[Bug] CJ() looses timezone of POSIXct vector

[MarkusBonsch]

I have a problem when using CJ() with POSIXct objects.

If CJ is used to construct a data.table from a POSIXct vector with timezone UTC and a second grouping vector, the resulting POSIXct column in the data.table has local timezone.
If the POSIXct vector is the only input to CJ, the timezone is preserved.
Also, plain data.table() preserves the timezone.

See below for a reproducible example and my sessionInfo(). Maybe related to (#1778)

library(data.table)

## get groups and dates as input for CJ()
input_groups <- letters[1:3]
input_times  <- as.POSIXct(sprintf("2016-%02d-01", 1:12), tz = "UTC")

## input_times timezone is UTC
all(lapply(input_times, attr, "tzone") == "UTC")
# [1] TRUE

## the following shows that CJ changes timezones to NULL (local time).
test <- CJ(group = input_groups, test_time = input_times)
## test$test_time timezone is not UTC but NULL
all(lapply(test$test_time, attr, "tzone") == "UTC")
# [1] FALSE
unique(lapply(test$test_time, attr, "tzone"))
# [[1]]
# NULL

## if only a single POSIXct vector is supplied, timezone is preserved.
test2 <- CJ(test_time = input_times)
all(lapply(test2$test_time, attr, "tzone") == "UTC")
# [1] TRUE

## also, original data.table() behaves as expected.
test2 <- data.table(group = rep(input_groups, length(input_times)), date = rep(input_times, length(input_groups)))
setkey(test2, group, date)
all(lapply(test2$date, attr, "tzone") == "UTC")
# [1] TRUE

Here is my sessionInfo()

R version 3.3.0 (2016-05-03)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

locale:
[1] LC_COLLATE=German_Germany.1252  LC_CTYPE=German_Germany.1252    LC_MONETARY=German_Germany.1252
[4] LC_NUMERIC=C                    LC_TIME=German_Germany.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] data.table_1.10.4

loaded via a namespace (and not attached):
[1] tools_3.3.0

Kind regards,
Markus Bonsch

[RoyalTS]

Can confirm this.

> sessionInfo()
R version 3.3.2 (2016-10-31)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: macOS Sierra 10.12.4

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] data.table_1.10.4

loaded via a namespace (and not attached):
[1] tools_3.3.2

As far as I can tell the reason behind this is mentioned explicitly in the comments for the CJ() function definition:

# using rep.int instead of rep speeds things up considerably (but attributes are dropped).

Couldn't one simply re-apply the attributes of the input columns to the output columns after the Cartesian product has been constructed, right around line 356 in https://github.com/Rdatatable/data.table/blob/fb03ad184de08535e28d772dfe172f5ddf384a52/R/setkey.R?