Vectorised force_tz()

[hadley]

At a minimum, this should be an error:

force_tz(rep(now(), 2), c("US/Eastern", "US/Central"))

But ideally, it would allow you to combine a vector of times with a vector of times, as when a dataset records local times and you want to standardise.

[vspinu]

Why is this needed? If you know your groups you can map by groups yourself.

This would go against all other functions which don't vectorize over TZ and will give wrong hints to folks who ask multiple tzs within a vector.

[hadley]

Obviously I can do that, but it's pretty tough challenge for most people. It's a not uncommon to have a vector of local times that you need to fix.

I think it's reasonable to vectorise tz here in the same way it's reasonable to vectorise hour, minute etc in the appropriate settors + update().

[vspinu]

This one is likely to be challenging. The update code is already quite complex as it tries to cut edges whenever possible.

[zyxdef]

Can this become a feature request?

[vspinu]

I am looking at this right now. Trying to put CCTZ to do some work here.

[vspinu]

Folks, vectorized force_tz doesn't make sense. tz argument indicates the time zone of the output vector, but in R vectors cannot have heterogeneus time zone. Could you please elaborate on the exact semantics that you had in mind?

[hadley]

I think what I actually want is more like local_time(datetimes, timezones), and feel like maybe it's more natural in hms.

[vspinu]

I see, that could be straightforwardly done with CCTZ. Will look into it. We need local_time function anyways.

[gregmacfarlane]

Here's a stack overflow questions detailing the precise problem and why a solution of some kind would be useful: https://stackoverflow.com/questions/33848563/with-tz-with-a-vector-of-timezones/33848764

[vspinu]

Just added two functions, force_tzs and local_time, both verctorized on the time zone argument. See the docs for more details and usage examples.

@gregmacfarlane, the SO questions looks for vectorized with_tz. Is that really what you want? with_tzs could be added for completeness, but one should be aware that time zone of the returned vector is meaningless.