Postgres Pro
Downloads
Home   >   mailing lists

Thread: .sequence_name != ?

.sequence_name != ?

From
"Ed L."
Date:
When a sequence is created in 7.3.2, it appears you get a new table for each
sequence object.  Is it ever possible for the sequence_name in a sequence
relation not to match the name of the relation itself?

For example, suppose I create a table:
CREATE TABLE t1(id serial);

A new relation called 't1_id_seq' is created where
t1_id_seq.sequence_name = 't1_id_seq'

Is that always true?

Ed

Re: .sequence_name != ?

From
Tom Lane
Date:
"Ed L." <pgsql@bluepolka.net> writes:
> When a sequence is created in 7.3.2, it appears you get a new table for each 
> sequence object.  Is it ever possible for the sequence_name in a sequence 
> relation not to match the name of the relation itself?

ALTER TABLE RENAME on a sequence doesn't update the sequence_name.

I think someone looked at doing that update, but we concluded it was too
messy (mainly because ALTER RENAME is transactional but updates to a
sequence tuple aren't).

In general I'd counsel that you should ignore the sequence_name field
anyway.  It's vestigial.
        regards, tom lane

Re: .sequence_name != ?

From
"Ed L."
Date:
On Friday April 4 2003 10:24, Tom Lane wrote:
> "Ed L." <pgsql@bluepolka.net> writes:
> > When a sequence is created in 7.3.2, it appears you get a new table for
> > each sequence object.  Is it ever possible for the sequence_name in a
> > sequence relation not to match the name of the relation itself?
>
> In general I'd counsel that you should ignore the sequence_name field
> anyway.  It's vestigial.

A related question:  Is there a single generalized SQL query which can yield
the set of (sequence_name, last_value) pairs for all sequence objects?  The
fact that each sequence is its own relation seems to block that, and the
query constructed from grabbing sequence names from pg_class gets quite
long for more than just a few sequence objects...

Ed