Counting distinct field in table2 from distinct ids in table1 in SQL (MySQL)

Question

I have 2 tables:

Table 1:

| order_id | shop_id | customer_id | total |  date    |
-------------------------------------------------------
|   9005   |    A    |      1      | 1205  | 20110210 |
|   9006   |    A    |      2      | 8591  | 20110212 |
|   9007   |    A    |      2      | 3472  | 20110216 |
|   9008   |    B    |      1      | 6310  | 20110218 |
-------------------------------------------------------

Table 2:

| shop_id | customer_id | reference |
-------------------------------------
|    A    |      1      | Friend    |
|    A    |      2      | Internet  |
|    B    |      1      | Friend    |
|    C    |      1      | Friend    |
-------------------------------------

I want to select distinct values (shop_id, customer_id) from table1 (filtered by date) and then count the references at table2.

Expected result:

| reference | count |
---------------------
|  Friend   |   2   |
|  Internet |   1   |
----------------------

So far, the Query that I'm using is:

SELECT reference,COUNT(*) as count FROM table1 JOIN table2 USING(shop_id,customer_id) GROUP BY reference

And the result is:

| reference | count |
---------------------
|  Friend   |   2   |
|  Internet |   2   |
---------------------

The problem is that is counting 2 times: shop_id=A and customer_id=2. That is why "Internet" is counted twice.

Could someone help me to identify what is wrong? If possible, I would like to achieve this without using sub-queries (technical limitation).

Thank you.

SQL Dump:

CREATE TABLE `table1` (
   `order_id` int(11),
   `shop_id` char(1),
   `customer_id` int(11),
   `total` smallint(6),
   `date` date
);

INSERT INTO `table1` (`order_id`, `shop_id`, `customer_id`, `total`, `date`) VALUES 
('9005', 'A', '1', '1205', '2011-02-10'),
('9006', 'A', '2', '8591', '2011-02-12'),
('9007', 'A', '2', '3472', '2011-02-16'),
('9008', 'B', '1', '6310', '2011-02-18');

CREATE TABLE `table2` (
   `customer_id` int(11),
   `shop_id` char(1),
   `reference` enum('Friend','Internet')
);

INSERT INTO `table2` (`customer_id`, `shop_id`, `reference`) VALUES 
('1', 'A', 'Friend'),
('2', 'A', 'Internet'),
('1', 'B', 'Friend'),
('1', 'C', 'Friend');

Answer 1

I think this might work for you - at least it returns the expected result

SELECT reference,COUNT(distinct(concat(shop_id,'_',customer_id))) as count 
FROM table1 JOIN table2 USING(shop_id,customer_id) GROUP BY reference;

The '_' is to avoid mixing store_id and customer_id, you can probably use a different hashing function to generate unique store/customer ID

Answer 2

This is one solution using sub-queries:

SELECT reference, COUNT(*) as count FROM (SELECT shop_id,customer_id,reference FROM table1 JOIN table2 USING(shop_id,customer_id) GROUP BY shop_id,customer_id) as filtered GROUP BY(reference)

But I would prefer not using sub-queries... If there is not such an alternative, this will become the correct answer.