MySQL: calculating localized MIN MAX with subselect

Question

I'm building a normalized dataset based in MySQL on on the relative values of the past five observations. To do this, I need to know what the maximum and minimum values for the past five datapoints have been. I'm working on a subselect that works as follows:

I need to get the MIN and the MAX of the prior five occurrences in a table. While my table is currently arranged in DESC order, I cannot rely on it maintaining that structure, as subsequent updates of RECENT_DATE for a specific MY_ID will change the most recent 5, as well as update the MY_VALUE.

Here's a bit of my base table that I am working on.

    MY_ID | RECENT_DATE | MY_VALUE
    392, 2013-10-06 12:00:00, 18332
    146, 2013-09-06 12:00:00, 5623
    72, 2013-09-02 12:00:00, 23242
    643, 2013-09-01 12:00:00, 15242
    492, 2013-08-06 12:00:00, 15332
    346, 2013-07-26 12:00:00, 17332
    172, 2013-07-22 12:00:00, 14001
    123, 2013-07-22 12:00:00, 13918
    243, 2013-07-11 12:00:00, 23229

At the end I need it to look like.

    MY_ID | RECENT_DATE | MY_VALUE | MAX_VALUE | MIN_VALUE
    392, 2013-10-06 12:00:00, 18332, 23242, 5623
    146, 2013-09-06 12:00:00, 5623, 23242, 5623
    72, 2013-09-02 12:00:00, 23242, 23242, 14001
    643, 2013-09-01 12:00:00, 15242, 17332, 13918
    492, 2013-08-06 12:00:00, 15332, 23229, 13918
    346, 2013-07-26 12:00:00, 17332, 23229, 13918
    172, 2013-07-22 12:00:00, 14001, 23229, 13918
    123, 2013-07-22 12:00:00, 13918, 23229, 13918
    243, 2013-07-11 12:00:00, 23229, 23229, 23229

So here's a quick sample table

CREATE TABLE IF NOT EXISTS dbo.baseTable
(
    MY_ID BIGINT(20) UNSIGNED, 
    RECENT_DATE DATETIME, 
    MY_VALUE BIGINT(20), 
);

INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (392, '2013-10-06 12:00:00', 18332);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (146, '2013-09-06 12:00:00', 5623);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (72, '2013-09-02 12:00:00', 23242);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (643, '2013-09-01 12:00:00', 15242);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (492, '2013-08-06 12:00:00', 15332);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (346, '2013-07-26 12:00:00', 17332);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (172, '2013-07-22 12:00:00', 14001);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (123, '2013-07-22 12:00:00', 13918);
INSERT INTO dbo.baseTable (MY_ID, RECENT_DATE, MY_VALUE) VALUES (243, '2013-07-11 12:00:00', 23229);

Now here's my slaughtered query attempt A:

SELECT DISTINCT t1.MY_ID,
        t1.RECENT_DATE,
        t1.MY_VALUE, MAX(t2.MY_VALUE) AS MAX_MY_VALUE, MIN(t2.MY_VALUE) AS MIN_MY_VALUE
FROM dbo.baseTable t1
    INNER JOIN (
        SELECT t.MY_ID, t1.RECENT_DATE, t.MY_VALUE
            FROM dbo.baseTable t
            WHERE t.RECENT_DATE <= t1.RECENT_DATE
            LIMIT 5) AS t2 ON t1.MY_ID = t2.MY_ID
/* WHERE OMITTED */
ORDER BY t1.RECENT_DATE DESC
;

Answer 1

This query should return the result that you need:

SELECT
  ta.MY_ID, ta.MY_ID, ta.RECENT_DATE, ta.MY_VALUE,
  MAX(tb.MY_VALUE),
  MIN(tb.MY_VALUE)
FROM (
  SELECT
    t1.MY_ID,
    t1.RECENT_DATE,
    t1.MY_VALUE,
    SUBSTRING_INDEX(GROUP_CONCAT(t2.MY_ID ORDER BY t2.RECENT_DATE DESC), ',', 5) as Last_5
  FROM
    baseTable t1 INNER JOIN baseTable t2
    ON t1.recent_date >= t2.recent_date
  GROUP BY
    t1.my_id
) ta INNER JOIN baseTable tb ON FIND_IN_SET(tb.MY_ID, Last_5)
GROUP BY
  ta.MY_ID
ORDER BY
  RECENT_DATE DESC

It's not standard SQL and uses some MySQL tricks. Please see fiddle here.