I thought folks wanted them to fail if they conflicted so that they
could know for sure how to derive such names definitively. Is that
accurate?
---------------------------------------------------------------------------
Christopher Kings-Lynne wrote:
> This patch makes all forms of autogenerated unique, primary key and sequence
> names avoid collisions. (Well, as much as you can do without "locking" your
> chosen name).
>
> This addresses a long standing annoyance I experience whenever I rename a
> table to table_old and then try to create a table with the same name. It
> also addresses a concern raised at linux.conf.au.
>
> I suggest that someone inspect the patch to determine if my "overloading" of
> CreateIndexName is appropriate (for sequences, etc.)
>
> All regression tests pass. The attached SQL script will run fine under the
> patch, whereas before it would have failed miserably.
>
> Chris
>
[ Attachment, skipping... ]
[ Attachment, skipping... ]
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073