查询键值对结构

Question

I am using SQL Server 2008 R2. I have a table structure as below :

Id (Identity): int  
GroupId:       int  
Field Id:      int  
Field Name:    nvarchar(1000)  
Value:         nvarchar(max)  

This table can be viewed as a key-value pair structure.

So if I have the following records in the UI :

Due Date    | Target Date       | Status  
-----------------------------------
5-Jun-2013  |       4-June-2013 | Pending  
10-Jun-2013 |       8-June-2013 | Completed  
15-Jun-2013 |      11-June-2013 | Pending  

In the database each column goes as a separate record. So the database will have:

Id  | Group Id  | Field Id  | Field Name    | Value  
-----------------------------------------------------------
1   | 1         | 1         | Due Date      | 5-Jun-2013  
2   | 1         | 2         | Target Date   | 4-Jun-2013  
3   | 1         | 3         | Status        | Pending  

4   | 2         | 1         | Due Date      | 10-Jun-2013  
5   | 2         | 2         | Target Date   | 8-Jun-2013  
6   | 2         | 3         | Status        | Completed  

7   | 3         | 1         | Due Date      | 15-Jun-2013  
8   | 3         | 2         | Target Date   | 11-Jun-2013  
9   | 3         | 3         | Status        | Pending  

While retrieving back from database I would like to get output in the same format as I see in UI. What is the best way to get this result, considering the table will have a huge amount of data?

Any help in this regards will be really useful.

Answer 1

Whether this is a good design or not, you can accomplish what you want with the data as presented in the OP. The key is to note that the difference between the ids identifies the rows. I must emphasize that an explicit "record" reference should be included in the design, so I would call this message a hack.

select (id - FieldId),
       max(case when FieldName = 'DueDate' then value end) as DueDate,
       max(case when FieldName = 'TargetDate' then value end) as TargetDate,
       max(case when FieldName = 'Status' then value end) as Status
from t
group by (id - FieldId)

By the way, the fact that you have to store dates as strings is further indication of a problem with the data structure. Even in an EAV structure, you can have "ValueStr", "ValueDate", "ValueInt" as separate columns to store different values (or use a sqlvariant type).

Answer 2

try this one

SELECT 
      MAX(CASE WHEN fieldName = 'Due Date' THEN fvalue ELSE NULL END) [Due Date],
      MAX(CASE WHEN fieldName = 'Target Date' THEN fvalue ELSE NULL END) [Target Date],
      MAX(CASE WHEN fieldName = 'Status' THEN fvalue ELSE NULL END) [Status] 
FROM #tmp GROUP BY (id - FieldId)