Обсуждение: Order of operations in postgreSQL.
Just a few notes on this issue:
PSQL:
Appears to be applying the filter to all the columns in the table after the join.
TSQL:
Tried to duplicate this behavior and got these results.
select t1.id as t1id,
cast(cast(t1.test_value as varchar(max)) as datetime) as tms
from table1 t1
join table2 t2
on t1.id = t2.id
where cast(cast(t1.test_value as varchar(max)) as datetime) > '2023-05-27'
These statements work as expected in SQL Server.
Let me know if you have any questions.
Вложения
From your stackoverflow example. (please don’t post links, they will be hard to use in future if link expires)
This is what is happening.
Your code: TO_TIMESTAMP( trim ( substring(m.text_value, 1, 19)), 'YYYY-MM-DD"T"HH24:MI:SS' ) as tms
Your ERROR: ERROR: invalid value "test" for "YYYY"
Postgres logs are very clear about what is going on, if you step back and read them.
Test is in ‘’, the error says ‘test’ is invalid for YYYY
Reproduced by:
SELECT TO_TIMESTAMP( trim ( substring('test00000000000000', 1, 19)), 'YYYY-MM-DD"T"HH24:MI:SS' )
Returns: ERROR: invalid value "test" for "YYYY" DETAIL: Value must be an integer.
From: Ken McClaren <ken.mcclaren@kipuhealth.com>
Sent: Wednesday, May 31, 2023 10:49 AM
To: pgsql-bugs@lists.postgresql.org
Cc: John Hall <john.hall@kipuhealth.com>
Subject: [EXTERNAL] Order of operations in postgreSQL.
Just a few notes on this issue:
PSQL:
Appears to be applying the filter to all the columns in the table after the join.
TSQL:
Tried to duplicate this behavior and got these results.
select t1.id as t1id,
cast(cast(t1.test_value as varchar(max)) as datetime) as tms
from table1 t1
join table2 t2
on t1.id = t2.id
where cast(cast(t1.test_value as varchar(max)) as datetime) > '2023-05-27'
These statements work as expected in SQL Server.
Let me know if you have any questions.
Вложения
Yes, the error is correct, but it should never have occurred.
The join should have eliminated the error condition before it was evaluated by the where clause. That is the case when this type of statement is executed in SQL Server.
From: Wetmore, Matthew (CTR) <Matthew.Wetmore@express-scripts.com>
Date: Wednesday, May 31, 2023 at 4:42 PM
To: Ken McClaren <ken.mcclaren@kipuhealth.com>, pgsql-bugs@lists.postgresql.org <pgsql-bugs@lists.postgresql.org>
Cc: John Hall <john.hall@kipuhealth.com>
Subject: RE: Order of operations in postgreSQL.
CAUTION: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe. |
From your stackoverflow example. (please don’t post links, they will be hard to use in future if link expires)
This is what is happening.
Your code: TO_TIMESTAMP( trim ( substring(m.text_value, 1, 19)), 'YYYY-MM-DD"T"HH24:MI:SS' ) as tms
Your ERROR: ERROR: invalid value "test" for "YYYY"
Postgres logs are very clear about what is going on, if you step back and read them.
Test is in ‘’, the error says ‘test’ is invalid for YYYY
Reproduced by:
SELECT TO_TIMESTAMP( trim ( substring('test00000000000000', 1, 19)), 'YYYY-MM-DD"T"HH24:MI:SS' )
Returns: ERROR: invalid value "test" for "YYYY" DETAIL: Value must be an integer.
From: Ken McClaren <ken.mcclaren@kipuhealth.com>
Sent: Wednesday, May 31, 2023 10:49 AM
To: pgsql-bugs@lists.postgresql.org
Cc: John Hall <john.hall@kipuhealth.com>
Subject: [EXTERNAL] Order of operations in postgreSQL.
Just a few notes on this issue:
PSQL:
Appears to be applying the filter to all the columns in the table after the join.
TSQL:
Tried to duplicate this behavior and got these results.
select t1.id as t1id,
cast(cast(t1.test_value as varchar(max)) as datetime) as tms
from table1 t1
join table2 t2
on t1.id = t2.id
where cast(cast(t1.test_value as varchar(max)) as datetime) > '2023-05-27'
These statements work as expected in SQL Server.
Let me know if you have any questions.
Вложения
"Ken McClaren" <ken.mcclaren@kipuhealth.com> writes: > Yes, the error is correct, but it should never have occurred. > The join should have eliminated the error condition before it was evaluated by the where clause. That is the case whenthis type of statement is executed in SQL Server. Postgres does not promise that JOIN conditions are evaluated before WHERE conditions. I rather doubt that SQL Server does either, because it'd cripple performance for a lot of real-world queries. If you need something like that, you should restructure the query into two levels with an optimization fence between them. One way is SELECT ... FROM (SELECT ... FROM t1 JOIN t2 ON join-condition OFFSET 0) ss WHERE risky-where-condition regards, tom lane
Thanks for your help.
This was a bad assumption on my part. I did not know that Postgres did not make that promise.
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Thursday, June 1, 2023 at 9:51 AM
To: Ken McClaren <ken.mcclaren@kipuhealth.com>
Cc: Wetmore, Matthew (CTR) <Matthew.Wetmore@express-scripts.com>, pgsql-bugs@lists.postgresql.org <pgsql-bugs@lists.postgresql.org>, John Hall <john.hall@kipuhealth.com>
Subject: Re: Order of operations in postgreSQL.
[You don't often get email from tgl@sss.pgh.pa.us. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
CAUTION: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
"Ken McClaren" <ken.mcclaren@kipuhealth.com> writes:
> Yes, the error is correct, but it should never have occurred.
> The join should have eliminated the error condition before it was evaluated by the where clause. That is the case when this type of statement is executed in SQL Server.
Postgres does not promise that JOIN conditions are evaluated before WHERE
conditions. I rather doubt that SQL Server does either, because it'd
cripple performance for a lot of real-world queries.
If you need something like that, you should restructure the query into two
levels with an optimization fence between them. One way is
SELECT ... FROM
(SELECT ... FROM t1 JOIN t2 ON join-condition OFFSET 0) ss
WHERE risky-where-condition
regards, tom lane
Вложения
Thanks for your help.
This was a bad assumption on my part. I did not know that Postgres did not make that promise.
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Thursday, June 1, 2023 at 9:51 AM
To: Ken McClaren <ken.mcclaren@kipuhealth.com>
Cc: Wetmore, Matthew (CTR) <Matthew.Wetmore@express-scripts.com>, pgsql-bugs@lists.postgresql.org <pgsql-bugs@lists.postgresql.org>, John Hall <john.hall@kipuhealth.com>
Subject: Re: Order of operations in postgreSQL.[You don't often get email from tgl@sss.pgh.pa.us. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
CAUTION: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
"Ken McClaren" <ken.mcclaren@kipuhealth.com> writes:
> Yes, the error is correct, but it should never have occurred.
> The join should have eliminated the error condition before it was evaluated by the where clause. That is the case when this type of statement is executed in SQL Server.
Postgres does not promise that JOIN conditions are evaluated before WHERE
conditions. I rather doubt that SQL Server does either, because it'd
cripple performance for a lot of real-world queries.
If you need something like that, you should restructure the query into two
levels with an optimization fence between them. One way is
SELECT ... FROM
(SELECT ... FROM t1 JOIN t2 ON join-condition OFFSET 0) ss
WHERE risky-where-condition
regards, tom lane