Postgres 按范围分组

Question

我有这样的数据

我想要实现的是直方图的数据，它将值计入特定范围。对于 A 类值范围 1-100 和 B 类值范围 0-125，其中 C 类值 = 5。我遇到的问题是多行数据，我需要先在 C 上过滤，然后将值计入要显示的范围直方图。要获得计数，可以说每 10 秒看起来像这样

生成数据的代码：

CREATE TEMP TABLE sample (
     ts timestamp
    ,category varchar(2)
    , val int)
    
insert into sample values 
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'A', 12),
(to_timestamp('01.01.2018 08:00:02', 'dd-mm-yyyy hh24:mi:ss'), 'A', 44),
(to_timestamp('01.01.2018 08:00:03', 'dd-mm-yyyy hh24:mi:ss'), 'C', 1),
(to_timestamp('01.01.2018 08:00:04', 'dd-mm-yyyy hh24:mi:ss'), 'B', 24),
(to_timestamp('01.01.2018 08:00:05', 'dd-mm-yyyy hh24:mi:ss'), 'B', 111),
(to_timestamp('01.01.2018 08:00:06', 'dd-mm-yyyy hh24:mi:ss'), 'C', 5),
(to_timestamp('01.01.2018 08:00:07', 'dd-mm-yyyy hh24:mi:ss'), 'A', 145),
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'B', 16),
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'C', 47),
(to_timestamp('01.01.2018 08:00:02', 'dd-mm-yyyy hh24:mi:ss'), 'C', 5),
(to_timestamp('01.01.2018 08:00:02', 'dd-mm-yyyy hh24:mi:ss'), 'B', 34),
(to_timestamp('01.01.2018 08:00:03', 'dd-mm-yyyy hh24:mi:ss'), 'B', 111),
(to_timestamp('01.01.2018 08:00:03', 'dd-mm-yyyy hh24:mi:ss'), 'C', 5),
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'A', 19),
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'B', 46),
(to_timestamp('01.01.2018 08:00:01', 'dd-mm-yyyy hh24:mi:ss'), 'C', 57)

我想如果我像这样旋转数据

s

elect 
    ts, 
    category,
    case when category = 'A' then val end as "A",
    case when category = 'B' then val end as "B",
    case when category = 'C' then val end as "C"
from sample
order by ts

那么主元空值有问题

Answer 1

这里是：

with periods(pts) as 
(
 select * 
 from generate_series
 (
   timestamp '2018-01-01 08:00:00', 
   timestamp '2018-01-01 08:01:00', 
   interval '10 seconds'
 ) ts
)
select  pts period_start, 
        pts + interval '10 seconds' period_end,
        lat.cat_a, 
        lat.cat_b, 
        lat.cat_c
from periods 
cross join lateral 
(
 select  count(1) filter (where category = 'A' and val between 0 and 100) as cat_a, 
         count(1) filter (where category = 'B' and val between 0 and 125) as cat_b, 
         count(1) filter (where category = 'C' and val = 5) as cat_c
 from sample 
 where ts >= pts and ts < pts + interval '10 seconds'
) lat;

period_start	period_end	cat_a	cat_b	cat_c
2018-01-01 08:00:00	2018-01-01 08:00:10	2	2	1
2018-01-01 08:00:10	2018-01-01 08:00:20	0	0	0
2018-01-01 08:00:20	2018-01-01 08:00:30	0	0	0
2018-01-01 08:00:30	2018-01-01 08:00:40	0	0	0
2018-01-01 08:00:40	2018-01-01 08:00:50	0	0	0
2018-01-01 08:00:50	2018-01-01 08:01:00	0	0	0
2018-01-01 08:01:00	2018-01-01 08:01:10	0	0	0

一排版很简单：

select  min(ts) period_start,
        max(ts) period_end,
        count(1) filter (where category = 'A' and val between 0 and 100) as cat_a, 
        count(1) filter (where category = 'B' and val between 0 and 125) as cat_b, 
        count(1) filter (where category = 'C' and val = 5) as cat_c
from sample;

在澄清评论后添加

select * from (<the first version of the query here>) t where cat_c > 0;

Postgres 按范围分组

Postgres grouping by range

sql

postgresql

grouping

hypertable

timescaledb