BUG #19651: to_date()/to_timestamp() silently accept out-of-range values for SSSSS, RM, IW and ID format fields
The following bug has been logged on the website:
Bug reference: 19651
Logged by: chunling qin
Email address: 303677365@qq.com
PostgreSQL version: 18.6
Operating system: x86_64
Description:
Four to_date()/to_timestamp() format fields whose documented value ranges
are narrow accept out-of-range input silently and produce a wrong date,
while sister fields in the same domain raise errors for the same kind of
input:
1. SSSSS (seconds since midnight, documented 0-86399) — negative value
silently negated:
SELECT to_timestamp('2024-01-01 -1', 'YYYY-MM-DD SSSSS');
-- 2024-01-01 00:00:01+00 (-1 second becomes +1 second)
2. RM (Roman month numeral, documented I..XII) — out-of-range numeral
accepted:
SELECT to_date('2024 XIII', 'YYYY RM');
-- 2024-12-01 (XIII parsed as month 12)
3. IW (ISO week, documented 01-53) — values 54-99 accepted, silently
spilling into the following year. The sister field WW validates the same
value correctly:
SELECT to_date('2024 54', 'IYYY IW');
-- 2025-01-06 (no error)
SELECT to_date('2024 99', 'IYYY IW');
-- 2025-11-17 (no error)
SELECT to_date('2024 54', 'YYYY WW');
-- ERROR: date/time field value out of range: "2024 54"
4. ID (ISO day of week, documented 1-7) — both 0 and 8 accepted, each mapped
to Sunday:
SELECT to_date('2024 01 8', 'IYYY IW ID');
-- 2024-01-07 (should be an error)
SELECT to_date('2024 01 0', 'IYYY IW ID');
-- 2024-01-07 (same Sunday result as 7)
Expected behavior
Each field should raise date/time field value out of range for input outside
its documented domain, consistently with how DDD (367+), WW (54+), MM (13+),
DD (32+) and others already behave.
чт, 3 сент. 2026 г. в 18:03, PG Bug reporting form <noreply@postgresql.org>:
The following bug has been logged on the website:
Bug reference: 19651
Logged by: chunling qin
Email address: 303677365@qq.com
PostgreSQL version: 18.6
Operating system: x86_64
Description:Four to_date()/to_timestamp() format fields whose documented value ranges
are narrow accept out-of-range input silently and produce a wrong date,
while sister fields in the same domain raise errors for the same kind of
input:1. SSSSS (seconds since midnight, documented 0-86399) — negative value
silently negated:SELECT to_timestamp('2024-01-01 -1', 'YYYY-MM-DD SSSSS');
-- 2024-01-01 00:00:01+00 (-1 second becomes +1 second)
2. RM (Roman month numeral, documented I..XII) — out-of-range numeral
accepted:SELECT to_date('2024 XIII', 'YYYY RM');
-- 2024-12-01 (XIII parsed as month 12)
3. IW (ISO week, documented 01-53) — values 54-99 accepted, silently
spilling into the following year. The sister field WW validates the same
value correctly:SELECT to_date('2024 54', 'IYYY IW');
-- 2025-01-06 (no error)
SELECT to_date('2024 99', 'IYYY IW');
-- 2025-11-17 (no error)SELECT to_date('2024 54', 'YYYY WW');
-- ERROR: date/time field value out of range: "2024 54"
4. ID (ISO day of week, documented 1-7) — both 0 and 8 accepted, each
mapped
to Sunday:SELECT to_date('2024 01 8', 'IYYY IW ID');
-- 2024-01-07 (should be an error)
SELECT to_date('2024 01 0', 'IYYY IW ID');
-- 2024-01-07 (same Sunday result as 7)
Expected behavior
Each field should raise date/time field value out of range for input
outside
its documented domain, consistently with how DDD (367+), WW (54+), MM
(13+),
DD (32+) and others already behave.Hi!
This also pertains to the [0]/messages/by-id/19650-d7ea430084cdc44a@postgresql.org thread; the patch is in there.
[0]: /messages/by-id/19650-d7ea430084cdc44a@postgresql.org
/messages/by-id/19650-d7ea430084cdc44a@postgresql.org
--
Regards,
Rachitskiy Andrey