Changelog • hms

Features

Sub-second value parsing accurately (@eitsupi, #127, #128).
is.hms() and as.hms() now warn even when called indirectly (@moodymudskipper, #121, #163).
Implement ceiling_hms() and floor_hms() (@moodymudskipper, #118, #160).

Bug fixes

Parsing now holds the date part fixed, removing unlikely errors during clock wraparound.

Chore

Spring cleaning (@schochastics, #174, #175).
Use rlang instead of ellipsis (#106).

Bug fixes

Fix method consistency, checked by R-devel.

Internal

Use rlang instead of ellipsis (#106).

Re-rendered documentation for compatibility with R-devel.

Avoid blanket import for lifecycle package for compatibility with upcoming rlang.
Establish compatibility with development version of testthat (#101, @lionel-).

Breaking changes

"hms" objects no longer have a common type with character and numeric. Combining such values previously threw a warning, now throws an error (#94).
vec_cast() and as_hms() throw error instead of a warning if input can’t be parsed (#68).

Features

New unique.hms() method (#98, @joethorley).
as_hms() is a generic again (#81).

Internal

Avoid LazyData in DESCRIPTION.
Bump required versions of ellipsis and vctrs to avoid warning during package load.
Using lifecycle package (#94).

Life cycle

hms is now marked as “stable”.

Breaking changes

"hms" objects no longer have a common type with character and numeric. Combining such values previously threw a warning, now throws an error (#94).
vec_cast() and as_hms() now throw error instead of a warning if input can’t be parsed (#68).

Features

as_hms() is a generic again (#81).
round_hms() and trunc_hms() gain digits argument (#78, @hglanz).

Bug fixes

as_hms() and vec_cast() now correctly treat objects of class "difftime" with integer mode (#84).

Internal

Using lifecycle package (#94).
hms has been re-licensed as MIT (#86).

Use vec_default_ptype2(), remove vec_ptype2.hms.unspecified() (#80, @romainfrancois).
vec_ptype2.hms.default() forwards to vec_default_ptype2() for compatibility with vctrs 0.2.1.
Remove as.data.frame.hms(), handeld by vctrs.

Internal changes only.

Work around parsing error that occurs on DST changeover dates (https://bugs.r-project.org/bugzilla/show_bug.cgi?id=16764).

Lossy casts from character vectors to hms now also trigger a warning if the cast succeeds in the first element of the vector but fails for other elements.

Breaking changes

Now based on vctrs >= 0.2.0 (#61). This adds support for vec_cast() and vec_ptype2(). Some operations (such as subset assignment) are now stricter. The new_hms() constructor permits safe construction of hms objects from a numeric vector.
The new is_hms() deprecates the existing is.hms(). as.hms() is deprecated in favor of vec_cast() or the new as_hms() (which is just a thin wrapper around vec_cast()).

Printing

Always show seconds in a pillar if they are different from zero (#54).
Values with nonzero hours, seconds and split-seconds are now displayed correctly in tibbles (#56), even with a very small distance to the full second (#64).

Internal

Don’t test colored output on CRAN.
Correct reference link on r4ds (#58, @evanhaldane).

Adapted tests to pillar 1.2.1.

Preserve NA when converting to character (#51, @jeroen).
Adapted tests to pillar 1.1.0.

Breaking changes

as.hms.POSIXt() now defaults to the current time zone, the previous default was "UTC" and can be restored by calling pkgconfig::set_config("hms::default_tz", "UTC").

New features

Pillar support, will display hms columns in tibbles in color on terminals that support it (#43).
New round_hms() and trunc_hms() for rounding or truncating to a given multiple of seconds (#31).
New parse_hms() and parse_hm() to parse strings in “HH:MM:SS” and “HH:MM” formats (#30).
as.hms.POSIXt() gains tz argument, default "UTC" (#28).
as.hms.character() and parse_hms() accept fractional seconds (#33).

Bug fixes

hms() now works correctly if all four components (days, hours, minutes, seconds) are passed (#49).
hms() creates a zero-length object of class hms that prints as "hms()".
hms(integer()) and as.hms(integer()) both work and are identical to hms().
Values with durations of over 10000 hours are now printed correctly (#48).
c() now returns a hms (#41, @qgeissmann).

Documentation and error messages

Fix and enhance examples in ?hms.
Documentation is in Markdown format now.
Improved error message if calling hms() with a character argument (#29).

Fix lubridate test for compatibility with 1.6.0 (#23, @vspinu).
NA values are formatted as NA (#22).

Minor fixes and improvements.

Subsetting keeps hms class (#16).
format.hms() right-justifies the output by padding with spaces from the left, as.character.hms() remains unchanged.
Times larger than 24 hours or with split seconds are now formatted correctly (#12, #13).
Sub-second part is printed with up to six digits, for even smaller values trailing zeros are shown (#17).

First CRAN release.

Values are stored as a numeric vector that contains the number of seconds since midnight.
- Inherits from difftime class.
- Updating units is a no-op, anything different from "secs" issues a warning.
Supports construction from time values, coercion to and from various data types, and formatting.
- Conversion from numeric treats input as seconds.
- Negative times are formatted with a leading -.
Can be used as a regular column in a data frame.
Full test coverage.
- Test for arithmetic with Date, POSIXt and hms classes.
- Test basic compatibility with lubridate package (#5).
Interface:
- hms() (with rigorous argument checks)
- as.hms() for character, numeric, POSIXct and POSIXlt
- as.xxx.hms() for character, numeric (implicitly), POSIXct and POSIXlt
- is.hms()
- as.data.frame.hms() (forwards to as.data.frame.difftime())
- format.hms()
- print.hms() (returns unchanged input invisibly)