BUG #17805: Unexpected results when bounds of FOR loop use non-integer numbers

From: PG Bug reporting form <noreply(at)postgresql(dot)org>
To: pgsql-bugs(at)lists(dot)postgresql(dot)org
Cc: jiangshan(dot)liu(at)tju(dot)edu(dot)cn
Subject: BUG #17805: Unexpected results when bounds of FOR loop use non-integer numbers
Date: 2023-02-22 14:23:11
Message-ID: 17805-274c2f3650367138@postgresql.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

The following bug has been logged on the website:

Bug reference: 17805
Logged by: Jiangshan Liu
Email address: jiangshan(dot)liu(at)tju(dot)edu(dot)cn
PostgreSQL version: 15.2
Operating system: Ubuntu 18.04
Description:

I got unexpected results when I ran this PL/pgSQL function:

CREATE FUNCTION test() RETURNS void AS $$
BEGIN
FOR i IN 1 .. 0.5*5 LOOP
raise notice '%', i;
END LOOP;
END;
$$ LANGUAGE plpgsql;
SELECT * FROM test();

The result of the execution is:
NOTICE: 1
NOTICE: 2
NOTICE: 3

I think this result is unexpected because the last round of loop variable is
beyond the specified upper bound of the loop. Is this a bug in a FOR loop?
I noticed that the documentation describes it this way:
The two expressions giving the lower and upper bound of the range are
evaluated once when entering the loop[1].
However, the documentation does not mention how the upper and lower bounds
were evaluated. Based on the analysis of the execution, it appears that the
FOR loop seems to implicitly convert the NUMERIC type upper bound value
(2.5) to INT type value (3) by rounding.
But this seems to cause incomprehensible results. If it is unsafe to take
values outside the upper and lower bounds of the loop, then this will lead
to unintended consequences. Is this implicit type conversion necessary? I
think there are two better ways to do it.
One is to report errors directly to the user. Pascal and Ada languages
support FOR loops of integer variant, and they throw errors to the user when
there are non-integer variant in the upper and lower bounds of the loop.
The second is to use an implicit type conversion from integer to
non-integer, followed by a comparison, to ensure that the loop does not
exceed the upper and lower bounds, just like other procedural languages
(e.g. C, Python, etc.).

[1]
https://www.postgresql.org/docs/current/plpgsql-control-structures.html#PLPGSQL-CONTROL-STRUCTURES-LOOPS:~:text=of%20the%20range-,are%20evaluated%20once,-when%20entering%20the

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message Justin Pryzby 2023-02-22 14:53:41 Re: Unlimited memory consumption with long-lived connection
Previous Message Dean Rasheed 2023-02-22 12:29:36 Re: BUG #17803: Rule "ALSO INSERT ... SELECT ..." fails to substitute default values