time_add: Arithmetics with periods

View source: R/addition.R

time_addR Documentation

Arithmetics with periods

Description

Add periods to date-time objects. Periods track the change in the "clock time" between two civil times. They are measured in common civil time units: years, months, days, hours, minutes, and seconds.

Usage

time_add(
  time,
  periods = NULL,
  year = NULL,
  month = NULL,
  week = NULL,
  day = NULL,
  hour = NULL,
  minute = NULL,
  second = NULL,
  roll_month = "preday",
  roll_dst = c("post", "pre"),
  ...
)

time_subtract(
  time,
  periods = NULL,
  year = NULL,
  month = NULL,
  week = NULL,
  day = NULL,
  hour = NULL,
  minute = NULL,
  second = NULL,
  roll_month = "preday",
  roll_dst = c("pre", "post"),
  ...
)

Arguments

time

date-time object

periods

a named list of the form list(year = 1, month = 2, ...).

year, month, week, day, hour, minute, second

Units to be added to time. Units except for seconds are converted to integer values. Components are replicated according to vctrs semantics, i.e vectors must be either of length 1 or same length as time vector.

roll_month

controls how addition of months and years behaves when standard arithmetic rules exceed limits of the resulting date's month. Possible values are "preday", "boundary", "postday", "full" and "NA". See "Details" or ⁠[(timechange::time_add())⁠ for further details.

roll_dst

is a string vector of length one or two. When two values are supplied they specify how to roll date-times when they fall into "skipped" and "repeated" DST transitions respectively. A single value is replicated to the length of two. Possible values are:

* `pre` - Use the time before the transition boundary.
* `boundary` - Use the time exactly at the boundary transition.
* `post` - Use the time after the boundary transition.
* `xfirst` - crossed-first: First time which occurred when crossing the
   boundary. For addition with positive units pre interval is crossed first and
   post interval last. With negative units post interval is crossed first, pre -
   last. For subtraction the logic is reversed.
* `xlast` - crossed-last.
* `NA` - Produce NAs when the resulting time falls inside the problematic interval.

For example 'roll_dst = c("NA", "pre") indicates that for skiped intervals return NA and for repeated times return the earlier time.

When multiple units are supplied the meaning of "negative period" is determined by the largest unit. For example time_add(t, days = -1, hours = 2, roll_dst = "xfirst") would operate as if with negative period, thus crossing the boundary from the "post" to "pre" side and "xfirst" and hence resolving to "post" time. As this might result in confusing behavior. See examples.

"xfirst" and "xlast" make sense for addition and subtraction only. An error is raised if an attempt is made to use them with other functions.

...

deprecated

Details

Arithmetic operations with multiple period units (years, months etc) are applied in decreasing size order, from year to second. Thus time_add(x, month = 1, days = 3) first adds 1 month to x, then ads to the resulting date 3 days.

Generally period arithmetic is undefined due to the irregular nature of civil time and complexities with DST transitions. 'timechange' allows for a refined control of what happens when an addition of irregular periods (years, months, days) results in "unclear" date.

Let's start with an example. What happens when you add "1 month 3 days" to "2000-01-31 01:02:03"? 'timechange' operates by applying larger periods first. First months are added1 + 1 = February which results in non-existent time of ⁠2000-02-31 01:02:03⁠. Here the roll_month adjustment kicks in. After the adjustment, the remaining 3 days are added.

roll_month can be one of the following:

  • boundary - if rolling over a month boundary occurred due to setting units smaller than month, the date is adjusted to the beginning of the month (the boundary). For example, ⁠2000-01-31 01:02:03 + 1 month = 2000-03-01 00:00:00⁠.

  • preday - roll back to the last valid day of the previous month (pre-boundary day) preserving the H, M, S units. For example, ⁠2000-01-31 01:02:03 + 1 month = 2000-02-28 01:02:03⁠. This is the default.

  • postday - roll to the first day post-boundary preserving the H, M, S units. For example, ⁠2000-01-31 01:02:03 + 1 month = 2000-03-01 01:02:03⁠.

  • full - full rolling. No adjustment is done to the simple arithmetic operations (the gap is skipped as if it's not there). For example, ⁠2000-01-31 01:02:03 + 1 month + 3 days⁠ is equivalent to ⁠2000-01-01 01:02:03 + 1 month + 31 days + 3 days⁠ resulting in ⁠2000-03-05 01:02:03⁠.

  • NA - if end result was rolled over the month boundary due to addition of units smaller than month (day, hour, minute, second) produce NA.

  • NAym - if intermediate date resulting from first adding years and months ends in a non-existing date (e.g. Feb 31) produce NA. This is how period addition in lubridate works for historical reasons.

Examples


# Addition

## Month gap
x <- as.POSIXct("2000-01-31 01:02:03", tz = "America/Chicago")
time_add(x, month = 1, roll_month = "postday")
time_add(x, month = 1, roll_month = "preday")
time_add(x, month = 1, roll_month = "boundary")
time_add(x, month = 1, roll_month = "full")
time_add(x, month = 1, roll_month = "NA")
time_add(x, month = 1, day = 3,  roll_month = "postday")
time_add(x, month = 1, day = 3,  roll_month = "preday")
time_add(x, month = 1, day = 3,  roll_month = "boundary")
time_add(x, month = 1, day = 3,  roll_month = "full")
time_add(x, month = 1, day = 3,  roll_month = "NA")

## DST gap
x <- as.POSIXlt("2010-03-14 01:02:03", tz = "America/Chicago")
time_add(x, hour = 1, minute = 50, roll_dst = "pre")
time_add(x, hour = 1, minute = 50, roll_dst = "boundary")
time_add(x, hour = 1, minute = 50, roll_dst = "post")
##' time_add(x, hours = 1, minutes = 50, roll_dst = "NA")

## DST repeated time with cross-first and cross-last
(tt <- as.POSIXct(c("2014-11-02 00:15:00", "2014-11-02 02:15:00"), tz = "America/New_York"))
time_add(tt, hours = c(1, -1), roll_dst = "pre")
time_add(tt, hours = c(1, -1), roll_dst = "post")
time_add(tt, hours = c(1, -1), roll_dst = "xfirst")
time_add(tt, hours = c(1, -1), roll_dst = "xlast")

## DST skip with cross-first and cross-last
cst <- as.POSIXlt("2010-03-14 01:02:03", tz = "America/Chicago")
cdt <- as.POSIXlt("2010-03-14 03:02:03", tz = "America/Chicago")
time_add(cst, hour = 1, roll_dst = "xfirst")
time_add(cst, hour = 1, roll_dst = "xlast")
time_add(cdt, hour = -1, roll_dst = "xfirst")
time_add(cdt, hour = -1, roll_dst = "xlast")

# WARNING:
# In the following example the overall period is treated as a negative period
# because the largest unit (hour) is negative. Thus `xfirst` roll_dst results in the
# "post" time. To avoid such confusing behavior either avoid supplying multiple
# units with heterogeneous sign.
time_add(cst, hour = -1, minute = 170, roll_dst = "xfirst")

# SUBTRACTION

## Month gap
x <- as.POSIXct("2000-03-31 01:02:03", tz = "America/Chicago")
time_subtract(x, month = 1, roll_month = "postday")
time_subtract(x, month = 1, roll_month = "preday")
time_subtract(x, month = 1, roll_month = "boundary")
time_subtract(x, month = 1, roll_month = "full")
time_subtract(x, month = 1, roll_month = "NA")
time_subtract(x, month = 1, day = 0, roll_month = "postday")
time_subtract(x, month = 1, day = 3, roll_month = "postday")
time_subtract(x, month = 1, day = 0, roll_month = "preday")
time_subtract(x, month = 1, day = 3, roll_month = "preday")
time_subtract(x, month = 1, day = 3, roll_month = "boundary")
time_subtract(x, month = 1, day = 3, roll_month = "full")
time_subtract(x, month = 1, day = 3, roll_month = "NA")

## DST gap
y <- as.POSIXlt("2010-03-15 01:02:03", tz = "America/Chicago")
time_subtract(y, hour = 22, minute = 50, roll_dst = "pre")
time_subtract(y, hour = 22, minute = 50, roll_dst = "boundary")
time_subtract(y, hour = 22, minute = 50, roll_dst = "post")
time_subtract(y, hour = 22, minute = 50, roll_dst = "NA")



timechange documentation built on May 29, 2024, 8:56 a.m.