参数注入装饰器与pytest fixtures冲突问题求解

pytest在测试收集阶段就会解析测试函数的参数，尝试为每个参数匹配对应的fixture，这个过程远早于装饰器的执行阶段。你的装饰器是在函数运行时才注入path_calibration参数，但pytest提前把这个参数当成fixture去查找，找不到就直接抛出FixtureLookupError，甚至没机会触发装饰器内的断点。

方案1：将装饰器逻辑转为pytest Fixture（推荐） #

把auto_asset的功能改成可参数化的fixture，让pytest能正确识别参数逻辑，避免冲突：

import pytest
from pathlib import Path, Optional

def load_asset(name: str, path: Optional[Path] = None) -> Path:
    # 保留原load_asset实现
    return path or Path(f"./assets/{name}")

@pytest.fixture
def auto_asset(request):
    # 读取测试函数上的标记参数
    marker = request.node.get_closest_marker("auto_asset")
    if not marker:
        pytest.skip("未标记auto_asset参数")
    
    name = marker.args[0]
    path = marker.kwargs.get("path")
    return_path = marker.kwargs.get("return_path", True)
    
    asset_path = load_asset(name=name, path=path)
    if return_path:
        return {f"path_{name}": asset_path}
    return asset_path

@pytest.fixture()
def my_fixture():
    return "test_fixture"

# 使用@pytest.mark替代原装饰器
@pytest.mark.auto_asset("calibration", return_path=True)
def test_something(auto_asset, my_fixture):
    path_calibration = auto_asset["path_calibration"]
    # 测试逻辑
    print(path_calibration, my_fixture)

方案2：调整测试函数参数写法 #

修改测试函数，用**kwargs接收注入的参数，让pytest不把path_calibration当作fixture查找：

# 原装饰器保持不变
@auto_asset("calibration", return_path=True)
def test_something(my_fixture, **kwargs):
    path_calibration = kwargs["path_calibration"]
    # 测试逻辑

这种方法无需修改装饰器，但会失去显式参数的可读性。

方案3：修改装饰器动态调整函数签名 #

通过inspect模块修改装饰后函数的签名，移除pytest会识别为fixture的参数，同时保留原函数的参数接收逻辑：

import inspect
from functools import wraps
from pathlib import Path, Optional

def load_asset(name: str, path: Optional[Path] = None) -> Path:
    return path or Path(f"./assets/{name}")

def auto_asset(name, path: Optional[Path] = None, return_path: bool = True):
    def wrap_fn(fn):
        param_name = f"path_{name}"
        # 获取原函数签名，移除要注入的参数
        sig = inspect.signature(fn)
        if param_name in sig.parameters:
            new_params = [p for p_name, p in sig.parameters.items() if p_name != param_name]
            new_sig = sig.replace(parameters=new_params)
        else:
            new_sig = sig

        @wraps(fn)
        def wrapped_fn(*args, **kwargs):
            asset_path = load_asset(name=name, path=path)
            if return_path:
                kwargs.update({param_name: asset_path})
            # 绑定参数并传递给原函数
            bound_args = new_sig.bind(*args, **kwargs)
            bound_args.apply_defaults()
            return fn(*bound_args.args, **bound_args.kwargs)

        # 替换包装函数的签名，让pytest忽略注入参数
        wrapped_fn.__signature__ = new_sig
        return wrapped_fn
    return wrap_fn

@pytest.fixture()
def my_fixture():
    return "test_fixture"

@auto_asset("calibration", return_path=True)
def test_something(path_calibration, my_fixture):
    print(path_calibration, my_fixture)

内容的提问来源于stack exchange，提问作者Haydnspass