Re: Moving sequences to another schema

I'm still playing around with the ALTER OBJECT SET SCHEMA stuff. I managed
to alter indexes and constraints as well for tables, but with SERIAL
sequences there is one little problem:

When altering a sequence created by a SERIAL column type (i do this by
examining pg_depend to avoid moving any other sequences that are
'foreign'), i need to recreate the default expression for the SERIAL column
(stored in pg_attrdef.adbin). Is there an API to do that, or do i have to
recreate the executable expression tree from scratch? Or am i missing
something completely...

TIA

--

Bernd

On Tue, Jun 28, 2005 at 01:43:27AM +0200, Bernd Helmle wrote:

> When altering a sequence created by a SERIAL column type (i do this by
> examining pg_depend to avoid moving any other sequences that are
> 'foreign'), i need to recreate the default expression for the SERIAL column
> (stored in pg_attrdef.adbin). Is there an API to do that, or do i have to
> recreate the executable expression tree from scratch? Or am i missing
> something completely...

I think this is done by AddRelationRawConstraints. You'd have to get
the parsetree of the default expression. I think you could get that by
applying raw_parser() to pg_attrdef.adsrc.

--
Alvaro Herrera (<alvherre[a]surnet.cl>)
"The Postgresql hackers have what I call a "NASA space shot" mentality.
Quite refreshing in a world of "weekend drag racer" developers."
(Scott Marlowe)

> When altering a sequence created by a SERIAL column type (i do this by
> examining pg_depend to avoid moving any other sequences that are
> 'foreign'), i need to recreate the default expression for the SERIAL
> column (stored in pg_attrdef.adbin). Is there an API to do that, or do i
> have to recreate the executable expression tree from scratch? Or am i
> missing something completely...

Does ALTER TABLE/RENAME code help you? You can rename sequences with
that...

Chris

Alvaro Herrera <alvherre(at)surnet(dot)cl> writes:
> I think this is done by AddRelationRawConstraints. You'd have to get
> the parsetree of the default expression. I think you could get that by
> applying raw_parser() to pg_attrdef.adsrc.

Not adsrc --- that's not trustworthy.

In practice I think you could just assume you know what the default
expression ought to be, and store a new one without looking at the old.

regards, tom lane

--On Dienstag, Juni 28, 2005 09:38:56 +0800 Christopher Kings-Lynne
<chriskl(at)familyhealth(dot)com(dot)au> wrote:

> Does ALTER TABLE/RENAME code help you? You can rename sequences with
> that...

Hmm, that doesn't cover pg_attrdef.adbin. I think the best way is to create
the default expressions from scratch, as Tom already mentioned.

--

Bernd

--On Dienstag, Juni 28, 2005 02:01:33 -0400 Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
wrote:

> Not adsrc --- that's not trustworthy.
>

Yes, that's documented in the docs, too.

> In practice I think you could just assume you know what the default
> expression ought to be, and store a new one without looking at the old.

I think i'll go for that (need to figure out how to do that first, but it
shouldn't be so hard). Whats the least that should go to -patches for
feature release on 1 July (only to know, if i can hold timeline or not).??

--

Bernd

Bernd Helmle <mailings(at)oopsware(dot)de> writes:
> Whats the least that should go to -patches for
> feature release on 1 July (only to know, if i can hold timeline or not).??

Something reasonably complete, working, credible. We'll cut you slack
on documentation changes and regression tests, and if it has a bug or
three that's what beta testing is for; but it has to at least look like
it will work.

regards, tom lane

--On Dienstag, Juni 28, 2005 01:43:27 +0200 Bernd Helmle
<mailings(at)oopsware(dot)de> wrote:

> When altering a sequence created by a SERIAL column type (i do this by
> examining pg_depend to avoid moving any other sequences that are
> 'foreign'), i need to recreate the default expression for the SERIAL
> column (stored in pg_attrdef.adbin).

I currently recognized that a SERIAL column doesn't only create an implicit
sequence, it creates an implicit composite type with the same name, too. I
think this is the same for CREATE SEQUENCE?

--

Bernd

On Jun 30, 2005, at 8:13 PM, Bernd Helmle wrote:

> I currently recognized that a SERIAL column doesn't only create an
> implicit sequence, it creates an implicit composite type with the
> same name, too. I think this is the same for CREATE SEQUENCE?

Sequences are just special types of tables. All tables have a
corresponding composite type, so sequences do as well.

CREATE TABLE foo (foo_id serial); is a shorthand for

CREATE SEQUENCE foo_foo_id_seq;
CREATE TABLE foo (foo_id integer default nextval
('foo_foo_id_seq'::text));

Using SERIAL implies CREATE SEQUENCE.

Does this help?

Michael Glaesemann
grzm myrealbox com