Обсуждение: BUG #17916: Expression IN list translates to unqualified operator

The following bug has been logged on the website:

Bug reference:      17916
Logged by:          RekGRpth
Email address:      rekgrpth@gmail.com
PostgreSQL version: 15.2
Operating system:   alpine in docker
Description:

```sql
create schema qwe;
create or replace function qwe.chartexteq(a char, b text) returns boolean
language plpgsql as $$begin perform 1/0;return true;end;$$;
create operator qwe.= (leftarg = char, rightarg = text, function =
qwe.chartexteq, commutator = operator(qwe.=), hashes, merges);
set search_path = qwe;
explain (costs off, verbose on) select i from generate_series(1, 10) i where
i::char in (2::text);
                  QUERY PLAN                   
-----------------------------------------------
 Function Scan on pg_catalog.generate_series i
   Output: i
   Function Call: generate_series(1, 10)
   Filter: ((i.i)::character(1) = '2'::text)
 Query Identifier: -7727870581584713193
(5 rows)
select i from generate_series(1, 10) i where i::char in (2::text);
ERROR:  division by zero
CONTEXT:  SQL statement "SELECT 1/0"
PL/pgSQL function chartexteq(character,text) line 1 at PERFORM
```
I expected, that IN list translates to pg_catalog.=

PG Bug reporting form <noreply@postgresql.org> writes:
> create operator qwe.= (leftarg = char, rightarg = text, function =
> qwe.chartexteq, commutator = operator(qwe.=), hashes, merges);
> set search_path = qwe;
> explain (costs off, verbose on) select i from generate_series(1, 10) i where
> i::char in (2::text);

> I expected, that IN list translates to pg_catalog.=

Why would you expect that?  It'd make it impossible to use IN
with user-defined data types.  In this case, you made an operator
that is a closer match to the given datatypes (ie, "char = text")
than the native "text = text" operator, so it used that one.

I've not checked the code, but my recollection is that X IN (Y) just
resolves to the same equality operator you'd get by writing X = Y.
There's been some discussion about allowing a schema qualifier to
be included in the syntax, but nothing's been done about that.

            regards, tom lane

Thank you for the clarification.

How can I safely use an expression IN list in extensions?

with bst regrds, RekGRpth

ср, 3 мая 2023 г. в 18:40, Tom Lane <tgl@sss.pgh.pa.us>:
>
> PG Bug reporting form <noreply@postgresql.org> writes:
> > create operator qwe.= (leftarg = char, rightarg = text, function =
> > qwe.chartexteq, commutator = operator(qwe.=), hashes, merges);
> > set search_path = qwe;
> > explain (costs off, verbose on) select i from generate_series(1, 10) i where
> > i::char in (2::text);
>
> > I expected, that IN list translates to pg_catalog.=
>
> Why would you expect that?  It'd make it impossible to use IN
> with user-defined data types.  In this case, you made an operator
> that is a closer match to the given datatypes (ie, "char = text")
> than the native "text = text" operator, so it used that one.
>
> I've not checked the code, but my recollection is that X IN (Y) just
> resolves to the same equality operator you'd get by writing X = Y.
> There's been some discussion about allowing a schema qualifier to
> be included in the syntax, but nothing's been done about that.
>
>                         regards, tom lane