Re: alter + preserving dependencies

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

Le 6 mai 08 =E0 19:44, Tom Lane a =E9crit :

> Andrew Dunstan <andrew@[EMAIL PROTECTED]
> writes:
>> Josh Berkus wrote:
>>> I don't follow you.  I can currently add a column, without breaking
>>> either foriegn keys or inheritance.  What's the problem?
>
>> not for a view at least.
>
> Yeah, the restrictions on replacing a view definition date from before
> we had any fancy ALTER TABLE stuff.  They could probably use a re-=20
> visit.

Could we add some other VIEWs features while revisiting?
We had some question on #postgresqlfr about updatable views and the=20=20
rewrite rule system, which I could chat about with Jan Wieck on=20=20
#postgresql too. The problem we had was related to DEFAULT versus NULL=20=
=20
handling from an insert or update RULE and how to avoid having an=20=20
absent column rewritten as NULL instead of DEFAULT.

Could we consider ALTER VIEW ALTER COLUMN ... SET DEFAULT ...;?

Bonus question: why is the rewriter unable to distinguish whether NULL=20=
=20
comes from the user or comes from the column was not part of the
user=20=20
query?

Regards,
- --
dim
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Darwin)

iEYEARECAAYFAkghQRIACgkQlBXRlnbh1bnWAQCeNF4RJwWLcGmmPPE8eUGn3/Yi
6NEAn1YOQ3Bz7L+tD01rQqqeyNt7djwA
=3DMXD8
-----END PGP SIGNATURE-----

--=20
Sent via pgsql-hackers mailing list (pgsql-hackers@[EMAIL PROTECTED]
)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers