huggingface / huggingface/datasets

function `load_dataset` can't solve folder path with regex characters like "[]"

Open
#7,468 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
22k
Forks
3.4k
Avg merge
5d 7h
Merged PRs (30d)
17

Description

### Describe the bug

When using the `load_dataset` function with a folder path containing regex special characters (such as "[]"), the issue occurs due to how the path is handled in the `resolve_pattern` function. This function passes the unprocessed path directly to `AbstractFileSystem.glob`, which supports regular expressions. As a result, the globbing mechanism interprets these characters as regex patterns, leading to a traversal of the entire disk partition instead of confining the search to the intended directory.

### Steps to reproduce the bug

just create a folder like `E:\[D_DATA]\koch_test`, then `load_dataset("parquet", data_dir="E:\[D_DATA]\\test", split="train")`
it will keep searching the whole disk.

I add two `print` in `glob` and `resolve_pattern` to see the path

### Expected behavior

it should load the dataset as in normal folders

### Environment info

- `datasets` version: 3.3.2
- Platform: Windows-10-10.0.22631-SP0
- Python version: 3.10.16
- `huggingface_hub` version: 0.29.1
- PyArrow version: 19.0.1
- Pandas version: 2.2.3
- `fsspec` version: 2024.12.0

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.