[R] Truncating dates (and other date-time manipulations)

[Whit Armstrong]

probably not pre-canned routines for that, but very easy to implement
with the tools provided in the library.

Looks like most of what you want to do is fairly simple and not worth
the trouble of involving c++.

but things like month_durations and year_durations make it clear that
the authors have been really thorough in their design:
http://www.boost.org/doc/libs/1_36_0/doc/html/date_time/gregorian.html#date_time.gregorian.date_duration

Reversibility of Operations Pitfall
A natural expectation when adding a number of months to a date, and
then subtracting the same number of months, is to end up exactly where
you started. This is most often the result the date_time library
provides but there is one significant exception: The
snap-to-end-of-month behavior implemented by the months duration type.
The months duration type may provide unexpected results when the
starting day is the 28th, 29th, or 30th in a 31 day month. The
month_iterator is not effected by this issue and is therefore included
in the examples to illustrate a possible alternative.

When the starting date is in the middle of a month, adding or
subtracting any number of months will result in a date that is the
same day of month (e.g. if you start on the 15th, you will end on the
15th). When a date is the last day of the month, adding or subtracting
any number of months will give a result that is also the last day of
the month (e.g if you start on Jan 31st, you will land on: Feb 28th,
Mar 31st, etc).

    // using months duration type
    date d(2005, Nov, 30); // last day of November
    d + months(1); // result is last day of December "2005-Dec-31"
    d - months(1); // result is last day of October "2005-Oct-31"

    // using month_iterator
    month_iterator itr(d); // last day of November
    ++itr; // result is last day of December "2005-Dec-31"
    --itr; // back to original starting point "2005-Nov-30"
    --itr; // last day of October "2005-Oct-31"

If the start date is the 28th, 29th, or 30th in a 31 day month, the
result of adding or subtracting a month may result in the
snap-to-end-of-month behavior kicking in unexpectedly. This would
cause the final result to be different that the starting date.

    // using months duration type
    date d(2005, Nov, 29);
    d += months(1); // "2005-Dec-29"
    d += months(1); // "2006-Jan-29"
    d += months(1); // "2006-Feb-28" --> snap-to-end-of-month behavior kicks in
    d += months(1); // "2006-Mar-31" --> unexpected result
    d -= months(4); // "2005-Nov-30" --> unexpected result, not where we started

    // using month_iterator
    month_iterator itr(date(2005, Dec, 30));
    ++itr; // "2006-Jan-30" --> ok
    ++itr; // "2006-Feb-28" --> snap-to DOES NOT kick in
    ++itr; // "2006-Mar-30" --> ok
    --itr; // "2006-Feb-28" --> ok
    --itr; // "2006-Jan-30" --> ok
    --itr; // "2005-Dec-30" --> ok, back where we started


Cheers,
Whit


On Thu, Sep 11, 2008 at 11:49 AM, hadley wickham <h.wickham at gmail.com> wrote:
>> I'm wrapping boost date_time into an R package.  I'll post it up to
>> cran shortly.
>>
>> http://www.boost.org/doc/libs/1_36_0/doc/html/date_time.html
>>
>> I'm not sure if that is what you are looking for, but there are a lot
>> of useful utilities in this library.
>
> Looks useful, but I didn't see any built in way to round a date to
> arbitrary precision.
>
> Hadley
>
>
> --
> http://had.co.nz/
>