Package with pip dep fails with Jupyter notebook on Windows

[madhavkrishnan]

I am writing a package with a pip depencency which works correctly in the REPL but fails in a Jupyter notebook on Windows. However, it works correctly on Ubuntu using WSL.

OS : Win 10
Julia 1.8.1
PythonCall 0.9.6

The error message I get from CondaPkg is

CondaPkg Installing Pip packages
failed to create process.

I've made MWE here https://github.com/madhavkrishnan/TestPkg.jl. Adding relavant files here for ease.

src/TestPkg.jl

module TestPkg

using PythonCall

const scipy = PythonCall.pynew() # initially NULL

function __init__()
	PythonCall.pycopy!(scipy, pyimport("scipy"))
end


end

CondaPkg.toml

[deps]
python = ">=3.5,<4"


[pip.deps]
scipy = ""

Failing notebook - https://github.com/madhavkrishnan/TestPkg.jl/blob/main/debug/test.ipynb

Working script
debug/test.jl

using TestPkg

println(TestPkg.scipy)

println(TestPkg.scipy.sin(π / 4))

Could someone check if this is reproducable or if it is a problem with my system?

[cjdoris]

Running a JupyterLab notebook on Windows 11 with the following code worked for me:

import Pkg
Pkg.activate(temp=true)
Pkg.add(url="https://github.com/madhavkrishnan/TestPkg.jl")
using TestPkg
TestPkg.scipy

[cjdoris]

I tried again in a VSCode notebook (since it appears that's what you're using) and that worked too.

[cjdoris]

Anyway you should use the conda scipy package instead of the pip one.

[madhavkrishnan]

Thanks for the response @cjdoris. The scipy package is just in the MWE, I actually needed a package only available on PyPI. The code snippet you left worked for me as well, I think because it activated a temp environment. I think the issue was perhaps my environments were somehow messed up. I removed all Julia installations completely and did a clean install and everything works fine now. I'll go ahead and close this issue. Thanks for the help.

[cjdoris]

OK cool. Environments shouldn't get messed up - if it happens again and you suspect this package is to blame, please open another issue.