Request: Declare miniconda pre-installed

[effigies]

If I want to make use neurodocker to create two images, one built on top of the other (as in nipype), and both should install conda packages, then each Dockerfile created will attempt to install miniconda. It would be nice to have an argument (e.g. preinstalled=True) to pass to the second image to prevent neurodocker from doubling up.

[kaczmarj]

Hi @effigies - this is a great idea. I'll add it to #143.

[kaczmarj]

Addressed by #151 with the preinstalled option.

use will look something like --miniconda env_name=neuro preinstalled=true conda_install="python=3.6 numpy". that will create a neuro environment with python3.5 and numpy and will not attempt to reinstall miniconda.

does that syntax look good to you?

[effigies]

Yup, looks great. And if I've already created the environment neuro, will this skip the conda create step, too?

[kaczmarj]

good point, i hadn't thought about that. that might require another flag. ideally there would be an option to conda create that would install into the environment if it exists or create the environment if it does not exist (sort of like the behavior of mkdir -p). but that doesn't seem to be the case.

what are you thoughts on the flag env_exists=true?

[effigies]

As a thought, you could have explicit options create_env=neuro and use_env=neuro, where env_name=neuro currently has create_env behavior on the first call and use_env on later. And perhaps use_env would imply preinstalled=true, for brevity.

On the other hand preinstalled=true env_exists=true has the advantage of being really explicit, so if that create/use doesn't seem instantly clear, maybe let's just go with that.

[kaczmarj]

@effigies - I'll go with create_env / use_env. the docs can explain the differences