pyproject.toml 完全指南：Python 项目配置现代化

特点：

• 简洁易读
• 支持注释
• 类型明确（字符串、数字、布尔值、数组、表）

三、pyproject.toml 的核心结构

一个完整的 pyproject.toml 通常包含以下几个部分：

[build-system] # 构建系统配置
[project]      # 项目元数据
[project.optional-dependencies] # 可选依赖
[tool.xxx]     # 各种工具的配置

让我们逐一解析。

四、详细解析各个部分

4.1 [build-system] - 构建系统配置

这是 pyproject.toml 的必需部分，告诉 Python 如何构建你的项目。

[build-system]
requires = ["setuptools>=45", "wheel"]
build-backend = "setuptools.build_meta"

字段说明：

常见构建后端：

hatchling（新兴工具）

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

flit（轻量级）

[build-system]
requires = ["flit_core >=3.2,<4"]
build-backend = "flit_core.buildapi"

poetry（现代化工具）

[build-system]
requires = ["poetry-core>=1.0.0"]
build-backend = "poetry.core.masonry.api"

setuptools（传统方式）

[build-system]
requires = ["setuptools>=45", "wheel"]
build-backend = "setuptools.build_meta"

4.2 [project] - 项目元数据

定义项目的基本信息和依赖关系。

[project]
name = "my-awesome-package"
version = "1.0.0"
description = "一个很棒的 Python 包"
readme = "README.md"
requires-python = ">=3.8"
license = {text = "MIT"}
authors = [
    {name = "张三", email = "[email protected]"}
]
maintainers = [
    {name = "李四", email = "[email protected]"}
]
keywords = ["web", "api", "framework"]
classifiers = [
    "Development Status :: 4 - Beta",
    "Intended Audience :: Developers",
    "License :: OSI Approved :: MIT License",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.8",
    "Programming Language :: Python :: 3.9",
    "Programming Language :: Python :: 3.10",
]
dependencies = [
    "requests>=2.28.0,<3.0.0",
    "click>=8.0.0",
    "pydantic>=2.0.0"
]

[project.urls]
Homepage = "https://github.com/username/my-awesome-package"
Documentation = "https://my-awesome-package.readthedocs.io"
Repository = "https://github.com/username/my-awesome-package"
Changelog = "https://github.com/username/my-awesome-package/blob/main/CHANGELOG.md"

字段详解：

4.3 [project.optional-dependencies] - 可选依赖

用于定义不同场景下的额外依赖，比如开发、测试、文档等。

[project.optional-dependencies]
# 开发依赖
dev = [
    "pytest>=7.0",
    "pytest-cov>=4.0",
    "black>=22.0",
    "ruff>=0.1.0",
    "mypy>=1.0"
]
# 测试依赖
test = [
    "pytest>=7.0",
    "pytest-asyncio>=0.21.0",
    "httpx>=0.24.0"
]
# 文档依赖
docs = [
    "sphinx>=4.0",
    "sphinx-rtd-theme>=1.0"
]
# 所有依赖
all = ["my-awesome-package[dev,test,docs]"]

使用方式：

# 安装基础依赖
pip install my-awesome-package
# 安装开发依赖
pip install my-awesome-package[dev]
# 安装多个可选依赖组
pip install my-awesome-package[dev,test]
# 安装所有依赖
pip install my-awesome-package[all]

4.4 [project.scripts] - 命令行入口

定义安装后可在命令行直接调用的命令。

[project.scripts]
my-cli = "my_package.cli:main"
my-tool = "my_package.tools:run"

示例场景：

假设你有以下代码结构：

my_package/
├── __init__.py
└── cli.py

cli.py 内容：

def main():
    print("Hello from my-cli!")

if __name__ == "__main__":
    main()

安装后，用户可以直接运行：

my-cli
# 输出：Hello from my-cli!

4.5 [tool.*] - 工具配置

这是 pyproject.toml 最强大的功能之一：集中管理各种开发工具的配置。

4.5.1 pytest（测试框架）

[tool.pytest.ini_options]
testpaths = ["tests"]
python_files = ["test_*.py", "*_test.py"]
python_classes = ["Test*"]
python_functions = ["test_*"]
addopts = [
    "--cov=my_package",
    "--cov-report=html",
    "--cov-report=term-missing",
    "-v"
]

4.5.2 black（代码格式化）

[tool.black]
line-length = 88
target-version = ['py38', 'py39', 'py310']
include = '\.pyi?$'
extend-exclude = '''
/(
    # 排除的目录
    .eggs |
    .git |
    .venv |
    build |
    dist
)/
'''

4.5.3 ruff（快速 Linter）

[tool.ruff]
line-length = 88
target-version = "py38"

[tool.ruff.lint]
select = [
    "E", # pycodestyle errors
    "W", # pycodestyle warnings
    "F", # pyflakes
    "I", # isort
    "B", # flake8-bugbear
    "C4", # flake8-comprehensions
]
ignore = [
    "E501", # 行过长（由 black 处理）
    "B008", # 函数调用中的参数默认值
]

[tool.ruff.lint.per-file-ignores]
"__init__.py" = ["F401"] # 允许未使用的导入

4.5.4 mypy（类型检查）

[tool.mypy]
python_version = "3.8"
warn_return_any = true
warn_unused_configs = true
disallow_untyped_defs = true
disallow_incomplete_defs = true
check_untyped_defs = true
no_implicit_optional = true

[[tool.mypy.overrides]]
module = "tests.*"
disallow_untyped_defs = false

4.5.5 coverage（代码覆盖率）

[tool.coverage.run]
source = ["my_package"]
omit = [
    "*/tests/*",
    "*/__init__.py"
]

[tool.coverage.report]
exclude_lines = [
    "pragma: no cover",
    "def __repr__",
    "raise AssertionError",
    "raise NotImplementedError",
    "if __name__ == .__main__.:",
]

五、实战案例：完整的 pyproject.toml

下面是一个生产环境级别的完整示例：

[build-system]
requires = ["setuptools>=68.0", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "data-processor"
version = "2.3.1"
description = "一个强大的数据处理库"
readme = "README.md"
requires-python = ">=3.8"
license = {text = "MIT"}
authors = [
    {name = "数据团队", email = "[email protected]"}
]
keywords = ["data", "processing", "etl", "pipeline"]
classifiers = [
    "Development Status :: 5 - Production/Stable",
    "Intended Audience :: Developers",
    "License :: OSI Approved :: MIT License",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.8",
    "Programming Language :: Python :: 3.9",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Topic :: Software Development :: Libraries :: Python Modules",
]
dependencies = [
    "pandas>=2.0.0,<3.0.0",
    "numpy>=1.24.0",
    "sqlalchemy>=2.0.0",
    "pydantic>=2.0.0",
    "click>=8.0.0",
]

[project.optional-dependencies]
dev = [
    "pytest>=7.4.0",
    "pytest-cov>=4.1.0",
    "pytest-asyncio>=0.21.0",
    "black>=23.0.0",
    "ruff>=0.1.0",
    "mypy>=1.5.0",
    "pre-commit>=3.3.0",
]
postgres = ["psycopg2-binary>=2.9.0"]
mysql = ["pymysql>=1.1.0"]
all = ["data-processor[dev,postgres,mysql]"]

[project.urls]
Homepage = "https://github.com/company/data-processor"
Documentation = "https://data-processor.readthedocs.io"
Repository = "https://github.com/company/data-processor"
Issues = "https://github.com/company/data-processor/issues"

[project.scripts]
dp = "data_processor.cli:main"

# pytest 配置
[tool.pytest.ini_options]
testpaths = ["tests"]
addopts = [
    "--cov=data_processor",
    "--cov-report=html",
    "--cov-report=term-missing:skip-covered",
    "-v",
    "--strict-markers",
]
markers = [
    "slow: 标记慢速测试",
    "integration: 集成测试",
    "unit: 单元测试",
]

# black 配置
[tool.black]
line-length = 100
target-version = ['py38', 'py39', 'py310', 'py311']
include = '\.pyi?$'

# ruff 配置
[tool.ruff]
line-length = 100
target-version = "py38"

[tool.ruff.lint]
select = ["E", "F", "I", "B", "C4", "UP"]
ignore = ["E501"]

[tool.ruff.lint.isort]
known-first-party = ["data_processor"]

# mypy 配置
[tool.mypy]
python_version = "3.8"
warn_return_any = true
warn_unused_configs = true
disallow_untyped_defs = true
plugins = ["pydantic.mypy"]

# coverage 配置
[tool.coverage.run]
source = ["data_processor"]
omit = ["*/tests/*", "*/__init__.py"]

[tool.coverage.report]
precision = 2
show_missing = true
skip_covered = false

六、pyproject.toml vs 传统方式对比

6.1 依赖管理对比

传统方式（requirements.txt）：

requests>=2.28.0
numpy>=1.20.0
pandas>=2.0.0

问题：

• 无法区分生产依赖和开发依赖
• 不支持可选依赖组
• 缺少项目元数据

现代方式（pyproject.toml）：

[project]
dependencies = [
    "requests>=2.28.0",
    "numpy>=1.20.0",
    "pandas>=2.0.0",
]
[project.optional-dependencies]
dev = ["pytest>=7.0", "black>=22.0"]

优势：

• ✅ 清晰分类
• ✅ 灵活安装
• ✅ 完整的项目信息

6.2 工具配置对比

传统方式：

需要多个配置文件：

• .flake8
• pytest.ini
• mypy.ini
• tox.ini

现代方式：

一个 pyproject.toml 搞定：

[tool.pytest.ini_options] # pytest 配置
[tool.black]              # black 配置
[tool.mypy]               # mypy 配置

七、最佳实践

7.1 版本管理

使用语义化版本（SemVer）：

[project]
version = "主版本。次版本。修订号" # 如 "2.3.1"

• 主版本：不兼容的 API 修改
• 次版本：向下兼容的功能新增
• 修订号：向下兼容的问题修正

7.2 依赖版本约束

dependencies = [
    "requests>=2.28.0,<3.0.0", # 推荐：指定范围
    "numpy>=1.20.0",           # 可以：仅指定最低版本
    "pandas==2.0.0",           # 避免：固定版本（除非必要）
]

7.3 项目结构建议

my_project/
├── pyproject.toml # 项目配置
├── README.md      # 项目说明
├── LICENSE        # 开源协议
├── .gitignore     # Git 忽略文件
├── src/
│   └── my_package/ # 源代码
│       ├── __init__.py
│       └── ...
└── tests/         # 测试代码
    └── ...

7.4 使用 src 布局

在 pyproject.toml 中配置：

[tool.setuptools]
package-dir = {"" = "src"}

[tool.setuptools.packages.find]
where = ["src"]

优势：

• 避免测试时导入本地未安装的包
• 确保测试的是安装后的版本

八、常见问题 FAQ

Q1: pyproject.toml 和 setup.py 可以共存吗？

答：可以，但不推荐。如果两者都存在，建议逐步迁移到 pyproject.toml。

Q2: 如何从 setup.py 迁移到 pyproject.toml？

答：

1. 创建 pyproject.toml
2. 将 setup() 中的参数转换为对应的 TOML 格式
3. 测试构建：pip install -e .
4. 确认无误后删除 setup.py

Q3: Poetry 和 setuptools 选哪个？

答：

• Poetry：现代化，功能丰富，适合新项目
• setuptools：传统，兼容性好，适合维护老项目

Q4: pyproject.toml 中的依赖和 requirements.txt 的关系？

答：

• pyproject.toml：定义抽象依赖和项目元数据
• requirements.txt：可选，用于锁定具体版本

推荐工作流：

# 从 pyproject.toml 生成 requirements.txt
pip-compile pyproject.toml -o requirements.txt

九、工具推荐

9.1 Poetry

一站式项目管理工具，自动生成 pyproject.toml。

# 安装
pip install poetry
# 初始化项目
poetry init
# 添加依赖
poetry add requests
# 安装依赖
poetry install

9.2 Flit

轻量级打包工具，配置简单。

# 安装
pip install flit
# 发布到 PyPI
flit publish

9.3 Hatch

现代化项目管理工具。

# 安装
pip install hatch
# 创建项目
hatch new my-project

十、总结

pyproject.toml 是 Python 项目管理的未来趋势，它带来了：

✅ 统一配置：一个文件管理所有配置 ✅ 标准化：Python 官方推荐的标准 ✅ 现代化：更清晰的语法和更强大的功能 ✅ 工具支持：主流工具都已支持

学习路径建议

1. 初级：理解基本结构，能读懂常见配置
2. 中级：为自己的项目编写 pyproject.toml
3. 高级：掌握各种工具配置，优化项目管理流程

参考资源

• PEP 518 - Specifying Minimum Build System Requirements
• PEP 621 - Storing project metadata in pyproject.toml
• TOML 官方规范
• Python Packaging User Guide