在基础R中快速有效地将年，朱利安，& 时间从多列转换为POSIXct。

Question

我有很多大型数据框（+500,000行），这些数据框中的日期时间信息存储在多列中。它的格式不是MMDDYYYY，而是一列是年，下一列是朱利安日历日，第三列是时间。数据的结构是这样的。

df<-data.frame(YEAR = sample(2000:2020,10000, replace=T), 
           JULIEN = sample(1:365,10000,replace=T),
           Time = sample(0:59,10000,replace = T),
           dataVar1 = runif(10000,1.0,10.0),
           dataVar2 = runif(10000,20.0,100.0))

到目前为止，我一直在用这个方法来处理:

timeR<-vector()
for (i in 1:dim(df)[1]){
currentTime<-paste(as.Date(df$JULIEN[i], origin=paste(df$YEAR[i]-1,"-12-31", sep = "")),formatC(df$Time[i], width = 4, format = "d", flag = "0"))
timeR<-c(timeR,currentTime)
}
df<-cbind(timeR,df[, ! names(df) %in% c("YEAR","JULIEN","Time")])
df$timeR<-as.POSIXct(df$timeR,format = "%Y-%m-%d %H%M", tz = "EST")
rm(timeR,i,currentTime)

但这需要大量的时间。有什么办法可以让它运行得更快吗？谢谢你。

Answer 1

paste 和 as.Date 矢量化

v1 <- as.Date(df$JULIEN, origin = paste0(df$YEAR-1,"-12-31"))
currentTime <- paste(v1, formatC(df$Time, width = 4, format = "d", flag = "0"))

基准

system.time({
 timeR<-vector()
 for (i in 1:dim(df)[1]){
  currentTime<-paste(as.Date(df$JULIEN[i], origin=paste(df$YEAR[i]-1,"-12-31", sep = "")),formatC(df$Time[i], width = 4, format = "d", flag = "0"))
  timeR<-c(timeR,currentTime)
 }})
#   user  system elapsed #
#  1.300   0.061   1.366 

system.time({
 v1 <- as.Date(df$JULIEN, origin = paste0(df$YEAR-1,"-12-31"))
 currentTime <- paste(v1, formatC(df$Time, width = 4, format = "d", flag = "0"))
 })
#  user  system elapsed 
#  0.076   0.004   0.080