每个组的SQL迭代

Question

i have this table:

|ID| GruopID | Status | Date        |
| 2| 1       | S1     | 7/29/2011   |
| 3| 1       | S2     | 7/30/2011   |
| 9| 1       | S1     | 8/02/2011   |
| 7| 1       | S1     | 8/03/2011   |
| 8| 1       | S1     | 8/04/2011   |
| 1| 2       | S1     | 7/28/2011   |
| 4| 2       | S2     | 7/30/2011   |
| 5| 2       | S3     | 8/01/2011   |
| 6| 3       | S1     | 8/02/2011   |

and also a specific date from an outside source: 7/31/2011

I need a query which will give me for each groupID the closest upper and lower date, so the result will be:

|ID| GruopID | Status | Date        |
| 3| 1       | S2     | 7/30/2011   |
| 9| 1       | S1     | 8/02/2011   |
| 4| 2       | S2     | 7/30/2011   |
| 5| 2       | S3     | 8/01/2011   |
| 6| 3       | S1     | 8/02/2011   |

Can someone please help me and show me the query?

Answer 1

Straightforward approach:

SELECT t1.ID, t1.GroupID, t1.Status, t1.Date
FROM MyTable t1
WHERE t1.Date IN (
    SELECT MAX(t2.Date)
    FROM MyTable t2
    WHERE t2.GroupID = t1.GroupID
    AND t2.Date <= '7/31/2011'
    UNION
    SELECT MIN(t3.Date)
    FROM MyTable t3
    WHERE t3.GroupID = t1.GroupID
    AND t3.Date >= '7/31/2011'
)

Please note that the number of rows per group may not always be two.

• If there is a record with precisely that date 7/31/2011, then a single record will be returned for that group.
• If all records of a group are either before or after 7/31/2011, then a single record will be returned for that group.
• If the table contains duplicate combinations of GroupID and Date, then more than two rows may be returned. This is inherent to the fact that no 'correct' way to choose between the duplicates has been defined.

Answer 2

This is a classic greatest-n-per-group query. I would use CROSS APPLY here.

Make sure that you have index on (GroupID, dt, ID) .

Most likely you have a table Groups that has a list of all GroupIDs . In the query below I use CTE to get a list of all distinct GroupIDs .

Sample data

I added few rows to show how the query works in various cases.

DECLARE @VarDate date = '2011-07-31';
DECLARE @T TABLE (ID int, GroupID int, Status varchar(2), dt date);
INSERT INTO @T (ID, GroupID, Status, dt) VALUES
(2, 1, 'S1', '2011-07-29'),
(3, 1, 'S2', '2011-07-30'),
(9, 1, 'S1', '2011-08-02'),
(7, 1, 'S1', '2011-08-03'),
(8, 1, 'S1', '2011-08-04'),
(1, 2, 'S1', '2011-07-28'),
(4, 2, 'S2', '2011-07-30'),
(5, 2, 'S3', '2011-08-01'),
(6, 3, 'S1', '2011-08-02'),

(11, 4, 'S1', '2011-08-04'),
(12, 4, 'S2', '2011-08-02'),
(13, 4, 'S3', '2011-08-02'),
(21, 4, 'S1', '2011-07-04'),
(22, 4, 'S2', '2011-07-04'),
(23, 4, 'S3', '2011-07-04'),

(31, 5, 'S1', '2011-07-31'),
(32, 5, 'S2', '2011-07-31'),
(33, 5, 'S3', '2011-07-31'),
(34, 5, 'S1', '2011-07-31'),
(35, 5, 'S2', '2011-07-31'),
(36, 5, 'S3', '2011-07-31'),

(41, 6, 'S1', '2011-07-31');

Query

For each GroupID we find the upper and lower row using CROSS APPLY , then UNION ALL upper and lower results together.

WITH
CTE_Groups
AS
(
    SELECT DISTINCT GroupID
    FROM @T
)
SELECT
    CA.ID
    ,Groups.GroupID
    ,CA.Status
    ,CA.dt
FROM
    CTE_Groups AS Groups
    CROSS APPLY
    (
        SELECT TOP(1)
            T.ID
            ,T.Status
            ,T.dt
        FROM @T AS T
        WHERE
            T.GroupID = Groups.GroupID
            AND T.dt >= @VarDate
        ORDER BY T.dt, ID
    ) AS CA

UNION ALL

SELECT
    CA.ID
    ,Groups.GroupID
    ,CA.Status
    ,CA.dt
FROM
    CTE_Groups AS Groups
    CROSS APPLY
    (
        SELECT TOP(1)
            T.ID
            ,T.Status
            ,T.dt
        FROM @T AS T
        WHERE
            T.GroupID = Groups.GroupID
            AND T.dt <= @VarDate
        ORDER BY T.dt DESC, ID DESC
    ) AS CA

ORDER BY GroupID, dt;

Result

+----+---------+--------+------------+
| ID | GroupID | Status |     dt     |
+----+---------+--------+------------+
|  3 |       1 | S2     | 2011-07-30 |
|  9 |       1 | S1     | 2011-08-02 |
|  4 |       2 | S2     | 2011-07-30 |
|  5 |       2 | S3     | 2011-08-01 |
|  6 |       3 | S1     | 2011-08-02 |
| 23 |       4 | S3     | 2011-07-04 |
| 12 |       4 | S2     | 2011-08-02 |
| 31 |       5 | S1     | 2011-07-31 |
| 36 |       5 | S3     | 2011-07-31 |
| 41 |       6 | S1     | 2011-07-31 |
| 41 |       6 | S1     | 2011-07-31 |
+----+---------+--------+------------+