`with_tz()` warns about invalid timezones, but returns them correctly anyway

[steffilazerte]

This is a problem I've noticed on Windows only (not my Ubuntu machine). The warnings are also popping up in the AppVeyor tests (but not on Travis CI for Ubuntu and Mac).

I think it's only in lubridate v0.7.0 (related to c10c8cf) but I had problems trying to install v0.6.0 to double check.

with_tz() warns about invalid timezones, but returns them correctly anyway...

> Sys.time()
[1] "2017-11-02 00:17:10 AST"

> lubridate::with_tz(Sys.time(), "UTC")
[1] "2017-11-02 04:17:26"

> lubridate::with_tz(Sys.time(), "America/Vancouver")
[1] "2017-11-01 21:17:28 PDT"
Warning message:
In lubridate::with_tz(Sys.time(), "America/Vancouver") :
  Unrecognized time zone 'America/Vancouver'

> lubridate::with_tz(Sys.time(), "Etc/GMT+8")
[1] "2017-11-01 20:18:25 -08"
Warning message:
In lubridate::with_tz(Sys.time(), "Etc/GMT+8") :
  Unrecognized time zone 'Etc/GMT+8'

If I try a non-legitimate name I get a second warning message (and UTC):

> lubridate::with_tz(Sys.time(), "America/Van")
[1] "2017-11-02 04:41:31 GMT"
Warning messages:
1: In lubridate::with_tz(Sys.time(), "America/Van") :
  Unrecognized time zone 'America/Van'
2: In as.POSIXlt.POSIXct(x, tz) : unknown timezone "America/Van"

The first two are both legitimate timezones (i.e. they're in OlsonNames())

> OlsonNames()[OlsonNames() == "America/Vancouver"]
[1] "America/Vancouver"

> OlsonNames()[OlsonNames() == "Etc/GMT+8"]
[1] "Etc/GMT+8"

> OlsonNames()[OlsonNames() == "America/Van"]
character(0)

[vspinu]

I don't see that on my windows machine. The internal check relies on same Olson database as OlsonNames but through the CCTZ lookup mechanism. Do you see this warning with all Timezones? "America/New_York", "Europe/Amsterdam"?

Could you please check if force_tz works as expected:

> lubridate::force_tz(Sys.time(), "America/Chicago")
[1] "2017-11-02 10:52:20 CDT"

If force_tz doesn't work, its probably some local installation problem.

[steffilazerte]

Thanks for your speedy reply,

Both with_tz and force_tz have problems on my Windows machine and on AppVeyor (in a clean testing package). However, as.POSIXct works fine.

The output of the failed tests from AppVeyor is here (the as.POSIXct tests passed): testthat.Rout.fail.txt

Considering that this is a problem on AppVeyor and that as.POSIXct works, do you still think it's a local installation problem?

Output from my Windows machine:

with_tz() does the conversion correctly but warns

> lubridate::with_tz(Sys.time(), tzone = "America/New_York")
[1] "2017-11-02 12:38:43 EDT"
Warning message:
In lubridate::with_tz(Sys.time(), tzone = "America/New_York") :
  Unrecognized time zone 'America/New_York'

> lubridate::with_tz(Sys.time(), tzone = "Europe/Amsterdam")
[1] "2017-11-02 17:39:15 CET"
Warning message:
In lubridate::with_tz(Sys.time(), tzone = "Europe/Amsterdam") :
  Unrecognized time zone 'Europe/Amsterdam'

force_tz() errors

> lubridate::force_tz(Sys.time(), tzone = "UTC")
[1] "2017-11-02 16:46:06 UTC"

> lubridate::force_tz(Sys.time(), tzone = "America/New_York")
Error in C_force_tz(time, tz = tzone, roll) : 
  CCTZ: Unrecognized output timezone: "America/New_York"

> lubridate::force_tz(Sys.time(), tzone = "America/Vancouver")
Error in C_force_tz(time, tz = tzone, roll) : 
  CCTZ: Unrecognized output timezone: "America/Vancouver"

> lubridate::force_tz(Sys.time(), tzone = "Europe/Amsterdam")
Error in C_force_tz(time, tz = tzone, roll) : 
  CCTZ: Unrecognized output timezone: "Europe/Amsterdam"

as.POSIXct() works fine

> as.POSIXct("2017-09-09 00:00:00", tz = "UTC")
[1] "2017-09-09 UTC"

> as.POSIXct("2017-09-09 00:00:00", tz = "Europe/Amsterdam")
[1] "2017-09-09 CEST"

> as.POSIXct("2017-09-09 00:00:00", tz = "America/New_York")
[1] "2017-09-09 EDT"

[vspinu]

Ok, thanks. It's the TZDIR env issue. Preparing a fix asap.

[steffilazerte]

Great!

[vspinu]

@steffilazerte Do you have the ability to test the dev version? I pushed a fix which works for me. I would be extremely grateful if you could check it asap. Lubridate was removed from CRAN because it failed on solaris, and I think it's the same issue.

[steffilazerte]

I'll give it a shot, get back to you soon.

[steffilazerte]

It works perfectly on my windows machine. To double check I uninstalled the dev version and reinstalled from cran and I got all the same errors. Switching back to the dev version was perfect again.

I couldn't test on AppVeyor (it had problems installing lubridate from the github site), but I can't see any reason why that wouldn't be fixed as well.

Let me know if you need any other tests! Thanks!

[vspinu]

Ok. Great. Nevermind for AppVeyor. I cloned your testing package and it's fixed. Thanks a bunch for you instant input!

[steffilazerte]

Awesome, glad to hear it worked out