Thread: BUG #6761: unexpected behaviour of 'now'::timestamp

The following bug has been logged on the website:

Bug reference:      6761
Logged by:          Bert Thomas
Email address:      bert@brothom.nl
PostgreSQL version: 9.1.3
Operating system:   Linux
Description:=20=20=20=20=20=20=20=20

Hi,

To reproduce what I mean, consider this function:

CREATE FUNCTION testbug() RETURNS character varying
    LANGUAGE plpgsql
    AS $$declare
  l_ts timestamp(0);

begin
  l_ts :=3D 'now'::timestamp(0);
  return l_ts::varchar;
end
$$;

If a program invokes this function multiple times on a single connection,
only the first time the correct date and time is produced. All other
invocations return the exact same value as the first invocation.

Changing the function to this fixes the problem:

CREATE FUNCTION testbug() RETURNS character varying
    LANGUAGE plpgsql
    AS $$declare
  l_ts timestamp(0);
  l_nu varchar;

begin
  l_nu :=3D 'now';
  l_ts :=3D l_nu::timestamp(0);
  return l_ts::varchar;
end
$$;

Appearently the expression is re-evaluated every time in this case, whilst
in the first case it is only evaluated once as the constant 'now' could not
change obviously. I'm not sure if this is a bug or not, but at least it is
suprising behaviour. To me it looks like a bad form of optimization.

Kind regards,
Bert Thomas
BroThom

Hello

this is not bug - it is consequence of plan cache

http://postgres.cz/wiki/Automatic_execution_plan_caching_in_PL/pgSQL

please, use CURRENT_TIMESTAMP instead - using 'now'::timestamp is
deprecated due this issue

Regards

Pavel

2012/7/25  <bert@brothom.nl>:
> The following bug has been logged on the website:
>
> Bug reference:      6761
> Logged by:          Bert Thomas
> Email address:      bert@brothom.nl
> PostgreSQL version: 9.1.3
> Operating system:   Linux
> Description:
>
> Hi,
>
> To reproduce what I mean, consider this function:
>
> CREATE FUNCTION testbug() RETURNS character varying
>     LANGUAGE plpgsql
>     AS $$declare
>   l_ts timestamp(0);
>
> begin
>   l_ts := 'now'::timestamp(0);
>   return l_ts::varchar;
> end
> $$;
>
> If a program invokes this function multiple times on a single connection,
> only the first time the correct date and time is produced. All other
> invocations return the exact same value as the first invocation.
>
> Changing the function to this fixes the problem:
>
> CREATE FUNCTION testbug() RETURNS character varying
>     LANGUAGE plpgsql
>     AS $$declare
>   l_ts timestamp(0);
>   l_nu varchar;
>
> begin
>   l_nu := 'now';
>   l_ts := l_nu::timestamp(0);
>   return l_ts::varchar;
> end
> $$;
>
> Appearently the expression is re-evaluated every time in this case, whilst
> in the first case it is only evaluated once as the constant 'now' could not
> change obviously. I'm not sure if this is a bug or not, but at least it is
> suprising behaviour. To me it looks like a bad form of optimization.
>
> Kind regards,
> Bert Thomas
> BroThom
>
>
> --
> Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-bugs

Pavel Stehule <pavel.stehule@gmail.com> writes:
> this is not bug - it is consequence of plan cache

FWIW, there is documentation of this issue near the end of
http://www.postgresql.org/docs/9.1/static/plpgsql-implementation.html#PLPGSQL-PLAN-CACHING

This exact case isn't covered in the examples, but the point is that the
expression 'now'::timestamp will get folded to a timestamp constant during
planning, and then not replanned later.  As Pavel says, it's a lot safer
to use one of the variants of the now() function.

            regards, tom lane

Thanks Tom and Pavel!