Default matrix number of columns and rows in R

Question

In R, I can create a matrix with matrix(), which has the function definition:

function (data = NA, nrow = 1, ncol = 1, byrow = FALSE, dimnames = NULL)
{ ... }

This suggests that the default value for number of rows (nrow) and columns (ncol) are 1.

So why does the following break? Why does specifying the same default values of 1 result in a different matrix?

> matrix(1:9)
      [,1]
 [1,]    1
 [2,]    2
 [3,]    3
 [4,]    4
 [5,]    5
 [6,]    6
 [7,]    7
 [8,]    8
 [9,]    9

> matrix(1:9, ncol=1, nrow=1)
     [,1]
[1,]    1

nrussell nrussell · Accepted Answer · 2016-11-30T20:09:09

The user-facing base::matrix calls an internal function:

matrix
# function (data = NA, nrow = 1, ncol = 1, byrow = FALSE, dimnames = NULL) 
# {
#     if (is.object(data) || !is.atomic(data)) 
#         data <- as.vector(data)
#     .Internal(matrix(data, nrow, ncol, byrow, dimnames, missing(nrow), 
#         missing(ncol)))
# }
# <bytecode: 0x000000000b802e88>
# <environment: namespace:base>

In particular, note that the last two arguments passed to the internal function are flags indicating if the row and / or column dimensions were specified: missing(nrow), missing(ncol). The fact that these parameters have default values does not preclude them from being "missing". For example,

f <- function(x = 1, y = 2) {
    cat(sprintf(
        "missing(x): %s\nmissing(y): %s\n", 
        missing(x), 
        missing(y)
    ))
}

f()
# missing(x): TRUE
# missing(y): TRUE

Since the "missing-ness" of these arguments is not affected by the fact that they have default values, that logic can be handled independently, as is done here.

Default matrix number of columns and rows in R

1 Answers