cloudpathlib中S3 Bucket本地实现未重置默认存储目录问题

按照cloudpathlib的S3本地实现模式编写测试，两个测试用例分别验证桶中有txt文件时返回True、空桶时返回False。批量运行pytest时第二个测试失败，单独运行则正常。请问本地S3桶的配置是否正确？

待测试代码 #

from cloudpathlib import CloudPath


def main(s3_bucket: str):
    s3_bucket_path = CloudPath(s3_bucket)
    if list(s3_bucket_path.glob('*.txt')):
        return True
    else:
        return False

测试代码 #

import pytest
from cloudpathlib import implementation_registry
from cloudpathlib.local import LocalS3Client, local_s3_implementation

from cloudpath import main


@pytest.fixture
def local_bucket_name():
    return 'cloudpathlib-test-bucket'


@pytest.fixture
def local_s3_bucket(monkeypatch, local_bucket_name):
    """
    Fixture that patches CloudPath dispatch and create bucket in LocalS3Client
    """
    monkeypatch.setitem(implementation_registry, 's3', local_s3_implementation)

    local_pathlib_path = LocalS3Client.get_default_storage_dir() / local_bucket_name
    local_pathlib_path.mkdir(exist_ok=True, parents=True)

    yield local_pathlib_path

    LocalS3Client.reset_default_storage_dir()  # clean up temp directory and replace with new one


class TestMain:
    def test_returns_true_if_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        (local_s3_bucket / 'test.txt').touch()

        assert main(f's3://{local_bucket_name}')

    def test_returns_false_if_bucket_does_not_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        assert not main(f's3://{local_bucket_name}')

你的配置存在两个核心问题：

1. LocalS3Client单例缓存：cloudpathlib的LocalS3Client是单例模式，第一个测试创建的client实例会被缓存，它指向的是旧的存储目录。第二个测试创建CloudPath时会复用这个旧实例，导致仍然读取第一个测试中创建的test.txt文件。
2. reset_default_storage_dir()的局限性：该方法仅更换默认存储目录的路径，但不会清除已缓存的client实例，也不会删除旧目录中的文件。

有两种简单有效的修复方式，任选其一即可：

方案1：使用pytest tmp_path实现完全隔离的测试目录 #

利用pytest内置的tmp_path fixture，为每个测试分配独立的临时存储目录，彻底避免测试间的污染：

import pytest
from cloudpathlib import implementation_registry
from cloudpathlib.local import LocalS3Client, local_s3_implementation

from cloudpath import main


@pytest.fixture
def local_bucket_name():
    return 'cloudpathlib-test-bucket'


@pytest.fixture
def local_s3_bucket(monkeypatch, local_bucket_name, tmp_path):
    monkeypatch.setitem(implementation_registry, 's3', local_s3_implementation)
    
    # 为当前测试设置独立的存储目录
    LocalS3Client.set_default_storage_dir(tmp_path)
    
    local_pathlib_path = tmp_path / local_bucket_name
    local_pathlib_path.mkdir(exist_ok=True, parents=True)
    
    yield local_pathlib_path
    
    # 无需手动清理，pytest会自动处理tmp_path


class TestMain:
    def test_returns_true_if_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        (local_s3_bucket / 'test.txt').touch()

        assert main(f's3://{local_bucket_name}')

    def test_returns_false_if_bucket_does_not_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        assert not main(f's3://{local_bucket_name}')

方案2：清除LocalS3Client单例缓存 #

在fixture的清理阶段，删除单例实例，确保下一个测试创建全新的client：

import pytest
from cloudpathlib import implementation_registry
from cloudpathlib.local import LocalS3Client, local_s3_implementation

from cloudpath import main


@pytest.fixture
def local_bucket_name():
    return 'cloudpathlib-test-bucket'


@pytest.fixture
def local_s3_bucket(monkeypatch, local_bucket_name):
    monkeypatch.setitem(implementation_registry, 's3', local_s3_implementation)

    local_pathlib_path = LocalS3Client.get_default_storage_dir() / local_bucket_name
    # 每次测试前清空并重建桶目录
    if local_pathlib_path.exists():
        import shutil
        shutil.rmtree(local_pathlib_path)
    local_pathlib_path.mkdir(exist_ok=True, parents=True)

    yield local_pathlib_path

    # 重置存储目录并清除单例缓存
    LocalS3Client.reset_default_storage_dir()
    if hasattr(LocalS3Client, '_instance'):
        delattr(LocalS3Client, '_instance')


class TestMain:
    def test_returns_true_if_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        (local_s3_bucket / 'test.txt').touch()

        assert main(f's3://{local_bucket_name}')

    def test_returns_false_if_bucket_does_not_contains_txt_files(self, local_s3_bucket, local_bucket_name):
        assert not main(f's3://{local_bucket_name}')

两种方案都能解决测试污染问题：

• 方案1更符合pytest的最佳实践，利用内置的临时目录管理，无需手动清理。
• 方案2则在原有代码基础上修改，适合不想调整存储目录策略的场景。

内容的提问来源于stack exchange，提问作者Lucie