BUG: Fix incorrect FutureWarning for logical ops on pyarrow bool Series (#62260)

[Tarun2605]

This pull request introduces support for Kleene's three-valued logic (handling True, False, and NA) in pandas logical operations on arrays containing missing values. The main changes include new helper functions to safely evaluate boolean logic with missing values, modifications to the logical operation implementation, and updates to tests to reflect the new behavior.

Enhancements to logical operations with missing values:

• Added is_nullable_bool, safe_is_true, and alignOutputWithKleene helper functions in pandas/core/ops/array_ops.py to enable elementwise logical operations that correctly handle NA values using Kleene logic.
• Modified logical_op in pandas/core/ops/array_ops.py to use Kleene logic when both operands are boolean arrays (possibly with NA), ensuring correct propagation of unknowns.

Test updates for new logic:

• Updated expected results in TestDataFrameLogicalOperators in pandas/tests/frame/test_logical_ops.py to match the new Kleene logic semantics, where logical operations with NA now return NA or propagate True/False according to Kleene's rules.
• Adjusted the test_logical_with_nas test to expect results consistent with Kleene logic, ensuring that logical operations involving NA and True/False yield the correct outcomes.- [x] closes BUG: Incorrect Future warning using a logical operation between two pyarrow boolean series #62260

• tests added / passed
• passes pre-commit run --all-files
• whatsnew entry

What does this PR change?

This pull request updates the Arrow extension array implementation in pandas to improve how missing values are handled during logical operations on bool.

Previously, operations like & or | between two bool series was not following Kleene's Logic.

How was this fixed?

• Implemented my own alignWithKleene function to align the output of the logical operation with Kleene principle

[Tarun2605]

Kindly review my PR for improving the logical operation of arrays and aligning them with Kleene's Principles. Please tell me if any issues. Finally heading out to touch some grass :)

[Tarun2605]

@simonjayhawkins Kindly review my PR and let me know 👍 (If in testing phase or not)

[Alvaro-Kothe]

This Wikipedia article may be useful to understand the changes in this PR: https://en.wikipedia.org/wiki/Three-valued_logic#Kleene_and_Priest_logics

It's also important to point-out that this is a breaking change and should have an entry in doc/source/whatsnew/v3.0.0.rst.

[Alvaro-Kothe]

Also, bool[pyarrow] already implements 3VL

import pandas as pd

index = list("FUT")
a = pd.Series([False, None, True], index=index, dtype="bool[pyarrow]")
t = pd.Series([True] * 3, index=index, dtype = "bool[pyarrow]")
u = pd.Series([None] * 3, index=index, dtype = "bool[pyarrow]")
f = pd.Series([False] * 3, index=index, dtype = "bool[pyarrow]")

print("negation")
print(~a)

methods = ["__and__", "__or__", "__xor__"]
for method in methods:
    print(method)
    fn = getattr(a, method)
    observed = pd.DataFrame(dict(F=fn(f), U=fn(u), T=fn(t)), index=index)
    print(observed)

Output

negation
F     True
U     <NA>
T    False
dtype: bool[pyarrow]
__and__
       F      U      T
F  False  False  False
U  False   <NA>   <NA>
T  False   <NA>   True
__or__
       F     U     T
F  False  <NA>  True
U   <NA>  <NA>  True
T   True  True  True
__xor__
       F     U      T
F  False  <NA>   True
U   <NA>  <NA>   <NA>
T   True  <NA>  False

[Tarun2605]

@Alvaro-Kothe Yes I agree, the pyarrow implementation follows the kleene principle whereas bool does not. Thank you for attaching the wiki article 👍 The core members of pandas lib are the ones who fill the whatsnew right?

[Alvaro-Kothe]

@Tarun2605 Usually, whoever creates the pull request should fill the whatsnew

[Tarun2605]

@Alvaro-Kothe Ohhhh!! Could you please tell me where do I fill it out? Thank you so much btw man

[Alvaro-Kothe]

Utilize your own discretion or await guidance from a core member.

[Tarun2605]

Okkk