Does dropping a column from a table mess up foreign keys?

Hi--

I'm getting the following error:

ERROR: Relation "accounts" with OID 72496 no longer exists

What I did was to drop a couple of columns using the example Bruce
provides in his book on page 264. Briefly, it's something like:

create table temp as select * from accounts;
drop table accounts;
create table accounts (etc with columns missing);
insert into accounts select <all except dropped cols> from temp;
drop table temp;
grant update,select,insert,delete on accounts to user;

And so on. I've also updated an "accounts_view" based on this table so
that the dropped columns wouldn't appear (the view being defined with an
asterisk field list).

I have lots of tables with "references accounts(id)" in them. Are these
going to be screwed up because of the drop/create above?

I even tried dumping the DB after the changes, then pg_restoring them, but
I get a message something like, Relation ACCOUNTS doesn't exist. I was
thinking that restoring the db in this way would recalculate the OIDs.

If I use pg_restore with --orig-order, I get a failure saying "Accounts"
doesn't exist. If I use pg_restore with --oid-order, I get an error about a
sequence that doesn't exist, or something like that. This tells me that
pg_dump --format=c is really screwed up, or that I fundamentally don't
understand how to backup/restore my database. I used to do:

pgdump > dump.file

and then

cat dump.file | psql database

but that no longer works either, as the pgdump dumps the grant statements
BEFORE the tables they're supposed to grant permissions to.

Is there some command I should do BEFORE the changes due to the fact that
other tables have contraints based on "accounts"? (Now that I think about
it, I'm amazed that I can drop a table with data dependent on it.)

Clues?

Keith

----
Keith Irwin
Development Ecosystem Team
Architecture & Development Services
E-Services Deployment & Operations
Email: keith_irwin(at)non(dot)hp(dot)com
Phone: 541.715.0032