Обсуждение: setseed accepts bad seeds

Why doesn't setseed complain when given a seed value outside of its
expected range?  The documentation claims that 0 -> 1 should be used,
but shouldn't it actually be -1 -> 1 to get the full range of the seed?
If passed a value outside this range you always get the same value which
isn't going to produce a good random number stream.

Kris Jurka

Kris Jurka <books@ejurka.com> writes:
> Why doesn't setseed complain when given a seed value outside of its
> expected range?

Why should it complain?  The use of the value is totally unspecified
anyway.

            regards, tom lane

On Wed, 11 Apr 2007, Tom Lane wrote:

> Kris Jurka <books@ejurka.com> writes:
>> Why doesn't setseed complain when given a seed value outside of its
>> expected range?
>
> Why should it complain?  The use of the value is totally unspecified
> anyway.
>

Because the user is likely using it incorrectly.  I'm not sure what you
mean by totally unspecified.  The documentation[1] states:

set seed for subsequent random() calls (value between 0 and 1.0)

When a user calls setseed(5), setseed(500), or setseed(-500) they get the
same seed value each time which is surely not what they intended.  At
minimum I think it should raise a warning.

Also I think that documentation should be corrected to indicate that
vaules -1 to 1 are the correct seed value range or it should it should map
0->1 to the entire seed space, not just half of it as is currently done.
The decision of which change to make is unclear because it's a change to
either the call signature or to the generated values for a given user
supplied seed.

Kris Jurka

[1] http://www.postgresql.org/docs/8.2/static/functions-math.html#FUNCTIONS-MATH-FUNC-TABLE

Kris Jurka <books@ejurka.com> writes:
> On Wed, 11 Apr 2007, Tom Lane wrote:
>> Kris Jurka <books@ejurka.com> writes:
>>> Why doesn't setseed complain when given a seed value outside of its
>>> expected range?
>>
>> Why should it complain?  The use of the value is totally unspecified
>> anyway.

> Because the user is likely using it incorrectly.

It's not really possible to use it "incorrectly", AFAICS.  Any value you
might pass to it will result in a specific new seed value.  Nowhere is
there any guarantee of what the mapping is, and it's obviously
impossible to guarantee that the mapping is one-to-one, so any user
assumptions about what a specific seed value might "mean" seem broken
regardless.

            regards, tom lane

On Wed, 11 Apr 2007, Tom Lane wrote:

> It's not really possible to use it "incorrectly", AFAICS.  Any value you
> might pass to it will result in a specific new seed value.  Nowhere is
> there any guarantee of what the mapping is, and it's obviously
> impossible to guarantee that the mapping is one-to-one, so any user
> assumptions about what a specific seed value might "mean" seem broken
> regardless.
>

Then please consider this patch which checks the range and maps the
provided value to the entire seed space.

Kris Jurka

This has been saved for the 8.4 release:

    http://momjian.postgresql.org/cgi-bin/pgpatches_hold

---------------------------------------------------------------------------

Kris Jurka wrote:
>
>
> On Wed, 11 Apr 2007, Tom Lane wrote:
>
> > It's not really possible to use it "incorrectly", AFAICS.  Any value you
> > might pass to it will result in a specific new seed value.  Nowhere is
> > there any guarantee of what the mapping is, and it's obviously
> > impossible to guarantee that the mapping is one-to-one, so any user
> > assumptions about what a specific seed value might "mean" seem broken
> > regardless.
> >
>
> Then please consider this patch which checks the range and maps the
> provided value to the entire seed space.
>
> Kris Jurka
Content-Description:

[ Attachment, skipping... ]

>
> ---------------------------(end of broadcast)---------------------------
> TIP 7: You can help support the PostgreSQL project by donating at
>
>                 http://www.postgresql.org/about/donate

--
  Bruce Momjian  <bruce@momjian.us>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +

Kris Jurka <books@ejurka.com> writes:
> On Wed, 11 Apr 2007, Tom Lane wrote:
>> It's not really possible to use it "incorrectly", AFAICS.  Any value you
>> might pass to it will result in a specific new seed value.  Nowhere is
>> there any guarantee of what the mapping is, and it's obviously
>> impossible to guarantee that the mapping is one-to-one, so any user
>> assumptions about what a specific seed value might "mean" seem broken
>> regardless.

> Then please consider this patch which checks the range and maps the
> provided value to the entire seed space.

I'm still not very happy about this.  It'll change the behavior of
existing applications, in the service of no goal that I consider
convincing.

I'd be inclined to leave the mapping alone and just insert a warning
(or hard error) for inputs outside the range -1 to 1.

            regards, tom lane

On Mon, 10 Mar 2008, Tom Lane wrote:

> I'd be inclined to leave the mapping alone and just insert a warning
> (or hard error) for inputs outside the range -1 to 1.
>

Here's a patch that errors out for out of range values.

Kris Jurka

Kris Jurka <books@ejurka.com> writes:
> On Mon, 10 Mar 2008, Tom Lane wrote:
>> I'd be inclined to leave the mapping alone and just insert a warning
>> (or hard error) for inputs outside the range -1 to 1.

> Here's a patch that errors out for out of range values.

Applied, thanks.

            regards, tom lane