vapply

vapply.R

testapply = list()
testvapply = list()
# First, we generate a 10001 x 10001 matrix of random numbers pulled from a gaussian distribution. Our goal is to sum each of the columns and return a vector of each sum. Here, we have two strategies:

# 1) Loop through each column, calculate the sum, and assign into a vector.
# 2) apply() the sum function to each column.

numbers = matrix(rnorm(10001^2,0,1),nrow=10001,ncol=10001)
iter=1

for(max in seq(1,601,25)) {
  print(max)
  nnumbers = numbers[,1:max,drop=FALSE]
  # First: the for loop.
  
  initialtime = proc.time()[3]
  totalsum = rep(0,max)
  for(i in 1:max) {
    totalsum[i] = sum(nnumbers[,i,drop=FALSE])
  }
  testapply$timeloop[iter] = proc.time()[3] - initialtime   
  
  # Now timing the apply function. The 2 in apply() tells R to sum the matrix nnumbers by column.
  
  initialtime = proc.time()[3]
  totalsum = apply(nnumbers,2,sum)
  testapply$timeapply[iter] = proc.time()[3] - initialtime
  
  testapply$iteration[iter] = max
  
  initialtime = proc.time()[3]
  totalsum = vapply(nnumbers,sum,FUN.VALUE = numeric(1))
  testvapply$timeapply[iter] = proc.time()[3] - initialtime
  testvapply$iteration[iter] = max
  iter = iter + 1 
}

forvsapplydf = as.data.frame(do.call(cbind,testapply) )
forvsvapplydf = as.data.frame(do.call(cbind,testvapply) )

Expand Plotting Code

ggplot() +
  geom_line(data=forvsapplydf,aes(x=iteration,y=timeloop),color="green") +
  geom_line(data=forvsapplydf,aes(x=iteration,y=timeapply),color="red") + 
  geom_line(data=forvsvapplydf,aes(x=iteration,y=timeapply),color="blue") + 
  theme(text=element_text(family="Helvetica"),
        panel.grid.major.y = element_line(color = "grey90"))