Ways to use a non-vectorized function on every row of a data.frame

applytorows.r

# Sample data
X <- data.frame(
  x = c(1, 2, 3),
  y = c(4, 5, 6),
  etc = c("a", "b", "c")
)

# Arbitrary stand-in for function that can't be vectorized (no pmax)
max.fun <- function(a, b) { max(c(a, b)) }


# Using dplyr
# First, tell dplyr how to group rows
# I'm using x, y here, but whatever uniquely identifies rows would work

library(dplyr)

Y <- group_by(X, x, y)


# This seems to do what you need without dropping any columns
Y <- mutate(Y, result = max.fun(x, y))



# Another option with data.table - for a long time, if you wanted
# the fastest possible ops, data.table was it. Not sure if dplyr
# has surpassed it, but for 5 million rows, it's worth a shot.
# Key advantage: no copying of objects in memory.

library(data.table)
Z <- as.data.table(X)

# No need to assign - this is the data.table magic
Z[, result := max.fun(x, y), by = list(x, y)]

# Check it
Z