Include Rows with Zero Count() Values

Question

I'm designing a tag-based pseudo-category system.

There I have two tables - connection and keyword. 'keyword' defines categories and 'connection' is an adjacency list.

When a category is clicked, the view shows subcategories.

SELECT kw.kw_id, kw.kw_name, kw.kw_desc
FROM keyword AS kw, connection AS conn
WHERE conn.kw1id = :kw1 AND conn.kw2id = kw.kw_id

This works just fine, but now I want to sort those subcategories by the number of their children.

SELECT  child.id AS child_id, keyword.kw_name AS child_name, keyword.kw_desc AS child_desc,
            COUNT(r1.conn_id) AS out_degree
FROM keyword
INNER JOIN connection AS r1 ON r1.kw1id = keyword.kw_id
INNER JOIN (
    SELECT kw2id AS id
    FROM connection AS conn
    WHERE conn.kw1id = [ID_CLICKED]
) AS child ON child.id = r1.kw1id

GROUP BY r1.kw1id

ORDER BY out_degree DESC

But this returns only categories with >0 children. How do I include leaf nodes?

Answer 1

simply replace the INNER with LEFT ought to do it...

by the way you don't really need that subquery

its hard to be sure, but from your description I think something like this is what you want??

 SELECT  k.id AS child_id,   
         k.kw_name AS child_name,  
         k.kw_desc AS child_desc,
         COUNT(rb.conn_id) AS out_degree
 FROM connection AS ra 
 LEFT JOIN keyword As k ON ra.kw2id = k.id AND ra.reltype = 'blah'
 LEFT JOIN connection As rb ON rb.kw1id = k.id AND rb.reltype = 'blah'
 WHERE ra.kw1id = [ID_CLICKED]
 GROUP BY k.id 
 ORDER BY out_degree DESC

This should count the children of the children of [ID_CLICKED] and should still include the childless

To avoid multiple counting, change this:

COUNT(rb.conn_id) AS out_degree

to

COUNT(distinct rb.kw2id) AS out_degree

so each grand child only gets counted once

Answer 2

as something closer to your original query - does this work? (note the 3 changes)

SELECT  child.id AS child_id, keyword.kw_name AS child_name, keyword.kw_desc AS child_desc,
            COUNT(r1.conn_id) AS out_degree
FROM keyword
**LEFT** JOIN connection AS r1 ON **r1.kw2id** = keyword.kw_id
**LEFT** JOIN (
    SELECT kw2id AS id
    FROM connection AS conn
    WHERE conn.kw1id = [ID_CLICKED]
) AS child ON child.id = r1.kw1id
GROUP BY r1.kw1id
ORDER BY out_degree DESC

yeah I know, two answers to one question... but they're different answers :-)