data.table 整表

# Loading package
library(plyr)
library(dplyr)
library(dtplyr)
library(data.table)
library(pipeR)

load('dat.RData')
st <- proc.time()
dat_list <- dat %>>% as.data.table %>>%
  `[`( , `:=`(IsOpened = !is.na(IsOpened), IsClick = !is.na(IsClick))) %>>%
  `[`( , grp := IsOpened * 3 + IsClick * 5) %>>%
  `[`( , grp_split := ifelse(all(grp == 0), 1, ifelse(all(grp == 3 | grp == 0), 2, 3)),
       by = "Account,Locale,Order_ID") %>>%
  split(.$grp_split)

dat_list[[1]] <- dat_list[[1]] %>>% setorder(Account, Locale, Order_ID, Send_Date) %>>%
  `[`(, .SD[1] , by = "Account,Locale,Order_ID") %>>% `[`( , IsClick := as.character(IsClick))
dat_list[[2]] <- dat_list[[2]] %>>%
  setorder(Account, Locale, Order_ID, Opened_Date) %>>%
  `[`(, .SD[1] , by = "Account,Locale,Order_ID") %>>% `[`( , IsClick := as.character(IsClick))
dat_list[[3]] <- dat_list[[3]] %>>% `[`(IsClick == 1) %>>%
  setorder(Account, Locale, Order_ID, Click_Date) %>>%
  `[`(, IsClick2 := paste(IsClick, collapse = ","), by = "Account,Locale,Order_ID") %>>%
  `[`(, .SD[1] , by = "Account,Locale,Order_ID") %>>% `[`( , IsClick := IsClick2) %>>%
  `[`( , IsClick2 := NULL)
dat_res_dt <- dat_list %>>% rbindlist
proc.time() - st
# user  system elapsed
# 0.05    0.05    0.06

table(dat_res_dt$grp_split) # 540 416 43



load('dat.RData')
st <- proc.time()
dat_res <- dat %>>% mutate(tmp=paste(Account,Locale,Order_ID)) %>>% split(.$tmp) %>>%
  llply(function(x){
    if(any(!is.na(x$IsOpened))) {
      x <- x %>>% filter(!is.na(IsOpened))
      if(any(!is.na(x$IsClick))) {
        x <- x %>>% filter(!is.na(x$IsClick))
        IsClick <- x$IsClick %>>% strsplit(split=',') %>>% (purrr::simplify(.)) %>>%
          unique %>>% na.omit %>>% paste(collapse=',')
        x <- x %>>% arrange(Click_Date) %>>% {.[1,]} %>>% mutate(grp = 3)
        x$IsClick = IsClick
      } else {
        x <- x %>>% arrange(Opened_Date) %>>% {.[1,]}  %>>% mutate(grp = 2)
      }
    } else {
      x <- x %>>% arrange(Send_Date) %>>% {.[1,]} %>>% mutate(grp = 1)
    }
    return(x)
  }) %>>% rbindlist
proc.time() - st
#  user  system elapsed
# 12.50   16.52   18.23

dat_res %>>% group_by(grp) %>>% summarise(n()) # 540 416 43