Обсуждение: BUG #18965: Issue with Short-Circuit Evaluation in Boolean Expressions

The following bug has been logged on the website:

Bug reference:      18965
Logged by:          Todd Brandys
Email address:      brandystodd@gmail.com
PostgreSQL version: 17.5
Operating system:   Linux
Description:

In the circumstance where a function evaluation is performed within Boolean
expression, the evaluation seems to continue past a function returning a
TRUE value.  Here is a very boiled down version of my code, but it results
in the same issue.  In the first SELECT statement, I get the expected
result, a single row with a TRUE value.  In the other two SELECT statements,
an EXCEPTION is thrown, which is unexpected.
CREATE OR REPLACE FUNCTION raise(
    IN          i_msg           text
)
RETURNS text AS $$
BEGIN
    RAISE EXCEPTION '%', i_msg;
    RETURN ''::text;
END;$$
LANGUAGE PLPGSQL
IMMUTABLE STRICT;
CREATE OR REPLACE FUNCTION test( )
RETURNS boolean AS $$
    SELECT      TRUE; $$
LANGUAGE SQL
IMMUTABLE STRICT;
SELECT      TRUE OR public.raise('this exception should not be
raised')::boolean;
SELECT      pg_catalog.PG_HAS_ROLE('postgres', 'postgres', 'member')
            OR public.raise('this exception should not be raised')::boolean;
SELECT      public.test() OR public.raise('this exception should not be
raised')::boolean;
Again, I am using PostgreSQL 17.5, compiled from source.  I have no
extensions installed in the database.  Here is the configure script I used
to build the distribution:
export PYTHON=/var/lib/pgsql/venv/bin/python3
./configure \
    --prefix=/usr/local/installed/postgresql-17.5 \
    --enable-atomics --enable-largefile --with-llvm --with-perl
--with-readline --with-python \
    --with-uuid=ossp --with-zlib --with-ssl=openssl --with-libxml
--with-libxslt

PG Bug reporting form <noreply@postgresql.org> writes:
> CREATE OR REPLACE FUNCTION raise(
>     IN          i_msg           text
> )
> RETURNS text AS $$
> BEGIN
>     RAISE EXCEPTION '%', i_msg;
>     RETURN ''::text;
> END;$$
> LANGUAGE PLPGSQL
> IMMUTABLE STRICT;

I think the fundamental problem you're having is that you marked
this function IMMUTABLE, which gives the planner license to
pre-evaluate it.  It had better be VOLATILE to discourage advance
evaluation.

https://www.postgresql.org/docs/current/xfunc-volatility.html

            regards, tom lane