Repository navigation
[BUG] create_dataset() raises IndexError instead of ValueError on empty NumPy array #1777
Description
Description:
When create_dataset() is called with a zero-row NumPy array, the
function raises an IndexError instead of a descriptive ValueError.
The error originates from an unchecked data[0] access inside the
format-detection block. The message gives zero indication of what
actually went wrong — a user passing an empty DataFrame export or a
filtered-down dataset with no rows gets a raw index crash with no
context.
No warning, no descriptive message, no hint that the root cause is
simply an empty array. The function just crashes silently with a raw
IndexError pointing deep into openml internals.
Expected Behavior:
A clear ValueError that tells the user exactly what the problem is:
ValueError: 'data' must contain at least one row,
but got an array with shape (0, 2).
Environment
- openml-python version: 0.15.1
- Python version: 3.12
- OS: Ubuntu 24
Error messages, stack traces, or logs
No descriptive error is raised. Just a raw IndexError with no context:
IndexError: index 0 is out of bounds for axis 0 with size 0
The traceback points inside openml internals with no hint that
the root cause is simply an empty array being passed in.
Verified with arrays of shape (0, 2), (0, 5), (0, 10) — all
produce the same IndexError with no useful message.Steps to reproduce
import numpy as np
import openml
openml.datasets.create_dataset(
name="empty_test",
description="Reproduces the IndexError bug",
creator=None,
contributor=None,
collection_date=None,
language=None,
licence=None,
attributes=[("feature1", "NUMERIC"), ("target", "NUMERIC")],
data=np.array([]).reshape(0, 2), # 0 rows — triggers the bug
default_target_attribute="target",
ignore_attribute=None,
citation="",
)
# Expected: ValueError: 'data' must contain at least one row,
# but got an array with shape (0, 2).
# Actual: IndexError: index 0 is out of bounds for axis 0 with size 0Additional context
Root cause is this unchecked data[0] access in datasets/functions.py:
# BEFORE (buggy)
if isinstance(data, (list, np.ndarray)):
if isinstance(data[0], (list, np.ndarray)): # crashes if data is empty
data_format = "arff"
elif isinstance(data[0], dict):
data_format = "sparse_arff"
# AFTER (suggested fix)
if isinstance(data, (list, np.ndarray)):
if len(data) == 0: # guard added
raise ValueError(
f"'data' must contain at least one row, "
f"but got an array with shape {np.array(data).shape}."
)
if isinstance(data[0], (list, np.ndarray)):
data_format = "arff"
elif isinstance(data[0], dict):
data_format = "sparse_arff"Confirmed across 3 consecutive runs on openml==0.15.1 / Python 3.12 / Ubuntu 24.
Happy to open a PR with the fix if this is confirmed.
Activity
RugvedBane commentedon Oct 15, 2026
@L-Agbaria, could you confirm whether this is a bug or intended behavior?