我可以在SQL Server中消除此子查询吗?

时间:2013-03-08 17:49:46

标签: sql sql-server join greatest-n-per-group

我有一个视图,我正在尝试合并两组以不同方式存储的时间序列数据。设置 D 在每个时间点都有存储的单个值。设置 R 的值只存储生效日期,< / em>并且它们一直有效直到被取代。这是一个示例表结构:

Table D
---------+------------------+--------
D_series | D_time           | D_value
---------+------------------+--------
1        | 2012-01-01 00:00 | 4.52
1        | 2012-01-01 01:00 | 2.41
1        | 2012-01-01 02:00 | 5.98
1        | 2012-01-01 03:00 | 3.51
2        | 2012-01-01 00:00 | 4.54
2        | 2012-01-01 01:00 | 6.41
2        | 2012-01-01 02:00 | 5.28
2        | 2012-01-01 03:00 | 3.11
3        | 2012-01-01 00:00 | 4.22
3        | 2012-01-01 01:00 | 9.41
3        | 2012-01-01 02:00 | 3.98
3        | 2012-01-01 03:00 | 3.53

Table L
---------+---------
D_series | R_series
---------+---------
1        | 1
2        | 1
3        | 2

Table RV
---------+----------+--------
R_series | R_header | R_value
---------+----------+--------
1        | 1        | 5.23
1        | 2        | 2.98
2        | 1        | 1.35

Table RH
---------+-----------------
R_header | R_start
---------+-----------------
1        | 2012-01-01 00:00
2        | 2012-01-01 01:49
3        | 2012-01-01 02:10

我希望该视图返回D_time中所有与其对应的D_value以及当前R_value无关的点:

---------+------------------+---------+--------
D_series | D_time           | D_value | R_value
---------+------------------+---------+--------
1        | 2012-01-01 00:00 | 4.52    | 5.23
1        | 2012-01-01 01:00 | 2.41    | 5.23
1        | 2012-01-01 02:00 | 5.98    | 2.98
1        | 2012-01-01 03:00 | 3.51    | 2.98
2        | 2012-01-01 00:00 | 4.54    | 5.23
2        | 2012-01-01 01:00 | 6.41    | 5.23
2        | 2012-01-01 02:00 | 5.28    | 2.98
2        | 2012-01-01 03:00 | 3.11    | 2.98
3        | 2012-01-01 00:00 | 4.22    | 1.35
3        | 2012-01-01 01:00 | 9.41    | 1.35
3        | 2012-01-01 02:00 | 3.98    | 1.35
3        | 2012-01-01 03:00 | 3.53    | 1.35

我知道如果我创建一个子查询并加入它,我就能做到这一点:

select D.D_series, D_time, D_value, RV1.R_value
from D
join L on L.D_series = D.D_series
join RV RV1 on RV1.R_series = L.R_series
join RH RH1 on RH1.R_header = RV1.R_header and RH1.R_start <= D.D_time
left join (
    select R_series, R_value, R_start
    from RV RV2
    join RH RH2 on RH2.R_header = RV2.R_header
) RZ on RZ.R_series = RV1.R_series and RZ.R_start > RH1.R_start
where RZ.R_start is null or RZ.R_start > D_time

但据我了解,此子查询将首先在RVRH中获取每个记录,即使该视图仅涉及少量R_series。有没有办法消除这个子查询并使其成为标准连接?

1 个答案:

答案 0 :(得分:0)

您是否考虑过使用交叉申请?他们在没有简单连接条件的情况下表现良好。

请参阅此答案:When should I use Cross Apply over Inner Join?

在你的情况下,我认为这些方面可能有用:

select D.D_series, D_time, D_value, current_r_values.R_value
from D
join L on L.D_series = D.D_series
cross apply (
   select top 1 R_series, R_value, R_start
   from RV RV2
   join RH RH2 on RH2.R_header = RV2.R_header
   where RH2.R_Start<D.D_time
   and RV2.R_series = L.R_series
   order by R.R_Start DESC) current_r_values

我知道这并不是你想要的,但我经常发现使用交叉应用可能比加入不等式要快得多。