How do I convert MySQL that uses group by and order by clauses to T-SQL query?

Question

I've got the following MySQL query that I would like to translate so it works in MSSQL:

SELECT  *
FROM 
(
    SELECT      *
    FROM        ranking
    ORDER BY    rank_no ASC
            ,   effective_dt DESC
) AS sorted_rank
WHERE       sorted_rank.rank_id = 1950
GROUP BY    sorted_rank.rank_no
LIMIT 10

I've spent an afternoon fiddling but I keep running into problems with GROUP BY and aggregates and all kinds of other errors.

To add a further spanner in the works, the ranking table DOES NOT have a primary key, and I think it may be required to get this working...

Server Version : Microsoft SQL Server 2000 - 8.00.2039

Schema :

rank_id         int(11)
week_id         int(11)
rider_id        int(11)
year_no         int(11)
rank_no         int(11)
effective_dt    datetime
lastupdate_dt   datetime
point_no        float
average_no      float  
result_qy       int(11)

Dataset :

INSERT INTO `ranking` (`id`, `rank_id`, `week_id`, `rider_id`, `year_no`, `rank_no`,  `effective_dt`, `lastupdate_dt`, `point_no`, `average_no`,  `result_qy`)
VALUES
(244, 1950, 417, 72253, 2007, 1, '2006-09-03 00:00:00', '2006-09-01 01:45:00', 2559.19, 426.53, 5),
(108, 1950, 426, 72253, 2007, 1, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 2559.19, 426.53, 5),
(340, 1950, 386, 21767, 2006, 1, '2006-01-29 00:00:00', '2006-11-29 13:31:00', 3256.25, 814.06, 4),
(178, 1950, 420, 60369, 2007, 2, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 2315.86, 385.98, 4),
(166, 1950, 417, 60369, 2007, 2, '2006-09-03 00:00:00', '2006-09-01 01:45:00', 2315.86, 385.98, 4),
(109, 1950, 426, 60369, 2007, 2, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 2315.86, 385.98, 4),
(110, 1950, 426, 49428, 2007, 3, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 2191.19, 365.2, 4),
(227, 1950, 417, 49428, 2007, 3, '2006-09-03 00:00:00', '2006-09-01 01:45:00', 2191.19, 365.2, 4),
(409, 1950, 388, 19570, 2006, 3, '2006-02-12 00:00:00', '2006-11-29 13:31:00', 3106.26, 776.57, 4),
(72, 1950, 399, 47036, 2006, 4, '2006-04-30 00:00:00', '2006-11-29 13:33:00', 1038.02, 346.01, 3),
(413, 1950, 388, 55533, 2006, 4, '2006-02-12 00:00:00', '2006-11-29 13:31:00', 2835.3, 708.83, 4),
(111, 1950, 426, 64517, 2007, 4, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 2001.68, 333.61, 4),
(112, 1950, 426, 72379, 2007, 5, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 1677.32, 279.55, 6),
(263, 1950, 420, 27123, 2007, 5, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1607.31, 267.88, 4),
(415, 1950, 388, 45738, 2006, 5, '2006-02-12 00:00:00', '2006-11-29 13:31:00', 2744.03, 686.01, 4),
(113, 1950, 426, 27123, 2007, 6, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 1607.31, 267.88, 4),
(575, 1950, 420, 50354, 2007, 6, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1148.84, 191.47, 2),
(422, 1950, 388, 39070, 2006, 6, '2006-02-12 00:00:00', '2006-11-29 13:31:00', 2626.51, 656.63, 4),
(114, 1950, 426, 65745, 2007, 7, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 1576.83, 262.8, 4),
(293, 1950, 388, 42127, 2006, 7, '2006-02-12 00:00:00', '2006-11-29 13:31:00', 2614.46, 653.62, 4),
(576, 1950, 420, 32669, 2007, 7, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1133.15, 188.86, 2),
(577, 1950, 420, 23242, 2007, 8, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1107.79, 184.63, 2),
(199, 1950, 420, 60322, 2007, 8, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1390.22, 231.7, 3),
(115, 1950, 426, 54984, 2007, 8, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 1532.8, 255.47, 3),
(578, 1950, 420, 41116, 2007, 9, '2006-09-24 00:00:00', '2006-09-26 06:31:00', 1096.64, 182.77, 2),
(116, 1950, 426, 72386, 2007, 9, '2006-11-05 00:00:00', '2006-11-01 05:59:00', 1442.49, 240.41, 6),
(74, 1950, 399, 32669, 2006, 9, '2006-04-30 00:00:00', '2006-11-29 13:33:00', 817.64, 272.55, 3),
(818890, 1950, 1167, 77510, 0, 10, '2012-01-13 00:00:00', '2012-01-13 11:11:00', 1464.43, 366.11, 4),
(825706, 1950, 1168, 70324, 0, 10, '2012-01-20 00:00:00', '2012-01-20 11:05:00', 1259.38, 314.85, 2),
(826752, 1950, 1170, 75911, 0, 10, '2012-02-01 00:00:00', '2012-02-01 12:58:00', 1237.95, 309.49, 3);

Correct output :

1950    426     72253   2007    1   2006-11-05 00:00:00  2006-11-01 05:59:00     2559.19    426.53      5
1950    426     60369   2007    2   2006-11-05 00:00:00  2006-11-01 05:59:00     2315.86    385.98      4
1950    426     49428   2007    3   2006-11-05 00:00:00  2006-11-01 05:59:00     2191.19    365.2       4
1950    426     64517   2007    4   2006-11-05 00:00:00  2006-11-01 05:59:00     2001.68    333.61      4
1950    426     72379   2007    5   2006-11-05 00:00:00  2006-11-01 05:59:00     1677.32    279.55      6
1950    426     27123   2007    6   2006-11-05 00:00:00  2006-11-01 05:59:00     1607.31    267.88      4
1950    426     65745   2007    7   2006-11-05 00:00:00  2006-11-01 05:59:00     1576.83    262.8       4
1950    426     54984   2007    8   2006-11-05 00:00:00  2006-11-01 05:59:00     1532.8     255.47      3
1950    426     72386   2007    9   2006-11-05 00:00:00  2006-11-01 05:59:00     1442.49    240.41      6
1950    1170    75911   0       10  2012-02-01 00:00:00  2012-02-01 12:58:00     1237.95    309.49      3

Answer 1

SQL Server 2000 and above

Based on your requirement, you need to find the most recent effective date within each rank. To achieve, you need to write a sub query that would fetch the maximum effective date for a given rank and then compare it with the date in the row to filter down to the required results. This will work in all versions from SQL Server 2000 and above.

Following query has been tested in SQL Server 2000.

Click here to view the demo in SQL Fiddle

Script :

SELECT  
TOP 10  id
    ,   rank_id
    ,   week_id
    ,   rider_id 
    ,   year_no
    ,   rank_no
    ,   effective_dt 
    ,   lastupdate_dt 
    ,   point_no  
    ,   average_no 
    ,   result_qy
FROM    ranking r_outer
WHERE   rank_id = 1950
AND     effective_dt = 
        (
            SELECT  MAX(effective_dt)
            FROM    ranking r_inner
            WHERE   r_inner.rank_no = r_outer.rank_no
            AND     r_inner.rank_id = 1950
        )
ORDER BY rank_no;

SQL Server 2005 and above

You need to use RANK function that would partition the results by rank_no column and sort each partition by rank_no followed by effective_dt in descending order. Each sorted partition section will be assigned a rank value like 1, 2, 3 etc. You are interested in only the first rank of each of these partition groups. So, hence the outer SELECT to filter the results by rank_num = 1 .

RANK function is supported only in SQL Server 2005 and above .

Click here to view demo in SQL Fiddle using SQL Server 2012.

Script:

SELECT
TOP 10      id 
        ,   rank_id
        ,   week_id
        ,   rider_id 
        ,   year_no
        ,   rank_no
        ,   effective_dt 
        ,   lastupdate_dt 
        ,   point_no  
        ,   average_no 
        ,   result_qy
FROM
(
    SELECT  id 
        ,   rank_id
        ,   week_id
        ,   rider_id 
        ,   year_no
        ,   rank_no
        ,   effective_dt 
        ,   lastupdate_dt 
        ,   point_no  
        ,   average_no 
        ,   result_qy
        ,   RANK() OVER(
                    PARTITION BY rank_no 
                    ORDER BY rank_no, effective_dt DESC) rownum
    FROM    ranking r
    WHERE   r.rank_id = 1950
) t1 
WHERE rownum = 1;

Answer 2

I would like to recommend to use sub-query in WHERE clause to fetch record with maximum effective_dt (assuming that values are unique within one rank_no, I guess it is a datetime):

SELECT TOP 10 *
FROM ranking R
WHERE R.rank_id = 1950 AND 
    R.effective_dt = (SELECT MAX(effective_dt) from ranking R2 WHERE R2.rank_no = R.rank_no AND R2.rank_id = R.rank_id)
ORDER BY R.rank_no

Answer 3

Just a note of caution without a primary key or clustered on this table it is going to be damned slow when the data grows. I would seriously recommend that you find out which fields are supposed to be keys. A table without a clustered index is a heap and they have no order(generally stored in the order of which they are inserted).

You can try the following query as well.

/*
    This CTE will sort the table into a usable structure. To see what it does run the select statement seperately.
*/
;WITH CTERankID
AS
(

SELECT  *
        ,ROW_NUMBER() OVER(PARTITION BY rank_no ORDER BY effective_Dt DESC) AS RowNo        
FROM ranking

)
SELECT  TOP 10   
    rank_id
    ,week_id
    ,year_no
   ,rank_no
   ,effective_dt
   ,lastupdate_dt
   ,point_no
   ,average_no
   ,result_qy
   ,rowno
FROM CTERankID AS rankTable
WHERE rankTable.RowNo=1
AND rankTable.rank_id = 1950

Hopefully it will help a little. PS This will not work in MySQL as MySQL does not support CTE's.