R 使用 as.Date() 转换带有 BST/GMT 标签的 POSIXct 日期

Question

I need to use as.Date on the index of a zoo object. Some of the dates are in BST and so when converting I lose a day on (only) these entries. I don't care about one hour's difference or even the time part of the date at all, I just want to make sure that the dates displayed stay the same. I'm guessing this is not very hard but I can't manage it. Can somebody help please?

class(xtsRet)
#[1] "xts" "zoo"

index(xtsRet)
#[1] "2007-07-31 BST" "2007-08-31 BST" "2007-09-30 BST" "2007-10-31 GMT"

class(index(xtsRet))
#[1] "POSIXt"  "POSIXct"

index(xtsRet) <- as.Date(index(xtsRet))

index(xtsRet)
#[1] "2007-07-30" "2007-08-30" "2007-09-29" "2007-10-31"

Minimally reproducible example (not requiring zoo package):

my_date <- as.POSIXct("2007-04-01") # Users in non-UK timezone will need to
                                    # do as.POSIXct("2007-04-01", "Europe/London")
my_date
#[1] "2017-04-01 BST"

as.Date(my_date)
#[1] "2017-03-31"

Answer 1

Suppose we have this sample data:

library(zoo)
x <- as.POSIXct("2000-01-01", tz = "GMT")

Then see if any of these are what you want:

# use current time zone
as.Date(as.character(x, tz = ""))

# use GMT
as.Date(as.character(x, tz = "GMT"))

# set entire session to GMT
Sys.setenv(TZ = "GMT")
as.Date(x)

Also try "BST" in place of "GMT" and note the article on dates and times in R News 4/1 .

Answer 2

You can offset the POSIX objects so its not based around midnight. 1 hour (3600 secs) should be sufficient:

d <- as.POSIXct(c("2007-07-31","2007-08-31","2007-09-30","2007-10-31"))
d
[1] "2007-07-31 BST" "2007-08-31 BST" "2007-09-30 BST" "2007-10-31 GMT"

as.Date(d)
[1] "2007-07-30" "2007-08-30" "2007-09-29" "2007-10-31"
as.Date(d+3600)
[1] "2007-07-31" "2007-08-31" "2007-09-30" "2007-10-31"

Answer 3

I would suggest using as.POSIXlt to convert to a date object, wrapped in as.Date:

d <- as.POSIXct(c("2007-07-31","2007-08-31","2007-09-30","2007-10-31"))

d
[1] "2007-07-31 BST" "2007-08-31 BST" "2007-09-30 BST" "2007-10-31 GMT"

as.Date(as.POSIXlt(d))
[1] "2007-07-31" "2007-08-31" "2007-09-30" "2007-10-31"

Achieves the same thing as the +3600 above, but slightly less of a hack