Thread: BUG #15208: COALESCE with CTE returns NULL
The following bug has been logged on the website: Bug reference: 15208 Logged by: Rick Gabriel Email address: klaxian@gmail.com PostgreSQL version: 10.4 Operating system: Ubuntu Linux 16.04 Description: When using COALESCE with an aggregate subquery to a CTE, NULL is returned instead of the first non-null value in the argument list. Proof-of-concept examples are below. I would expect COALESCE to return 0 in these cases. Am I missing something? Thanks. WITH test_cte AS (SELECT 1 AS id, 2 AS qty) SELECT COALESCE(SUM(qty), 0) FROM test_cte WHERE id=2; WITH test_cte AS (SELECT 1 AS id, 2 AS qty) SELECT COALESCE((SELECT SUM(qty) FROM test_cte WHERE id=2), 0);
Hello I can not reproduce psql (10.4 (Debian 10.4-1.pgdg80+1)) Type "help" for help. postgres=# WITH test_cte AS (SELECT 1 AS id, 2 AS qty) SELECT COALESCE(SUM(qty), 0) FROM test_cte WHERE id=2; coalesce ---------- 0 (1 row) postgres=# WITH test_cte AS (SELECT 1 AS id, 2 AS qty) SELECT COALESCE((SELECT SUM(qty) FROM test_cte WHERE id=2), 0); coalesce ---------- 0 (1 row) Can you give more complete example? regards, Sergei
=?utf-8?q?PG_Bug_reporting_form?= <noreply@postgresql.org> writes: > When using COALESCE with an aggregate subquery to a CTE, NULL is returned > instead of the first non-null value in the argument list. Proof-of-concept > examples are below. I would expect COALESCE to return 0 in these cases. Am I > missing something? Thanks. Um ... I get zero from each of your test cases, in all PG versions back to 9.3. regards, tom lane
I'm very sorry to have bothered you. I thought this was a pg problem, but it's not. It appears to be a bug in a GUI wrapper I'm using to issue the queries. When using the psql client, everything works as expected. I will file a separate bug report for the GUI. Please disregard this bug report.
On Fri, May 25, 2018 at 12:24 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
PG Bug reportingform <noreply@postgresql.org> writes:
> When using COALESCE with an aggregate subquery to a CTE, NULL is returned
> instead of the first non-null value in the argument list. Proof-of-concept
> examples are below. I would expect COALESCE to return 0 in these cases. Am I
> missing something? Thanks.
Um ... I get zero from each of your test cases, in all PG versions back
to 9.3.
regards, tom lane