factor_performance.R

library(microbenchmark)

a <- sample(letters, 1e6, TRUE)
b <- as.factor(a)

microbenchmark(c <- a[a == "d"],
               d <- b[b == "d"],
               e <- {
                 mtch <- match("d", levels(b))
                 b[as.integer(b) == mtch]
               })

## Unit: milliseconds
##                                                                     expr
##                                                         c <- a[a == "d"]
##                                                         d <- b[b == "d"]
##  e <- {     mtch <- match("d", levels(b))     b[as.integer(b) == mtch] }
##        min        lq      mean    median        uq      max neval cld
##  19.685260 19.747577 20.588849 19.836498 20.275723 27.79781   100  b 
##  43.150390 43.381784 45.982250 43.683979 48.716526 70.98689   100   c
##   4.003705  4.106672  4.466901  4.149938  4.223226 12.29115   100 a

all.equal(c, as.character(d), as.character(e))
## [1] TRUE

object.size(a)
## 8001288 bytes

object.size(b)
## 4001856 bytes

object.size(as.integer(b))
## 4000040 bytes

object.size(c)
## 308216 bytes

object.size(d)
## 155920 bytes

object.size(e)
## 155920 bytes