Randomly associate elements of two vectors given conditions

Question

I have a data.table of capitals

capitals<-data.table(capital=c(100,50,25,5))
capitals
   capital
1:     100
2:      50
3:      25
4:       5
         


        

Answer 1

First off, thank you everyone for your attempts. I've implemented a simple algorithm which is quicker than the answers thus far (and easier to understand, I think).

ffben<-function(capitals, losses){ #note, the inputs here are vectors, not data.tables
  lossSamples<-numeric()
  capitals<-sort(capitals)
  for(i in 1:(length(capitals)-1)){
    lossSamples[i]<-sample(x=losses[losses<=capitals[i]],1)
    losses<-losses[-which(losses==lossSamples[i])[1]]
  }
  lossSamples[i+1]<-losses[1]
  return(data.table(capitals=capitals, losses=lossSamples))
}

Benchmark against alexis's solution

cap2 = 1:10000; los2 = pmax(0,1:10000-10)  #10 capitals and losses
microbenchmark::microbenchmark(ffalex(cap2, los2), ffben(cap2, los2), times = 5)

Unit: seconds
               expr   min    lq median    uq   max neval
 ffalex(cap2, los2) 3.725 3.775  3.792 3.977 5.606     5
  ffben(cap2, los2) 2.680 2.868  2.890 2.897 3.056     5

However, I recognize that my solution still has much room for improvement, so I won't accept it as the best answer unless it's still the quickest solution in a week or so. In particular, I am hoping someone can develop a data.table based solution that takes advantage of data.table's inherent binary searching algorithms.