Consistency with math?

[mikmart]

As André Gillibert wrote on r-devel, it would seem desirable for ifelse to behave similarly to it's mathematical cousin expression test * yes + (1 - test) * no.

Currently at least the handling of various combinations of scalar, matrix and vector inputs is quite different:

library(ifelse)

math_ifelse <- function(test, yes, no) {
  storage.mode(test) <- "logical"
  test * yes + (1 - test) * no
}

scalar <- 0
vector <- 0:1
matrix <- matrix(0:3, 2, 2)

ifelse1(scalar, vector, matrix)
#> [1] 0
math_ifelse(scalar, vector, matrix)
#>      [,1] [,2]
#> [1,]    0    2
#> [2,]    1    3

ifelse1(scalar, matrix, vector)
#> [1] 0
math_ifelse(scalar, matrix, vector)
#>      [,1] [,2]
#> [1,]    0    0
#> [2,]    1    1

ifelse1(vector, matrix, scalar)
#> [1] 0 1
math_ifelse(vector, matrix, scalar)
#>      [,1] [,2]
#> [1,]    0    0
#> [2,]    1    3

ifelse1(vector, scalar, matrix)
#> [1] 0 0
math_ifelse(vector, scalar, matrix)
#>      [,1] [,2]
#> [1,]    0    2
#> [2,]    0    0

ifelse1(matrix, vector, scalar)
#> [1] 0 1 0 1
math_ifelse(matrix, vector, scalar)
#>      [,1] [,2]
#> [1,]    0    0
#> [2,]    1    1

ifelse1(matrix, scalar, vector)
#> [1] 0 0 0 0
math_ifelse(matrix, scalar, vector)
#>      [,1] [,2]
#> [1,]    0    0
#> [2,]    0    0

^{Created on 2025-08-02 with reprex v2.1.1}

Related to #1.

[moodymudskipper]

As said on #1 I am for a recycle = FALSE argument to prevent silent mistakes, but with recycle = TRUE since we've secured the basic usage and force users to think about what they're doing I think it would make sense to allow "3 way recycling", where we recycle to the longest element (test, yes, no, na) if its length is a multiple of all the other lengths.