Question

I have data in two tables.

The first table has a Primary Key called PKID

PKID  DATA
0    myData0
1    myData1
2    myData2

The second table has the PKID column from table 1 as a foreign key

PKID_FROM_TABLE_1  U_DATA
       0          unique0
       0          unique1        
       0          unique2
       1          unique3
       1          unique4
       1          unique5
       2          unique6
       2          unique7
       2          unique8

The basic SELECT statement I am making now is

SELECT a.PKID, a.DATA, b.U_DATA
FROM table1 as a
INNER JOIN table2 as b
ON a.PKID = b.PKID_FROM_TABLE_1

This produces a table like this:

PKID   DATA     U_DATA
 0   myData0    unique0
 0   myData0    unique1
 0   myData0    unique2
 1   myData1    unique3
 1   myData1    unique4
 1   myData1    unique5
 2   myData2    unique6
 2   myData2    unique7
 2   myData2    unique8

What I would like is the following table:

PKID   DATA    U_DATA1    U_DATA2    U_DATA3
 0     myData0 unique0    unidque1   unique2
 1     myData1 unique3    unidque4   unique5
 2     myData2 unique6    unidque7   unique8

If it helps, each PKID will have exactly 3 entries in table2.

Is something like this possible in MySQL?

Answer 1

This is one way to get the result.

This approach uses correlated subqueries. Each subquery uses an ORDER BY clause to sort the related rows from table2, and uses the LIMIT clause to retrieve the 1st, 2nd and 3rd rows.

SELECT a.PKID
     , a.DATA
     , (SELECT b1.U_DATA FROM table2 b1
         WHERE b1.PKID_FROM_TABLE_1 = a.PKID 
         ORDER BY b1.U_DATA LIMIT 0,1
       ) AS U_DATA1
     , (SELECT b2.U_DATA FROM table2 b2
         WHERE b2.PKID_FROM_TABLE_1 = a.PKID 
         ORDER BY b2.U_DATA LIMIT 1,1
       ) AS U_DATA2
     , (SELECT b3.U_DATA FROM table2 b3
         WHERE b3.PKID_FROM_TABLE_1 = a.PKID 
         ORDER BY b3.U_DATA LIMIT 2,1
       ) AS U_DATA3
  FROM table1 a
 ORDER BY a.PKID  

---

FOLLOWUP

@gliese581g points out that there may be performance issues with this approach, with a large number of rows returned by the outer query, since each subquery in the SELECT list gets executed for each row returned in the outer query.

It should go without saying that this approach cries out for an index:

ON table2 (PKID_FROM_TABLE_1, U_DATA)

-or, at a minimum-

ON table2 (PKID_FROM_TABLE_1)

It’s likely the latter index already exists, if there’s a foreign key defined. The former index would allow the query to be satisfied entirely from the index pages (“Using index”), without the need for a sort operation (“Using filesort”).

@glies581g is quite right to point out that performance of this approach can be problematic on “large” sets.