The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
This vignette is designed to introduce some of the common terminology used in almanac to get you up to speed on how to use the package. Along the way, we will see example usage of a number of the building blocks that will allow you to construct more complex recurrence objects.
A recurrence rule is a structured object that determines if a date should be counted as an event or not. At the most basic level, the job of a recurrence rule is to search through a pre-specified range of dates and flag any event dates in that range.
To build a recurrence rule, you start with a base recurrence frequency. There are 4 frequencies to choose from:
daily()
weekly()
monthly()
yearly()
Take the yearly()
frequency, for example. By default,
this will declare that 1 value per year is an event.
on_yearly <- yearly()
on_yearly
#> <rrule>
#> • frequency: yearly
#> • range: [1900-01-01, 2100-01-01]
The return value of yearly()
is a rrule object,
short for “recurrence rule”. This base object is all you need to start
doing something useful. For example, you can pass this rrule to
alma_search()
along with a from
and
to
date, and it will return all of the events in that date
range.
alma_search(from = "1990-01-01", to = "1995-12-31", on_yearly)
#> [1] "1990-01-01" "1991-01-01" "1992-01-01" "1993-01-01" "1994-01-01"
#> [6] "1995-01-01"
What if we want a yearly value, but we want it on January 5th every
year, rather than on the 1st? yearly()
has an important
argument called since
that controls two things: the start
date of the recurrence rule, and information such as the month, or the
day of the month to use if no other conditions have been specified to
override those.
The default of since
is set to 1900-01-01
,
but this is arbitrary (see almanac_since()
). It is because
of this default that in the above example with
alma_search()
we get values on January 1st. Let’s change
that.
on_yearly_jan_5 <- yearly(since = "1990-01-05")
alma_search("1990-01-01", "1995-12-31", on_yearly_jan_5)
#> [1] "1990-01-05" "1991-01-05" "1992-01-05" "1993-01-05" "1994-01-05"
#> [6] "1995-01-05"
Now that the since
date has been set to 1990, if we try
and find yearly dates before 1990, they will not be included.
# Same result as above, because the 1988 and 1989 dates are not included.
alma_search("1988-01-01", "1995-12-31", on_yearly_jan_5)
#> [1] "1990-01-05" "1991-01-05" "1992-01-05" "1993-01-05" "1994-01-05"
#> [6] "1995-01-05"
There is also an until
argument to yearly()
that controls the upper bound of the range to look in. This is
arbitrarily set to 2100-01-01
, but can be expanded or
contracted as required (see almanac_until()
).
I mentioned earlier that the job of a recurrence rule is to flag dates in a pre-specified range to be events or not. The dates that are flagged as events are known as the event set.
In the previous example, we used alma_search()
to
extract a subset of dates from the event set that were between
from
and to
. You can get the entire event set
with alma_events()
. Notice that this is bounded by our
custom since
date, and the default until
upper
bound. Otherwise we’d have an infinite event set, which is nice in
theory but bad in practice.
alma_events(on_yearly_jan_5)
#> [1] "1990-01-05" "1991-01-05" "1992-01-05" "1993-01-05" "1994-01-05"
#> [6] "1995-01-05" "1996-01-05" "1997-01-05" "1998-01-05" "1999-01-05"
#> [11] "2000-01-05" "2001-01-05" "2002-01-05" "2003-01-05" "2004-01-05"
#> [16] "2005-01-05" "2006-01-05" "2007-01-05" "2008-01-05" "2009-01-05"
#> [21] "2010-01-05" "2011-01-05" "2012-01-05" "2013-01-05" "2014-01-05"
#> [26] "2015-01-05" "2016-01-05" "2017-01-05" "2018-01-05" "2019-01-05"
#> [31] "2020-01-05" "2021-01-05" "2022-01-05" "2023-01-05" "2024-01-05"
#> [36] "2025-01-05" "2026-01-05" "2027-01-05" "2028-01-05" "2029-01-05"
#> [41] "2030-01-05" "2031-01-05" "2032-01-05" "2033-01-05" "2034-01-05"
#> [46] "2035-01-05" "2036-01-05" "2037-01-05" "2038-01-05" "2039-01-05"
#> [51] "2040-01-05" "2041-01-05" "2042-01-05" "2043-01-05" "2044-01-05"
#> [56] "2045-01-05" "2046-01-05" "2047-01-05" "2048-01-05" "2049-01-05"
#> [61] "2050-01-05" "2051-01-05" "2052-01-05" "2053-01-05" "2054-01-05"
#> [66] "2055-01-05" "2056-01-05" "2057-01-05" "2058-01-05" "2059-01-05"
#> [71] "2060-01-05" "2061-01-05" "2062-01-05" "2063-01-05" "2064-01-05"
#> [76] "2065-01-05" "2066-01-05" "2067-01-05" "2068-01-05" "2069-01-05"
#> [81] "2070-01-05" "2071-01-05" "2072-01-05" "2073-01-05" "2074-01-05"
#> [86] "2075-01-05" "2076-01-05" "2077-01-05" "2078-01-05" "2079-01-05"
#> [91] "2080-01-05" "2081-01-05" "2082-01-05" "2083-01-05" "2084-01-05"
#> [96] "2085-01-05" "2086-01-05" "2087-01-05" "2088-01-05" "2089-01-05"
#> [101] "2090-01-05" "2091-01-05" "2092-01-05" "2093-01-05" "2094-01-05"
#> [106] "2095-01-05" "2096-01-05" "2097-01-05" "2098-01-05" "2099-01-05"
You can also check if an existing date is included in a recurrence
rule’s event set with alma_in()
.
# Uses the 10th of the month, pulled from `since`
on_monthly <- monthly(since = "1990-01-10")
x <- as.Date("2000-01-08") + 0:5
x
#> [1] "2000-01-08" "2000-01-09" "2000-01-10" "2000-01-11" "2000-01-12"
#> [6] "2000-01-13"
x_in_set <- alma_in(x, on_monthly)
x_in_set
#> [1] FALSE FALSE TRUE FALSE FALSE FALSE
x[x_in_set]
#> [1] "2000-01-10"
almanac attempts to be smart by caching the event set of a recurrence rule the first time that it is queried. This means that the first usage of a recurrence rule is generally slower than repeated uses.
since <- "1990-01-01"
on_weekly <- weekly(since = since)
# The first time is "slow"
system.time(alma_search(since, "2000-01-01", on_weekly))
#> user system elapsed
#> 0.058 0.002 0.058
# Repeated access is fast
system.time(alma_search(since, "2000-01-01", on_weekly))
#> user system elapsed
#> 0 0 0
# The entire event set is cached, so even if you change the arguments,
# the operation is still fast.
system.time(alma_search(since, "1990-05-01", on_weekly))
#> user system elapsed
#> 0 0 0
So far we have worked with the base recurrence rules. Things get
much more interesting when we start adding extra conditions to
these rules. Conditions are ways to limit or expand a
given recurrence rule to hone in on recurring dates that you are
particularly interested in. All condition functions in almanac start
with recur_*()
. For example, let’s take a monthly rule,
which defaults to give us 1 day per month, and expand it to
give us every 4th and 16th day of the month.
on_4th_and_16th <- monthly(since = "2000-01-01") %>%
recur_on_day_of_month(c(4, 16))
alma_search("2000-01-01", "2000-06-01", on_4th_and_16th)
#> [1] "2000-01-04" "2000-01-16" "2000-02-04" "2000-02-16" "2000-03-04"
#> [6] "2000-03-16" "2000-04-04" "2000-04-16" "2000-05-04" "2000-05-16"
An important thing to note here is that even though our
since
date is on the first of the month, we are
“overriding” that with the recurrence condition, so that information is
not used.
Recurrence rules can continually be added to further refine your rule. When you add a condition to a rule, you get another rule back. Let’s try creating a rule for the recurring holiday, Labor Day. This recurs on the first Monday of September, yearly. To do this, we will:
Use a yearly()
base since this happens 1 time per
year.
Use recur_on_month_of_year()
to hone in on
September.
Use recur_on_day_of_week()
to hone in on the first
Monday of the month.
on_labor_day <- yearly() %>%
recur_on_month_of_year("Sep") %>%
recur_on_day_of_week("Monday", nth = 1)
alma_search("2000-01-01", "2005-01-01", on_labor_day)
#> [1] "2000-09-04" "2001-09-03" "2002-09-02" "2003-09-01" "2004-09-06"
The nth
argument of recur_on_day_of_week()
is especially useful for selecting from the end of the month. If we
wanted the last Monday in September instead, we could do:
Recurrence rules are powerful tools on their own, but they aren’t enough to solve every task. Say you want to construct a rule that includes both Christmas and Labor Day as events. It would be impossible to construct this kind of event set using a single rule, but if you could bundle multiple rules together, one for Christmas and one for Labor Day, then it would be possible.
An rset is a bundle of recurrence schedules. A recurrence schedule, or rschedule, is an overarching term for both rrules and rsets. There are three types of rsets in almanac. Each create their event set by performing some kind of set operation on the event sets of the underlying rschedules that you added to the set.
runion()
takes the union.
rintersect()
takes the intersection.
rsetdiff()
takes the set difference.
The most useful rset is runion, as this allows you construct an event set that, for example, falls on multiple holidays and all weekends. The following creates an runion from rrules based on Christmas and Labor Day.
on_christmas <- yearly() %>%
recur_on_month_of_year("Dec") %>%
recur_on_day_of_month(25)
christmas_or_labor_day <- runion(
on_christmas,
on_labor_day
)
alma_search("2000-01-01", "2002-01-01", christmas_or_labor_day)
#> [1] "2000-09-04" "2000-12-25" "2001-09-03" "2001-12-25"
christmas_or_labor_day_except_2000_labor_day <- rsetdiff(
christmas_or_labor_day,
rcustom("2000-09-04")
)
alma_search("2000-01-01", "2002-01-01", christmas_or_labor_day_except_2000_labor_day)
#> [1] "2000-12-25" "2001-09-03" "2001-12-25"
A recurrence set is a critical data structure in almanac. It serves as a general container to dump all of your company’s holiday and weekend recurrence rules.
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.