!

[pradyunsg]

I agree with @pfmoore here. I'm also not in favor of this.

Anyways, the ship has sailed on this. I don't see users being helped by this change because nearly all of the reported issues are due to users messing up their OS package manager-managed pip installations.

[pradyunsg]

The underlying issue here IMO was pip's upgrade being done incorrectly, partly due to pip printing a bad message for users to do the upgrade. We're working on that in #5346.

[pradyunsg]

Closing since I don't see anything actionable on this PR anymore.

Feel free to comment here if I'm wrong on that. :)

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.