SELECT id where min or max value from multiple columns

Question

I'm looking for a fast MySQL query that returns the id of the product with the lowest ( min ) or highest ( max ) price, compared to all price categories ( a, b, c and d ).

I have a product table called chocolate_stock with several price categories. It's pretty easy to receive the lowest ( min ) or highest ( max ) price from a specific category ( a or b or c or d ).

id |      name       | price_a | price_b | price_c | price_d |
--------------------------------------------------------------
1  |   Chips Ahoy    |   250   |   530   |   720   |   120 
--------------------------------------------------------------
2  | Chocolate Chunk |   250   |   90    |  32.92  |   110   
--------------------------------------------------------------
3  |      Oreo       |   103   |  44.52  |   250   |   850   
--------------------------------------------------------------

The price categories are decimal(10,2) . Here's a sample that returns the highest price from the categories, but not the id:

$t = 'chocolate_stock';
$arrIds = array(1, 3);

$strQuery = "SELECT id, 
             MAX(price_a) AS price_a, 
             MAX(price_b) AS price_b, 
             MAX(price_c) AS price_c, 
             MAX(price_d) AS price_d 
             FROM $t WHERE id IN(". implode(',', array_map('intval', $arrIds)) .")";

What is the fastest way to retrieve this information?

Answer 1

It might help if you were to tabulate what you want your output to look like, but I think the piece you are missing is the HAVING clause.

first - try this

select min(id), max(price_a) from $t having price_a = max(price_a)

Then try

select min(id), min(price_a) from $t having price_a = min(price_a)
union
select min(id), max(price_a) from $t having price_a = max(price_a)

Answer 2

This query does what you want:

(select t.*
 from $t t
 where . . .
 order by price_a desc
 limit 1) union all
 (select t.*
 from $t t
 where . . .
 order by price_b desc
 limit 1) union all
(select t.*
 from $t t
 where . . .
 order by price_c desc
 limit 1) union all
(select t.*
 from $t t
 where . . .
 order by price_d desc
 limit 1)

If you have an index on id it should perform reasonably well.

That approach requires four passes through the table (although the index on id should greatly reduce that). The following approach requires only one pass through the table:

select MAX(price_a),
       substring_index(group_concat(id order by price_a desc), ',', 1),
       max(price_b),
       substring_index(group_concat(id order by price_b desc), ',', 1),
       max(price_c),
       substring_index(group_concat(id order by price_c desc), ',', 1),
       max(price_d),
       substring_index(group_concat(id order by price_d desc), ',', 1)
from $t
where . . .

It uses a trick with group_concat() and substring_index() to get the max id for each of the columns.

Answer 3

The first thing you would want to do is normalise your data, for ease of later querying I would create the following view:

CREATE VIEW NormalT
AS
    SELECT  ID, Name, 'Price_a' AS Type, Price_a AS Price
    FROM    T
    UNION ALL
    SELECT  ID, Name, 'Price_b' AS Type, Price_b AS Price
    FROM    T
    UNION ALL
    SELECT  ID, Name, 'Price_c' AS Type, Price_c AS Price
    FROM    T
    UNION ALL
    SELECT  ID, Name, 'Price_d' AS Type, Price_d AS Price
    FROM    T;

Then I am not sure of the format you want, if you want the min an max for each price you could use this:

SELECT  mt.Type2,
        mt.Type,
        mt.Price,
        t.ID,
        t.Name
FROM    (   SELECT  Type, MIN(Price) AS Price, 'MIN' AS Type2
            FROM    NormalT
            GROUP BY Type
            UNION ALL
            SELECT  Type, MAX(Price) AS Price, 'MAX' AS Type2
            FROM    NormalT
            GROUP BY Type
        ) mt
        INNER JOIN NormalT T
            ON mt.Type = T.Type
            AND mt.Price = t.Price
ORDER BY mt.Type2, mt.Type, t.ID;

Which will output the following from your sample data:

TYPE2   TYPE        PRICE   ID  NAME
MAX     Price_a     250     1   Chips Ahoy
MAX     Price_a     250     2   Chocolate Chunk
MAX     Price_b     530     1   Chips Ahoy
MAX     Price_c     720     1   Chips Ahoy
MAX     Price_d     850     3   Oreo
MIN     Price_a     103     3   Oreo
MIN     Price_b     44.52   3   Oreo
MIN     Price_c     32.92   2   Chocolate Chunk
MIN     Price_d     110     2   Chocolate Chunk

However, if it is just the min and max of all prices (a, b, c and d) then you could use this:

SELECT  mt.Type2,
        t.Type,
        mt.Price,
        t.ID,
        t.Name
FROM    (   SELECT  MIN(Price) AS Price, 'MIN' AS Type2
            FROM    NormalT
            UNION ALL
            SELECT  MAX(Price) AS Price, 'MAX' AS Type2
            FROM    NormalT
        ) mt
        INNER JOIN NormalT T
            ON mt.Price = t.Price;

Which will output this:

TYPE2   TYPE    PRICE       ID  NAME
MIN     Price_c     32.92   2   Chocolate Chunk
MAX     Price_d     850     3   Oreo

Examples on SQL Fiddle

Answer 4

Try this , It's emulating Analytics as MYSQL doesn't have them by default :

SELECT id, 
             ( select MAX(price_a) from $t t2 where  t2.id = t1.id ) AS price_a, 
             ( select MAX(price_b) from $t t2 where  t2.id = t1.id ) AS price_b, 
             ( select MAX(price_c) from $t t2 where  t2.id = t1.id ) AS price_c, 
             ( select MAX(price_d) from $t t2 where  t2.id = t1.id ) AS price_d 
             FROM $t t1 WHERE id IN(". implode(',', array_map('intval', $arrIds)) .")

Source From : http://www.oreillynet.com/pub/a/mysql/2007/03/29/emulating-analytic-aka-ranking-functions-with-mysql.html?page=3

Answer 5

You are not getting id because, MAX returns one value. But it is not so with id.You can use seperate queries like

SELECT id,MAX(price_a) FROM $t WHERE id IN (". implode(',', array_map('intval', $arrIds)).")";
SELECT id,MAX(price_b) FROM $t WHERE id IN (". implode(',', array_map('intval', $arrIds)).")";

etc