Re: ALTER command reworks

2012/10/2 Alvaro Herrera <alvherre@2ndquadrant.com>:
> Excerpts from Alvaro Herrera's message of vie sep 28 18:17:32 -0300 2012:
>> Excerpts from Kohei KaiGai's message of lun sep 10 08:08:32 -0300 2012:
>>
>> > As attached, I split off the original one into three portions; for set-schema,
>> > set-owner and rename-to. Please apply them in order of patch filename.
>>
>> Hmm, in the first patch, it seems to me we can simplify
>> AlterObjectNamespace's signature: instead of passing all the details of
>> the object class (cache Ids and attribute numbers and so on), just do
>>
>> AlterObjectNamespace(catalog-containing-object, objectId, newNamespaceOid)
>
> Like in the attached reworked version, in which I renamed the function
> to AlterObjectNamespace_internal because the other name seemed a bit
> wrong in the fact of the existing AlterObjectNamespace_oid.
>
> I also made the ALTER FUNCTION case go through
> AlterObjectNamespace_internal; it seems pointless to have a separate
> code path to go through when the generic one does just fine (also, this
> makes functions identical to collations in implementation).  That's one
> less hook point for sepgsql, I guess.
>
Thanks for your reviewing, and sorry for my late response.

I definitely agree with your solution. The reason why my original patch
had separate code path for function and collation was they took
additional elements (such as argument-list of function) to check
duplicate names. So, I think it is a wise idea to invoke the common
code after name duplication checks.

Best regards,
-- 
KaiGai Kohei <kaigai@kaigai.gr.jp>