Re: Fix for edge case in date_bin() function

I wrote:
> Hmm, yeah.  The "stride_usecs > 1" test seems like it's a partial
> attempt to account for this that is probably redundant given the
> additional condition.  Also, can we avoid computing tm_diff %
> stride_usecs twice?  Maybe the compiler is smart enough to remove the
> common subexpression, but it's a mighty expensive computation if not.

I think we could do it like this:

    tm_diff = timestamp - origin;
    tm_modulo = tm_diff % stride_usecs;
    tm_delta = tm_diff - tm_modulo;
    /* We want to round towards -infinity when tm_diff is negative */
    if (tm_modulo < 0)
        tm_delta -= stride_usecs;

Excluding tm_modulo == 0 from the correction is what fixes the
problem.

> I'm also itching a bit over whether there are integer-overflow
> hazards here.  Maybe the range of timestamp is constrained enough
> that there aren't, but I didn't look hard.

Hmm, it's not.  For instance this triggers the overflow check in
timestamp_mi:

regression=# select '294000-01-01'::timestamp - '4714-11-24 00:00:00 BC';
ERROR:  interval out of range
regression=# \errverbose
ERROR:  22008: interval out of range
LOCATION:  timestamp_mi, timestamp.c:2832

So we ought to guard the subtraction that produces tm_diff similarly.
I suspect it's also possible to overflow int64 while computing
stride_usecs.

                        regards, tom lane