OVER PARTION BY 没有订购

Question

我有一个 table，我只想从中检索每组记录中的最新记录，由特定字段中的值标记。

table 内容看起来像这样：

state   date
512     2021-03-09 11:31:38.300
512     2021-03-09 11:31:38.300
512     2021-03-09 11:31:31.693
512     2021-03-09 11:31:31.693
512     2021-03-08 12:49:10.753
512     2021-03-08 12:35:47.357
514     2021-03-08 12:35:01.030
512     2021-03-08 12:33:48.050
514     2021-03-08 12:14:29.537
514     2021-03-08 12:14:29.537
514     2021-03-08 12:14:18.760
512     2021-03-08 12:14:05.597

我想像这样使用 OVER 和 PARTITION 到 SELECT 输出：

row state   date
1   512     2021-03-09 11:31:38.300
2   512     2021-03-09 11:31:38.300
3   512     2021-03-09 11:31:31.693
4   512     2021-03-09 11:31:31.693
5   512     2021-03-08 12:49:10.753
6   512     2021-03-08 12:35:47.357
1   514     2021-03-08 12:35:01.030
1   512     2021-03-08 12:33:48.050
1   514     2021-03-08 12:14:29.537
2   514     2021-03-08 12:14:29.537
3   514     2021-03-08 12:14:18.760
1   512     2021-03-08 12:14:05.597

如您所见，行按 date DESC 排序，state 字段根据每个更改从 1 开始的 row 字段分组在 state 字段中。

目前，我的代码如下所示：

with query as
(
    select state, date, row = row_number() over (partition by state order by date desc)
    from table
)
select t.*
from table t
inner join query q on t.state = q.state and t.date = q.date
where row = 1
order by t.date desc

不幸的是，这似乎是在按 date DESC 对记录进行排序之前按 state 对记录进行分组，因此结果只有两个结果集记录，因为 [=17] 中只有两个不同的值=] 字段。应该（对于上面的示例数据）有 5 个结果集记录。

如何正确编号分区组？

Answer 1

这可行，但可能有更简单的方法。

WITH
cte1 AS
(
    SELECT 
        ROW_NUMBER() OVER(ORDER BY date_ DESC) as dateRow,
        ROW_NUMBER() OVER(PARTITION BY state ORDER BY date_ DESC) as stateRow,
        state,
        date_
    FROM    StateDate
)
SELECT 
    ROW_NUMBER() OVER(PARTITION BY state, (dateRow - stateRow)  ORDER BY date_ DESC) as row,
    state, date_
FROM cte1
ORDER BY date_ DESC, state

这是我使用的数据设置：

CREATE TABLE StateDate( state INT, date_ DATETIME)
GO

--state   date
INSERT INTO StateDate VALUES (512, '2021-03-09 11:31:38.300');
INSERT INTO StateDate VALUES (512, '2021-03-09 11:31:38.300');
INSERT INTO StateDate VALUES (512, '2021-03-09 11:31:31.693');
INSERT INTO StateDate VALUES (512, '2021-03-09 11:31:31.693');
INSERT INTO StateDate VALUES (512, '2021-03-08 12:49:10.753');
INSERT INTO StateDate VALUES (512, '2021-03-08 12:35:47.357');
INSERT INTO StateDate VALUES (514, '2021-03-08 12:35:01.030');
INSERT INTO StateDate VALUES (512, '2021-03-08 12:33:48.050');
INSERT INTO StateDate VALUES (514, '2021-03-08 12:14:29.537');
INSERT INTO StateDate VALUES (514, '2021-03-08 12:14:29.537');
INSERT INTO StateDate VALUES (514, '2021-03-08 12:14:18.760');
INSERT INTO StateDate VALUES (512, '2021-03-08 12:14:05.597');
GO

OVER PARTION BY 没有订购

OVER PARTITION BY without ordering

tsql

row-number

data-partitioning