I am packaging an application that does a bit of environment management using pip. In essence, it does this:
import subprocess
import sys
def main():
own_python = sys.executable
print(own_python)
# this is silly and not what the original application does.
# the point is to invoke `python -m pip` and not fail with "pip module does not exist"
process = subprocess.run([own_python, "-m", "pip", "install", "foo"],
capture_output=True,
text=True
)
print(process.returncode)
print(process.stderr)
if __name__ == "__main__":
main()
The application relies on the fact that python interpreter that’s invoking it has pip module available for a certain functionality. The rest of the application successfully gets all dependencies during runtime as they are passed through buildPythonApplication.
I have extracted the relevant bits into this flake runnable as:
nix run github:VTimofeenko/python-pip-nix-scratchpad
Question is: is it possible to use one of the parameters of buildPythonApplication or overrides of python to make python -m pip work inside the script?
I’d imagine it is possible to use an override of the python configure flag to set --ensurepip but that would require recompiling python. Or I could write a patch to replace subprocess call with the proper pip location but I’d rather keep the code as close to upstream as possible.