Skip to content

Rounding

Rounding down or up a nanotime type

Description

The functions nano_floor and nano_ceiling round down or up, respectively. Although the underlying implementation of nanotime has negative numbers for values before 1970-01-01 UTC, the rounding is always done backward in time for nano_floor and forward in time for nano_ceiling. The functions take a nanotime argument x which is the instance to round, together with a second argument precision which indicates an arbitrary precision to which the rounding should be performed. This argument can be either a nanoduration or or a nanoperiod. In the latter case, the argument tz must also be specified in order to give the nanoperiod a meaning. Finally, the nanotime argument origin can be optionally specified to fix the rounding to a specific point in time.

Usage

nano_ceiling(x, precision, ...)

nano_floor(x, precision, ...)

## S4 method for signature 'nanotime,nanoduration'
nano_ceiling(x, precision, origin = nanotime())

## S4 method for signature 'nanotime,nanoduration'
nano_floor(x, precision, origin = nanotime())

## S4 method for signature 'nanotime,nanoperiod'
nano_ceiling(x, precision, origin = nanotime(), tz)

## S4 method for signature 'nanotime,nanoperiod'
nano_floor(x, precision, origin = nanotime(), tz)

Arguments
x a nanotime object which must be sorted
precision a nanoduration or nanoperiod object indicating the rounding precision
... for future additional arguments
origin a nanotime scalar indicating the origin at which the rounding is considered
tz a character scalar indicating the time zone in which to conduct the rounding

Details

This flexible rounding must be understood in the context of a vector. The rounding precision can then be considered as an interval that defines a grid over which the elements are either assigned to the starting value of the interval to which they belong (nano_floor) or the ending value of the interval to which they belong (nano_ceiling). This allows for a grouping of a nanotime vector on which a statistic may then be run. In the examples below, such a use case is shown in the context of a data.table object.

If "business" concepts such as month or days are needed, the argument precision must be of type period. It is then mandatory to specify the timezone argument tz as this ensures timezone correctness of the intervals including for example for the rare hourly transitions of some countries going from a timezone with a whole hour difference with UTC to one with a fractional hour difference. In the case of a period, the functions align the rounding if the precision is an integer divisor of a larger quantity. For instance, if one specifies a rounding of 6 hours, a divisor of a day, the hours are aligned on days and the rounding is made to a grid at hours 0, 6, 12 and 18 in the specified timezone. If the precision is not a divisor, the grid is aligned to the nearest hour before the first element of the vector to round.

The argument origin controls the reference point of the rounding, allowing arbitrary specification of the reference point of the rounding.

Examples

## Not run: 
## "classic" rounding:
nano_floor(as.nanotime("2010-10-10 11:12:15 UTC"), as.nanoduration("01:00:00"))
## rounding with arbitrary precision:
nano_floor(as.nanotime("2010-10-10 11:12:15 UTC"), as.nanoduration("06:00:00"))
nano_floor(as.nanotime("2010-10-10 11:23:15 UTC"), as.nanoduration("00:15:00"))
nano_ceiling(as.nanotime("2010-10-10 11:23:15 UTC"), as.nanoduration("01:15:23"))
## controlling the reference point via the 'origin' argument:
nano_ceiling(as.nanotime("2010-10-10 11:23:15 UTC"),
             as.nanoduration("01:15:23"),
             origin=as.nanotime("2010-10-10 11:23:15 UTC"))
## using business concepts and rounding across a daylight saving change:
v <- seq(as.nanotime("2020-03-08 America/New_York"),
         by=as.nanoperiod("06:00:00"), length.out=8, tz="America/New_York")
print(nano_floor(v, as.nanoperiod("1d"), tz="America/New_York"), tz="America/New_York")
## using the concept in a 'data.table':
library(data.table)
n <- 3 * 24
idx <- seq(as.nanotime("2020-03-07 America/New_York"),
           by=as.nanoperiod("01:00:00"), length.out=n, tz="America/New_York")
dt <- data.table(idx, a=1:n, b=2:(n+1))
dt_mean <- dt[, list(mean = mean(a)),
              by=nano_ceiling(idx, as.nanoperiod("1d"), tz="America/New_York")]

## End(Not run)