SQL Server 查询以获取第一个和最后一个值答案

【问题标题】：SQL Server query to get first and last valueSQL Server 查询以获取第一个和最后一个值
【发布时间】：2018-10-20 20:29:57
【问题描述】：

我有一张带有FieldID, ChangeField, OldValue, NewValue and ChangeDate 的表格，如下所示：

FieldID   ChangeField  OldValue   NewValue   ChangeDate
1         interest     1.5        1.2        2018-05-01 13:00:00
1         interest     1.2        1.3        2018-05-01 14:00:00
1         quantity     2          1          2018-05-01 15:00:00
1         quantity     1          2          2018-05-01 16:00:00
1         quantity     2          3          2018-05-01 17:00:00
2         quantity     10         20         2018-05-01 18:00:00
2         quantity     20         30         2018-05-01 19:00:00

是否可以获得每个 FieldID 的每个 ChangeField 的第一个和最后一个更改，例如下面的结果？

FieldID ChangeField     OldValue    NewValue    dtChangeDate
1       interest        1.5         1.3     2018-05-01 14:00:00  
--> The original value for interest was 1.5, the last value for interest is 1.3
1       quantity        2           3       2018-05-01 17:00:00  
--> The original value for quantity was 2, the last value for interest is 3
2       quantity        10          30      2018-05-01 19:00:00  
--> The original value for quantity was 10, the last value for interest is 30

请注意，ChangeDate 始终是最新的 ChangeDate

【问题讨论】：

是的，您可以使用适当命名的FIRST_VALUE 和LAST_VALUE 函数。

标签： sql sql-server tsql

【解决方案1】：

使用我的标量聚合比较技术（根据我在此处的回答 Get other columns that correspond with MAX value of one column?），您可以非常简单地在没有 CTE 或分区的单个分组查询中执行此操作。

这是查询：

SELECT FieldID, ChangeField,
    MIN(ChangeDate) AS FirstChangeDate,
    SUBSTRING(MIN(CONCAT(ChangeDate, OldValue)), 20) AS FirstOldValue,
    SUBSTRING(MIN(CONCAT(ChangeDate, NewValue)), 20) AS FirstNewValue,
    MAX(ChangeDate) AS LastChangeDate,
    SUBSTRING(MAX(CONCAT(ChangeDate, OldValue)), 20) AS LastOldValue,
    SUBSTRING(MAX(CONCAT(ChangeDate, NewValue)), 20) AS LastNewValue
FROM MyTable
GROUP BY FieldID, ChangeField;

与上面提供的其他答案相比，性能可能提高了 3-4 倍。

这里有更多关于这种技术的信息：https://www.stevenmoseley.com/high-performance-correlated-aggregate-sql-queries-without-ctes

【讨论】：

【解决方案2】：

您可以尝试以下代码，我使用FIRST_VALUE() 获取 OldValue 列中的第一个值，使用 LAST_VALUE() 获取 NewValue 和 dtChangeDate 组的最后一个值。

    SELECT
    FieldID,
    ChangeField,
    FIRST_VALUE(OldValue) OVER (PARTITION BY
                                    fieldID, ChangeField
                                ORDER BY FieldID
                               )  AS OldValue,
    LAST_VALUE(NewValue) OVER (PARTITION BY
                                   fieldID, ChangeField
                               ORDER BY FieldID
                              )   AS NewValue,
    LAST_VALUE(ChangeDate) OVER (PARTITION BY
                                     fieldID, ChangeField
                                 ORDER BY FieldID
                                ) AS dtChangeDate
    FROM YourTable;

【讨论】：

谢谢大家。我尝试了 Aureate 的解决方案，它有效！

【解决方案3】：

您可以使用ROW_NUMBER() 获取最旧/最新。

;WITH Boundaries AS
(
    SELECT
        T.FieldID,
        T.ChangeField,
        T.OldValue,
        T.NewValue,
        OldestRanking = ROW_NUMBER() OVER (
            PARTITION BY
                T.FieldID,
                T.ChangeField
            ORDER BY
                T.ChangeDate ASC),
        NewestRanking = ROW_NUMBER() OVER (
            PARTITION BY
                T.FieldID,
                T.ChangeField
            ORDER BY
                T.ChangeDate DESC)
    FROM
        YourTable AS T
)
SELECT
    T.FieldID,
    T.ChangeField,
    T.OldValue,
    OldValueDate = T.ChangeDate,
    N.NewValue,
    NewValueDate = N.ChangeDate,
FROM
    Boundaries AS T
    INNER JOIN Boundaries AS N ON 
        T.FieldID = N.FieldID AND
        T.ChangeField  = N.ChangeField
WHERE
    T.OldestRanking = 1 AND
    N.NewestRanking = 1

【讨论】：

【解决方案4】：

您可以尝试如下查询：

see live demo

;with cte as
( 
    select
    *,
    rn1= row_number() over (partition by FieldID,ChangeField order by ChangeDate asc),
    rn2= row_number() over (partition by FieldID,ChangeField order by ChangeDate desc)
    from
    Tbl
)

select 
    FieldId,
    ChangeField,
    ChangeDate=MAX(ChangeDate),
    Oldvalue= MAX(Case when rn1=1 then Oldvalue end),
    NewValue= MAX(Case when rn2=1 then NewValue end)
from cte 
    where  rn1=1 or rn2=1
group by 
    FieldId,ChangeField

【讨论】：

【解决方案5】：

使用first_value() 和last_value ()

select distinct * 
from (
  select FieldID, ChangeField  
       first_value(OldValue) over (partition by FieldID, ChangeField order by ChangeDate) OldValue,
       last_value(OldValue) over (partition by FieldID, ChangeField order by ChangeDate) NewValue,
       last_value(OldValue) over (partition by FieldID, ChangeField order by ChangeDate) dtChangeDate
  from table t
) t;

【讨论】：