Isoweek() problem in European timezones

[johandicap]

Hi, it appears that there is a problem with the isoweek() function for European timezones.

For instance tomorrow, April 14th 2015, it is Tuesday of week 16 here in Copenhagen, Denmark, but isoweek says it is week 15:

> ymd_hms("2015-04-14 16:45:00", tz="Europe/Copenhagen")
[1] "2015-04-14 16:45:00 CEST"
> isoweek(ymd_hms("2015-04-14 16:45:00", tz="Europe/Copenhagen"))
[1] 15  # Wrong

For UTC, it correctly returns week 16:

> ymd_hms("2015-04-14 16:45:00", tz="UTC")
[1] "2015-04-14 16:45:00 UTC"
> isoweek(ymd_hms("2015-04-14 16:45:00", tz="UTC"))
[1] 16  # Correct

Note that CEST (Central European Summer Time) is UTC + 2, so this should not influence the week number on a Tuesday afternoon.

It does not work for CET either:

> ymd_hms("2015-04-14 16:45:00", tz="CET")
[1] "2015-04-14 16:45:00 CEST"
> isoweek(ymd_hms("2015-04-14 16:45:00", tz="CET"))
[1] 15  # Wrong

So this appears to be a bug in lubridate or in the general handling of timezones.

(I am using lubridate 1.3.3 with 32-bit R 3.1.3 in RStudio 0.98.1103 on 64-bit Windows 7 SP1.)

Best regards,
Johan

ps: Finding the isoweek() function in the first place took me actually quite a while, since I did not see it mentioned when I skimmed the week() documentation via ?week. I suggest adding a "See also" section or an example mentioning isoweek() to the week() documentation to highlight its existence.

[vspinu]

Note that CEST (Central European Summer Time) is UTC + 2, so this should not influence the week number on a Tuesday afternoon.

Yep. It's a bug. Thanks.

ps: Finding the isoweek() function in the first place took me actually quite a while, since I did not see it mentioned when I skimmed the week() documentation via ?week. I suggest adding a "See also" section or an example mentioning isoweek() to the week() documentation to highlight its existence.

Indeed. The docs should be refactored and isoweek should be at least mentioned in the week docs.

[vspinu]

The problem was the small difference introduced by daylight saving time. To avoid this issue the isoweek is computed now exclusively in UTC.

It turns out that the documentation for week and isoweek is is now the same page. I forgot that fixed that some time ago. You should probably upgrade to the github version.