RGenerics.md

R Generic Functions

R is both a functional language and an object-oriented one. One feature is polymorphism. R has a number of generic functions, such as print(), plot(() and summary(). They are not functions in the usual sense of performing actual work. Instead, their job is simply to call other functions.

As an example, consider R's ecdf() function, which produces the empirical cdf from the usual data set X₁,...,X_n. The empirical CDF is the estimate of the population CDF, F_X(t). For instance, the empirical CDF of the Nile dataset, evaluated at t = 1129.2, is the proportion of data points below or equal to that number, 0.85:

mean(Nile <= 1129.2)  # 0.85

We can get that using the ecdf() function:

> z <- ecdf(Nile)
> z(1129.2)
[1] 0.85

Obviously, the object z here is a function. Let's confirm:

> class(z)
[1] "ecdf"     "stepfun"  "function"

Ah, yes, z is indeed of class 'function'. (All class names in R are quoted strings.) But there's more: It turns out that the 'function' class has as subclass, 'stepfun', not too surprising since empirical CDFs are step functions. And in turn, one of the subclasses of 'stepfun' (there are probably many) is 'ecdf'.

Now, what happens in the call

plot(z)

A graph of the empirical CDF of Nile does pop up (not shown here), but the drawing on the screen is NOT done by the plot() function. Instead, the R interpreter dispatches the plot() call to a function plot.ecdf(), which does the actual work of drawing on the screen.

We can check that by a direct call():

plot.ecdf(z)

which produces the same graph.

Note carefully that plot() and plot.ecdf() are two entirely separate functions. The convention is to give the name plot.x() to the function that plot() dispatches to, where x is the name of the class. There are many, many plot.x functions in both base R and the tens of thousands of contributed R packages in the CRAN repository.

In other words, plot() is merely a go-between.

You'll find on the other hand that for instance there is no function summary.ecdf(). A more sophisticated mechanism is used here. But there are summary.data.frame() and print.data.frame():

> summary.data.frame
function (object, maxsum = 7L, digits = max(3L, getOption("digits") - 
    3L), ...) 
{
    ncw <- function(x) {
        z <- nchar(x, type = "w")
        if (any(na <- is.na(z))) {
            z[na] <- nchar(encodeString(z[na]), "b")
        }
        z
    }
    z <- lapply(X = as.list(object), FUN = summary, maxsum = maxsum, 
        digits = 12L, ...)
...