Bug: char.isnumeric returns True for empty strings (NumPy mismatch)

[ajpotts]

Bug: char.isnumeric returns True for empty strings (NumPy mismatch)

Summary

arkouda.numpy.char.isnumeric does not match NumPy/Python semantics for empty strings.
Specifically, Arkouda returns True for "", while NumPy (np.char.isnumeric) and Python (str.isnumeric) return False.

This was discovered via a new NumPy-alignment verification test.

---

Expected Behavior (NumPy / Python)

"".isnumeric()              # False
np.char.isnumeric([""])    # array([False])

---

Actual Behavior (Arkouda)

ak.char.isnumeric(ak.array([""]))  
# pdarray([True])

This causes a mismatch in NumPy-alignment tests.

---

Failing Test

tests/numpy/alignment_verification/char_alignment.py::
TestArkoudaNumpyCharAlignment::test_isnumeric_matches_numpy_char

Mismatch at index 0:
value=''
got=True
expected=False

---

Root Cause (Likely)

The current implementation treats strings as numeric if all characters are numeric.
For empty strings, this condition is vacuously true, resulting in an incorrect True.

---

Proposed Fix

Update Strings.isnumeric() (or arkouda.numpy.char.isnumeric) to explicitly treat empty strings as non-numeric, matching NumPy semantics.

Conceptually:

isnumeric(s) = (len(s) > 0) AND all_characters_numeric(s)

Possible implementation:

• AND the existing result with (self != ""), or
• AND with (string_lengths > 0) if lengths are readily available.

---

Why This Matters

• Aligns Arkouda string semantics with NumPy and Python
• Prevents subtle downstream bugs in pandas/NumPy compatibility layers
• Strengthens the new NumPy-alignment test suite

---

Environment

• Arkouda client: 2025.12.16+
• Arkouda server: 2025.12.16+ (minor version mismatch warning present but not believed to be causal)
• Tests executed against a live Arkouda server