Error: R is already initialized about reticulate HOT 5 CLOSED

I think the problem is that rpy2 wants to initialize it's own instance of the R runtime, and when it's called via reticulate the R runtime is of course already initialized. Unfortunately I don't think we can find a way around this one.

from reticulate.

This sounds reasonable - though it's strange to me that the error is not always triggered by rpy2.

For example, I don't have any problems sourcing numpy, matplotlib or pandas in the following environment :

conda create -n testenv3 python=3 scipy numpy pandas matplotlib
source activate testenv3
pip install rpy2
Rscript run_validate_setup2.R

yields output:

(testenv3) jacquelineburos@jacki1:~/projects/test-reticulate$ Rscript run_validate_setup2.R
python:         /home/jacquelineburos/miniconda3/envs/testenv3/bin/python
libpython:      /home/jacquelineburos/miniconda3/envs/testenv3/lib/libpython3.6m.so
pythonhome:     /home/jacquelineburos/miniconda3/envs/testenv3:/home/jacquelineburos/miniconda3/envs/testenv3
version:        3.6.1 |Continuum Analytics, Inc.| (default, May 11 2017, 13:09:58)  [GCC 4.4.7 20120313 (Red Hat 4.4.7-1)]
numpy:          /home/jacquelineburos/miniconda3/envs/testenv3/lib/python3.6/site-packages/numpy
numpy_version:  1.12.1

python versions found:
 /home/jacquelineburos/miniconda3/envs/testenv3/bin/python
 /usr/bin/python
 /usr/bin/python3
[1] "checking that numpy is available"
[1] "checking that matplotlib is available"
[1] "checking that rpy2 is available"
[1] "checking that python is initialized"
[1] "loading numpy"
[1] "loading matplotlib"
[1] "loading rpy2"
[1] "setup OK"
[1] "Finished"

Where run_validate_setup2.R looks like the following:

#!/usr/bin/env Rscript

# execute script
source('validate_setup.function.R')
validate_setup(verbose = TRUE, test_packages = c('numpy', 'matplotlib', 'rpy2'))

print('Finished')

If it's OK with you, I'm going to keep working through dependencies of cohorts to try to isolate the issue. My thinking is, in the long run, this might help make reticulate more reliable overall.

from reticulate.

from reticulate.

Nope you were right! the error occurs on installing PyVCF, one of the dependencies of cohorts, since that package uses rpy2 to initializes R in the code, but only if rpy2 is installed.

e.g. here:

try:
    from rpy2 import robjects
except:
    robjects = None

In general it seems safer to not have rpy2 installed, since otherwise would be hard to know which packages use it or depend on it. But simply installing it doesn't cause a problem unless one of the python packages uses it.

Anyways I feel better (at least) having tracked this down. Agreed with your earlier statement above that it's unlikely that anything can be done to avoid the issue. Circular r->py->r dependencies seem like a bad idea, in general.

from reticulate.

from reticulate.