Reshape a Table to Convert Rows to Columns

I’m going to add a somewhat longer and more detailed explanation of the steps to take to solve this problem. I apologize if it’s too long.


I’ll start out with the base you’ve given and use it to define a couple of terms that I’ll use for the rest of this post. This will be the base table:

select * from history;

+--------+----------+-----------+
| hostid | itemname | itemvalue |
+--------+----------+-----------+
|      1 | A        |        10 |
|      1 | B        |         3 |
|      2 | A        |         9 |
|      2 | C        |        40 |
+--------+----------+-----------+

This will be our goal, the pretty pivot table:

select * from history_itemvalue_pivot;

+--------+------+------+------+
| hostid | A    | B    | C    |
+--------+------+------+------+
|      1 |   10 |    3 |    0 |
|      2 |    9 |    0 |   40 |
+--------+------+------+------+

Values in the history.hostid column will become y-values in the pivot table. Values in the history.itemname column will become x-values (for obvious reasons).


When I have to solve the problem of creating a pivot table, I tackle it using a three-step process (with an optional fourth step):

  1. select the columns of interest, i.e. y-values and x-values
  2. extend the base table with extra columns — one for each x-value
  3. group and aggregate the extended table — one group for each y-value
  4. (optional) prettify the aggregated table

Let’s apply these steps to your problem and see what we get:

Step 1: select columns of interest. In the desired result, hostid provides the y-values and itemname provides the x-values.

Step 2: extend the base table with extra columns. We typically need one column per x-value. Recall that our x-value column is itemname:

create view history_extended as (
  select
    history.*,
    case when itemname = "A" then itemvalue end as A,
    case when itemname = "B" then itemvalue end as B,
    case when itemname = "C" then itemvalue end as C
  from history
);

select * from history_extended;

+--------+----------+-----------+------+------+------+
| hostid | itemname | itemvalue | A    | B    | C    |
+--------+----------+-----------+------+------+------+
|      1 | A        |        10 |   10 | NULL | NULL |
|      1 | B        |         3 | NULL |    3 | NULL |
|      2 | A        |         9 |    9 | NULL | NULL |
|      2 | C        |        40 | NULL | NULL |   40 |
+--------+----------+-----------+------+------+------+

Note that we didn’t change the number of rows — we just added extra columns. Also note the pattern of NULLs — a row with itemname = "A" has a non-null value for new column A, and null values for the other new columns.

Step 3: group and aggregate the extended table. We need to group by hostid, since it provides the y-values:

create view history_itemvalue_pivot as (
  select
    hostid,
    sum(A) as A,
    sum(B) as B,
    sum(C) as C
  from history_extended
  group by hostid
);

select * from history_itemvalue_pivot;

+--------+------+------+------+
| hostid | A    | B    | C    |
+--------+------+------+------+
|      1 |   10 |    3 | NULL |
|      2 |    9 | NULL |   40 |
+--------+------+------+------+

(Note that we now have one row per y-value.) Okay, we’re almost there! We just need to get rid of those ugly NULLs.

Step 4: prettify. We’re just going to replace any null values with zeroes so the result set is nicer to look at:

create view history_itemvalue_pivot_pretty as (
  select 
    hostid, 
    coalesce(A, 0) as A, 
    coalesce(B, 0) as B, 
    coalesce(C, 0) as C 
  from history_itemvalue_pivot 
);

select * from history_itemvalue_pivot_pretty;

+--------+------+------+------+
| hostid | A    | B    | C    |
+--------+------+------+------+
|      1 |   10 |    3 |    0 |
|      2 |    9 |    0 |   40 |
+--------+------+------+------+

And we’re done — we’ve built a nice, pretty pivot table using MySQL.


Considerations when applying this procedure:

  • what value to use in the extra columns. I used itemvalue in this example
  • what “neutral” value to use in the extra columns. I used NULL, but it could also be 0 or "", depending on your exact situation
  • what aggregate function to use when grouping. I used sum, but count and max are also often used (max is often used when building one-row “objects” that had been spread across many rows)
  • using multiple columns for y-values. This solution isn’t limited to using a single column for the y-values — just plug the extra columns into the group by clause (and don’t forget to select them)

Known limitations:

  • this solution doesn’t allow n columns in the pivot table — each pivot column needs to be manually added when extending the base table. So for 5 or 10 x-values, this solution is nice. For 100, not so nice. There are some solutions with stored procedures generating a query, but they’re ugly and difficult to get right. I currently don’t know of a good way to solve this problem when the pivot table needs to have lots of columns.

Leave a Comment

deneme bonusu veren sitelerbahis casino