BUG: `pandas.array` works fine when 2-D array contains string

[antareepsarkar]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd

a = pd.array(np.array([[1], [2]])) #Raises an error

a = pd.array(np.array([[1], ['a']])) #Does not raise an error

Issue Description

pandas.array API mentions that it expects 1-D arrays as input. But, in some cases, it works fine with 2-D arrays.
Should not it raise an error?

Expected Behavior

It should show the same behaviour for same kind of input.

Installed Versions

Installed Versions

python : 3.12.12
OS : Windows
OS-release : 11

pandas : 2.3.3
numpy : 2.3.4

[Alvaro-Kothe]

On main, pd.array(np.array([[1], ['a']])) raises

ArrowInvalid: only handle 1-dimensional arrays

I checked this problem on main against several data types using nested lists:

import pandas as pd
import pandas as pd
import numpy as np

int_2d = [[1], [2]]
float_2d = [[1.0], [2.0]]
string_2d = [["a"], ["b"]]
mixed_2d = [[1, 2], ["a", "b"]]
np_object21 = np.array([[1], ["a"]])
np_object22 = np.array([[1, 2], ["a", "b"]])

# raises in Cython
# ValueError: Buffer has wrong number of dimensions (expected 1, got 2)
# print(pd.array(int_2d)) # raises
# print(pd.array(float_2d)) # raises
# print(pd.array(string_2d)) # raises
# print(pd.array(mixed_2d)) # raises

# print(pd.array(int_2d, dtype=pd.Int64Dtype())) # TypeError: values must be a 1D list-like
# print(pd.array(float_2d, dtype=pd.Float64Dtype())) # TypeError: values must be a 1D list-like
# print(pd.array(np_object21, dtype=pd.StringDtype())) # pyarrow.lib.ArrowInvalid: only handle 1-dimensional arrays
# print(pd.array(np_object22, dtype=pd.StringDtype())) # pyarrow.lib.ArrowInvalid: only handle 1-dimensional arrays

print(pd.array(string_2d, dtype=pd.StringDtype()))
# <ArrowStringArray>
# ['['a']', '['b']']
# Length: 2, dtype: string

print(pd.array(mixed_2d, dtype=pd.StringDtype()))
# <ArrowStringArray>
# ['[1]', '['a']']
# Length: 2, dtype: string

print(pd.array(mixed_2d, dtype=object))
# <NumpyExtensionArray>
# [[1], ['a']]
# Length: 2, dtype: object

print(pd.array(np_object21, dtype=object))
# <NumpyExtensionArray>
# [
# ['1'],
# ['a']
# ]
# Shape: (2, 1), dtype: object
print(a := pd.array(np_object22, dtype=object))
# <NumpyExtensionArray>
# [
# ['1', '2'],
# ['a', 'b']
# ]
# Shape: (2, 2), dtype: object
print(a[:, 1])
# <NumpyExtensionArray>
# ['2', 'b']
# Length: 2, dtype: object

The object case with mixed2d doesn't seem problematic, because it can be considered an 1D array of python objects. The string case seems weird to convert everything to the string representation of the object. The other object cases seems wrong that it's returning a 2D object.

In the docs states:

Raises:
ValueError
When data is not 1-dimensional.

[rhshadrach]

The object case with mixed2d doesn't seem problematic, because it can be considered an 1D array of python objects. The string case seems weird to convert everything to the string representation of the object.

Agreed, string case should be fixed here.

The other object cases seems wrong that it's returning a 2D object.

pandas/pandas/core/arrays/numpy_.py

Line 122 in 27a3271

# Technically we support 2, but do not advertise that fact.