将 PostgreSQL 数据分组以在输出中包含不超过 N 个值

Question

我需要快速加载图表数据。数据位于日期 Y 表中：

CREATE TABLE (
    id serial primary key,
    date timestamp not null,
    y1 double precision,
    y2 double precision,
    y3 double precision,
    y4 double precision,
    ....................
    yN double precision
);

为了简化这个问题，假设我只加载日期（X 轴）和 Y1（Y 轴）。

图表包含水平 N 个点并具有可变缩放。 X 轴包含从 mindate 到 maxdate 的点。 mindate 和 maxdate 的值取决于当前的缩放和平移位置。

为了快速加载数据，我需要使用任何聚合函数（MIN、MAX、AVG）执行动态数据分组，以便在从 mindate 到 maxdate 的范围内获得不超过 N 条聚合记录。

数据应按相等的时间间隔分组，即每 5 秒、每 3 分钟，具体取决于当前图表比例。

如何使用 PostgreSQL 实现这一点？

样本

select date, "1_2_pt100_x_1" from ob_data where "1_2_pt100_x_1" is not null order by date

13414 条记录：

"2016-12-08 12:13:41";44.5
"2016-12-08 12:16:42";44.7000007629395
"2016-12-08 12:19:43";44.9000015258789
"2016-12-08 12:22:44";45.2999992370606
"2016-12-08 12:25:44";45.5
"2016-12-08 12:30:11";45.9000015258789
"2016-12-08 12:33:12";45.5999984741211
"2016-12-08 12:36:12";45.9000015258789
"2016-12-08 12:39:13";46.2000007629395
"2016-12-08 12:42:14";46.5
"2016-12-08 12:45:15";46.5999984741211
"2016-12-08 12:48:15";46.9000015258789
"2016-12-08 12:51:16";47
"2016-12-08 12:54:16";47.0999984741211
"2016-12-08 12:57:17";47.4000015258789
"2016-12-08 13:00:17";47.4000015258789
"2016-12-08 13:03:18";47.7000007629395
"2016-12-08 13:06:18";47.7999992370606
"2016-12-08 13:09:19";47.7999992370606
"2016-12-08 13:12:19";47.7999992370606
"2016-12-08 13:15:20";48.0999984741211
"2016-12-08 13:18:20";48.0999984741211
"2016-12-08 13:21:21";48
"2016-12-08 13:24:21";48.4000015258789
"2016-12-08 13:27:22";48.2999992370606
"2016-12-08 13:30:22";48.2000007629395
"2016-12-09 08:58:54";20.2999992370606
"2016-12-09 09:01:55";20.3999996185303
"2016-12-09 09:04:55";20.6000003814697
"2016-12-09 09:07:56";20.6000003814697
"2016-12-09 09:10:56";20.8999996185303
"2016-12-09 09:13:57";20.7000007629395
"2016-12-09 09:16:57";21.2999992370606
"2016-12-09 09:19:58";21.5
"2016-12-09 09:22:58";22
"2016-12-09 09:25:59";22.2000007629395
"2016-12-09 09:28:59";22.5
"2016-12-09 09:32:00";22.8999996185303
"2016-12-09 09:35:00";23.1000003814697
"2016-12-09 09:38:01";23.2999992370606
"2016-12-09 09:41:01";23.6000003814697
"2016-12-09 09:44:02";24.1000003814697
"2016-12-09 09:47:02";24.2999992370606
"2016-12-09 09:50:03";24.6000003814697
"2016-12-09 09:53:03";24.7999992370606
"2016-12-09 09:56:04";25.2000007629395
"2016-12-09 09:59:04";25.2999992370606
"2016-12-09 10:02:05";25.7000007629395
"2016-12-09 10:05:05";25.8999996185303
"2016-12-09 10:08:06";25.8999996185303
"2016-12-09 10:11:06";26.2999992370606
"2016-12-09 10:14:07";26.3999996185303
"2016-12-09 10:17:07";26.7999992370606

预期输出

对于前 6 行，比例因子为 2，使用平均函数：

"2016-12-08 12:13:41";44,6000003814698
"2016-12-08 12:19:43";45,1000003814698
"2016-12-08 12:25:44";45,7000007629395

（我是自己用计算器算出来的，所以不要问我对应的SQL查询）。

为每个汇总结果取 2 行实际上并不是正确的方法。 mindate 和 maxdate 之间的整个数据空间应该被分成 N 个相等的间隔，每个间隔应该被聚合。作为日期的汇总值，我应该取每个间隔的第一个日期。

Answer 1

我导入了你的数据：

create table so6 (d timestamp,f float);
copy so6 from stdin delimiter ',';

这里ntile(27) 意味着你必须将 53 除以 2 才能得到 27 - 如果你想每两行聚合，如果你想要每五行，那就是 ntile(11)：

with p as (
         select *, ntile(27) over(order by d) n
         from so6
         order by d
) select distinct min(d) over (partition by n),
         avg(f) over (partition by n)
from p
order by min;

         min         |       avg
---------------------+------------------
 2016-12-08 12:13:41 | 44.6000003814698
 2016-12-08 12:19:43 | 45.1000003814697
 2016-12-08 12:25:44 | 45.7000007629395
 2016-12-08 12:33:12 |            45.75
 2016-12-08 12:39:13 | 46.3500003814698
 2016-12-08 12:45:15 |            46.75
 2016-12-08 12:51:16 | 47.0499992370605
 2016-12-08 12:57:17 | 47.4000015258789
 2016-12-08 13:03:18 |            47.75
 2016-12-08 13:09:19 | 47.7999992370606
 2016-12-08 13:15:20 | 48.0999984741211
 2016-12-08 13:21:21 | 48.2000007629395
 2016-12-08 13:27:22 |            48.25
 2016-12-09 08:58:54 | 20.3499994277954
 2016-12-09 09:04:55 | 20.6000003814697
 2016-12-09 09:10:56 | 20.8000001907349
 2016-12-09 09:16:57 | 21.3999996185303
 2016-12-09 09:22:58 | 22.1000003814697
 2016-12-09 09:28:59 | 22.6999998092652
 2016-12-09 09:35:00 | 23.1999998092652
 2016-12-09 09:41:01 | 23.8500003814697
 2016-12-09 09:47:02 | 24.4499998092652
 2016-12-09 09:53:03 |               25
 2016-12-09 09:59:04 |             25.5
 2016-12-09 10:05:05 | 25.8999996185303
 2016-12-09 10:11:06 | 26.3499994277954
 2016-12-09 10:17:07 | 26.7999992370606
(27 rows)