format.POSIXct to as.POSIXct conversion issue [duplicate] - r

Question

This question already has an answer here:
R conversion of week number (UK) to POSIXct issue
(1 answer)
Closed 9 years ago.
I have a POSIXct variable with the value "2012-04-15 16:49:36 CEST". The format function returns the year, week of the year and the weekday in decimal numbers, for this example 2012 15 0. The description of the format for those less familiar with it:
%Y: Year with century.
%W: Week of the year as decimal number
(00–53) using Monday as the first day of week (and typically with the
first Monday of the year as day 1 of week 1). The UK convention.
%w: Weekday as decimal number (0–6, Sunday is 0).
Then, I try to convert the values back to a POSIXct variable and something unexpected happens. When I read the values, the functions seems to interpret a wrong date (2012-04-08). However, the surprise comes when I do the same procedure with a second example using Sys.time() and it works as expected. Can someone explain me why it does not work in the first example?
(TS <- structure(1334501376, class = c("POSIXct", "POSIXt")))
(TS_YWw <- format(TS,format="%Y %W %w"))
as.POSIXct(TS_YWw,format="%Y %W %w")
(TS <- Sys.time())
(TS_YWw <- format(TS,format="%Y %W %w"))
as.POSIXct(TS_YWw,format="%Y %W %w")
Output
> (TS <- structure(1334501376, class = c("POSIXct", "POSIXt")))
[1] "2012-04-15 16:49:36 CEST"
> (TS_YWw <- format(TS,format="%Y %W %w"))
[1] "2012 15 0"
> as.POSIXct(TS_YWw,format="%Y %W %w")
[1] "2012-04-08 CEST"
>
> (TS <- Sys.time())
[1] "2013-05-16 15:27:44 CEST"
> (TS_YWw <- format(TS,format="%Y %W %w"))
[1] "2013 19 4"
> as.POSIXct(TS_YWw,format="%Y %W %w")
[1] "2013-05-16 CEST"
By the way, I ran the code on a Windows XP 32bit machine with R 2.15.3. Thank you all!

Answer 1

It seems like a bug. Below I create a sequence of the days in 2012 (dtimes) and convert to strings and back again using the '%Y %W %w' format. The two series are compared and the head output shows which datetimes weren't preserved in the conversion. There's an obvious weekly pattern. Note also that as.POSIXct('2012 0 0', '%Y %W %w') returns NA.
dtimes <- seq(as.POSIXct('2012-1-1'), as.POSIXct('2013-1-1'), by=as.difftime(1, units='days'))
convert.YWw <- function(dtime) {
fmt <- "%Y %W %w"
string <- format(dtime, format=fmt)
as.POSIXct(string, format=fmt)
}
converted <- lapply(dtimes, convert.YWw)
preserved <- dtimes == converted
dtimes.and.converted <- mapply(function(d, c) c(dtime=d, convert=c), dtimes, converted, SIMPLIFY=FALSE)
head(dtimes.and.converted[! preserved])
# [[1]]
# NULL
#
# [[2]]
# dtime convert
# "2012-01-08 EST" "2012-01-01 EST"
#
# [[3]]
# dtime convert
# "2012-01-15 EST" "2012-01-08 EST"
#
# [[4]]
# dtime convert
# "2012-01-22 EST" "2012-01-15 EST"
#
# [[5]]
# dtime convert
# "2012-01-29 EST" "2012-01-22 EST"
#
# [[6]]
# dtime convert
# "2012-02-05 EST" "2012-01-29 EST"