| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Andrew Dunstan <andrew(at)dunslane(dot)net> |
| Cc: | pgsql-hackers(at)postgresql(dot)org, Jonathan Gardner <jgardner(at)jonathangardner(dot)net> |
| Subject: | Re: ALTER TABLE ... TO ... to change related names |
| Date: | 2003-08-30 16:38:06 |
| Message-ID: | 3889.1062261486@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
> There is no guarantee that a given sequence is used only for one column
> in one table, as I understand it. So renaming it could screw you up badly.
Yeah, I would recommend having a discussion about the details of the
proposed behavior before you start coding, not after.
It'd probably be reasonable to rename only those sequences that are
connected to the target table/column by internal dependencies --- this
indicates that they were created by a SERIAL column definition and not
by manual operations. IIRC, pg_dump already uses this same cue to
decide whether to dump the sequence definition separately or say SERIAL.
It'd still be possible for someone's schema to break, if they made a
sequence via SERIAL and then referred to it by name in defaults for
other columns. But I think we could say they were being unreasonably
intimate with implementation details in that case.
Similar rules need to be agreed to about when to rename indexes.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Dennis Björklund | 2003-08-30 16:52:41 | Re: ALTER TABLE ... TO ... to change related names |
| Previous Message | Bruce Momjian | 2003-08-30 16:33:42 | Re: FE/BE Protocol - Specific version |