new function earlyCloseNYSE gives datetimes of early closings of NYSE, including scheduled and unscheduled ones. The information is incomlete. This answers issue [#6757], see also the related issue [#1356] mentioned below in the news for a previous version.
all functions returning public and eclessiastical holidays get new arguments value, na_drop, and ..., controlling the type of the return value and the handling of missing values. Argument ... is for things like ‘format’ when applicable. Not all honour the new arguments though, see below.
all functions returning public and eclessiastical holidays now honour the new argument value.
some holiday functions were amended or rewritten to honour the new argument na_drop. These include the England holidays (‘GBxxx’), some Japan holidays (JPxxx), and a number of other functions (e.g., `CAFamilyDay). Contributions and/or information needed to have them do so is wellcome (e.g. dates when a holiday was introduced and/or moved).
midnightStandard2() (and hence timeDate()) was throwing a puzzling error when the input character vector contained one or more NAs mixed with non-NA’s.
whichFormat() was throwing error when the first element of the input vector was NA. This could be considered a feature but that’s inconsistent since NAs at other positions were not causing trouble. Now fixed.
the ‘character’ method for timeDate() was passing only the first element of the input vector to whichFormat(). This was preventing it from inferring the format when argument ‘format’ was missing.
new function specialHolidayGB gives the special UK one-off holidays for the requested years.
MilleniumDay is now deprecated, use specialHolidayGB(1999).
fixed omissions in holidayLONDON and refactored it completely. It became easier to maintain and orders of magnitude faster.
fixed a bug in an internal function which was throwing errors from holidayLONDON for some ranges including years before 1916.
GBEarlyMayBankHoliday, GBSpringBankHoliday and GBSummerBankHoliday now give dates according to historical rules, not simply according to current ones. See the remarks above about the new arguments of holiday functions.
major overhaul and update of the Japan’s holidays. Previously fixed dates were returned for each holiday. Now moving holidays are calculated (e.g., second month of January) and some exceptions handled (like holidays moved due to the Olympics in 2020 and 2021). Historical changes are (partially) taken into account. Further corrections and historical amendments are welcome.
Thanks to Sylvie Lachaud for reporting the issues with Japan holiday functions, as well as providing current correct definitions and extensive list of links.
new function JPMountainDay gives the relatively recently introduced Japan holiday ‘Mountain Day’.
JPKeirouNoHi replaces JPKeirouNOhi. The latter is an aberration from all other xxxNoHi Japanese holiday functions and a source of difficult to spot errors. The old one is now deprecated and will be removed in the future.
modified the ‘timeDate’ method for summary to return an object from class “timeDate_summary” (rather than print directly) and created a print method for the latter.
deprecated the timeDate method for cut, see deprecation note for v4021.105 below.
stopped exporting all but one (.endpoints) functions starting with a ‘.’. They all seem for internal purposes and/or have similarly named functions with normal names.
removed .whichFormat and .midnightStandard, not officially deprecated but had been renamed to whichFormat() and midnightStandard(), respectively, a long time ago.
removed several internal functions that are no longer used anywhere in the package.
removed file ‘namespace-tags.R’ as it had not been updated recently, see revision r83578 or earlier if you need it.
added the 2023 UK Bank holiday for the coronation of King Charles III.
axis.timeDate was not handling properly the case when x was missing, leading to errors from R-devel check (2023-01-07 r83578). Fix suggested by Uwe Lieges.
refactored file NAMESPACE to facilitate maintenance (that revealed the two omissions listed below).
CAFamilyDay (Canada Family Day) was missing from the list returned by listHolidays(). It was missing only from that list, holidayTSX() was including it when applicable.
JPVernalEquinox was missing from the list returned by listHolidays().
the financial centers are now updated to reflect changes in time zones in recent years. The list returned by listFinCenter() is synchronised with current time zone names. Previous names supported by timeDate are available as aliases.
import selectively from ‘stats’ and ‘utils’.
London financial centre holidays - fixed and/or included non-standard holidays (e.g., Early May Bank holiday was moved in 2020 to VE day; Spring Bank holiday was moved in Queen’s Jubilee years; state funeral of the Queen). Millenium day now is included in the result of holidayLONDON(1999). The London holidays should now be complete up to the time of writing this (1 Oct 2022).
renamed GBMayDay to GBEarlyMayBankHoliday and GBBankHoliday to GBSpringBankHoliday. The old names are somewhat ambiguous and strongly discouraged but still available. listHolidays() gives the new names.
the generic timeDate() gets argument ‘…’ to allow methods for it to have additional arguments (e.g., for DST gaps).
the ‘character’ method for timeDate() gets a new argument dst_gap to control what to do with non-existent DST times at the requested FinCenter with options to add/subtract (“+”, “-”) the DST shift or set them to NA.
timeDate() was not handling correctly some times just after the switch to/from DST. This was visible mostly for time zones away from GMT and GMT+1.
In timeSequence(), if any of the generated times would fall in DST gaps, they are moved by “+1 hour”, corresponding to dst_gap = "+" in timeDate. This is consistent with seq for other time objects. Currently there is no option to change this behaviour of timeSequence.
Previously timeSequence was moving DST gaps down by 1 hour (for by = ‘DSTday’ and similar). This was not consistent similar time functions in R and was actually due to a bug (or unfinished DST handling) in timeDate, see remarks for timeDate above.
timeSequence() now throws error if argument from is in a DST gap. It seems desirable to have a default action for this case. Rolling the faulty time by an hour in the case of ‘DSTday’ may be suitable in most cases but for other values of by it might be totally wrong.
updated the DST rules.
internally, refactored the way the DST rules are generated (not visible to users).
rulesFinCenter() now looks for a financial center starting from the namespace of timeDate. Previously it was starting from the environment of the caller which could result in using an unrelated object or, if timeDate was loaded but not attached, not finding it.
whichFormat() to accommodate a change in R-devel after which as.character(Sys.time()) contains fractional seconds. (format(Sys.time()) doesn’t; before this change in R-devel both were dropping the fractional seconds). (fixed by Martin Maechler, see timeDate rev 6286)the list returned by holidaysNYSE() was missing the special closing days of the New York stock exchange (NYSE). Now it should be complete (though there may be ommissions after 2011). This fixes issue #1356 reported by Corwin Joy. Thanks to him and Ian E for the insigthful discussion and useful links.
See also below. Contributions for the other exchanges and corrections are welcome.
holidaysNYSE() gets a new argument, type, to select what type of the exchange’s closing days to return. The default is to return all days in the requested years when NYSE was closed for whatever reason. Use type = "standard" and type = special to get the standard holidays and the special closings, respectively.
Returning any closing day by default might be considered a breaking change. However, not returning all closing days was perceived as erroneous by users (eg issue #1356). In fact, the package itself calculates business days by dropping weekends and days returned by holidayXXXX.
Note that holiday() returns the actual dates of the public holidays, while the corresponding days returned by holidayXXXX are the resulting non-weekend closing days, if any.
holidayTSX() now correctly calculates Christmas and Boxing day closures when Christmas is on Monday. Fixes part (2) of issue #1288 reported by Stefan Wilhelm (part (1) was fixed in a previous release). The fix is really a patch for the specific issue, maybe the same should be done when Christmas is on Sunday, for example. Information/contribution on Canadian holidays is welcome.
now holiday() accepts also a function or a list of functions for argument ‘Holiday’.
timeNthNdayInMonth could return a value in the following month. Now fixed. This is bug #1463 reported with a fix by Manny C. Note that the bug was not present for dates in the first day of a month.
timeLastNdayInMonth could return a value in the following month, e.g. ‘1996-06-04’ for the last Tuesday in May 1996. Now fixed. The check of this function was prompted by the bug report for #1463 (see above) for timeNthNdayInMonth but the error was different.
the data.frame methods for kurtosis() and skewness() now set attribute method as for the other methods and as documented.
removed .holidayList() which had been replaced by listHolidays() a long time ago and was not exported in recent versions of timeDate.
updated documentation files.
timeDate method for cut has been discouraged in the sources for a long time with a recommendation to use window instead (just replace cut(x, from = xx , to = yy) with window(x, start = xx, end = yy). The cut method will be deprecated in the next release and later removed or replaced by a method that is consistent with the methods for cut in base R.new maintainer: Georgi N. Boshnakov.
updated DESCRIPTION with links and moved all Depends: to Imports:.
removed the line LazyData: yes from DESCRIPTION to fix the NOTE on CRAN.
added the new US holiday, Juneteenth National Independence Day. Fixes #6755 by Ian E (ene100).
holidayTSX() now includes the Labour Day. Fixes part (1) of issue #1288 reported by Stefan Wilhelm.
created a first version of _pkgdown.yml for more organised view of the large number of objects in the package. Unpack the tarball and run pkgdown::build_site() to build the site locally. Don’t know if this could work directly off the R-forge repository.
See file ChangeLog for changes before 4021.104.