``disallowed-name`` check no longer checks non-constants

[mattclay]

Commit 3422e4a changed the behavior of the invalid-name check, but also changed the behavior of blacklisted-name, which appears to be unintentional (and undocumented).

Steps to reproduce

demo.py:

foo = {}.keys()
foo = None

pylint demo.py -d all -e blacklisted-name

Current behavior

From pylint 2.5.0 and later:

demo.py:2:0: C0102: Black listed name "foo" (blacklisted-name)

Expected behavior

From pylint 2.4.4:

demo.py:1:0: C0102: Black listed name "foo" (blacklisted-name)
demo.py:2:0: C0102: Black listed name "foo" (blacklisted-name)

pylint --version output

pylint 2.5.0
astroid 2.4.2
Python 3.7.4 (default, Oct 16 2019, 10:35:50)
[Clang 11.0.0 (clang-1100.0.33.8)]

[relrod]

@PCManticore I'm willing to look at this issue, as it's something that's blocking us in a number of ways. Would you be willing to set me on the right path to get started here?

It looks like the function that was modified is used by multiple of the lint checks. So off the bat I can think of two possible ways to solve it:

• Have visit_assignname() check and see if the check that is being run is blacklisted-name and adjust its logic accordingly to still emit like it used to if it is. I don't know if the function has any way to get at the currently-running check name or not, though, and I don't know if hardcoding the check name like that in the function body/logic is a feasible approach.
• Create a new visit_assignname_const_blacklisted_name() and have it only be used for blacklisted-name, and copy the chain of ifs up to the "const" check, reverting the final one to what it was before 3422e4a.

There might be a better option than both of these, I'm not familiar enough with the pylint codebase yet. If you can point me in the right direction, I'd be able to try to tackle this regression and send a patch!

[relrod]

/cc @Pierre-Sassoulas @hippo91

Maybe one of you could point me in the right direction here? We're still hitting this and it's still blocking us on some things, any help is much appreciated :)

[Pierre-Sassoulas]

Hello @relrod, thank you for being willing to contribute, much appreciated !

The way to fix this would be to first add:

foo = {}.keys()  # [blacklisted-name]
foo = None  # [blacklisted-name]

in blacklisted_name.py

Then regarding the implementation itself maybe moving the blacklist check to it's own checker would be more reasonable than adding another condition or a specific case. In my opinion blacklisted-name is a very simple check that should happen on all name so it does not have to be mixed with other checkers (it just is right now for some reason).

I don't have a strong opinion on this, I would merge any reasonable looking change that add the functional tests you gave as an example while still passing the old tests.

[relrod]

@Pierre-Sassoulas I took a different approach, just splitting out blacklisted-name and non-ascii-name (also affected) into helper functions and calling those when we can't call the whole _check_name() helper. It seemed like a way less invasive change.

CI is currently broken (independently of this change) due to this change in astroid. I didn't attempt to debug it, mainly because I'm not sure if this is an astroid regression or that something needs to be updated in pylint to account for the breaking change.