Question

I have a table (let’s call it tableA) that looks something like this:

    id    |   dateA    |   dateB    
----------|------------|-------------
    1     | 2011-01-01 | 2010-05-01
    1     | 2011-05-01 | 2010-06-12
    2     | 2011-01-11 | 2010-01-31
    2     | 2011-01-31 | 2010-02-01
    3     | 2011-05-11 | 2010-08-02
    3     | 2011-08-02 | 2010-09-10

My goal is to group rows by id, and get minimum of dateA and maximum of dateB
and get a table that looks something like this:

    id    | min(dateA) | max(dateB)
----------|------------|-------------
    1     | 2011-01-01 | 2010-06-12
    2     | 2011-01-11 | 2010-02-01
    3     | 2011-05-11 | 2010-09-11

Right now, I’m using LEFT join approach:

SELECT
    id,
    tableB.dateA,
    tableC.dateB
FROM tableA as a

    LEFT JOIN (
        SELECT id, min(dateA)
        FROM tableA
        GROUP BY id
    )tableB ON a.id = tableB.id

    LEFT JOIN (
        SELECT id, max(dateB)
        FROM tableA
        GROUP BY id
    )tableC ON a.id = tableC.id

However, my approach is way too slow. I have pretty big table, and it takes about 7seconds to get the desired outcome.

Could anyone suggest me a good optimizing technique to apply to my situation?

Thank you.

J

Answer 1

What’s wrong with a simple GROUP BY using one pass through the table and no self-joins?

SELECT id, MIN(dateA), MAX(dateB)
  FROM TableA
 GROUP BY id;