Environment file dependencies outdated

[fedem-p]

Bug -> sashimi doesn't install following the described installation procedure:

to reproduce the error:

• create env from environment file
• activate new env
• install pyvista and pyvista-py
• pip install .

error msg:

Attempting uninstall: llvmlite Found existing installation: llvmlite 0.38.0 ERROR: Cannot uninstall 'llvmlite'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall.

[diegoasua]

llvmlite is a dependency of numba. It seems that installation fails because the manager is afraid of removing it. This is likely happening because numba is installed twice during this process, once here and then again here. The difference is that packages in the first file are managed by condaand the second by pip. Can you repeat the whole process in a clean environment after removing numba from requirements.txt?

[fedem-p]

llvmlite is a dependency of numba. It seems that installation fails because the manager is afraid of removing it. This is likely happening because numba is installed twice during this process, once here and then again here. The difference is that packages in the first file are managed by condaand the second by pip. Can you repeat the whole process in a clean environment after removing numba from requirements.txt?

Thanks! I'll try again and see if that solves it, I haven't though about that!

[fedem-p]

@diegoasua Sorry for the long delay, I tried removing it from the req.txt but that doesn't solve the problem since when installing sashimi pip tries to re-install the correct version of llvmlite.

Also removing it from the req.txt file doesn't seem to be a great solution in my opinion since some users may prefer using their environment and install the dependencies from this file.

As such the solution I suggest - see PR #135 - is to specify the correct version both in the env file and in the req file. Let me know what you think about it!

[diegoasua]

I can't reproduce your error. Please provide theconda, pip and python versions you used.

I have seen the PR. I am not in favor of locking a llvmlite version. Locking numba version is ok though.

[fedem-p]

to reproduce the error:

I have conda 4.10.3

I run the following commands:

conda env create -f environment.yml
conda activate sashimi
pip install pyvisa
pip install pyvisa-py
pip install .

I have nothing against locking Numba version, I'll test both that and updating my conda version

---

Update:

I've did various tests:

• even updating the conda version didn't solve the initial problem
• locking the numba version to numba >= 0.55.1 works and seems to me a good solution since it doesn't fully lock it and doesn't lock llvmlite

[diegoasua]

Thanks for the report @fedem-p. If locking numba works then it's all good for me and we can move this to your PR #135

[fedem-p]

Thanks @diegoasua for your suggestions and supervision :)