SQL Complex Counting - sql

Question

I want to get all the possible type from a table and then count the rows under a group who has this type.
To better illustrate consider the following table.
Object (o)
id name group_id type
-----------------------------------------------------------
1 Computer 100 A
2 Monitor 100 A
3 Chair 100 B
4 Table 100 B
5 Telephone 100 C
6 Notebook 200 D
7 Pen 200 D
8 Wire 100 E
What I want to get is the following result:
group_id type total
-------------------------------
100 A 2
100 B 2
100 C 1
100 D 0
100 E 1
The group_id doesn't have to be included in the result, I just wanted to let you see here that I'm using a WHERE clause in my query. Basically I only want the results for group | 100 but I want the counts for all type.
Using my query below
SELECT o.group_id, o.type, COUNT(*) AS total
FROM object o
WHERE o.group_id = 100
GROUP BY o.type
I only get the following result:
group_id type total
-------------------------------
100 A 2
100 B 2
100 C 1
100 E 1
You will notice that I am missing the row for type | D. I also want to include that.

Answer 1

You can use the following query:
SELECT 100 AS group_id,
o.type,
COUNT(CASE WHEN o.group_id = 100 THEN 1 END) AS total
FROM object o
GROUP BY o.type
This query groups by type and uses conditional aggregation so as to count the rows under each group who have type = 100.

Answer 2

The problem is that Type D is not returned in the query for that ID, so it is not included in the groupings. To do this, we need to get the list of ALL Types in the table, then do the counts for your ID in the table. Something like this:
SELECT o_list.type, COUNT(o.id) AS total
FROM object o
RIGHT OUTER JOIN ( SELECT DISTINCT type from object ) o_list on o_list.type = o.type
WHERE o.group_id = 100
GROUP BY o_list.type
Giorgos's version is, arguably, cleaner to read and will always run in one full scan of the table. Mine is more complex, but if there are indexes on group_id and type, will run on index scans and so may be significantly faster on a large data set.
And you get to see that there are always options on how to solve a problem. :)

Answer 3

Since Type isn't in group 100 I first get a distinct set of types, then join back to objects since you want 100 for type D, I always show 100 when the type doesn't exist.
SELECT coalesce(O.Group_ID,100) as Group_ID, T.Type, count(O.Group_ID) as Total
FROM (Select Distinct Type from object) T
LEFT JOIN Object O
on T.Type = O.Type
GROUP BY coalesce(O.Group_ID,100), T.Type