almanac

library(almanac)

almanac provides tools for working with recurrence rules, the fundamental building blocks used to identify calendar “events”, such as weekends or holidays.

Additionally, it provides a full suite of tools for working with holidays and calendars. It includes a number of built in holidays, such as hol_christmas(), but you can also add your own custom holidays through rholiday(). Once you have a set of holidays specific to your business, you can aggregate them into a calendar with rcalendar(), which has specialized tooling like cal_events() to generate the holiday dates for a particular year.

Installation

Install the released version of almanac from CRAN with:

install.packages("almanac")

Install the development version from GitHub with:

# install.packages("pak")
pak::pak("DavisVaughan/almanac")

Recurrence rules

Constructing recurrence rules looks like this:

# Thanksgiving = "The fourth Thursday in November"
on_thanksgiving <- yearly() %>% 
  recur_on_month_of_year("November") %>%
  recur_on_day_of_week("Thursday", nth = 4)

on_thanksgiving
#> <rrule>
#> • frequency: yearly
#> • range: [1900-01-01, 2100-01-01]
#> • month of year: Nov
#> • day of week: Thu[4]

This is the underlying recurrence rule for the Thanksgiving holiday represented by hol_us_thanksgiving().

After constructing a recurrence rule, it can be used to generate dates that are in the “event set”. For example, you can search for all Thanksgivings between 2000-2006.

alma_events(on_thanksgiving, year = 2000:2006)
#> [1] "2000-11-23" "2001-11-22" "2002-11-28" "2003-11-27" "2004-11-25"
#> [6] "2005-11-24" "2006-11-23"

Determine if a particular date is a part of the event set with alma_in().

# Is this a Thanksgiving?
alma_in(c("2000-01-01", "2000-11-23"), on_thanksgiving)
#> [1] FALSE  TRUE

You can also shift an existing sequence of dates, “stepping over” dates that are part of the event set.

wednesday_before_thanksgiving <- as.Date("2000-11-22")

# Thanksgiving was on 2000-11-23.
# This steps over Thanksgiving to 2000-11-24.
# Then steps 1 more day to 2000-11-25.
alma_step(wednesday_before_thanksgiving, n = 2, on_thanksgiving)
#> [1] "2000-11-25"

There is an additional “stepper” object you can create for more intuitive stepping. Combine it with %s+% to perform the same step done by alma_step(). Create a stepper function with stepper(), and then use it by supplying the number of days to step.

step_over_thanksgiving <- stepper(on_thanksgiving)
wednesday_before_thanksgiving %s+% step_over_thanksgiving(2)
#> [1] "2000-11-25"

Holidays and calendars

The above example just scratches the surface of what almanac can do. Practically speaking, you’ll probably have multiple holidays that you’d like to combine into one big calendar. almanac provides a full API for working with holidays and calendars.

This example creates a calendar containing Christmas, Thanksgiving, and New Year’s Day:

cal <- rcalendar(
  hol_christmas(),
  hol_us_thanksgiving(),
  hol_new_years_day()
)

cal
#> <rcalendar[3]>
#> • Christmas
#> • US Thanksgiving
#> • New Year's Day

We can ask for the next upcoming holiday with cal_next():

x <- as.Date(c("2019-12-05", "2020-02-05"))
cal_next(x, cal)
#>              name       date
#> 1       Christmas 2019-12-25
#> 2 US Thanksgiving 2020-11-26

Or for holidays that belong to a particular year with cal_events():

events <- cal_events(cal, year = 2028)
events$weekday <- lubridate::wday(events$date, label = TRUE)
events
#>              name       date weekday
#> 1  New Year's Day 2028-01-01     Sat
#> 2 US Thanksgiving 2028-11-23     Thu
#> 3       Christmas 2028-12-25     Mon

Note that New Year’s Day occurred on Saturday. If your business observes New Year’s Day on the nearest weekday, you can adjust the holiday to respect that observance rule before adding it into the calendar:

on_weekends <- weekly() %>%
  recur_on_weekends()

cal <- rcalendar(
  hol_christmas(),
  hol_us_thanksgiving(),
  hol_observe(
    hol_new_years_day(), 
    adjust_on = on_weekends, 
    adjustment = adj_nearest
  )
)

# Now it returns the previous Friday for the observed New Year's date.
# Note that this fell in 2027, but was included in the 2028 set of dates
# since most people would consider that part of the 2028 holiday calendar.
events <- cal_events(cal, year = 2028)
events$weekday <- lubridate::wday(events$date, label = TRUE)
events
#>              name       date weekday
#> 1  New Year's Day 2027-12-31     Fri
#> 2 US Thanksgiving 2028-11-23     Thu
#> 3       Christmas 2028-12-25     Mon

We can union our calendar with the on_weekends rule to get a recurrence set that represents days when our business is closed. Then we can create a stepper out of that so we can step forwards by “a business day.”

business_day <- stepper(runion(cal, on_weekends))

For example, Christmas was on a Monday in 2006. If you wanted to step 1 business day forward from the Friday before Christmas, you’d probably like it to step over the weekend and the Christmas Monday to finally land on Tuesday:

# Christmas was on a Monday in 2006.
# This is the Friday before Christmas
friday <- as.Date("2006-12-22")

# Step forward 1 business day, going over the weekend and Christmas
friday %s+% business_day(1)
#> [1] "2006-12-26"

Learning more

View the vignettes on the website to learn more about how to use almanac.

vignette("almanac")
vignette("holidays-calendars")
vignette("adjust-and-shift")
vignette("quarterly")
vignette("icalendar")

Acknowledgements

almanac has developed as a composite of ideas from multiple different libraries.

First off, it directly embeds the amazing JavaScript library rrule for the core event set calculations. To do this, it uses the equally awesome R package, V8, from Jeroen Ooms.

The date shifting / adjusting functions are modeled after similar functions in QuantLib.

The fast binary search based implementations of alma_next() and alma_step() are inspired by Pandas and the implementation of Numpy’s busday_offset().

The author of gs, James Laird-Smith, has been a great collaborator as we have bounced ideas off of each other. gs attempts to solve a similar problem, but with a slightly different implementation.