需要一个 SQL 查询，它将与另一个 table 左联接，后者又会 return 基于时间的最新值，分组到一行中

Question

我需要有关 SELECT SQL 查询的帮助，该查询将在 id 列上与另一个 table 左连接，后者又将 return 最新值基于时间，分组为一行？

基本上，加入两个 table 的方式使得用户 table 中存在于 time_series table、return 中的每条记录基于时间的最新值分组为一行。

用户table：

id  |  name
----+--------
 1  | "Joe" 
 2  | "Ron" 

time_series table:

time| id  |  a  |  b  | c  | d
----+-----+-----+-----+----+----
 1  |  1  |  a1 |     |    |
 2  |  1  |     |  b1 |    |
 3  |  1  |     |     | c0 |
 4  |  1  |  a3 |     | c3 |
 5  |  1  |  a0 |     |    |
 6  |  2  |  a3 |     | c3 | d1
 7  |  2  |  a2 |     |    | d3

结果应如下所示：

id  |  a  |  b  | c  | d  | name
----+-----+-----+----+----+------
 1  |  a0 |  b1 | c3 |    | "Joe"
 2  |  a2 |     |    | d3 | "Ron"

Answer 1

一个选项使用几个子查询：

select u.*,
    (select ts.a from time_series ts where ts.id = u.id and ts.a is not null order by ts.time desc limit 1) a,
    (select ts.b from time_series ts where ts.id = u.id and ts.b is not null order by ts.time desc limit 1) b,
    (select ts.c from time_series ts where ts.id = u.id and ts.c is not null order by ts.time desc limit 1) c,
    (select ts.d from time_series ts where ts.id = u.id and ts.d is not null order by ts.time desc limit 1) d
from users u

另一种解决方案是对每列row_number()的行进行排序，优先考虑非null值，然后进行条件聚合：

select u.id,
    max(case when ts.rn_a = 1 then ts.a end) a,
    max(case when ts.rn_b = 1 then ts.b end) a,
    max(case when ts.rn_c = 1 then ts.c end) a,
    max(case when ts.rn_d = 1 then ts.d end) d
from users u
inner join (
    select ts.*,
        row_number() over(order by (case when a is null then 1 else 0 end), time desc) rn_a,
        row_number() over(order by (case when b is null then 1 else 0 end), time desc) rn_b,
        row_number() over(order by (case when c is null then 1 else 0 end), time desc) rn_c,
        row_number() over(order by (case when d is null then 1 else 0 end), time desc) rn_d
    from time_series ts
) ts on ts.id = u.id
group by u.id

Answer 2

我认为您要查找的函数是 coalesce。

SELECT u.id
    , coalesce(a)
    , coalesce(b)
    , coalesce(c)
    , coalesce(d)
    , name
FROM users u 
LEFT JOIN time_series ts
    ON u.id = ts.id
GROUP BY 1, 6
ORDER BY ts.time DESC;