Using the R API Standalone

University of Minnesota, Twin Cities     School of Statistics     Charlie's Home Page

This page still under construction. Don't look yet. Or you can look, but be warned it's unfinished.

Almost all of the R API can be used standalone, that is, from a C main program with no help from the R main program. This just uses R as a C math library.

Example for 32 Bit Machines

Get the files

Then do

    make
    export LD_LIBRARY_PATH=/APPS/32/lib
    foo

in the directory where you put the files.

The example uses the qbeta function (the same C function that the R qbeta function calls) to calculate the median of a beta distribution.

Example for 64 Bit Machines

Get the files

Then do

    make
    export LD_LIBRARY_PATH=/APPS/64/lib64
    foo

in the directory where you put the files.

More Stuff

There's lots more stuff in this R standalone library. For example,

• random number generation (uniform and non-uniform)
• special functions (gamma, digamma, Bessel)
• math constants (pi, e)

To use the random number generators, you have to set the seeds as described in the section on this library in Writing R Extensions.

To find out what's included, look at the header file

To actually find out how a function is called, one must RTFS. To find the source, you must look in the directory where the source is. On our machines it is

    /APPS/src/R-2.2.1/src

right now. The directory name changes as the version number changes.

For example, to find out what the calling sequence was for the qbeta function is, I did

    find /APPS/src/R-2.2.1/src -name qbeta.c

It shows two copies of the same file (look at either), one of which is

A perusal of the source code shows the function signature

    double qbeta(double alpha, double p, double q, int lower_tail, int log_p)

so the function takes 5 arguments (but we already knew that from looking at Rmath.h) and the argument names are a bit informative. Looking at the on-line help for the R function qbeta tells us the arguments are

• the quantile to look up
• the first parameter of the beta distribution
• the second parameter of the beta distribution
• a logical (TRUE or FALSE) that says whether the lower tail or upper tail is wanted.
• a logical that says whether the probability is supplied on the log scale.

So that's how the example works.