Python from R

R has an advanced python API called reticulate. If you primarily work in R and Rstudio, here is an RMD document that outlines calling python from R in RMD documents.

Here we’ll more focus on direct use of the reticulate API. You can install reticulate with install.packages("reticulate"). Or, in conda you can do conda install r-reticulate. Do that first.

library(reticulate)
pd = import("pandas")
url = "https://raw.githubusercontent.com/bcaffo/ds4bme_intro/master/data/kirby127a_3_1_ax_283Labels_M2_corrected_stats.csv"
dat = pd$read_csv(url)
head(dat)

A data.frame: 6 × 10

	Unnamed: 0	rawid	roi	volume	min	max	mean	std	type	level
	<dbl>	<chr>	<chr>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>	<dbl>
1	1	kirby127a_3_1_ax.img	Telencephalon_L	531111	0	374	128.3013	51.8593	1	1
2	2	kirby127a_3_1_ax.img	Telencephalon_R	543404	0	300	135.0683	53.6471	1	1
3	3	kirby127a_3_1_ax.img	Diencephalon_L	9683	15	295	193.5488	32.2733	1	1
4	4	kirby127a_3_1_ax.img	Diencephalon_R	9678	10	335	193.7051	32.7869	1	1
5	5	kirby127a_3_1_ax.img	Mesencephalon	10268	55	307	230.8583	29.2249	1	1
6	6	kirby127a_3_1_ax.img	Metencephalon	159402	2	299	138.5200	52.2241	1	1

Python functions can be used as R functions. Here’s a simple example.

npr = import("numpy.random")
normalGenerator = npr$normal
normalGenerator(size=as.integer(5))

1. 0.483189688515033
2. 1.15551860776896
3. -0.230891758864093
4. 0.349830581235137
5. -2.15526446076882

(Of course, this example isn’t necessary since R already has rnorm.) Notice that you actually have to do the type conversion, as.integer, since python is more strongly typed than R.

In addition, you can have python code blocks within RMD blocks. In addition, you can call python scripts

In general, it’s only worth doing this if you mostly work in R and there’s a python function that you really need. Similarly, for calling R from within python.