groupby_benchmark.R

library(data.table)
library(dtplyr)
library(dplyr, warn.conflicts = FALSE)
library(microbenchmark)
library(ggplot2)

set.seed(123)
df <- data.frame(
  letters = sample(c('A', 'B', 'C'), size = 3e6, replace = TRUE, prob = c(.5, .4, .1)),
  number = sample(c(1:5), size = 3e6, replace = TRUE)
)
dpt <- lazy_dt(df)
dt <- data.table(df)


mm <-
  microbenchmark(
    base = aggregate(df["number"], df["letters"], mean),
    dplyr = df %>% group_by(letters) %>% summarise(mean(number)),
    DT = dt[, mean(number), by = letters],
    dtplyr = dpt %>% group_by(letters) %>% summarise(mean(number)) %>% as.data.table()
  )

autoplot(mm)