Strange behaviour for equal variable and function name

[SirJoeth3rd]

Bug report

Bug description:

Hello, at my job we recently encountered a strange bug that resulted in a UnboundLocalError.

The following is from a ipython terminal.

In [2]: def example_one():
   ...:     x = x()

In [3]: dis.dis(example_one)
  2           0 LOAD_FAST                0 (x)
              2 CALL_FUNCTION            0
              4 STORE_FAST               0 (x)
              6 LOAD_CONST               0 (None)
              8 RETURN_VALUE

In [4]: def example_two():
   ...:     x = y()

In [5]: dis.dis(example_two)
  2           0 LOAD_GLOBAL              0 (y)
              2 CALL_FUNCTION            0
              4 STORE_FAST               0 (x)
              6 LOAD_CONST               0 (None)
              8 RETURN_VALUE

As you can see in example_one python is attempting to fetch the x function locally. Wouldn't expected behaviour be like example_two where the function is loaded from global and the name is then reassigned in the local scope?

It's obviously stupid rare (and bad practice) to shadow a function name like this, but still worth a mention.

CPython versions tested on:

3.10, 3.14

Operating systems tested on:

Windows, Linux

[SirJoeth3rd]

Forgot to mention, example_one will result in a UnboundLocalError even if the x function is defined globally.

[ZeroIntensity]

I think this would require some weird special-casing to fix, and would be a breaking change to some degree. I'm probably used to Python's quirks, but this actually makes perfect sense to me; the interpreter knows the local variables of a function in advance, and as such, if you want to lookup from the global namespace to shadow a local, you need to use the global keyword. That's always been the practice.

I guess we could make the error nicer, somehow? What could be improved in the error message?

[serhiy-storchaka]

This works as expected. x is a local variable, because there is an assignment, y is global. The variable becames local not after assignment, but if there is any assignment (or modification) to it in the function. Alternative behavior would lead to writing more error prone code.