Background

I want to write a package overloading +, -, %% on matrices, to make it easier to use R for the quick matrix calculations one might do in Matlab:

• +X computes t(X) (binary X+Y works as normal)
• -X computes solve(X) (binary X-Y works as normal)
• -I(X) computes -X
• X%%Y computes Y%*%X (write multiplications right-to-left, for reasons beyond scope of question)

Example use of the package:

library(mymatrixpackage)
X <- matrix(1:6, nrow = 3)
y <- matrix(1:3, nrow = 3)
+X == t(X)
X+X == 2*x

# soln to lm coefficients (remember: read right-to-left)
y%%+X%%-(X%%+X) == solve(t(X)%*%X)%*%t(X)%*%y

Main Question

For a matrix X, how can I make +X dispatch to +.matrix (a function I can define). +X uses the internal method and seems to ignore +.matrix.

I think I need to use Ops, or something to do with group generics.

Alternative solution

I have successfully implemented this idea using an s3 class “tmatrix” and:

`+.tmatrix` <- function(X,Y) {
    if (missing(Y)) {
        t(X)
    } else {
        structure(unclass(X)+unclass(Y), class = oldClass(X))
    }
}

It would be better to not use an s3 class because:

• Need to do: attr(X, "class") <- "tmatrix"
• many operations (eg [) lose X‘s class attribute, and so I need to overload those operations as well.