I have a function f(x,y) that takes two vector arguments as input and returns a vector as output (all of equal length n). I am trying to apply this function to every row of two m x n matrices M1 and M2 to return an m x n matrix, i.e. so that every row of M1 and of M2 correspond to input arguments x,y.

I used the following syntax:

M3 = mapply(f, M1, M2)

Which gives an error message stating that f is taking NA as input arguments. In my test example,

M1 = rbind(c(5,0,0,6),c(20,0,0,0))
M2 = rbind(c(5,15,25,55),c(0,0,5,40))

and f(M1[1,],M2[1,]), f(M1[2,],M2[,2]) return valid results. What is the problem with my application of mapply? I thought that perhaps it was an issue of row vs. column applications, but mapply(f, t(M1), t(M2)) returns the same error, so obviously f is being fed something other than the respective rows (or columns) of M1,M2 as input.

CodePudding user response：

If we want to apply on each row, then split the matrix by row and pass as a list

mapply(f, asplit(M1, 1), asplit(M2, 1))

Note that mapply on a matrix (or a vector) will loop over each element i.e. here the unit is a single element whereas in data.frame/data.table/tibble, the single unit is a column. By splitting by row (asplit - MARGIN = 1), we get a list of vectors and here the unit is a list element

As @Adam mentioned in the comments, it may needs to be transposed (Not clear without testing with f)