djnavarro edited a comment on pull request #12154:
URL: https://github.com/apache/arrow/pull/12154#issuecomment-1063692453
Maybe it's easiest to illustrate with actual R code. The behaviour I need to
emulate with regards to the `change_on_boundary` argument looks different
depending on whether the data are stored as a datetime (e.g. POSIXct in R,
timestamp in Arrow) versus a Date (e.g., Date in R, date32 in Arrow):
```r
# a datetime object at the midnight boundary, and a date object with no time
midnight_time <- as.POSIXct(strptime("2022-03-10 00:00:00", format = "%F
%T"))
timeless_date <- as.Date("2022-03-10")
# ceiling date applied to the datetime
ceiling_date(midnight_time, unit = "day", change_on_boundary = NULL) #
returns the 10th
ceiling_date(midnight_time, unit = "day", change_on_boundary = TRUE) #
returns the 11th
ceiling_date(midnight_time, unit = "day", change_on_boundary = FALSE) #
returns the 10th
# ceiling date applied to the date
ceiling_date(timeless_date, unit = "day", change_on_boundary = NULL) #
returns the 11th
ceiling_date(timeless_date, unit = "day", change_on_boundary = TRUE) #
returns the 11th
ceiling_date(timeless_date, unit = "day", change_on_boundary = FALSE) #
returns the 10th
```
This pattern only affects times and dates deemed to be on the boundary with
respect to the rounding unit. Any time past midnight and
`lubridate::ceiling_date()` would always return the 11th for the datetime
object.
Currently, the `ceil_temporal` C++ function would return the 10th for the
analogous cases. ~~If I can work out how to add +1 time unit to the boundary
cases then I have no problems exactly mimicking lubridate behaviour. I can get
this to work for timestamp objects. This code works, for instance, and adds 1
day to the Arrow timestamp that corresponds to `midnight_time`:~~
```r
call_function("add_checked", Scalar$create(midnight_time),
Scalar$create(as.difftime("24:00:00")))
```
~~But when I try the same thing for the date, however, it throws this
error:~~
```
Error: NotImplemented: Function 'add_checked' has no kernel matching input
types (scalar[date32[day]], scalar[duration[s]])
```
~~That seems odd to me and I'm not sure why I'm getting that error because
I'm pretty sure you implemented this in
[ARROW-14901](https://github.com/apache/arrow/pull/12377). So I must be doing
something wrong?~~
EDIT: Never mind, I worked it out. 🤦🏻♀️ I can handle the
`change_on_boundary` behaviour on the R side!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
