On Thu, Nov 27, 2014 at 11:43 PM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On Thu, Nov 27, 2014 at 8:18 PM, Sawada Masahiko <sawada.mshk@gmail.com> wrote:
>> On Thu, Nov 27, 2014 at 1:07 PM, Michael Paquier
>> <michael.paquier@gmail.com> wrote:
>>> On Thu, Nov 27, 2014 at 12:55 AM, Sawada Masahiko <sawada.mshk@gmail.com> wrote:
>>>> +1 to define new something object type and remove do_user and do_system.
>>>> But if we add OBJECT_SYSTEM to ObjectType data type,
>>>> system catalogs are OBJECT_SYSTEM as well as OBJECT_TABLE.
>>>> It's a bit redundant?
>>> Yes, kind of. That's a superset of a type of relations, aka a set of
>>> catalog tables. If you find something cleaner to propose, feel free.
>>
>> I thought we can add new struct like ReindexObjectType which has
>> REINDEX_OBJECT_TABLE,
>> REINDEX_OBJECT_SYSTEM and so on. It's similar to GRANT syntax.
> Check.
>
>>>>> Another thing, ReindexDatabaseOrSchema should be renamed to ReindexObject.
>>>>> So, I think that we need to think a bit more here. We are not far from
>>>>> smth that could be committed, so marking as "Waiting on Author" for
>>>>> now. Thoughts?
>>>>
>>>> Is the table also kind of "object"?
>>> Sorry, I am not sure I follow you here. Indexes and tables have
>>> already their relkind set in ReindexStmt, and I think that we're fine
>>> to continue letting them go in their own reindex code path for now.
>>
>> It was not enough, sorry.
>> I mean that there is already ReindexTable() function.
>> if we renamed ReindexObject, I would feel uncomfortable feeling.
>> Because table is also kind of "object".
> Check.
>
> If you get that done, I'll have an extra look at it and then let's
> have a committer look at it.
Attached patch is latest version.
I added new enum values like REINDEX_OBJECT_XXX,
and changed ReindexObject() function.
Also ACL problem is fixed for this version.
Regards,
-------
Sawada Masahiko
