CODEWITHSUNDEEP
arraysrdynamic
Bear Bile Farming is Torture
Bear Bile Farming is Torture

Reputation: 5113

In R what is an efficient way to fill a vector of unknown size?

v <- c()
i <- 1
while (some_condition) {
    v[i] <- some_value
    i <- i + 1
}

So I am aware that each time v is modified a copy is made and v is moved. This is obviously very slow. This appears to be a trivia problem in other languages. What is the efficient way to do this in R?

Upvotes: 1

Views: 565

Answers (1)

AkselA
AkselA

Reputation: 8846

As far as I'm aware there is no native method for populating a vector of unknown length without constantly rewriting it to memory. Maybe data.table has some tricks?

Lacking any sophisticated solutions, something as simple as preallocating an oversized vector, as 12b345b6b78 suggest, can save you a fair bit of time.

unk <- 1e5

f1 <- function(unk) {
    v <- c()
    for (i in 1:unk) {
        v[i] <- i
    }
    v
}

f2 <- function(unk) {
    v <- vector(length=unk*2)
    for (i in 1:unk) {
        v[i] <- i
    }
    v[1:i]
}

f10 <- function(unk) {
    v <- vector(length=unk*10)
    for (i in 1:unk) {
        v[i] <- i
    }
    v[1:i]
}

library(microbenchmark)
mb <- microbenchmark(f1(unk), f2(unk), f10(unk), times=50)
mb
# Unit: milliseconds
#      expr       min        lq     mean    median       uq       max neval cld
#   f1(unk) 27.177678 28.364024 32.65618 29.896673 36.18994  48.37088    50   c
#   f2(unk)  8.075867  9.025156 10.87335  9.271589 10.07932  35.29222    50 a  
#  f10(unk) 11.132773 13.071857 20.46808 15.059086 21.53610 187.00786    50  b

Upvotes: 1

Related Questions