pylr vs doBy speed test

gistfile1.txt

# 分類用
types <- c("A","B","C","D","E")
# 行数
obs <- 4e+06
# データフレームの作成
dat <- data.frame(type = as.factor(sample(types, obs, replace=TRUE)))

# 列数を増やしながら時間を計測していく
Nmax <- 10
plyr_time <- 0
doBy_time <- 0

for (N in 1:Nmax){
  dat[,N+1] <- round(runif(obs, min=0, max = 1), digits = 2)
  names(dat)[N+1] <- paste("value", N, sep="")

  plyr_time[N] <- system.time(
    plyr_res <- ddply(dat, .(type), summarize, 
                      mean_percent = mean(value1))
    )[3]
  
  doBy_time[N] <- system.time(
    doBy_res <- summaryBy(value1 ~ type, data = dat, FUN = mean)
    )[3]
}