问题
I want to subtract 1 day from a POSIX date and end up at the same time around DST.
For example, when I add a day:
> as.POSIXct('2009-03-08 23:00:00.000') + 86400
[1] "2009-03-09 23:00:00 EDT"
But when I go past, it offsets:
> as.POSIXct('2009-03-08 23:00:00.000') - 86400
[1] "2009-03-07 22:00:00 EST"
What's the best way to deal with absolute time differences around DST? Usually I deal with this by converting the times into strings and dealing with them separately so that DST isn't applied.
回答1:
Your code is doing exactly what you asked it to do, because you didn't ask it to add or subtract one day, you asked it to add or subtract 24 hours. 24 hours (86400 seconds) before 2009-03-08 23:00:00 EDT is 2009-03-07 22:00:00 EST. I'm not familiar with the R libraries, but I am familiar with the POSIX functions that they wrap. If you take a POSIXct, decrease its day property by 1, and then "re-cast" it to POSIXct via POSIXlt ( to ensure that e.g. February -1st becomes January 31st) then you should be able to subtract one day reliably.
回答2:
Store datetimes as UTC - or in this case convert to UTC subtract a day and convert back to local time. (This ignores leap seconds)
回答3:
Thanks hobbs! I need to do a little more work with it, but subtracting from the day slot works in POSIXlt:
> a <- as.POSIXct('2009-03-08 23:00:00.000')
> as.POSIXlt(a)
[1] "2009-03-08 23:00:00 EDT"
> a <- as.POSIXlt(a)
> a$mday <- a$mday -1
> a
[1] "2009-03-07 23:00:00 EDT"
回答4:
If you just want to count the days, you can use trunc to just move the day:
> trunc(Sys.time(), "day") + 86400
[1] "2009-09-13 PDT"
> trunc(Sys.time(), "day") - 86400
[1] "2009-09-11 PDT"
回答5:
The package lubridate provides two additional timespan classes that specify what is somewhat unclear in the base package. From tha manual:
Durations
Durations measure the exact amount of time that occurs between two instants. This can create unexpected results in relation to clock times if a leap second, leap year, or change in daylight savings time (DST) occurs in the interval.
Periods
Periods measure the change in clock time that occurs between two instants. Periods provide robust predictions of clock time in the presence of leap seconds, leap years, and changes in DST.
来源：https://stackoverflow.com/questions/1395499/posixct-times-around-dst