Bruce Momjian <bruce@momjian.us> writes:
> On Fri, Aug 16, 2024 at 11:37:55AM -0400, Tom Lane wrote:
>> But I'm starting to despair of reaching a solution that's actually
>> self-consistent. Maybe we should leave the DTK_QUARTER behavior
>> alone, and content ourselves with adding DTK_WEEK.
> Well, I liked that -4 months actually was in -2 quarter.
Yeah. On further reflection, I agree it's a bad idea for the
DTK_QUARTER computation to depend on anything but the months field.
So that lets out v3. However, what we have historically is
regression=# select n, extract(quarter from interval '1 mon' * n) from generate_series(-12,12) n;
n | extract
-----+---------
-12 | 1
-11 | -2
-10 | -2
-9 | -2
-8 | -1
-7 | -1
-6 | -1
-5 | 0
-4 | 0
-3 | 0
-2 | 1
-1 | 1
0 | 1
1 | 1
2 | 1
3 | 2
4 | 2
5 | 2
6 | 3
7 | 3
8 | 3
9 | 4
10 | 4
11 | 4
12 | 1
(25 rows)
which is fine on the positive side but it's hard to describe the
results for negative months as anything but wacko. The v2 patch
gives
regression=# select n, extract(quarter from interval '1 mon' * n) from generate_series(-12,12) n;
n | extract
-----+---------
-12 | -1
-11 | -4
-10 | -4
-9 | -4
-8 | -3
-7 | -3
-6 | -3
-5 | -2
-4 | -2
-3 | -2
-2 | -1
-1 | -1
0 | 1
1 | 1
2 | 1
3 | 2
4 | 2
5 | 2
6 | 3
7 | 3
8 | 3
9 | 4
10 | 4
11 | 4
12 | 1
(25 rows)
which is a whole lot saner. So let's run with v2.
regards, tom lane
