One to one left join across two tables - postgresql

Question

So.. This is the query I have:
WITH cat AS (
SELECT day, domain
FROM table1
GROUP BY day, domain
), dog AS (
SELECT day, domain, SUM(count)
FROM table2
GROUP BY day, domain
)
SELECT c.domain,
COALESCE(SUM(d.count),0) AS count
FROM cat c
LEFT JOIN dog d
ON c.domain = d.domain
AND c.day <= d.day
GROUP BY c.domain;
Here's what the "cat" returns:
day | domain
------------+-----------
2015-10-01 | nba.com
2015-10-02 | nba.com
And here's what the "dog" returns:
day | domain | count
------------+----------------+--------
2015-10-03 | nba.com | 2
And here is what the full query returns:
domain | count
------------+-------
nba.com | 4
Count is 4 because the LEFT JOIN satisfies both of the rows in "cat". However, I want the left join only being applied ONCE.. (i.e. have a count of 2 rather than 4). That is the "dog" count of 2 be only applied ONCE (if it is ever satisfied).. and not more than that. Is this possible? I hope I made sense here

Answer 1

Your question is not quite clear.
If you want to find the first row from dog:
WITH cat AS (
SELECT day, domain
FROM table1
GROUP BY day, domain
), dog AS (
SELECT day, domain, SUM(count) count
FROM table2
GROUP BY day, domain
)
SELECT DISTINCT on (c.domain) c.domain,
COALESCE(d.count, 0) AS count
FROM cat c
LEFT JOIN dog d
ON c.domain = d.domain
AND c.day <= d.day
ORDER BY c.domain, c.day, d.day;
If you want to sum all rows from dog (which meet the conditions):
WITH cat AS (
SELECT day, domain
FROM table1
GROUP BY day, domain
), dog AS (
SELECT day, domain, SUM(count) count
FROM table2
GROUP BY day, domain
), cat_and_dogs as (
SELECT DISTINCT ON(c.domain, d.day) c.domain, d.count
FROM cat c
LEFT JOIN dog d
ON c.domain = d.domain
AND c.day <= d.day
)
SELECT domain,
COALESCE(sum(count), 0) AS count
FROM cat_and_dogs
GROUP BY domain;

Answer 2

Basically, I agree with #klin that question is not clear. If you don NOT want to apply LEFT JOIN each times for all domain entries from cat result, domain column should be unique, so you will NOT hit SUM(count) in dog result twice.
You can use count column just in cat table to simply your query, but if you have a reason to keep them separate, it might be a good idea to rely on some other unique column to join on.
WITH cat(day,domain) AS (
VALUES
('2015-10-01'::DATE,'nba.com'),
('2015-10-02'::DATE,'nba.com')
),dog(day,domain,count) AS (
VALUES
('2015-10-03'::DATE, 'nba.com', 1),
('2015-10-03'::DATE, 'nba.com', 1)
),dog_data AS (
SELECT day, domain, SUM(count) count
FROM dog
GROUP BY day, domain
), result_to_normilize AS (
SELECT DISTINCT ON(c.domain,c.day) c.domain,c.day,
COALESCE(SUM(d.count),0) AS count
FROM cat c
LEFT JOIN dog_data d
ON c.domain = d.domain
AND c.day <= d.day
GROUP BY c.domain,c.day
-- | domain | day | count
-- | nba.com | 2015-10-01 | 2
-- | nba.com | 2015-10-02 | 2
)
SELECT DISTINCT ON (r.domain) r.domain, r.count
FROM result_to_normilize r;
-- | domain | count
-- | nba.com | 2