Re: ALTER TABLE ... IF EXISTS feature?

On Fri, Nov 5, 2010 at 4:48 PM, Daniel Farina <drfarina(at)acm(dot)org> wrote:
> On Fri, Nov 5, 2010 at 1:31 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Daniel Farina <drfarina(at)acm(dot)org> writes:
>>> On Fri, Nov 5, 2010 at 11:04 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>>>> Daniel Farina <drfarina(at)acm(dot)org> writes:
>>>>> pg_dump --clean will successfully and silently wipe out a foreign key
>>>>> right now, should it exist,
>>>>
>>>> No, it will not, because we don't use CASCADE in the drop commands.
>>
>>> I know it does not use CASCADE, but if I understand it correctly,
>>> foreign keys are dropped between tables, and then the tables are
>>> dropped. (effectively a manual cascade)
>>
>> You're missing the point.  The scenario I'm concerned about is:
>>
>>        source database contained table foo
>>
>>        target database contains table foo, and table bar, and
>>        bar has an FK reference to foo
>>
>
> I think that's intended and okay to fail, and would continue to fail
> post-patch, if I understand what I am doing correctly (always
> suspect).
>
> The only condition where this should be emitted is when all the
> dependent objects are going to be dropped anyway.

Dan,

Can you give us a self-contained example of the problem you're talking about?

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company