Make --ignore-missing-imports less strict

[ilevkivskyi]

Fixes #5226

Currently, import existing_package.non_existing_mod (with --ignore-missing-imports) allows subsequent

from existing_package import non_existing_mod
x = non_existing_mod.f()

but doesn't allow a direct call

x = existing_package.non_existing_mod.f()

With this change both are allowed. This doesn't solve the problem of partial packages, but makes --ignore-missing-imports more consistent with the flag name IMO.

[JukkaL]

Looks good to me. This makes import semantics across different forms of imports more consistent and predictable.

Have you verified that doesn't break anything in internal Dropbox codebases?

[ilevkivskyi]

Have you verified that doesn't break anything in internal Dropbox codebases?

Not yet, I will check this in next few hours, I will merge if testpr passes.

[gvanrossum]

Of course testpr will fail, since it already fails with the current master. :-( Maybe you can compare the errors you get without and with this PR.

…

On Mon, Jun 18, 2018, 06:37 Ivan Levkivskyi ***@***.***> wrote: Have you verified that doesn't break anything in internal Dropbox codebases? Not yet, I will check this in next few hours, I will merge if testpr passes. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#5228 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACwrMoHxrSY9iCkTBt6vD47QthDt0N10ks5t960ogaJpZM4UqsHF> .

[emmatyping]

LGTM

[msullivan]

Is there any documentation that should be updated along with this?

[ilevkivskyi]

Is there any documentation that should be updated along with this?

This is a good idea, I will add some docs.

[ilevkivskyi]

I was finally able to check, this passes testpr.