I'm using miniconda to manage my installation of data science packages. It is a workflow that I have somewhat established now, so I would like it to work in this case too. I also would assume it to work, since it is supposed to help in situations like this - where more dependencies than pure python is needed.
I would like to install the python cdt toolbox. It is a pip-installable package available on PyPI but not in any conda channels. It requires PyTorch, available easily in PyPI and in their own conda channel. But it further requires some r-packages, most available on CRAN and some only on github. My dream scenario is to have a single environment.yml file looking something like this:
name: my_env
channels:
- defaults
- pytorch
dependencies:
- pytorch
- cpuonly
- pip
- pip:
- cdt
- -e .
- r: # this line and below don't work...
- pcalg # available on CRAN, but not in conda channel r
- kpcalg # available on CRAN, but not in conda channel r
- github:https://github.com/Diviyan-Kalainathan/RCIT # R package not in CRAN nor on conda
I assume there is not a simple direct way to do it like above, based on this question on SO about similar npm installs. I cannot possibly be the first persion with a need to install both r and python packages... so what is the 'standard' workarounds?