Re: Possibly a bug - Mailing list pgsql-bugs

From surya poondla
Subject Re: Possibly a bug
Date
Msg-id CAOVWO5oT0x9snrGFPgoWc+-2sBcYSgm+aU=DbiZHWM_HXTbfYg@mail.gmail.com
Whole thread
In response to Re: Possibly a bug  (jian he <jian.universality@gmail.com>)
Responses Re: Possibly a bug
List pgsql-bugs
Hi Jian,

These current.* functions always use the "C" collation for their
output, so in this case, they are not related to session variables, in
my view.

Thank you, I did some more testing and see the below.

postgres=# SELECT collation for (user);
 pg_collation_for
------------------
 "C"
(1 row)

postgres=# SELECT collation for (current_user);
 pg_collation_for
------------------
 "C"
(1 row)

postgres=#   SELECT to_upper_varchar('тест'::varchar);
 to_upper_varchar
------------------
 ТЕСТ
(1 row)

postgres=#   SELECT collation for ('тест');
 pg_collation_for
------------------

(1 row)

postgres=#   CREATE OR REPLACE FUNCTION debug_func(param1 text, param2 text) RETURNS text AS $$ SELECT 'param1 collation: ' || pg_collation_for(param1) || ', param2 collation: ' || pg_collation_for(param2); $$ LANGUAGE sql;
CREATE FUNCTION
postgres=#   SELECT debug_func('тест', user::text);
                 debug_func
----------------------------------------------
 param1 collation: "C", param2 collation: "C"
(1 row)

postgres=#   CREATE OR REPLACE FUNCTION debug_func2(param1 text, param2 text) RETURNS text AS $$ SELECT 'param1 collation: ' || pg_collation_for(param1) || ', param2 collation: ' || pg_collation_for(param2); $$ LANGUAGE sql;
CREATE FUNCTION
postgres=# SELECT debug_func2('тест'::text, user::text);
                  debug_func2
----------------------------------------------
 param1 collation: "C", param2 collation: "C"
(1 row)

postgres=#

My observation is the user, current* functions collation is "C" collation and when a "C" collation is passed as a parameter, it converts all the parameters to use the "C" collation.

I was looking at sql_fn_make_param() function and the comments over their say
"/*
* If we have a function input collation, allow it to override the
* type-derived collation for parameter symbols.  (XXX perhaps this should
* not happen if the type collation is not default?)
*/
"
From the above code comment it looks like the "C" collation is overriding the type-driven collation, and this might be causing the issue.

I am currently using gdb to debug more and will work on a patch based on my findings.

Regards,
Surya Poondla

pgsql-bugs by date:

Previous
From: Andres Freund
Date:
Subject: Re: BUG #19400: Memory leak in checkpointer and startup processes on PostgreSQL 18
Next
From: Tom Lane
Date:
Subject: Re: Possibly a bug