从所有列中获取最后一个NOT NULL值

Question

我有一种情况，我试图从所有列中获取最后一个非空值。这里是一些数据来演示puposes。

测试数据

DECLARE @t TABLE (ID INT, Col1 INT, Col2 INT, Col3 INT, Dated DATETIME) 
INSERT INTO @t VALUES 
(1, NULL, 100 , NULL, '20131210'), --<-- Column2 from this row 
(1, 20 , 200 , NULL, '20131209'), --<-- Column1 from this row 
(1, 30 , NULL , 300 , '20131208'), --<-- Column3 from this row 
(1, 40 , 400 , NULL, '20131207') 

╔════╦══════╦══════╦══════╦═════════════════════════╗ 
║ ID ║ Col1 ║ Col2 ║ Col3 ║   Dated   ║ 
╠════╬══════╬══════╬══════╬═════════════════════════╣ 
║ 1 ║ NULL ║ 100 ║ NULL ║ 2013-12-10 00:00:00.000 ║ 
║ 1 ║ 20 ║ 200 ║ NULL ║ 2013-12-09 00:00:00.000 ║ 
║ 1 ║ 30 ║ NULL ║ 300 ║ 2013-12-08 00:00:00.000 ║ 
║ 1 ║ 40 ║ 400 ║ NULL ║ 2013-12-07 00:00:00.000 ║ 
╚════╩══════╩══════╩══════╩═════════════════════════╝ 

我建立了一个查询返回所需的结果。但是想知道是否有更高性能的这种做法。由于这种方式，我每次希望找到获得这些结果的更有效方法时，都在嘲笑整个表格。

我的查询

SELECT ColumnName, Col1 AS Value, Dated 
FROM 
(
SELECT TOP 1 'Column1' AS ColumnName 
      , Col1 
      , Dated 
FROM @t 
WHERE Col1 IS NOT NULL 
ORDER BY Dated DESC 
)Q1 

UNION ALL 

SELECT * FROM 
(
SELECT TOP 1 'Column2' AS ColumnName 
      , Col2 
      , Dated 
FROM @t 
WHERE Col2 IS NOT NULL 
ORDER BY Dated DESC 
)Q2 

UNION ALL 

SELECT * FROM 
(
SELECT TOP 1 'Column3' AS ColumnName 
      , Col3 
      , Dated 
FROM @t 
WHERE Col3 IS NOT NULL 
ORDER BY Dated DESC 
)Q3 

结果集

╔════════════╦═══════╦═════════════════════════╗ 
║ ColumnName ║ Value ║   Dated   ║ 
╠════════════╬═══════╬═════════════════════════╣ 
║ Column1 ║ 20 ║ 2013-12-09 00:00:00.000 ║ 
║ Column2 ║ 100 ║ 2013-12-10 00:00:00.000 ║ 
║ Column3 ║ 300 ║ 2013-12-08 00:00:00.000 ║ 
╚════════════╩═══════╩═════════════════════════╝ 

它返回正确的结果。但我确信它可以通过更简单/高效的查询完成。任何帮助或指针是多少appericiated

Answer 1

差不多一个简单的UNPIVOT操作和一个合适的ROW_NUMBER来选择最近的值。该UNPIVOT自动消除NULL值：

;With Numbered as (
    select *,ROW_NUMBER() OVER (PARTITION BY Col ORDER BY Dated desc) rn 
    from @t 
    unpivot (Value for Col in (Col1,Col2,Col3)) p 
) 
select * from Numbered where rn = 1 

结果：

ID   Dated     Value  Col  rn 
----------- ----------------------- ----------- -------- -------------------- 
1   2013-12-09 00:00:00.000 20   Col1  1 
1   2013-12-10 00:00:00.000 100   Col2  1 
1   2013-12-08 00:00:00.000 300   Col3  1 

Answer 2

我认为，在垂直行格式的UNION选择是好，但如果你需要它在水平列的格式，那么你可以使用自加入像下面

SELECT top 1 t1.Col1, t2.Col2, t3.col3 
FROM @t t1 
inner join @t t2 on t2.col2 is not null 
inner join @t t3 on t3.col3 is not null 
where t1.col1 is not null 
order by t1.dated desc