with_tz with a vector of timezones

问题:

I have a dataframe like so:

library(dplyr) data <- data_frame(   timestamp_utc = c('2015-11-18 03:55:04', '2015-11-18 03:55:08',                      '2015-11-18 03:55:10'),   local_tz = c('America/New_York', 'America/Los_Angeles',                 'America/Indiana/Indianapolis')   ) 

I need to create a new variable that converts the UTC timestamp to the local time as defined in the local_tz column. However, both format and with_tz (from lubridate) expect only one timezone, not a vector of timezones. I'm looking for something like this:

mutate(data, timestamp_local = with_tz(timestamp_utc, tzone = local_tz)) 

Any ideas?

回答1:

First make sure your data is loaded as a date - I had to convert to date first:

data$timestamp_utc <- as.POSIXct(data$timestamp_utc, tz = "UTC") 

Then you can use the function rowwise from dplyr, in combination with do:

library(lubridate) library(dplyr) z <- data %>% rowwise() %>%               do(timestamp_local = with_tz(.$timestamp_utc, tzone = .$local_tz)) data$timestamp_local <- z$timestamp_local  data$timestamp_local [[1]] [1] "2015-11-17 22:55:04 EST"  [[2]] [1] "2015-11-17 19:55:08 PST"  [[3]] [1] "2015-11-17 22:55:10 EST" 

We need to make the timestamp_local column a list, as otherwise all the timezones are converted back to one you can only have one timezone in a vector).

回答2:

Here is one method. With this, the result has to be a string, otherwise unlist() or c() will turn the result back to the system timezone for every element in the list.

It's still slow though because it is not vectorized.

> get_local_time <- function(timestamp_utc, local_tz) {     l <- lapply(seq(length(timestamp_utc)),                  function(x) {format(with_tz(timestamp_utc[x], local_tz[x]), "%FT%T%z")})     unlist(l)     }  > mutate(data, timestamp_local = get_local_time(timestamp_utc, tzone = local_tz))  Source: local data frame [3 x 3]          timestamp_utc                     local_tz          timestamp_local                (time)                        (chr)                    (chr) 1 2015-11-18 03:55:04             America/New_York 2015-11-17T22:55:04-0500 2 2015-11-18 03:55:08          America/Los_Angeles 2015-11-17T19:55:08-0800 3 2015-11-18 03:55:10 America/Indiana/Indianapolis 2015-11-17T22:55:10-0500 

Update 2015-11-24

Using dplyr::combine() rather than unlist() allows the variable to remain datetimes with the right timezone attributes rather than converting to strings.

> get_local_time <- function(timestamp_utc, local_tz) {     l <- lapply(seq(length(timestamp_utc)),                  function(x) {with_tz(timestamp_utc[x], local_tz[x])})     combine(l)     }  > mutate(data, timestamp_local = get_local_time(timestamp_utc, tzone = local_tz))  Source: local data frame [3 x 3]          timestamp_utc                     local_tz     timestamp_local                (time)                        (chr)              (time) 1 2015-11-18 03:55:04             America/New_York 2015-11-17T22:55:04 2 2015-11-18 03:55:08          America/Los_Angeles 2015-11-17T19:55:08 3 2015-11-18 03:55:10 America/Indiana/Indianapolis 2015-11-17T22:55:10 

文章来源: with_tz with a vector of timezones