Mysql query to dynamically convert rows to columns on the basis of two columns

Moon picture Moon · Jul 31, 2013 · Viewed 28.7k times · Source

I have followed a question here to use a Mysql query to dynamically convert rows to columns. This works fine, but i need to convert this on basis of two columns,

The query mentioned in above link works for a single column "data", but i want to work for two columns which are "data" and "price".

i have added an example here,

Given a table A, which look like

Table A

|  id|order|data|item|Price|
-----+-----+----------------
|   1|    1|   P| 1  | 50  |
|   1|    1|   P| 2  | 60  |
|   1|    1|   P| 3  | 70  |
|   1|    2|   Q| 1  | 50  |
|   1|    2|   Q| 2  | 60  |
|   1|    2|   Q| 3  | 70  |
|   2|    1|   P| 1  | 50  |
|   2|    1|   P| 2  | 60  |
|   2|    1|   P| 4  | 80  |
|   2|    3|   S| 1  | 50  |
|   2|    3|   S| 2  | 60  |
|   2|    3|   S| 4  | 80  |

I like to write a query that looks like the following:

Result Table

|  id|order1|order2|order3|item1|item2|item3|item4|
-----+-----+---------------------------------------
|   1|    P |    Q |      | 50  | 60  | 70  |     |
|   2|    P |      |    S | 50  | 60  |     | 80  |

I have tried to create two different queries and then a join to achieve this, but that may not be a good solution. Can any one suggest a solution same like mentioned in the link above.

Thanks

Answer

Taryn picture Taryn · Jul 31, 2013

If you had a known number of values for both order and item, then you could hard code the query into:

select id,
  max(case when `order` = 1 then data end) order1,
  max(case when `order` = 2 then data end) order2,
  max(case when `order` = 3 then data end) order3,
  max(case when item = 1 then price end) item1,
  max(case when item = 2 then price end) item2,
  max(case when item = 3 then price end) item3,
  max(case when item = 4 then price end) item4
from tableA
group by id;

See Demo. But part of the problem that you are going to have is because you are trying to transform multiple columns of data. My suggestion to get the final result would be to unpivot the data first. MySQL does not have an unpivot function but you can use a UNION ALL to convert the multiple pairs of columns into rows. The code to unpivot will be similar to the following:

select id, concat('order', `order`) col,  data value
from tableA
union all
select id, concat('item', item) col, price value
from tableA;

See Demo. The result of this will be:

| ID |    COL | VALUE |
-----------------------
|  1 | order1 |     P |
|  1 | order1 |     P |
|  1 | order1 |     P |
|  1 |  item1 |    50 |
|  1 |  item2 |    60 |
|  1 |  item3 |    70 |

As you can see this has taken the multiple columns of order/data and item/price and convert it into multiple rows. Once that is completed, then you can convert the values back into columns using an aggregate function with a CASE:

select id, 
  max(case when col = 'order1' then value end) order1,
  max(case when col = 'order2' then value end) order2,
  max(case when col = 'order3' then value end) order3,
  max(case when col = 'item1' then value end) item1,
  max(case when col = 'item2' then value end) item2,
  max(case when col = 'item3' then value end) item3
from
(
  select id, concat('order', `order`) col,  data value
  from tableA
  union all
  select id, concat('item', item) col, price value
  from tableA
) d
group by id;

See Demo. Finally, you need to convert the above code into a dynamic prepared statement query:

SET @sql = NULL;
SELECT
  GROUP_CONCAT(DISTINCT
    CONCAT(
      'max(case when col = ''',
      col,
      ''' then value end) as `', 
      col, '`')
  ) INTO @sql
FROM
(
  select concat('order', `order`) col
  from tableA
  union all
  select concat('item', `item`) col
  from tableA
)d;

SET @sql = CONCAT('SELECT id, ', @sql, ' 
                  from
                  (
                    select id, concat(''order'', `order`) col,  data value
                    from tableA
                    union all
                    select id, concat(''item'', item) col, price value
                    from tableA
                  ) d
                  group by id');

PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;

See SQL Fiddle with demo. This gives a result:

| ID | ORDER1 | ORDER2 | ORDER3 | ITEM1 | ITEM2 |  ITEM3 |  ITEM4 |
-------------------------------------------------------------------
|  1 |      P |      Q | (null) |    50 |    60 |     70 | (null) |
|  2 |      P | (null) |      S |    50 |    60 | (null) |     80 |