Sum on distinct and group by

Question

I have this :

CREATE TABLE tickets(id serial PRIMARY KEY, source text NOT NULL, user_id int NOT NULL, shop_id int NOT NULL, created_at time DEFAULT current_time);
INSERT INTO tickets VALUES (0, 'web-mobile', 1, 2);
INSERT INTO tickets VALUES (1, 'web-destkop', 1, 2);
INSERT INTO tickets VALUES (2, 'web-destkop', 2, 2);

SELECT shop_id, source, COUNT(DISTINCT(user_id)) AS user_count
FROM tickets
WHERE shop_id = 2
GROUP BY shop_id, source;

Rexster : http://rextester.com/TDS9913

But what I would like is to get the sum of distinct user_id on web-*.

shop_id source  user_count
2       web     2

Answer 1

You need to group by a substring of source:

SELECT shop_id, 
       substring(source,1,3), 
       COUNT(DISTINCT user_id) AS user_count
FROM tickets
WHERE shop_id = 2
GROUP BY shop_id, substring(source,1,3);

http://rextester.com/EFZZSG27212

Answer 2

What you want is something like:

SELECT   
  shop_id, 
  user_id,  
  COUNT(user_id)  AS user_count, 
  LEFT(source, 4) AS new_source
FROM tickets
GROUP BY (user_id, shop_id, new_source)
ORDER BY user_id;

Result: (check the db-fiddle here)

shop_id     user_id     user_count  new_source
      2           1              2        web-
      2           2              1        web-

Of course you can use SUBSTRING instead of LEFT as suggested.