as.Date does not handle the case where tz is a vector

[lsilvest]

We handle tz as a vector it correctly in the extract functions (nano_month, etc.), but in our as.Date we convert first to POSIXct and then call the base as.Date . So the idea here is to use our own functions instead of relying on POSIXct.

[eddelbuettel]

Could do. Do you have a use case for that?

[lsilvest]

Yes, I recently had a set of observations over a couple of months with UTC times, all with different geographical locations and needed to group the observations by day. I converted the geographical locations to timezones, but couldn't directly use nanotime's as.Date as it doesn't accept a vector of timezones. Got out of trouble with nano_month, etc., but we could easily fix our as.Date to handle that out of the box.

[eddelbuettel]

Do you have a clever idea for how to do 'maybe-vectorization' ? Or maybe we already do elsewhere? (Asking because I thought 20 years ago it was an important feature over several args so I did a pretty ham-fisted set of nested loop and always wondered about 'better' way....)

[lsilvest]

Fortunately, we've already got some vectorization for this at C level, so it turns out to be fairly straightforward. See if this new code seems OK to you.

[eddelbuettel]

Really nice. I had one gut-instinct idea now but not strong feeling either way. With/without automagic tz works for me.