Preallocation in r

Question

In Matlab it is drilled into you - preallocate, preallocate, preallocate. If you fail to do this, gremlins will eat cpu cycles and you will be a bad person. Is it as important to preallocate in r as it is in Matlab?

Answer 1

Since in R we tend to avoid explicit loops, it is not as important. Many functions do it under the hood for us. Of course, if you insist on using for loops, you should pre-allocate to avoid growing an object in a loop (which is one of the slowest operations you can do). Relevant reading material: The R Inferno.

Answer 2

some examples

test1=function(){
  l=list()
  for(i in 1:10000){
    l=append(l,"abc")
  }
  return(l)
}
system.time(test1()) # 2.367 sec

test2=function(){
  l=vector("list", 10000)
  for(i in 1:10000){
    l[i]="abc"
  }
  return(l)
}
system.time(test2()) # 0.015 sec

test3=function(){
  l=list()
  for(i in 1:10000){
    l[i]="abc"
  }
  return(l)
}
system.time(test3()) # 0.309 sec

test4=function(){
  return(lapply(1:10000, function(x) "abc"))   
}
system.time(test4()) # 0.003

R for loops suck indeed :)

which is problematic, cause its not always readable to change it into a lappy