Re: [boost] Quick thought on Date-Time

From: Jeff Garland (jeff_at_[hidden])
Date: 2004-04-24 08:35:49

On Fri, 23 Apr 2004 23:22:34 -0400, Daryle Walker wrote
> [NOTE: I haven't worked on this library; I'm just going by the posts 
> I see here.]
> 
> For this increment/decrement by a month problem, is there any way to 
> store a date with only a month resolution?  Nothing smaller will be 
> stored, so day crossovers won't be a problem.
 
I think what you mean is a 'duration' with month resolution.  Ala
  months m1(3);
  months m2(2); 
  months m3 = m2 + m1; 
  //...

This can be done, but the issue is that to work with the date timepoint
calculation is done at single day resolution. And there isn't a fixed number
of days in a month, so you CANNOT say:
  days dd(28) == months(1);  //proposition 1

Similarly the following is NOT true:
  days dd(365) == years(1);  //proposition 2

Conversly, you can always say this:
  days dd(28) == weeks(4);

So the problem is while a week is a fixed length that is well known to be 7
days, a month and year are not.  The length of a month and year depends on the
context of the date.  So people want the following sort of behavior:
  date d1 += months(12);  //if d1 is a leap year add 366 days instead of 365

Unfortunately, as soon as you start accepting these sorts of adjustments you
lose the mathmatical properties that you might expect:
  date d1 += months(1);
  date d2 = d1 - months(1);
  d2 == d1   //Sorry, not always true...

If instead you are willing to accept proposition 1 and 2 above then this is
all trivial...
  
> Wouldn't something 
> similar have/already-had been done when skipping days but ignoring 
> the exact hour?  

No, because a day is a fixed length duration while a month isn't -- see above.

>(Or does every date-time object work at the 
> fractional-second level?  If so, maybe that's what needs fixing.)

No, boost::gregorian::date has a resolution of 1 day -- it can't represent
hours, minutes, etc...

Jeff