Question

I am using mysql. I don’t care how many groups get returned back but if a single group has more then 4 items i only want the first 4. How do i write a statement that only returns 4rows per group? As a temporary fix i am just returning them all and filtering it out in code. Its still pretty fast although it would be easier if i knew syntax

Answer 1

If I understand your question correctly, I believe the following answer should do approximately what you need (note: I’ve included two test tables and their associated inserts as an example since your table structure was not provided):

Given these tables and data:

DROP TABLE IF EXISTS tCustomer;
CREATE TABLE tCustomer (
     customerId INT(11) UNSIGNED NOT NULL auto_increment,
     name       VARCHAR(8),
     PRIMARY KEY (customerId)
) AUTO_INCREMENT=1;

INSERT INTO tCustomer VALUES
(NULL, 'Alex'),
(NULL, 'Bob'),
(NULL, 'Carl')
;

DROP TABLE IF EXISTS tPurchases;
CREATE TABLE tPurchases (
    purchaseId   INT(11) UNSIGNED NOT NULL auto_increment,
    customerId   INT(11) UNSIGNED NOT NULL,
    amount       DECIMAL(9,2),
    purchaseDate DATETIME,
    PRIMARY KEY (purchaseId),
    CONSTRAINT fk_customer FOREIGN KEY (customerId) REFERENCES tCustomer (customerId)
) AUTO_INCREMENT=1;

INSERT INTO tPurchases VALUES
(NULL, 1, 1.00, '2011-01-01 08:00'),
(NULL, 1, 1.01, '2011-01-02 08:00'),
(NULL, 1, 1.02, '2011-01-03 08:00'),
(NULL, 1, 1.03, '2011-01-04 08:00'),
(NULL, 1, 1.04, '2011-01-05 08:00'),
(NULL, 1, 1.05, '2011-01-06 08:00'),
(NULL, 2, 1.01, '2011-01-01 08:00'),
(NULL, 2, 1.02, '2011-01-02 08:00'),
(NULL, 3, 1.01, '2011-01-02 08:00'),
(NULL, 3, 1.02, '2011-01-04 08:00'),
(NULL, 3, 1.03, '2011-01-08 08:00')
;

The following SQL will select purchase data by customer returning no more than the 4 most recent purchases:

SELECT
    tC.customerId, tC.name, tP.purchaseDate, tP.amount, COUNT(tPNewer.customerId) newerCt
FROM
                    tCustomer  tC
    INNER JOIN      tPurchases tP      ON tC.customerId = tP.customerId
    LEFT OUTER JOIN tPurchases tPNewer ON tP.customerId = tPNewer.customerId AND tPNewer.purchaseId > tP.purchaseId
GROUP BY
    tC.customerId,
    tC.name,
     tP.purchaseDate,
     tP.amount
HAVING
    newerCt < 4 -- Ignore rows that have more than 3 newer records
ORDER BY
     tC.customerId,
     tP.purchaseDate desc
;

Here is the resulting output from this select (note that Alex’s oldest transactions are absent):

+------------+------+---------------------+--------+---------+
| customerId | name | purchaseDate        | amount | newerCt |
+------------+------+---------------------+--------+---------+
|          1 | Alex | 2011-01-06 08:00:00 |   1.05 |       0 |
|          1 | Alex | 2011-01-05 08:00:00 |   1.04 |       1 |
|          1 | Alex | 2011-01-04 08:00:00 |   1.03 |       2 |
|          1 | Alex | 2011-01-03 08:00:00 |   1.02 |       3 |
|          2 | Bob  | 2011-01-02 08:00:00 |   1.02 |       0 |
|          2 | Bob  | 2011-01-01 08:00:00 |   1.01 |       1 |
|          3 | Carl | 2011-01-08 08:00:00 |   1.03 |       0 |
|          3 | Carl | 2011-01-04 08:00:00 |   1.02 |       1 |
|          3 | Carl | 2011-01-02 08:00:00 |   1.01 |       2 |
+------------+------+---------------------+--------+---------+