[<<]Message[>>]    [<<]Author[>>]    [<<]Subject[>>]    [<<]Thread[>>]

Number   :  4493
Date     :  2003-05-01
Author   :  Garry Deane
Subject  :  Re: Date Macro
Size(KB) :  3


--- In xxcopy@yahoogroups.com, Kan Yabumoto  wrote:
> Thanks Garry and Chuck for your pointers. But, I was
> busy coding my own. As to Garry's comments on the mess
> with the partial week problem, it does not apply to XXCOPY's
> situation. That is, at this point, the macro always
> converts from the date value (which originates yyyy-mm-dd)
> to kk value. Therefore, for the date 2003-01-01,
> /$yyyy-wkk-k$ will always be expanded to 2003-W01-2 even
> though it could be also written as 2002-W53-2, we need not
> give the option.

No, this is incorrect. 2003-W01-2 (2003-12-31) is not equivalent to 
or an alternate form of 2002-W53-2. Every calendar day has a unique 
ISO designation.

> Here's my code:
> 
> I often use the unix timevalue as a convenient standard even in
> the Microsoft environment. So, the current time is already
> available as unix-time and unix-date was obtained by (unix_time
> / 86400L).
> I don't normally put this many comments but this function really
> needs extra comments for me to ever go back to it. I used to
> code in much more convoluted way for efficiency and "cleverness". 
> But, lately, as I age, I tend to use more straightforward coding
> with ample intermediate values which make it easier to follow for
> later revisions.
> 
> --------------------------------------------------------------
> 
> ushort udate2wkno( // Iso week number of the year (1st wk = 01)
> ulong udate) // unix date (relative to 1970-01-01)
> {
> ulong ydate; // date index of the year (Jan 1st is always 0)
> ulong jan1; // day of the week of Jan1 of the year
> 
> // this scheme uses a 4-yr cycle which puts a leap year as the 
4th yr of
> // the 4-yr cycle. The first 4-yr cycle therefore is 1969-1972
> // note that the ISO 8601 std defines that a week starts on 
Mon, ends Sat.

This fails to take account of some years having 52 weeks and others 
having 53 weeks (because a calendar year is not exactly 52 weeks). A 
4 year cycle takes account of leap years but not 53 week years. Since 
each year has an average of 365.25 days (ignoring non-leap years at 
the turn of the century) with 52.1786 weeks, the cycle needs to be 28 
years.

> 
> udate += 365; // date index relative to 1969-01-01
> ydate = udate % 1461; // 1461 = 365 * 4 + 1;
> ydate = (ydate == 1460) ? 365 : (ydate % 365);
> jan1 = (udate - ydate + 2) % 7; / day of the week of jan1
> return((ushort)(((ydate + jan1) / 7) + 1)); // week number of 
the year
> }
> ----------------------------------------------------------------
> 
> This ISO-8601 thing was added just as an individual who strongly
> supports the standard. However, I don't expect this feature to
> be used widely. It's there for the sake of completeness; not more
> than anything else. For the same reason, I decided to retain the
> original /$W$ definition as it was.
> 
> The revised macro feature is posted at XXTB #24
> 
> http://www.xxcopy.com/betatest/xxcopy24.htm
> 
> It is now posted at
> 
> http://www.xxcopy.com/betatest/ (v.2.84.1)
> 
> See if anyone can find a bug in the new macros....

The ISO week number macro gives the wrong week no for all weeks of 
the year when Jan 1 is in the last week of the previous year. It also 
gives the wrong week for Dec 29, 30, 31 when they are in week 1 of 
next year e.g.

date ISO date Xxcopy \$yyyy-wkk-k$

2000-01-01 1999-W52-6 2000-W01-6
2000-01-08 2000-W01-6 2000-W02-6
2001-12-31 2002-W01-1 2001-W53-1

Garry



This message if part of XXCOPY's message Archive.
The archive contains all the messages posted at Yahoo!Groups: XXCOPY.