Why pmax(dataFrame, int) would introduce NAs?

Question

I'm running into a behavior of pmax that I can't quite understand:

pmax(data.frame(matrix(1:16, nrow=4)), c(6))

would return

  X1 X2 X3 X4
1  6 NA  9 13
2  6  6 10 14
3  6  7 11 15
4  6  8 12 16

What I don't understand is why only the entries smaller than 6 and are not in the first column got turned into NA - if we recycle c(6, NA, NA, NA) through the rows, wouldn't all of X2, X3, X4 be NA since max(NA, anything) = NA? Why only the entries that are not in the first column and is smaller than 6 is changed into NAs?

Answer 1

pmax is not designed to be used with data.frame input.

The error is introduced in line 35 of pmax:

mmm[change] <- each[change]

because each is defined to be as long as the length of the input, which for a data.frame is the number of columns. Therefore when it tries to address the 5th element, it gets NA.

each
[1] 6 6 6 6
each[change]
[1]  6  6  6  6 NA

The obvious workaround is to convert to data.frame after using pmax:

data.frame(pmax(matrix(1:16, nrow=4), c(6)))
  X1 X2 X3 X4
1  6  6  9 13
2  6  6 10 14
3  6  7 11 15
4  6  8 12 16

Or convert back and forth as required.