apache / apache/iceberg-python

Implementation classes named in table properties are imported and instantiated without a type check

オープン
#3,932 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る
bug
主要言語
Python
スター
1.1k
フォーク
581
平均マージ
1日 17時間
マージ済み PR(30日)
78

説明

Three properties name a class that PyIceberg imports and calls:

| Property | Resolved by | Called as |
|---|---|---|
| `py-io-impl` | `_import_file_io` (`pyiceberg/io/__init__.py`) | `class_(properties)` |
| `write.py-location-provider.impl` | `_import_location_provider` (`pyiceberg/table/locations.py`) | `class_(table_location, table_properties)` |
| `s3.retry-strategy-impl` | `_import_retry_strategy` (`pyiceberg/io/pyarrow.py`) | `class_()` |

Each follows the same pattern:

```python
module = importlib.import_module(module_name)
class_ = getattr(module, class_name)
return class_(...)
```

None of the three checks that the resolved object is the type it is about to be used as — there is no `issubclass` against `FileIO`, `LocationProvider`, or `S3RetryStrategy`. Any importable dotted name resolves and is called, with the property map passed as an argument in two of the three cases.

All three properties are read from the merged table property map, so their values can originate in a table's metadata rather than in the operator's catalog configuration.

---
Issue investigation generated via claude, reviewed by Sung, Kevin, Fokko.

コントリビューションガイド

このリポジトリのコントリビューションガイドは索引されていません

調査の方向性

Start with _import_file_io in pyiceberg/io/__init__.py, _import_location_provider in pyiceberg/table/locations.py, and _import_retry_strategy in pyiceberg/io/pyarrow.py. Trace how each property is resolved and invoked, then verify that only subclasses of FileIO, LocationProvider, or S3RetryStrategy are accepted before instantiation; done means all three paths reject unrelated importable classes.

索引モデルが issue の本文から書いたものです。

評価

技術スタック
python
領域
security
issue の種類
バグ
難易度
3/5
見積もり時間
1〜2日
活発さ
活発
明瞭さ
おおむね明確
初心者へのやさしさ
68/100

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。