Thread: pgsql: Document that age() adds days, then full months.

Log Message:
-----------
Document that age() adds days, then full months.

Modified Files:
--------------
    pgsql/doc/src/sgml:
        func.sgml (r1.382 -> r1.383)
        (http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/src/sgml/func.sgml.diff?r1=1.382&r2=1.383)

momjian@postgresql.org (Bruce Momjian) writes:
> Document that age() adds days, then full months.

AFAICT, this documentation "improvement" is outright wrong.

            regards, tom lane

Tom Lane wrote:
> momjian@postgresql.org (Bruce Momjian) writes:
> > Document that age() adds days, then full months.
>
> AFAICT, this documentation "improvement" is outright wrong.

I am now thinking no documentation paragraph is even needed. We have to
use the number of months in the earlier date or simple computations
would not work like:

    test=> select age('2004-05-29', '2004-06-28');
       age
    ----------
     -30 days
    (1 row)

The end of the earlier month is part of the interval between the two
timestamps, while the end of the later month is not.  Of course with a
multi-months span there are more chances for variance, but we certainly
should give the right answer for an interval < 1 month.  I can just add
a C comment.

--
  Bruce Momjian  <bruce@momjian.us>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +

Bruce Momjian wrote:
> Tom Lane wrote:
> > momjian@postgresql.org (Bruce Momjian) writes:
> > > Document that age() adds days, then full months.
> >
> > AFAICT, this documentation "improvement" is outright wrong.
>
> I am now thinking no documentation paragraph is even needed. We have to
> use the number of months in the earlier date or simple computations
> would not work like:
>
>     test=> select age('2004-05-29', '2004-06-28');
>        age
>     ----------
>      -30 days
>     (1 row)
>
> The end of the earlier month is part of the interval between the two
> timestamps, while the end of the later month is not.  Of course with a
> multi-months span there are more chances for variance, but we certainly
> should give the right answer for an interval < 1 month.  I can just add
> a C comment.

I have updated the documentation to read:

   Note there can be ambiguity in the <literal>months</> returned by
   <function>age</> because different months have a different number of
   days.  <productname>PostgreSQL</>'s approach uses the month from the
   earlier of the two dates when calculating partial months.  For example,
   <literal>age('2004-06-01', '2004-04-30')</> uses April to yield
   <literal>1 mon 1 day</>, while using May would yield <literal>1 mon 2
   days</> because May has 31 days, while April has only 30.

--
  Bruce Momjian  <bruce@momjian.us>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +