Jim Nasby <Jim.Nasby@BlueTreble.com> writes:
> On 11/17/15 2:09 AM, Vitaly Burovoy wrote:
>> Proposed patch has that behavior: ±infinity for epoch, julian,
>> century, decade, isoyear, millennium and year; NULL for other fields.
> What's the logic behind NULL here? Infinity is infinity, whether it's
> minutes or years.
Didn't you follow the upthread discussion? Fields such as "minutes"
are cyclic, so it's impossible to say either that they converge to
a defined limit or diverge to infinity as x increases. NULL, in the
sense of "unknown", seems like a reasonable representation of that.
Infinity doesn't.
> My specific fear is that now people will have to do a bunch of IF
> timestamp IS NOT NULL THEN ... to get the behavior they need.
Considering that the old behavior is to return zero, and we've had
relatively few complaints about that, I doubt very many people are
going to care.
regards, tom lane