BUG: Series.clip fails with 0-dimensional numpy arrays

[maliktafheem]

Summary

Fixes #59053

Series.clip() raises TypeError: len() of unsized object when passed 0-dimensional numpy arrays as lower or upper bounds.

Problem

import numpy as np
import pandas as pd

pd.Series([-1, 2, 3]).clip(lower=np.array(0))
# TypeError: len() of unsized object

Root Cause

In Series._flex_method, when other is a numpy array, the code immediately calls len(other) without checking if it's a 0-dimensional array. 0-dimensional arrays have no length and should be treated as scalars.

Solution

Before the length check, detect 0-dimensional numpy arrays and convert them to scalars using .item():

elif isinstance(other, (np.ndarray, list, tuple, ExtensionArray)):
    # Handle 0-dimensional numpy arrays as scalars (GH#59053)
    if isinstance(other, np.ndarray) and other.ndim == 0:
        return op(self, other.item())
    if len(other) != len(self):
        raise ValueError("Lengths must be equal")

Testing

Added test_clip_with_zerodim_ndarray test case covering:

• clip(lower=np.array(0))
• clip(upper=np.array(1))
• clip(lower=np.array(0), upper=np.array(2))

[biplavbarua]

Changes look good!

I verified that Series.clip relies on _clip_with_one_bound, which calls Series.ge/le, eventually hitting _flex_method. Intercepting the 0-d array here and unboxing it with .item() is the correct fix at the root level.

One minor suggestion: Since the current test uses np.array(0) (integer), maybe strictly add a case for np.array(1.5) (float) to ensure type casting behaviors remain consistent?

LGTM.

[maliktafheem]

Thanks for the review and suggestion @biplavbarua! Added a float 0-d array test case (np.array(0.5), np.array(3.0)) to ensure type casting consistency.

[biplavbarua]

Thanks @maliktafheem, the tests in b7f9678 look perfect. Appreciate the quick update! LGTM.