> You probably right. I belive that Thomas say more about it...
to_char() is compatible with Oracle. date_part() is compatible with
Ingres (or should be). I've got the Ingres docs somewhere, but
presumably I looked at them when implementing this in the first place.
Maybe not, but what I have is compatible with Unix date formatting.
> For PG date_part/trunc will SET (or anything like this) good.
Let's decide what these functions are for; in this case they are each
cribbed from an existing database product, and should be compatible with
those products imho.
btw, the "week of year" issue is quite a bit more complex; it is defined
in ISO-8601 and it does not correspond directly to a "Jan 1" point in
the calendar. In fact, there can be 53 weeks in a year, and some days
early in the calendar year will fall into the preceeding year for
purposes of this week calculation.
- Thomas