Counting rows in referencing parents and children - mysql

Question

I have a table that looks like this:
Categories:
cId | Name | Parent
----+-------------------------+-------
1 | Parent One | NULL
2 | Child of 1st Parent | 1
3 | Parent Two | NULL
4 | Child of 1st Parent | 1
5 | Child of 2nd Parent | 2
The table does not represent a heirarchy: Every item is either a child or a parent, but not both.
And one table like this:
Posts:
pId | Name | cID
----+-------------------------+-------
1 | Post 1 | 1
2 | Post 2 | 2
3 | Post 3 | 2
4 | Post 4 | 3
I'd like to run a query on it that returns this:
cId | Count
---+---------
1 | 3
2 | 2
3 | 1
4 | 0
5 | 0
Count is the number of posts connected to the category.
All categories should be returned.
Parent categories should have the count of the category + child categories sum. (this is one of the things I'm having problem with)
Child categories should have the category sum.
How should I do this?

Answer 1

From your expected results, it looks like you don't care about grandchildren and lower, in which case, this should work. To get the correct parent count, I'm checking for Parent IS NULL or Count(children) > 0, in which case, I'm adding 1:
SELECT c.cId, CASE WHEN C.Parent IS NULL OR COUNT(C2.cId) > 0 THEN 1 ELSE 0 END +
COUNT(C2.cId) TotalCount
FROM Categories C
LEFT JOIN Categories C2 on c.cId = c2.Parent
GROUP BY c.cId
Here is some sample fiddle: http://www.sqlfiddle.com/#!2/b899f/1
And the results:
CID TOTALCOUNT
1 3
2 2
3 1
4 0
5 0
---EDIT---
From reading your comments, it looks like you want something like this:
SELECT c.cId,
COUNT(DISTINCT P.pId) + COUNT(DISTINCT P2.pId) TotalCount
FROM Categories C
LEFT JOIN Posts P ON C.CId = P.CId
LEFT JOIN Categories C2 on c.cId = c2.Parent
LEFT JOIN Posts P2 ON C2.CId = P2.CId
GROUP BY c.cId
http://www.sqlfiddle.com/#!2/eb0d2/3

Answer 2

This is general hint. I do not know if analytic functions and partitioning available in MySQL but you can partition your output by categories then count and sum up within categories. Do some research about analytic functions and partition by clause. General example of what I meant - output is partitioned by deptno and ordered. Also, max hiredate determined within partition - replace max with count, sum etc... in your case:
SELECT * FROM
(
SELECT deptno
, empno
, ename
, sal
, RANK() OVER (PARTITION BY deptno ORDER BY sal desc) rnk
, ROW_NUMBER() OVER (PARTITION BY deptno ORDER BY sal desc) rno
, MAX(hiredate) OVER (PARTITION BY deptno ORDER BY deptno) max_hire_date
FROM emp_test
ORDER BY deptno
)
--WHERE rnk = 1
ORDER BY deptno, sal desc
/
DEPTNO EMPNO ENAME SAL RNK RNO MAX_HIRE_DATE
--------------------------------------------------------------------
10 7839 KING 5000 1 1 1/23/1982
10 7782 CLARK 2450 2 2 1/23/1982
10 7934 MILLER 1300 3 3 1/23/1982
20 7788 SCOTT 3000 1 1 1/28/2013
20 7902 FORD 3000 1 2 1/28/2013
20 7566 JONES 2975 3 3 1/28/2013