rstats
v0.3.1
Published
A node.js interface for statistical programming language R
Downloads
25
Readme
node-Rstats
An interface for node.js to statistical programming language R based on the fabulous Rcpp package
Installation
Currently, rstats
is ONLY supported for Unix operating systems.
Also, it is required that the R packages RInside
, Rcpp
and RJSONIO
are installed inside R. Additionally, building the package using node-gyp
requires
python
(v2.7
,v3.x.x
is not supported)make
- A C/C++ compiler toolchain, such as GCC
With these prerequisites satisfied, one can simply install rstats
using npm
npm install rstats
Getting Started
After installation, the package can be loaded as follows:
var rstats = require('rstats');
Once the package is loaded, we can create an R session by the command
var R = new rstats.session();
Important Functions
parseEvalQ
Evaluating R expressions is easy. We can use the parseEvalQ function as follows:
R.parseEvalQ("cat('\n Hello World \n')");
parseEval
To evaluate an R expression and directly capture its return value, one can use the parseEval function.
var x = R.parseEval("c(1,2,3)");
The variable x
is now equal to the array [1,2,3]
.
assign
Numeric values can be easily assigned to variables in the current R session:
R.assign('x', 17);
R.assign('y', 3);
// calculate the sum of x+y and print the result
R.parseEvalQ("res = x + y; print(res);");
get
To retrieve an object from the R session, we use the get command. For example, let us create a 2x2 matrix in R and retrieve it in JavaScript as a nested array:
R.parseEvalQ("mat = matrix(1:4,ncol=2,nrow=2)");
var mat = R.get('mat');
Internally, the get function uses JSON in order to convert the R data types to JavaScript data types.
We can also run much more complicated calculations and expose the R objects to JavaScript. Consider a linear regression example:
R.parseEvalQ('x = rnorm(100); y = 4x + rnorm(100); lm_fit = lm(y~x);');
var lm_fit = R.get('lm_fit');
var coefs = lm_fit.coefficients;
var residuals = lm_fit.residuals;
Unit Tests
Run tests via the command npm test