如何使用时间戳作为参考合并SQL Server 2014中同一表中的记录

Question

Currently I have a table after using PIVOT() with the format and data values as follows

NameID   DocumentName   Time  Value1  Value2  Value3  Value4  Value5
------   ------------   ----  ------  ------  ------  ------  ------
2221        Doc1        1053    23      24      25     NULL    NULL     
2221        Doc1        1153    31      32      NULL    28      30
2221        Doc2        1253    NULL    NULL    NULL    40      41  
2222        Doc3        1053    03      06      09      12      15

I need to merge record 1,2,3 with the latest values into one single row, where the format would be like this.

NameID       Value1   Value2  Value3  Value4  Value5
------       ------   ------  ------  ------  ------
2221         31       32      25      40      41    
2222         03       06      09      12      15

It would use timestamp as reference to decide what values to update and which values will be kept. Any help or start would be appreciated for me to carry on my work!

Answer 1

This is tricky.

Without knowledge about the pivot query itself, perhaps the simplest method is correlated subqueries or apply :

select nameId,
       (select top (1) t2.documentname
        from t t2
        where t2.nameId = t.nameId and t2.documentname is not null
        order by t2.time desc
       ) as documentname,
       max(time) as time,
       (select top (1) t2.value1
        from t t2
        where t2.nameId = t.nameId and t2.value1 is not null
        order by t2.time desc
       ) as value1,
       (select top (1) t2.value2
        from t t2
        where t2.nameId = t.nameId and t2.value2 is not null
        order by t2.time desc
       ) as value2,
       . . .
from t;

Answer 2

I can not see any difference between this and your question here...

How to transpose rows of data into a single row with different columns in SQL Server

The answer I gave you is already taking the last values before do the pivot, if this is a new requirement you probably just need to adapt the partitions to it...

Anyway, If that is not possible and assuming that you are using SQL Server 2014 for this requirement too, then you can use LAST_VALUE over your current approach

WITH YourPivot AS (
    <Put your current pivot code here>
)
SELECT DISTINCT 
    NameID
    ,LAST_VALUE(Value1) OVER (PARTITION BY NameID ORDER BY CASE WHEN Value1 IS NOT NULL THEN 1 ELSE 0 END, [Time] ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) Value1
    ,LAST_VALUE(Value2) OVER (PARTITION BY NameID ORDER BY CASE WHEN Value2 IS NOT NULL THEN 1 ELSE 0 END, [Time]  ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) Value2
    ,LAST_VALUE(Value3) OVER (PARTITION BY NameID ORDER BY CASE WHEN Value3 IS NOT NULL THEN 1 ELSE 0 END, [Time]  ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) Value3
    ,LAST_VALUE(Value4) OVER (PARTITION BY NameID ORDER BY CASE WHEN Value4 IS NOT NULL THEN 1 ELSE 0 END, [Time]  ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) Value4
    ,LAST_VALUE(Value5) OVER (PARTITION BY NameID ORDER BY CASE WHEN Value5 IS NOT NULL THEN 1 ELSE 0 END, [Time]  ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) Value5
FROM YourPivot