Re: relkind check in info.c

On 03/19/2014 01:17 AM, Michael Paquier wrote:
> On Wed, Mar 19, 2014 at 8:07 AM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> On Wed, Mar 19, 2014 at 3:39 AM, Alvaro Herrera
>> <alvherre(at)2ndquadrant(dot)com> wrote:
>>> Hi,
>>>
>>> I see that PGAPI_Tables() has a relkind check that only seems to include
>>> tables and views. Does it make sense to have it consider materialized
>>> views as well?
>>>
>>> I don't know anything about psqlodbc -- I'm just going by a question in
>>> the spanish list:
>>> http://www.postgresql.org/message-id/53287159.7040009@gmail.com
>>
>> Yeah, it would make sense to list it by adding 'm' to the relkind
>> queries. I am noting as well that we should list it when invocating
>> SQL_ALL_TABLE_TYPES. A patch to do that is simple, and not worth an
>> additional regression test IMO, just to avoid the pain to have
>> version-related regression tests...
> Just adding that a similar thing should be done with foreign tables.

We need to decide to what to return in the table_type column for them.
"FOREIGN TABLE" and "MATERIALIZED VIEW" would be natural choices.

But it would be good to return the same string that other DBMS's return
for those things. What does SQL Server's driver return for materialized
views? I'm not aware of any other DBMS that implements foreign tables
per the spec, but e.g DB2 has "remote tables" and Oracle has "database
links", which are similar features. I wonder what table type do they return.

- Heikki