SQL选择MAX查找组的单个记录

Question

I have many records that represent the same information, but with different keys to other tables. I need to group these records by their common attributes (Att1, Att2, Att3), then select a single RecordID of all the common records to be the masterID for those common records. That masterID needs to be added to each record that it represents, only if the masterID is null. I cannot change the data structure and all table values are GUID. I have tried MAX, but I dont get the grouping.

TableA (Current State)

RecordID          Att1         Att2     Att3     MasterID
1                 A            B        C        
2                 A            B        C
3                 A            B        C
4                 D            E        F
5                 D            E        F
6                 D            E        F
7                 G            H        I         7
8                 G            H        I         7
9                 G            H        I         7

UPDATE: Added expected result.

TableA (Expected Result)

RecordID          Att1         Att2     Att3     MasterID
1                 A            B        C        1
2                 A            B        C        1
3                 A            B        C        1
4                 D            E        F        4
5                 D            E        F        4
6                 D            E        F        4
7                 G            H        I        7
8                 G            H        I        7
9                 G            H        I        7

Answer 1

Yuck, I dislike this setup. You're violating good normalization practices, and there's some inherent assumptions that require 'outside' knowledge. However, you've said you can't change it so...

I believe the following should work (vendor neutral):

UPDATE TableA as a SET masterId = (SELECT MIN(b.recordId) 
                                   FROM TableA as b
                                   WHERE b.att1 = a.att1
                                   AND b.att2 = a.att2
                                   AND b.att3 = a.att3)
WHERE masterId IS NULL

---

EDIT:

Upon the revelation that GUIDs can be ordered, but cannot be passed for MIN() or MAX() (What?) -

You have (at least) three options:

1. Cast the guid to a passable type (ie - SELECT MIN(CAST(b.recordId as CHAR(36))) ). However, this is likely to be poor performing, simply because it's going to cast every row (see, this is why it's preferrable for internal ids to be simple integers).

2. Attempt this vendor-neutral statement:

    UPDATE TableA as a SET masterId = (SELECT b.recordId FROM TableA as b LEFT JOIN TableA as c ON c.att1 = b.att1 AND c.att2 = b.att2 AND c.att3 = b.att3 AND c.recordId < b.recordId WHERE b.att1 = a.att1 AND b.att2 = a.att2 AND b.att3 = a.att3 AND b.recordId <= a.recordId AND c.recordId IS NULL) WHERE masterId IS NULL 

3. There's also this more idiomatic version for SQL Server:

    UPDATE Updating SET Updating.masterId = Origin.recordId FROM TableA Updating JOIN (TableA Origin LEFT JOIN TableA Exclusion ON Exclusion.att1 = Origin.att1 AND Exclusion.att2 = Origin.att2 AND Exclusion.att3 = Origin.att3 AND Exclusion.recordId < Origin.recordId) ON Exclusion.recordId IS NULL AND Origin.att1 = Updating.att1 AND Origin.att2 = Updating.att2 AND Origin.att3 = Updating.att3 WHERE Updating.masterId IS NULL 

   SQL Fiddle example

Answer 2

Try this:

UPDATE m
SET
    MasterID = s.RecordID
FROM TableA AS m
INNER JOIN (
    SELECT Att1, Att2, Att3, MIN(RecordID) AS RecordID          
    FROM TableA 
    GROUP BY Att1, Att2, Att3
) AS s
    ON m.Att1 = s.Att1 AND m.Att2 = s.Att2 AND m.Att3 = s.Att3
WHERE
    MasterID  IS NULL

Answer 3

If you want to select only then you can use the following query -

select recordid ,
       Att1,
       Att2,
       Att3,
       COALESCE( MasterID , (select min(a2.recordid) from tableA a2
                            where a2.Att1 = a1.Att1
                            and a2.Att2= a1.Att2
                            and a1.Att3 = a2.Att3))
from TableA a1

Answer 4

You could use a cross apply to make a selection. The cross apply basically selects the lowest RecordID that has matching Att1, Att2, and Att3 and joins the RecordID as a column named MasteRID.

WITH temp as (SELECT RecordID, Att1, Att2, Att3, c.MasterID FROM dbo.TableA b
    CROSS APPLY (SELECT TOP 1 RecordID as MasterID 
    FROM dbo.TableA a 
    WHERE a.Att1=b.Att1 
    AND a.Att2=b.Att2 AND a.Att3=b.Att3 
    ORDER BY RecordID) c)

UPDATE TableA 
SET MasterID=t.MasterID
FROM temp t
WHERE RecordID=t.RecordID

The update will update each RecordID entry's MasterID column with the appropriate RecordID.