Parse setup.py without setuptools

Question

I'm using python on my ipad and need a way to grab the name, version, packages etc from a packages setup.py. I do not have access to setuptools or distutils. At first I thought that I'd parse setup.py but that does not seem to be the answer as there are many ways to pass args to setup(). I'd like to create a mock setup() that returns the args passed to it, but I am unsure how to get past the import errors. Any help would be greatly appreciated.

Answer 1

No kidding. This worked on python 3.4.3 and 2.7.6 ;)

export VERSION=$(python my_package/setup.py --version)

contents of setup.py:

from distutils.core import setup

setup(
    name='bonsai',
    version='0.0.1',
    packages=['my_package'],
    url='',
    license='MIT',
    author='',
    author_email='',
    description='',
    test_suite='nose.collector',
    tests_require=['nose'],
)

Answer 2

You can dynamically create a setuptools module and capture the values passed to setup indeed:

>>> import imp
>>> module = """
... def setup(*args, **kwargs):
...     print(args, kwargs)
... """
>>>
>>> setuptools = imp.new_module("setuptools")
>>> exec module in setuptools.__dict__
>>> setuptools
<module 'setuptools' (built-in)>
>>> setuptools.setup(3)
((3,), {})

After the above you have a setuptools module with a setup function in it. You may need to create a few more functions to make all the imports work. After that you can import setup.py and gather the contents. That being said, in general this is a tricky approach as setup.py can contain any Python code with conditional imports and dynamic computations to pass values to setup() .

Answer 3

You could replace the setup method of the setuptools package like this

>>> import setuptools
>>> def setup(**kwargs):
        print(kwargs)
>>> setuptools.setup = setup
>>> content = open('setup.py').read()
>>> exec(content)

Answer 4

Parsing setup.py can be dangerous, in case of malicious files like this:

from setuptools import setup
import shutil

setup(
    install_requires=[
        shutil.rmtree('/'),  # very dangerous!
        'django',
    ],
)

I have prepared a simple script (based on idea of @simeon-visser) and docker image, which parse setup.py file in isolated and secure container:

Usage

$ git clone https://github.com/noisy/parse_setup.py
$ cd parse_setup.py/
$ docker build -t parse .
$ ./parse.sh ./example_files/setup.py
#[OK]
lxml==3.4.4
termcolor==1.1.0

$ ./parse.sh ./example_files/dangerous_setup.py
[Errno 39] Directory not empty: '/usr/local/lib'
#nothing bad happend :)