Resolve used-before-assignment false negative when a function is defined under TYPE_CHECKING

[zenlyj]

Type of Changes

	Type
✓	🐛 Bug fix
	✨ New feature
	🔨 Refactoring
	📜 Docs

Description

Raise used-before-assignment when a function is defined under TYPE_CHECKING and used later.

Also extending the nonlocal handling to resolve the following false positive that was discovered by the primer tests, where a variable usage references a nonlocal binding from an outer, enclosing frame.

These changes mainly include a search for nonlocal declarations in all enclosing frames of the variable usage. The idea is to rely on existing nonlocal checks and to only check outer frames if all nodes are filtered out by the consumer.

Refs #10028

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (main@15a5ac0). Learn more about missing BASE report.
Report is 5 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #10034   +/-   ##
=======================================
  Coverage        ?   95.80%           
=======================================
  Files           ?      174           
  Lines           ?    18956           
  Branches        ?        0           
=======================================
  Hits            ?    18160           
  Misses          ?      796           
  Partials        ?        0           

Files with missing lines	Coverage Δ
pylint/checkers/variables.py	97.24% <100.00%> (ø)

[jacobtylerwalls]

Thanks so much @zenlyj, is this ready for review?

[zenlyj]

i don't think the current approach is correct, it silences the check in many cases and something like this is not flagged even though it would blow up in runtime:

def outer(callback):
    def inner():
        nonlocal callback
        def inner2():
            callback()  # this should raise used-before-assignment
            def callback():
                pass
        inner2()
    inner()

def foo():
    print('foo')

outer(foo)

[zenlyj]

@jacobtylerwalls this is ready for review. let me know if there is anything i missed here, thank you!

[Pierre-Sassoulas]

LGTM, thank you, great work. Do you mind adding a fragment for the changelog ? (I'll also let Jacob review before merging).

[jacobtylerwalls]

I'll handle the Home Assistant false positive in another PR.

[zenlyj]

hey @jacobtylerwalls, i was just thinking about this case:

def nonlocal_in_outer_frame_ok(callback, condition_a, condition_b):
    def outer():
        nonlocal callback
        def inner():
            if condition_a:
                def inner2():
                    callback()  # possibly-used-before-assignment?
                inner2()
            else:
                if condition_b:
                    def callback():
                        pass
        inner()
    outer()

we should probably raise a message here?

[jacobtylerwalls]

we should probably raise a message here?

Good catch, I'm fine with opening a new issue for it. I think it should be real UBA, not "possibly".

[zenlyj]

it is worth noting that prior to my changes in this PR, we're emitting possibly-used-before-assignment for this snippet. but i feel like this will be tough to crack as it requires some form of control flow analysis?

[github-actions]

🤖 Effect of this PR on checked open source code: 🤖

Effect on home-assistant:
The following messages are now emitted:

1. possibly-used-before-assignment:
   Possibly using variable 'touch_event_callback' before assignment
   https://github.com/home-assistant/core/blob/b09e54c961db279785b75b5c3d192624b3d65664/homeassistant/components/nanoleaf/__init__.py#L77

This comment was generated for commit 3808c4e

[Pierre-Sassoulas]

Hello @zenlyj. You've been doing a lot of great work for pylint lately, including this one, would you be interested in becoming a triager ? (I would normally have asked you privately but your commit mail is a no reply :) )

[zenlyj]

Hello @zenlyj. You've been doing a lot of great work for pylint lately, including this one, would you be interested in becoming a triager ? (I would normally have asked you privately but your commit mail is a no reply :) )

hello, i just dropped you an email, let's talk there