Hi David,
On Thursday, 2007-02-22 10:56:03 -0500, David A. Wheeler wrote:
> Wheeler:
> > > 1. Simple/absolute weeks
> > > We can cover the "absolute/simple" week number system with a trivial modification to WEEKNUM. Here's a proposal: if WEEKNUM's second parameter is "0", then the simple week number is returned. That would mean that WEEKNUM(d;0) == WEEKNUM(x; WEEKDAY(DATE(1;1;YEAR(x))).
>
> Eike:
> > Apart from that it would be DATE(YEAR(x);1;1) instead, I don't see why
> > WEEKNUM(d;0) would be equal to WEEKNUM(x;n) with n!=0.
>
> Ooops, I changed the unbound variable midways through. Bad me! Let me try again; what I meant was:
> WEEKNUM(x;0) == WEEKNUM(x; WEEKDAY(DATE(1;1;YEAR(x)))
I still don't see how that's meant. Maybe a misunderstanding from mixing
absolute and simple weeks. From my understanding the definitions are:
1. Absolute week, as used by US military:
The first 7 days of a year are the first week, regardless of the
day-of-week the year started with, days 8-14 are the second week, and
so on. This would be covered with a 2nd argument of 0.
2. Simple weeks, as used in many countries not following ISO 8601 and
implemented by Excel:
Week 1 always is the week containing January 1, week number changes
to week 2 as soon as the start-day-of-week is reached. Remaining days
of the first week in the previous year are always in week 53
respectively 54 of that year.
> Eike:
> > Actually quite a few locales not using ISO 8601 week numbering use
> > minimum-number-of-days 1, resulting in the first week of the year being
> > the week where the first start-of-week-day of that year is, which would
> > be covered by your 1..7 proposal above.
> >
> > Some, similar to ISO 8601, use minimum-number-of-days 4, but with Sunday
> > as start-of-week-day. These would need extra handling. Currently the
> > locales are *-CA Canada and *-MT Malta, according to the CLDR
> > supplementalData.xml
>
> Eek. Okay, I stand corrected.
>
> Okay, I think there's wide agreement that we need to _AT LEAST_ add support for 1..7 as the second parameter.
>
> Should we add the optional 3rd parameter for "minimum number of days in a week??
Yes, we should, so we can support
> "0" for simple years?
No, 1; see definition of simple week #2 and my previous explanation
above. For absolute weeks (argument2 = 0) the parameter is meaningless
and must be ignored.
Eike
--
Automatic string conversions considered dangerous. They are the GOTO statements
of spreadsheets. --Robert Weir on the OpenDocument formula subcommittee's list.
