SQL: Split a column into multiple

Question

I have the following table:

CREATE TABLE yow(
  userid INT,
  itemid INT,
  feedback INT,
  value INT)

(userid,itemid,feedback) can be considered a primary key, where each of these tuples contains a value.

I want a query which returns a table with the following columns: userid | itemid | col0 | col1 | col2

Where col0 contains value for all rows in yow where feedback = 0, and col1 contains value where feedback = 1 and so on.

I have a somewhat working query:

SELECT
yow.userid AS uid, 
yow.itemid AS iid,
isNull(col0.value, 0) AS col0,
IsNull(col1.value, 0) AS col1,
IsNull(col2.value, 0) AS col2
FROM yow 

LEFT JOIN yow AS col0 ON col0.userid=yow.userid AND col0.itemid=yow.itemid
LEFT JOIN yow AS col1 ON col1.userid=yow.userid AND col1.itemid=yow.itemid
LEFT JOIN yow AS col2 ON col2.userid=yow.userid AND col2.itemid=yow.itemid

WHERE col0.feedback = 0 
AND col1.feedback = 1
AND col2.feedback = 2
GROUP BY uid, iid

The problem is that I can have a value for (userid,itemid) in col1 or col2 but not the others. With this query, those rows are filtered out instead of the missing cells defaulting to 0.

As an example, I am getting something like this:

+-------+-------+--------+--------+--------+
| UID   | IID   | COL0   | COL1   | COL2   |
+-------+-------+--------+--------+--------+
|     1 |   101 |     23 |     22 |    241 |
|     1 |   101 |     51 |     13 |    159 |
|     2 |   102 |     22 |     55 |    152 |
|     3 |   103 |     14 |     41 |    231 |
+-------+-------+--------+--------+--------+

But instead I want something like this, where the missing values of col0 are defaulted to 0.

+-------+-------+--------+--------+--------+
| UID   | IID   | COL0   | COL1   | COL2   |
+-------+-------+--------+--------+--------+
|     1 |   101 |     23 |     22 |    241 |
|     1 |   101 |     51 |     13 |    159 |
|     1 |   102 |      0 |     15 |    142 |
|     2 |   102 |     22 |     55 |    152 |
|     2 |   103 |      0 |     45 |     92 |
|     3 |   103 |     14 |     41 |    231 |
+-------+-------+--------+--------+--------+

Can anyone suggest a fix to my query or perhaps propose a better one? I'm running this on H2, so I reckon the query should be somewhat standard. Thanks:)

Answer 1

The where condition will filter out all the null (missing) entries. To avoid that, you need to move the feedback = x to the join condition. Try the following instead:

SELECT
yow.userid AS uid, 
yow.itemid AS iid,
isNull(col0.value, 0) AS col0,
IsNull(col1.value, 0) AS col1,
IsNull(col2.value, 0) AS col2
FROM yow 
LEFT JOIN yow AS col0 ON col0.feedback = 0 
AND col0.userid=yow.userid AND col0.itemid=yow.itemid
LEFT JOIN yow AS col1 ON col1.feedback = 1 
AND col1.userid=yow.userid AND col1.itemid=yow.itemid
LEFT JOIN yow AS col2 ON col2.feedback = 2 
AND col2.userid=yow.userid AND col2.itemid=yow.itemid
GROUP BY uid, iid

---