Question

I have below table with Boolean column has_object which indicate each row has associated digital object or not.

+------------+--------------+------+-----+---------+-------+
| Field      | Type         | Null | Key | Default | Extra |
+------------+--------------+------+-----+---------+-------+
| pid        | varchar(255) | NO   | PRI |         |       |
| title      | text         | YES  |     | NULL    |       |
| owner_uid  | int(11)      | YES  |     | NULL    |       |
| has_object | int(11)      | YES  |     | NULL    |       |
+------------+--------------+------+-----+---------+-------+

I have tried this query to obtain statistical information about each owner_uid. but in my table it returns wrong result:

SELECT 
  a.owner_uid, 
  count(b.pid) as count1, 
  count(c.pid) as count2  
FROM
  islandora_report a
  JOIN islandora_report b ON b.owner_uid = a.owner_uid AND b.has_object = 0
  JOIN islandora_report c ON c.owner_uid = a.owner_uid AND c.has_object = 1
GROUP BY a.owner_uid;

The result:

Answer 1

Since the BOOLEAN is merely a 0 or 1, you can actually do one pair of SUM() without any joins to add up the column.

SELECT
  owner_uid,
  /* SUM() adds up all the 1 values */
  SUM(has_object) AS count_true,
  /* Invert the boolean with a case statement to get the inverse */
  SUM(CASE WHEN has_object = 1 THEN 0 ELSE 1 END) AS count_false
FROM islandora_report
GROUP BY owner_uid

There are other ways to invert the boolean than the method above. It’s just the first that came to mind. You could also subtract the true sum from the total count, for example:

SUM(has_object) AS count_true,
COUNT(*) - SUM(has_object) AS count_false