Question

I have a MySQL table with the following schema:

+--------------+-------------+------+-----+---------+----------------+
| Field        | Type        | Null | Key | Default | Extra          |
+--------------+-------------+------+-----+---------+----------------+
| id           | int(11)     | NO   | PRI | NULL    | auto_increment |
| country_dest | int(11)     | YES  |     | NULL    |                |
| type         | varchar(56) | YES  |     | NULL    |                |
| version      | varchar(56) | YES  |     | NULL    |                |
+--------------+-------------+------+-----+---------+----------------+

For every country_dest there will be several type – business, tourist, etc (but each country_dest may have different or totally unique types).

I want to make sure that for every type, in every country there is a particular value of version (the same value eg ‘complete’).

So for instance:

+----+--------------+------------+----------+
| id | country_dest | type       | versions |
+----+--------------+------------+----------+
|  1 |            8 | business   | single   |
|  2 |            8 | business   | complete |
|  3 |            8 | tourist    | single   |
|  4 |            8 | tourist    | complete |
|  5 |            8 | diplomatic | single   |
|  6 |           14 | business   | single   |
|  7 |           14 | business   | complete |
|  8 |           14 | private    | single   |
|  9 |           31 | business   | double   |
| 10 |           31 | business   | complete |
+----+--------------+------------+----------+

If that was the table contents, then the query would return that:

country_dest: 8 with type: diplomatic, and
country_dest: 14 with type: private

don’t have the complete value.

Output should be:

+---------------+---------------+
| country_dest  |      type     |
+---------------+---------------+
|             8 |   diplomatic  |
|            14 |   private     |
+---------------+---------------+

It seems like it would be great to loop through this in a for loop, but that’s not how it’s done. I’ve tried lots of solutions but I can’t get them to work and I’m sure there must be a tried and tested way of doing this. This is my latest effort:

SELECT form_dest, form_vtype, form_ispack FROM _forms f
WHERE form_dest IN 
    (SELECT DISTINCT dest.form_dest FROM _forms) dest
AND form_vtype IN 
    (SELECT DISTINCT form_vtype FROM _forms as type)
AND NOT EXISTS  
    (SELECT * FROM _forms 
    WHERE f.form_dest = dest.form_dest 
    AND f.form_type = type.form_type 
    AND f.form_ispack = "Yes") 

but I get errors.

Grateful for any help on this. Cheers.

Answer 1

Using a NOT EXISTS should be enough to get the results you require.

To get the entire _forms record, you can wrap the statement into a subselect and join it again with _forms.

SQL Statement

SELECT  country_dest
        , type
FROM    _forms invalid
WHERE   NOT EXISTS (        
          SELECT  country_dest
                  , type
          FROM    _forms valid
          WHERE   versions = 'complete'
                  AND valid.country_dest = invalid.country_dest
                  AND valid.type = invalid.type
        )          
GROUP BY
        country_dest
        , type

or equivalent using a DISTINCT

SELECT  DISTINCT country_dest
        , type
FROM    _forms invalid
WHERE   NOT EXISTS (        
          SELECT  country_dest
                  , type
          FROM    _forms valid
          WHERE   versions = 'complete'
                  AND valid.country_dest = invalid.country_dest
                  AND valid.type = invalid.type
        )          

Test script (SQL Server)

;WITH _forms (country_dest, type, versions) AS (
  SELECT  8, 'business', 'single'
  UNION ALL SELECT  8, 'business', 'complete'
  UNION ALL SELECT  8, 'tourist', 'single'
  UNION ALL SELECT  8, 'tourist', 'complete'
  UNION ALL SELECT  8, 'diplomatic', 'single'
  UNION ALL SELECT 14, 'business', 'single'
  UNION ALL SELECT 14, 'business', 'complete'
  UNION ALL SELECT 14, 'private', 'single'
  UNION ALL SELECT 31, 'business', 'double'
  UNION ALL SELECT 31, 'business', 'complete'
)
SELECT  country_dest
        , type
FROM    _forms f
WHERE   NOT EXISTS (        
          SELECT  country_dest
                  , type
          FROM    _forms f2
          WHERE   versions = 'complete'
                  AND f2.country_dest = f.country_dest
                  AND f2.type = f.type
        )          
GROUP BY
        country_dest
        , type