Stop using for loops in R

Many have mentioned the use of “for loops” in R is suboptimal. We are all taught about for loops when we begin coding because they work and are easy to follow. However, computers hate it.

I decided to check this out by benchmarking execution time and comparing a for loop with a do.call alternative in R.

library(microbenchmark)

loop_f <- function(x){
  opt<-c()
  for (i in seq(1:x)){
    opt[i] <-i+1
  }
  return(opt)
}


docall_f <-function(x){
  x<-list(1:x)
  f <-function(x) return(x+1)
  out<-do.call(f,x)
  return(out)
}

Moment of truth, run the benchmark test.

microbenchmark(
  loop_f(100),
  docall_f(100)
)

Results are quite shocking. Even for a very simple function, the performance is 10x (median) better when we use do.call instead of looping through the data.

Conclusion

Stop using for loops when possible.