Implement date_now() and date_today()

[AndrewKostandy]

I've looked for equivalent functions for lubridate's now() and today() functions that default to system time but also have a time zone argument but couldn't find any.

zoned_time_now() was the closest I could find to now(), but it doesn't have a precision argument to enable it to be equivalent. Not sure if I'm missing something. Thanks for the great package!

[DavisVaughan]

zoned_time_now() is what you want. If you really want a Date or POSIXct, you can convert to them and that will convert precisions automatically. Otherwise if you want to stay with clock types then you convert to a time point, floor that, then convert back to a zoned-time (this is the general pattern with manipulating zoned-times).

library(lubridate)
library(clock)
library(magrittr)

l_now <- now("")
l_now
#> [1] "2021-04-01 12:09:59 EDT"

c_now <- zoned_time_now("")
c_now
#> <zoned_time<nanosecond><America/New_York (current)>[1]>
#> [1] "2021-04-01 12:09:59.678911000-04:00"

# if you want to get back to base R types (which changes precisions automatically)
as.POSIXct(c_now)
#> [1] "2021-04-01 12:09:59 EDT"

# you can do this, but be warned that it first converts from the `""` time zone
# to UTC, then floors to day precision and returns a Date. occasionally that
# will produce unexpected results depending on your local offset from UTC
as.Date(c_now)
#> [1] "2021-04-01"

# swap to a sys-time and back to adjust precision if you want to keep a zoned-time
c_now %>%
  as_sys_time() %>%
  time_point_floor("second") %>%
  as_zoned_time(zoned_time_zone(c_now))
#> <zoned_time<second><America/New_York (current)>[1]>
#> [1] "2021-04-01 12:09:59-04:00"

I could add date_now() and date_today() to the high level API if that seems useful? It would just be:

library(clock)

date_now <- function(zone = "") {
  as.POSIXct(zoned_time_now(zone))
}
date_today <- function(zone = "") {
  date_floor(date_now(zone), "day")
}

date_now()
#> [1] "2021-04-01 12:20:08 EDT"
date_today()
#> [1] "2021-04-01 EDT"

Note that date_today() would return a POSIXct in your requested time zone, not a Date (which is always assumed by base R to be UTC). This is different from lubridate::today(), but I think it is more appropriate, as with today() you can get into very odd scenarios like this one due to the fact that the way today() is computed sort of assumes that Date is time zone naive, which isn't the case.

library(lubridate)

# assume that now() returned this
now <- as.POSIXct("2019-01-01 23:00:00", "America/New_York")
now
#> [1] "2019-01-01 23:00:00 EST"

# today() calls `as_date(now())` like this.
# this glosses over the fact that we went from "America/New_York" -> "UTC",
# and sort of makes the assumption that Date is time zone naive, which isn't the case
today <- as_date(now)
today
#> [1] "2019-01-01"

# making this very confusing
as_datetime(today, "America/New_York")
#> [1] "2018-12-31 19:00:00 EST"

I probably would just expose date_now(), as date_today() is a bit redundant

[AndrewKostandy]

Thanks for the detailed & clear response.

I think adding both might be worthwhile (I tend to use them frequently), but I find the name "date_now()" confusing as I wouldn't automatically realize it gives the time up to the second and not just the date as of now. date_today() would be great.

[DavisVaughan]

@AndrewKostandy I've added these two in #210:

• date_now() for the current date-time as a POSIXct
• date_today() for the current date as a Date

While thinking about these, I learned that R does NOT always treat Date as UTC. Instead, it bounces between treating Date as UTC (i.e. sys-time in clock terms) and treating it as naive-time (i.e. a yet-to-be-specified time zone), depending on the function. You can see my full thoughts on this here #203

Since R is actually not consistent about this, I've ended up changing clock to consistently treat Date as a naive-time, rather than as UTC, as that seems to make more sense with how people use Dates. This had the side-effect of making date_today() well-defined and useful in my mind.

library(clock)

# system time zone
date_today("")
#> [1] "2021-04-07"

# current Date in Europe/London
date_today("Europe/London")
#> [1] "2021-04-07"

date_now("America/Los_Angeles")
#> [1] "2021-04-07 12:39:25 PDT"

^{Created on 2021-04-07 by the reprex package (v1.0.0)}

[AndrewKostandy]

Thanks for the update, Davis.