是否有任何工具可以为主要用 Python 编写的 GitHub 私有存储库创建依赖关系图 - 仅显示私有依赖关系？

Question

我在一家拥有几个私有存储库的公司工作 - 所有这些都主要用 Python 编写。
每个 repo 在其根目录下都有一个 requirements.txt 文件。

我们有兴趣列出每个存储库的依赖项和依赖项，而不是公共依赖项，而只是我们的私有依赖项和依赖项。

例如，假设 our_private_package1 的 requirements.txt 文件如下所示：

public_package1
public_package2
public_package3
our_private_package2
our_private_package3

还假设有另一个包our_private_package4 具有以下requirements.txt 文件：

public_package2
public_package4
our_private_package1

所以我们想构建一个依赖关系图，我们可以在其中看到 our_private_package1 的以下内容：
依赖：

• our_private_package2
• our_private_package3
  

家属：

• our_private_package4

理想情况下，我们希望用图表来可视化依赖关系。

2 个注释：

• GitHub's solution 不适合这里有两个原因。首先，它 仅显示公共依赖项。也不能确定 私有存储库的依赖项（依赖项是否是公共的 或私人）。
• 依赖项不应是可传递的。在上面的例子中，our_private_package2 不应该被列为 our_private_package4 的依赖项（尽管我们 应该在视觉表示中看到一条从 our_private_package2 到 our_private_package4，具有从 our_private_package2 到 our_private_package1 的定向分支和 从our_private_package1 到our_private_package4 的另一个分支）。

如果没有针对此类问题的可用工具，我至少想要一些指导如何使用一些 Python 代码来解决（几乎可以肯定它是可行的）。

Answer 1

正如之前在 cmets 中提到的，我也有类似的需求。我使用与此类似的方法解决了它：

from collections import defaultdict
from typing import List, Tuple, Dict


def main():
    requirements_file_content_per_package = {
        "our_private_package1": """\
public_package1
public_package2
public_package3
our_private_package2
our_private_package3
        """,

        "our_private_package2": """\
public_package2
public_package4
our_private_package1
        """
    }

    private_packages_names = ["our_private_package1",
                              "our_private_package2",
                              "our_private_package3",
                              "our_private_package4"]

    # process descending dependency
    private_packages_dependencies: Dict[str, Tuple[List[str], List[str]]] = {}
    for private_package_name in private_packages_names:
        requirements_file_content = requirements_file_content_per_package.get(private_package_name, "")
        private_packages_dependencies[private_package_name] = classify_dependencies(requirements_file_content, private_packages_names)

    # display
    for private_package_name, (package_private_dependencies_names, _) in private_packages_dependencies.items():
        print(f"package {private_package_name!r} depends on private {package_private_dependencies_names!r}")

    # process ascending dependencies
    private_packages_dependents: Dict[str, List[str]] = defaultdict(list)
    for private_package_name, (package_private_dependencies_names, _) in private_packages_dependencies.items():
        for package_private_dependency_name in package_private_dependencies_names:
            private_packages_dependents[package_private_dependency_name].append(private_package_name)

    # display
    for private_package_name, package_private_dependents_names in private_packages_dependents.items():
        print(f"package {private_package_name!r} is depended on by private {package_private_dependents_names!r}")


def classify_dependencies(requirements_file_content: str, private_packages_names: List[str]) -> Tuple[List[str], List[str]]:
    private_dependencies: List[str] = []
    public_dependencies: List[str] = []
    for dependency_name in requirements_file_content.splitlines():
        is_private = dependency_name in private_packages_names
        if is_private:
            private_dependencies.append(dependency_name)
        else:
            public_dependencies.append(dependency_name)
    return private_dependencies, public_dependencies


main()

这有点粗糙，我使用的是非常基本的 Python（字符串、列表和集合），但可以使用有向彩色图代替。另外，名字很长，因为我想要描述性，但我认为使用简洁的定义词汇会更好。

至于可视化：

    import graphviz  # installed with pip

    dot = graphviz.Digraph()
    # at first, create all the nodes
    for private_package_name in private_packages_names:
        dot.node(private_package_name)
    # then create the edges
    for private_package_name, package_private_dependents_names in private_packages_dependents.items():
        for package_private_dependent_name in package_private_dependents_names:
            dot.edge(package_private_dependent_name, private_package_name)
    dot.render("deps.gv", format="png")

出品deps.gv.png :