Requested feature condacolab.install(yml="path/to/file.yml")

[rsanchezgarc]

It would be super nice if we could get a one-liner to create an environment from a .yml file.

So far I am doing this

!pip install -q condacolab
import condacolab
condacolab.install() #ignore message about session crashing, this is intended

import condacolab
condacolab.check()
!wget -c https://raw.githubusercontent.com/XXXXXXXXX/master/environment.yml
!conda env update --file environment.yml

But I guess it should be easy to wrap it into
condacolab.install(yml = "environment.yml")

Thanks!

[jaimergp]

Ooops, I somehow never replied to this.

So this is tricky to implement. The install() function invokes a kernel restart, so we lose the connection to the server for a moment and then the process is restarted. That means that the environment installation cannot happen from the notebook until we are back alive.

We could prepare a script and set it to auto-launch after the restart or something like that, but it looks very fragile from my perspective. I would be happy to discuss this more thoroughly in a potential PR though!

[rsanchezgarc]

Thanks @jaimergp for your answer. I am afraid I am not brave enough to start working on this.

[proppy]

That means that the environment installation cannot happen from the notebook until we are back alive.

Does the installations needs to happen with the environment activated? Or would executing $CONDA_PREFIX/bin/conda env update --file environment.yml also work?

[jaimergp]

That should work but you might need to add -p <PREFIX> to your command (I think it's /usr/local for the currently released version).

That said, @ssurbhi560 is working on this feature request at #38!