Query to check if column value appears more than once - sql

Question

I have a PostgreSQL query:
SELECT DISTINCT ON ("contract"."contract_id") "contract"."id"
FROM "contract_versions" "contract"
WHERE "contract"."client_id" = 1
GROUP BY "contract"."contract_id", "contract"."id"
ORDER BY "contract"."contract_id", "contract"."change_effective_date" DESC
I want to add something like if contract_id occurs more than once then change_effective_date >= now()
Dataset:
id | contract_id | client_id | change_effective_date
-----+-------------+-----------+-----------------------
100 | 10 | 1 | 2020-05-17 00:00:00
200 | 10 | 1 | 2020-05-16 00:00:00
300 | 10 | 1 | 2020-05-14 00:00:00
400 | 20 | 1 | 2020-05-17 00:00:00
500 | 30 | 1 | 2020-05-13 00:00:00
600 | 30 | 1 | 2020-05-14 00:00:00
Expected result:
id | contract_id | client_id | change_effective_date
-----+-------------+-----------+-----------------------
200 | 10 | 1 | 2020-05-16 00:00:00
400 | 20 | 1 | 2020-05-17 00:00:00
600 | 30 | 1 | 2020-05-14 00:00:00
If the count of contract_id is more than 1, I want a row with change_effective_date less than or equals today
I tried using:
SELECT DISTINCT ON ("contract"."contract_id") "contract"."id",
COUNT("contract"."contract_id") AS cnt
FROM "contract_versions" "contract"
WHERE "contract"."client_id" = 1 AND
CASE WHEN "cnt" > 1 THEN "contract"."change_effective_date" <= now() END
GROUP BY "contract"."contract_id", "contract"."id"
ORDER BY "contract"."contract_id", "contract"."change_effective_date" DESC
but its throwing an error column "cnt" does not exist
Thanks

Answer 1

I have never seen DISTINCT ON combined with GROUP BY. It is possible (aggregate first, then pick rows from that aggregation result), but this is not what you are doing and it is not what you want either. You want the ranking applied by the ORDER BY clause for the DISTINCT ON to take the current date into account.
SELECT DISTINCT ON (contract_id) *
FROM contract_versions
WHERE client_id = 1
ORDER BY
contract_id,
CASE WHEN change_effective_date <= CURRENT_DATE THEN 1 ELSE 2 END,
change_effective_date DESC;
Demo: https://dbfiddle.uk/?rdbms=postgres_12&fiddle=f709d18b23504dfaf9f586ace231be1d

Answer 2

There are 2 problems in your queries:
In the first query you have:
ERROR: column "contract.change_effective_date" must appear in the
GROUP BY clause or be used in an aggregate function
In the second you cannot directly reference a column alias ("cnt") in the query: you can do this in another query that must reference the original query as derived table or inline view.
Here is something that could be a solution:
select * from contract_versions;
id | contract_id | client_id | change_effective_date
-----+-------------+-----------+-----------------------
100 | 10 | 1 | 2020-05-14 14:00:00
100 | 10 | 1 | 2020-05-14 14:00:00
100 | 10 | 1 | 2020-05-14 14:00:00
100 | 20 | 1 | 2020-05-16 09:00:00
(4 rows)
Null display is "NULL".
SELECT DISTINCT ON (contract_id) contract_id,
change_effective_date
FROM contract_versions contract
WHERE client_id = 1
GROUP BY contract_id, id, change_effective_date
ORDER BY contract_id, change_effective_date DESC;
contract_id | change_effective_date
-------------+-----------------------
10 | 2020-05-14 14:00:00
20 | 2020-05-16 09:00:00
(2 rows)
SELECT DISTINCT ON (contract_id) contract_id,
COUNT(contract_id) AS cnt,
change_effective_date
FROM contract_versions contract
WHERE client_id = 1
GROUP BY contract_id, id, change_effective_date
ORDER BY contract_id, change_effective_date DESC;
contract_id | cnt | change_effective_date
-------------+-----+-----------------------
10 | 3 | 2020-05-14 14:00:00
20 | 1 | 2020-05-16 09:00:00
(2 rows)
SELECT
contract_id,
CASE WHEN cnt > 1 THEN change_effective_date <= now()
END
FROM
(
SELECT DISTINCT ON (contract_id) contract_id,
COUNT(contract_id) AS cnt,
change_effective_date
FROM contract_versions contract
WHERE client_id = 1
GROUP BY contract_id, id, change_effective_date
ORDER BY contract_id, change_effective_date DESC
) v;
contract_id | case
-------------+------
10 | t
20 | NULL
(2 rows)

Answer 3

With row_number() window function:
select t.id, t.contract_id, t.client_id, t.change_effective_date
from (
select *,
row_number() over (
partition by contract_id
order by (change_effective_date > now())::int, change_effective_date desc
) rn
from contract_versions
) t
where t.rn = 1
order by t.id
It is not clear in your question, because your sample data contains only 1 client_id, if a contract_id's value belongs to only one client_id.
If this is not the case, then you must change in the above query:
partition by contract_id
to:
partition by contract_id, client_id
See the demo.
Results:
| id | contract_id | client_id | change_effective_date |
| --- | ----------- | --------- | ------------------------ |
| 200 | 10 | 1 | 2020-05-16 00:00:00.000 |
| 400 | 20 | 1 | 2020-05-17 00:00:00.000 |
| 600 | 30 | 1 | 2020-05-14 00:00:00.000 |