How to make RCall.jl use a CondaPkg.jl environment?

[schlichtanders]

Hi,
this package is really really awesome! :)

I would like to also use RCall on top of it, however for this I seem to need Conda.jl. Hence I would like to let it make Conda.jl use my current CondaPkg.jl environment.

Conda.jl documentation tells:

1. set ENV["CONDA_JL_HOME"] to the environment (which will become Conda.jl root environment, which should be the one picked up by RCall / PyCall)
2. set ENV["CONDA_JL_CONDA_EXE"] to the executable in order to not install two different condas
3. rebuild Conda.jl and respective dependencies

My status:

• 1. I guess this is corresponds to CondaPkg.jl ENV["JULIA_CONDAPKG_ENV"], which I already set
• 2. I guess this is corresponds to CondaPkg.MicroMamba.executable()
• 3. rebuilding is okay for me, because I am using a fixed conda environment per julia process

So it happens by writing this issue, I have solved kind of a couple of steps myself :)
Maybe someone can check whether this makes sense, or whether I overlooked something.

I will try it out myself soon and report in this issue.

[schlichtanders]

Short remark: When simple using R, it is probably sufficient to just set

ENV["R_HOME"] = CondaPkg.envdir() * "/lib/R"

and rebuild RCall (Probably a similar approach also works for PyCall)

[schlichtanders]

I tried running the following

julia> using CondaPkg

julia> ENV["R_HOME"] = CondaPkg.envdir() * "/lib/R"

julia> import Pkg

julia> Pkg.build("RCall")

julia> import RCall
┌ Warning: RCall.jl: During startup - Warning messages:
│ 1: package ‘utils’ in options("defaultPackages") was not found 
│ 2: package ‘stats’ in options("defaultPackages") was not found 
└ @ RCall ~/.julia/packages/RCall/gOwEW/src/io.jl:172

The CondaPkg environment does not have a proper R setup... I am confused. Any help is highly appreciated

---

Edit: CondaPkg.status() looks good

julia> CondaPkg.status()
CondaPkg Status /home/ssahm/Tmp/tmp-julia/condapkg/CondaPkg.toml
Environment
  /home/ssahm/Tmp/tmp-julia/condapkg/.CondaPkg/env
Packages
  r-base v4.3.1
  r-codedepends v0.6.5
  r-globals v0.16.2
Channels
  bioconda
  conda-forge

[schlichtanders]

@simonbyrne maybe you can help?

EDIT: This last error rather looks like this issue could be migrated to RCall.jl actually

[simonbyrne]

Unfortunately, I'm afraid I haven't looked at RCall.jl for some time, and my bandwidth is rather limited at the moment.

[cjdoris]

Perhaps you need to activate the Conda environment in your Julia session first like PythonCall does here: https://github.com/JuliaPy/PythonCall.jl/blob/0056de8711e274b3d8084d19a23fe54adf5d0640/src/cpython/context.jl#L68

[schlichtanders]

Thank you for your comments. It turned out it was a linux distribution specific problem which is already reported at anaconda. Concretely you need to have /usr/bin/which available on your system for R...

It works now that I've symlinked /usr/bin/which

[schlichtanders]

related issue: JuliaInterop/RCall.jl#480 (tries to improve RCall.jl for better integration with CondaPkg.jl without the need of Pkg.build("RCall"))

[schlichtanders]

Short remark: When simple using R, it is probably sufficient to just set

ENV["R_HOME"] = CondaPkg.envdir() * "/lib/R"

and rebuild RCall (Probably a similar approach also works for PyCall)

this is the final way I am using

import CondaPkg
ENV["R_HOME"] = CondaPkg.envdir() * "/lib/R"
import Pkg
Pkg.build("RCall")
using RCall