Re: [HACKERS] Re: [SQL] Column name's length

Philip Warner <pjw(at)rhyme(dot)com(dot)au> writes:
>> Also, this will break pg_dump, which will have no good way to restore
>> the state of a serial sequence object. (CREATE SEQUENCE pg_xxx will
>> fail, no?)

> I know I'm probably out of my depth here, but couldn't pg_dump ignore
> everything with a pg_* prefix?

It does, for the most part. The trouble is that if we rename SERIAL
sequences to pg_xxx, and pg_dump then ignores them, then dump and
reload will fail to restore the next-serial-number state of a SERIAL
column. (Actually, given no other code changes, the serial column
would fail entirely because its underlying sequence wouldn't be
recreated at all. I was pointing out that it's not even *possible*
for pg_dump to restore the sequence's state if the sequence is given
a protected name.)

> As a user with about 20000 blobs to load, the output of a \d is pretty
> cumbersome.

Hmm, I suppose \d ought to ignore xinv relations ...

regards, tom lane