Here's a list of a few things I find unelegant in the date_time package:

1. Order or parameters

Consider these declarations:

date d(year, month, day); partial_date pd(day, month);

So date() uses month before day, while partial_date() uses day before month.

Yes, this was recently changed. The reason is that partial_date is a date generator like 'first_kday_after' (eg: 'First Monday in January') and I was able to write time adjustment code more cleanly by keeping he month as the second parameter in all date generators.

2. No += and -=

I can say date1 = date1 + date_duration(3); but not date1 += date_duration(3);

Good point. Apparently we aren't using boost::operators aggressively enough here. I'll put it on the todo list.

3. Namespace problems

Date generators are in different namespaces: Both partial_date and first_kday_of_month are found in boost::date_time, but partial_date is also carried over in boost::gregorian, whereas first_kday_of_month is not.

Fixed in CVS. All the generators are now available in boost::gregorian.

4. Inconsistent abbreviations

If I want the first Thursday in September, I must write:

boost::date_time::first_kday_of_month<date> fts(Thursday, Sep);

'Thursday' must not be abbreviated to 'Thu'. 'Sep' must not be unabbreviated to 'September'.

True. If you really need this you can just update boost/date_time/date_defs.hpp to the way you want it.

5. Misleading exception string

This code: try { date d(2003, Feb, 29); } catch (std::out_of_range exc) { std::cout << "Caught exception: " << exc.what() << std::endl; }

surprisingly outputs this text:

Caught exception: Day of month value is out of range 1..31

So, it seems that 29 is not in the range 1..31.

Fixed in CVS. Now you will get "Day of month is not valid for year" instead. Jeff