r-test-matplotlib

The goal of r-test-matplotlib is testing that Python matplotlib and PyQt5 work seemlessly from within RStudio 1.2-preview.

Motivation

Making matplotlib run from within RStudio using the R package reticulate and Python Anaconda is a very difficult task. PyQt is not recognized by RStudio, and requires from the user to provide a path to the Qt plugins adding instability to the R-Python installation.

Sometimes, this fix will not work, specially with PyQt5.

I found that portable WinPython 3.6 runs smoothly with RStudio 1.2-preview without crashing it. And without providing a path for PyQt.

Examples

I have included some working examples of Rmarkdown notebooks with Python matplotlib plots. All of them work without providing a path for PyQt or without crashing RStudio.

Requisites

1. Install WinPython 3.6.7 64-bit in any folder in your machine. WinPython is portable; it could be installed in a pen drive.

2. Clone this repo under the notebooks folder that can be found under WPy-3670, the WinPython installation folder.

3. Be sure of loading the reticulate package and indicating the Python interpreter. In the notebook you will see a R chunk with reticulate::use_python("..\\..\\python-3.6.7.amd64\\python.exe"). This is key to make R use WinPython.

WinPython

I downloaded WinPython 3.6 64-bit from this page. The specific version I tested was WinPython64-3.6.7.0Qt5.exe which can be driectly obtained from here. This installation may take around 2 GB of disk space because it is a full fledged Python environment.

The alternative is installing a minimum Python installing WinPython64-3.6.7.0Zero.exe. This file can be downloaded from here. After installation, this zero Python will take around 200 MB. Afterwards, you can install only the packages that you require. I tested this configuration as well, after installing matplotlib, PyQt5, Jupyter Notebook, pandas, numpy, and few others, taking my imstallation to 400 MB. It run Python matplotlib and PyQt5 from RStudio and `reticulate, flawlessly.