Path dependencies with setup.py cannot be installed if setup.py imports `numpy` (or more general, imports any site-packages)

• I am on the latest Poetry version.
• I have searched the issues of this repo and believe that this is not a duplicate.
• If an exception occurs when executing a command, I executed it again in debug mode (-vvv option).

• OS version and name: Windows 10
• Poetry version: 1.1.4

Issue

I have searched in here and found #2638 is similar but not the same problem as I am facing. I have also tried a possible solution of [build-system].requires and consulted its documentation (as specified in #744) but no success. Here is a snapshot of the error message:

C:\Users\XXX\Desktop\test-module>poetry update -vvv
Updating dependencies
Resolving dependencies...

  PackageInfoError

  Unable to determine package info for path: C:\Users\XXX\Desktop\test-module\lib\my-dev-module

  Fallback egg_info generation failed.

  Command C:\Users\XXX\AppData\Local\Temp\tmpg5w9jw0j\.venv\Scripts\python.exe setup.py egg_info errored with the following return code 1, and output:
  Traceback (most recent call last):
    File "setup.py", line 11, in <module>
      import numpy
  ModuleNotFoundError: No module named 'numpy'

  at ~\.poetry\lib\poetry\inspection\info.py:503 in _pep517_metadata
      499│                     venv.run("python", "setup.py", "egg_info")
      500│                     return cls.from_metadata(path)
      501│                 except EnvCommandError as fbe:
      502│                     raise PackageInfoError(
    → 503│                         path, "Fallback egg_info generation failed.", fbe
      504│                     )
      505│                 finally:
      506│                     os.chdir(cwd.as_posix())
      507│

I will reuse its structure and show how the problem is reproduced.

The project structure is something like

test-module/
  + pyproject.toml
  + testmodule/
      + __init__.py
  + lib/
      + my-dev-module/
          + setup.py
          + mydevmodule/
              + __init__.py

The toml file is

[tool.poetry]
name = "testmodule"
version = "0.1.0"

[tool.poetry.dependencies]
python = "^3.8"
numpy = "^1.19.1"
mydevmodule = {path = "./lib/my-dev-module", develop = true}

[build-system]
requires = ["numpy>=1.19.1,<1.20.0","poetry-core>=1.0.0"]
build-backend = "poetry.core.masonry.api"

And ./lib/my-dev-module/setup.py is like

from distutils.core import setup
import numpy

if __name__ == "__main__":
    setup(
        name="mydevmodule",
        version="0.2",
        description="",
        author="me",
    )

The setup() in setup.py is a placeholder and the script does not get executed to setup(). It erred out at line 2 when importing numpy. In my actual application, module numpy is used to provide its C library for non-Python extension.

My question is: why does it fail at importing numpy? Shouldn’t specifying [build-system].requires fix this? If not, I got that to fix this problem requires a [build-system].requires and a pyproject.toml in mydevmodule but that is not an ideal solution. Is there any other ways to specify the installation time requirements of a path-dependency module? Note that I have put numpy in both the dependencies and build-system requirement of the top-level project testmodule.

[finswimmer]

To be honest, I do not know if poetry will forever be able to install projects that have a setup.py and are not PEP517 conform (i.e. no pyproject.toml). I am not a maintainer, so it is not my place to talk on this topic.

There are currently no plans to drop the support for non-pyproject.toml projects. Furthermore PEP517 says “If the pyproject.toml file is absent, or the build-backend key is missing, the source tree is not using this specification, and tools should revert to the legacy behaviour of running setup.py (either directly, or by implicitly invoking the setuptools.build_meta:legacy backend).”

[sinoroc]

@hyliu1989 You can add the minimal pyproject.toml without changing anything else. No modification to setup.py. Developers of the project can still use python setup.py develop or whatever other setuptools commands they need. There should be no drawbacks when done correctly. But the big advantage of making sure that numpy is available when pip (or poetry or any other build frontend) need to build that project.