Optimizing for loop in R

Question

I have been doing a lot of research and I think I am missing something when it comes to nested for loops in R. I have two dataframes - one that contains observations and locations where I want to write the outputs and another that has the variable names I am looping through. Right now the loop works, but it is taking 14+ hours to loop through 200 rows which seems a bit excessive. Granted I am preforming 12 separate permutations (100 times) at each row, though I would ideally like to do >1000+ permutations. Is there a more efficient way of preforming this for loop? When I run a single observation it takes vey little time to complete (sub 2 seconds), which makes me beg the question that there should be a better way to accomplish this task. Any help you can give in optimizing this code would be greatly appreciated! thanks!

The main dataset is attached(fbfm.xlsx) which is called fm.std https://www.dropbox.com/s/vmd8d05yxds93j6/fbfm.xlsx?dl=0

library(rothermel)
u.val<-c(5,10,15,25,35,45,55,65,75,85,95,100)
unames <- data.frame(u=u.val,ros.nam=paste("u",u.val,"_ROS",sep=""), stringsAsFactors = FALSE)
ros.out<-data.frame(fm.std)
for (i in 1:dim(unames)[1]){
     ros.out[,unames[i,'ros.nam']]<-999
          }
ros.out <- as.vector(ros.out)
fm.std <- as.vector(fm.std)
for (i in 1:dim(ros.out)[1]){
   ros.out[i,1:32]
     for (u in 1:dim(unames)[1]){
         ros.out[i,unames[u,'ros.nam']]<-mean(rosunc(modeltype=fm.std[i,'Fuel_Model_Type'], #Dyanmic or static model
                                            w=fm.std[i,4:8], # fuel loads (1, 10, 100, herb, and shrub)
                                            s=fm.std[i,9:13], # SAV measurements
                                            delta=fm.std[i,14], #fuel bed depth
                                            mx.dead=fm.std[i,15], # dead fuel mositure of extinction
                                            h=fm.std[i,16:20], # heat content for fuel classes
                                            m=fm.std[i,c(25,24,23,26,30)], #percent moisture of fuel classes
                                            u = unames[u,'u'],
                                            slope=0,
                                            sdm=0.3,
                                            nsim=100) ) #wind and slope of 0 }}

Answer 1

Consider a more vectorized sapply() approach passing in two vectors, u.val and 1:nrow(fm.std) . This will build a 200-row, 12-column matrix that you can convert to a dataframe and then cbind to original dataframe.

ucols <- sapply(u.val, 
                function(x, y){
                   mean(rosunc(modeltype=fm.std[y,'Fuel_Model_Type'],  # Dyanmic or static model
                               w=fm.std[y,4:8],       # fuel loads (1, 10, 100, herb, and shrub)
                               s=fm.std[y,9:13],      # SAV measurements
                               delta=fm.std[y,14],    # fuel bed depth
                               mx.dead=fm.std[y,15],  # dead fuel mositure of extinction
                               h=fm.std[y,16:20],     # heat content for fuel classes
                               m=fm.std[y,c(25,24,23,26,30)],     # percent moisture of fuel classes
                               u=x,
                               slope=0,
                               sdm=0.3,
                               nsim=100))
                }, 1:nrow(fm.std))

# CONVERT MATRIX TO DATA FRAME
ucols <- data.frame(ucols)
# RENAME COLUMNS
names(test) <- paste("u",u.val,"_ROS",sep="")

# BIND COLUMNS TO ORIGINAL DATA FRAME
ros.out <- cbind(fm.std, ucols)

Alternatively, consider using outer() with transpose, t() to achieve the 200-row and 12-col matrix.

ucols <- t(outer(u.val, 1:nrow(fm.std), 
                    function(x, y){
                         mean(rosunc(...))
                    }
           ))
...