Re: BUG #7920: Sequence rename leave stale value for sequence_name

On 2013-03-06 09:27:55 -0500, Tom Lane wrote:
> Andres Freund <andres(at)2ndquadrant(dot)com> writes:
> > I don't find this particularly suprising. Nothing looks at that field in
> > sequences, there imo is no point on having the name inside at all.
>
> Yeah, and we really can't update the name there because there is no
> provision for transactional updates of sequence tuples.

True.

> > I personally don't see any way to nicely fix that. We can add code to
> > also change the contents, but currently thats generic code. Or we could
> > just remove the column in the next release?
>
> This has been discussed before, and the general opinion has been to
> leave things alone until we get around to doing a wholesale refactoring
> of sequence support. There has been talk for example of merging all
> sequences into one catalog, instead of the current very wasteful
> technique of having a whole relation to store (in essence) one counter.
> That would probably break existing code that tries to select from a
> sequence, but at least there would be objective benefits from it.
> Removing the sequence_name column alone would also break existing code,
> for ... um ... not much.

The only argument I see is reduced chance of people making errors. Code
that actually uses sequence_name is broken.

If we had something like columns that are computed on output, we could
use that. What we could do is invent a new pseudo-column type like
tableoid that renders as text..

In the end it doesn't seem worth bothering.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services