On Fri, Jun 3, 2011 at 10:59 AM, Tom Lane <t...@sss.pgh.pa.us> wrote: > "Daniel Cristian Cruz" <danielcrist...@gmail.com> writes: >> CREATE TABLE a ( >> id_a serial primary key, >> v text >> ); >> CREATE TABLE b ( >> id_b serial primary key, >> id_a integer REFERENCES a (id_a), >> v text >> ); >> CREATE TABLE c ( >> id_c serial primary key, >> id_b integer references b (id_b), >> v text >> ); > >> CREATE VIEW cba AS >> SELECT c.v AS vc, b.v AS vb, a.v AS va >> FROM c >> JOIN b USING (id_b) >> JOIN a USING (id_a); > >> ALTER TABLE c ADD id_a integer; > >> [ view definition now fails due to multiple "id_a" columns ] > > I'm inclined to write this off as "so don't do that". There's nothing > that pg_dump can do to make this work: it has to use the USING syntax > for the join, and that doesn't offer any way to qualify the column name > on just one side. The only possible fix would be to try to make ALTER > TABLE reject the addition of the conflicting column name to "c" in the > first place. That doesn't seem very practical; it would require ALTER > TABLE to do a tremendous amount of analysis, and exclusively lock all > the dependent views, and then lock all the other tables used in the > views, and so on. > > Personally my advice is to avoid USING: it wasn't one of the SQL > committee's better ideas.
I don't understand why we can't just translate the USING into some equivalent construct that doesn't involve USING. I proposed that a while ago and you shot it down, but I didn't find the reasoning very compelling. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company -- Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-bugs
