[SOLVED] Is there a faster way to write code to draw many random samples from 1:N?

Is there a faster way to write code to draw many random samples from 1:N?

I'm drawing many sorted samples of size k (typically ~5) from 1:35. The code I am using is:

replicate(1000000, sort(sample(1:35, size = 5)) )

But this is rather slow. I assume that since I am using replicate, no parallelisation is happening. Is there a faster way to write this code?

Solution

TTBOMK, replicate is a wrapper for sapply, which itself is a wrapper for lapply. There are plenty of options; simple might be:

 library(future.apply)
#> Loading required package: future
 plan(multisession, workers = 6L)
 i = seq(35L)
 system.time({ X = future_replicate(1e6, sample(i, 5L, replace = TRUE)) |> t() })
#>    user  system elapsed 
#>   3.122   0.141   4.619
 head(X)
#>      [,1] [,2] [,3] [,4] [,5]
#> [1,]   26   18   31    9   35
#> [2,]   25   10   10   24   16
#> [3,]    9    9   35    4   25
#> [4,]   27   32   26   15   27
#> [5,]   13    2    6   26   24
#> [6,]   26   23   12    7   26
 library(Rfast)
 rowSort(head(X))
#>      [,1] [,2] [,3] [,4] [,5]
#> [1,]    9   18   26   31   35
#> [2,]   10   10   16   24   25
#> [3,]    4    9    9   25   35
#> [4,]   15   26   27   27   32
#> [5,]    2    6   13   24   26
#> [6,]    7   12   23   26   26