c.POSIXct and c.Date can deal with heterogeneous object types (e.g c(date, datetime) works as expected)#994 Subtracting two duration or two period objects no longer results in an ambiguous dispatch note.
c.Date and c.POSIXct correctly deal with empty vectors.
as_datetime(date, tz=XYZ) returns the date-time object with HMS set to 00:00:00 in the corresponding tz
type argument to quarter() for more control over the returned classfast_strptime() and parse_date_time2() now accept multiple formats and apply them in turn%z formatrollforward() functionym() and my()as.period() on intervals now returns valid Periods with double fields (not integers)vctrs supportquarter(..., with_year = TRUE)with_tz() for POSIXlt objectsmonth()dyears(1) == years(1) is now TRUE.dmonths() to complement other duration constructors.duration() constructor now accepts months and years arguments.format_ISO8601() methods.as_date() now ignores the tz argumentforce_tz(), with_tz(), tz<- convert dates to date-timesNA_Date_ and NA_POSIXct_ which parallel built-in primitive constants.Date() and POSIXct() which parallel built-in primitive constructors.as_datetime() always returns a POSIXct()generics dependency package.-.%within% throws more meaningful messages when applied on unsupported classesfiscal_starts.leap_year() works with objects supported by year().duration()/period()/make_difftime() work with repeated unitsc.Period concatenation doesn’t fail with empty components.exact = TRUE argument in parse_date_time2, which was so far ignored.%within% now accepts a list of intervals, in which case an instant is checked if it occurs within any of the supplied intervals.%% on intervals relies on %m+ arithmetic and doesn’t produce NAs when intermediate computations result in non-existent dates.tz() always returns “UTC” when tzone attribute cannot be inferred.interval constructor accepts start and end character vectors in ISO 8601 formatformat argument to as_date and as_datetime (regression in v1.7.1)ymd_hms strings by as_date.ymd strings by as_datetime.trunc_multi_unit so that it overwrites the vector argument x versus making a new vector y.force_tzs for “enforcement” of heterogeneous time zones.local_time for the retrieval of local day time in different time zones.cutoff_2000 for parsing functions to indicate 20th century cutoff for y format.week_start parameter in wday and wday<- to set week start.locale in wday. Labels of the returned factors (when label=TRUE) now respect current locale.quarter gained a new argument fiscal_start to address the issue of different fiscal conventions.epiweek and epiyear.locale in month. Labels of the returned factors (when label=TRUE) now respect current locale.week_start to floor_date, ceiling_date and round_date.season is now supported.as_date and as_datetime now understand character vectors.parse_date_time - train=TRUE and drop=FALSE which allow more refined control of the format guessing. Formats are no longer dropped in the process by default, process which resulted in surprising behavior on several occasions (#516,#308,#307).wday and month are now in current locale. The abbreviated labels in English locales have been changed to standard abbreviations (Tues -> Tue, Thurs -> Thu etc.).with_tz on invalid timezone.B and b formats no longer match numeric months. This corresponds to the original intent, and was always documented as such.days, dhours, round_date work when the methods package is not loaded.wday work on character inputs as it is the case with all other day accessors.y format when it’s the last in format order (as in mdy).m and b formats now works correctly.date() when called with missing argument.NA properly.make_datetime respects tz argument and is much faster now.day<-, minute<-, etc. from producing an error when length(x) is 0 (issue #517)parse_date_time now throw warnings only for actual parsing errors (input with all NAs are silent)semester to extract semesters form date-time objects.period and duration constructors now accept string as first argument. Same parsing rules apply to the ‘unit’ parameter in rounding functions.hms, hm, ms gained new argument roll=TRUE which rolls minutes and seconds bigger than 59 towards higher units.round_date, floor_date and ceiling_date now support rounding to multiple of units.yq for parsing most common version of quarter strings.q for parsing quarters in all lubridate parsing functions.as_datetime to coerce to POSIXct object. A counterpart of as_date.make_date to produce Date objects. A counterpart of make_datetime.ceiling_date for Date objects was changed to what most of the users expect. Rounding up by months now produces first day of the next months even for first day of the month.round_date, ceiling_date, and floor_date now accept “quarter”, “bimonth”, and “halfyear” as unit options.parse_date_time and related functions are now deprecated.pretty.* family of functions were renamed and are no longer exported. If you need to use them, use lubridate:::pretty_* versions.change_on_boundary argument in ceiling_date does not allow for global option anymore.as.duration, as.numeric don’t show “only estimate” messages on conversion from periods. The occasional approximate conversion is documented and deemed common knowledge.as.numeric with unit="month" now works on duration objects.Date objects now return POSIXct instead of POSIXlt.mdy or myd beginning with "January" or "Jan" now parsing correctlyhere and olson_time_zones were deprecated in favor of new and base OlsonNames respectively.ceiling_date when unit = "week"period and duration constructors.date_decimal and decimal_date.as.numeric correctly converts periods to (aproximate) numeric time lengths.ceiling_date gains new argument change_on_boundary to allow ceiling on boundary of date-time objects.parse_date_time2 and fast_strptime gain new lt argument to control type of output.date and date<- additions to the year, month etc family of accessors.make_datetime (dropin replacement of ISOdatetime).force_tz and with_tz can handle data.frames component-wise.as_date replacement of as.Date with more intuitive behavior with non-UTC timezones.hms parsers now handle negative components.ymd family of functions return Date object when tz argument is NULL (new default) and POSIXct otherwise.base:strptime fast_strptime now returns POSIXlt object. That is, its lt argument defaults to TRUE.interval constructor treats timezones correctly and works with UTC whenever meaningful.as.period correctly computes months with intervals spanning multiple years.time_length and add_with_rollback now work correctly with missing intervals.fast_strptime and parse_date_time2 correctly treat non-UTC time zones.floor_date and round_date are not preserving tz component for larger than day unitstime_length method.isoyear function to line up with isoweek.exact = TRUE option to parse_date_time for faster and much more flexible specification of formats.simple argument to fit_to_timeline and update methods mostly intended for internal use.unique method for interval class.preserve_hms and roll_to_first in rollback function.quarter option in floor_date and friends.as.list.Interval S3 method.qday (quarter day).new_interval, new_period, new_duration, new_difftime were deprecated in favour of the more powerful interval, period, duration and make_difftime functions.eseconds, eminutes etc. were deprecated in favour of dsecons, dminutes etc.test/testthat.isodate was replaced with a much faster parse_date_time2(paste(...)) alternativeLubridate’s trunc, ceiling and floor functions have been optimised and now are relying on R’s trunc.POSIXct whenever possible.fst objects.enc2utf8.int_standardizeceiling_date skip day light gap.preproc_wday=T.parse_date_time.fit_to_timeline function.isoweek computationseconds_to_period.$<- assignment for periods are now properly recycled.round_date.time-zones.R in coercion.R.standardise_lt_names.as.period.period if unit != NULL.as.period.interval without recurring to modulo arithmetic.hms, hm and ms for new printing style.guess_formats now matches flex regexp first.dmilliseconds now correctly returns a Duration object.New low level C parser for numeric formats and two new front-end R functions parse_date_time2 and fast_strptime. The achieved speed up is 50-100x as compared to standard as.POSIXct and strptime functions.
The user level parser functions of ymd_hms family drop to these C routines whenever plain numeric formats are detected.
ymd("2013-01-31") + months(1) will return NA instead of 2013-03-04 as in v1.2.0. ymd("2012-02-29") + years(1) will also return an NA. This rule change helps ensure that date + timespan - timespan = date (or NA). If you’d prefer that such arithmetic just returns the last day of the resulting month, see %m+% and %m-%.force_tz, with_tz, floor_date, ceiling_date, second<-, minute<-, hour<-, day<-, month<-, year<-, and other functions that rely on update (such as math with Periods).browseVignettes("lubridate") to view it.second, minute, hour, day, month, year and the settors second<-, minute<-, hour<-, day<-, month<-, year<- now work on Period class objectsoptions(lubridate.verbose = TRUE) to turn parsing messages on. Run options(lubridate.verbose = FALSE) to turn estimation and coercion messages off.%m+% and %m-% methods for Interval and Duration class objects that throw useful errors.olson_time_zones retreives a character vector is Olson-style time zone names to use in lubridateas.interval, int_overlaps, %within% and the interval methods of c, intersect, union, setdiff, and summary.as.interval, as.period and as.duration now handlevectors of NA’s without returning errors.round_date, ceiling_date and update now correctly handle input of length zerodecimal_date no longer returns NaN for first second of the yearlubridate 1.2.0 is significantly faster than lubridate 1.1.0. This is largely thanks to a parser rewrite submitted by Vitalie Spinu. Thank you, Vitalie. Some metrics:
with_tz speed up - 15x fasterforce_tz speed up - 3x fasterDevelopment for 1.2.0 has also focused on improving the way we work with months. rollback rolls dates back to the last day of the previous month. provides more options for working with months. days_in_month finds the number of days in a date’s month. And, %m+% and %m-% provide a new way to ### handle unequal month lengths while doing arithmetic. See NEW FEATURES for more details
date parsing can now parse multiple date formats within the same vector of date-times. Parsing can also recognize a greater variety of date-time formats as well as incomplete (truncated) date-times. Contributed by Vitalie Spinu. Thank you, Vitalie.
1.2.0 introduces a new display format for periods. The display is more math and international friendly.
1.2.0 transforms negative intervals into periods much more gracefully (e.g, - 3 days instead of -1 years, 11 months, and 27 days)
S3 update methods are now exported
stamp allows users to print dates in whatever form they like. Contributed by Vitalie Spinu. Thank you, Vitalie.
periods now handle fractional seconds. Contributed by Vitalie Spinu. Thank you, Vitalie.
date parsing can now parse multiple date formats within the same vector of date-times. Parsing can also recognize a greater variety of date-time formats as well as incomplete (truncated) date-times. Contributed by Vitalie Spinu. Thank you, Vitalie.
sort, order, rank and xtfrm now work with periods
as.period.Interval accepts a unit argument. as.period will convert intervals into periods no larger than the supplied unit.
days_in_month takes a date, returns the number of days in the date’s month. Contributed by Richard Cotton. Thank you, Richard.
%m+% and %m-% perform addition and subtraction with months (and years) without rollover at the end of a month. These can be used in place of + and -. These can’t be used with periods smaller than a month, which should be handled separately. An example of the new behavior:
ymd(“2010-01-31”) %m+% months(1) # “2010-02-28 UTC”
ymd(“2010-01-31”) + months(1) # “2010-03-03 UTC”
ymd(“2010-03-31”) %m-% months(1) # “2010-02-28 UTC”
ymd(“2010-01-31”) - months(1) # “2010-03-03 UTC”
rollback rolls a date back to the last day of the previous month.
quarter returns the fiscal quarter that a date occurs in. Like quartes in base R, but returns a numeric instead of a character string.
date parsers now handle NAs
periods now handle NAs
[<- now correctly updates all elements of a period inside a vector, list, or data.frame
period() now works with unit = “weeks”
ceiling_date no longer rounds up if a date is already at a ceiling
the redundant (i.e, repeated) hour of fall daylight savings time now displays with the correct time zone
update.POSIXct and update.POSIXlt handle vectors that sum to zero in the days argument
the format method for periods, intervals and duration now accurately displays objects of length 0.
lubridate no longer overwrites base R methods for +, - , *, /, %%, and %/%. To recreate the previous experience of subtracting two date times to create an interval, we’ve added the interval creation function %–%.
lubridate has moved to an S4 object system. Timespans, Intervals, Durations, and Periods have each been redefined as an S4 class with its own methods.
arithmetic operations will no longer perform implicit class changes between timespans. Users must explicitly state how and when they wish class changes to occur with as.period(), as.duration(), and as.interval(). This makes code written with lubridate more robust, as such implicit changes often did not produce consistent behavior across a variety of operations. It also allows lubridate to be less chatty with fewer console messages. lubridate does not need to explain what it is doing, because it no longer attempts to do things whose outcome would not be clear. On the other hand, arithmetic between multiple time classes will produce informative error messages.
the internal structure of lubridate R code has been reorganized at https://github.com/tidyverse/lubridate to make lubridate more development friendly.
intervals are now more useful and lubridate has more ways to manipulate them. Intervals can be created with %–%; modified with int_shift(), int_flip(), and int_standardize(); manipulated with intersect(), union(), and setdiff(); and used in logical tests with int_aligns(), int_overlaps(), and %within%. lubridate will no longer perform arithmetic between two intervals because the correct results of such operations is no more obvious than the correct result of adding two dates. Instead users are encouraged to use the new set operations or to directly modify intervals with int_start() and int_end(), which can also be used as settors. lubridate now supports negative intervals as well as positive intervals. Intervals also now display with a time zone.
Modulo methods for timespans have been changed to return a timespan. this allows modulo methods to be used with integer division in an intuitive manner, e.g. a = a %/% b * b + a %% b
Users can still acheive a numerical result by using as.numeric() on input before performing modulo.
Periods, durations, and intervals can now all be put into a data frame.
Periods, durations, and intervals can be intuitively subset with $ and []. These operations also can be used as settors with <-.
The parsing functions and the as.period method for intervals are now slightly faster.
month<- and wday<- settors accept names as well as numbers
parsing functions now have a quiet argument to parse without messages and a tz argument to directly parse times into the desired time zone.
logical comparison methods now work for period objects.
use test_package to avoid incompatibility with current version of testthat
other minor fixes to pass R CMD check
removed bug in parsing dates with “T” in them
modified as.period.interval() to display periods in positive units
ymd_hms(), hms(), and ms() functions can now parse dates that include decimal values in the seconds element.
milliseconds(), microseconds(), nanoseconds(), and picoseconds() create period objects of the specified lengths. dmilliseconds(), dmicroseconds(), dnanoseconds(), and dpicoseconds() make duration objects of the specified lengths.
lubridate no longer overwrites months(), start(), and end() from base R. Start and end have been replaced with int_start() and int_end().
lubridate imports plyr and stringr packages, instead of depending on them.
made division, modulo, and integer division operations compatible with difftimes
created c() methods for periods and durations
division between timespans: each timespan class (durations, periods, intervals) can be divided by other timespans. For example, how many weeks are there between Halloween and Christmas?: (christmas - halloween) / weeks(1)
modulo operations between timespans
duration objects now have their own class and display format separate from difftimes
interval objects now use an improved data structure and have a cleaner display format
lubridate now loads its own namespace
math operations now automatically coerce interval objects to duration objects. Allows intervals to be used “right out of the box” without error messages.
created start() and end() functions for accessing and changing the boundary date-times of an interval
rep() methods for periods, intervals, and durations
added a package help page with functions listed by purpose
eseconds(), eminutes(), etc. are aliased to dseconds(), dminutes(), etc. to make it easier to remember they are duration objects.
changed leap.years() to leap_years() to maintain consistent naming scheme
rewrote as.period() to create only positive periods.
fixed rollover bug in update.POSIXct()
edited make_diff() to display in days when approporiate, not weeks