Re: ALTER TYPE 2: skip already-provable no-work rewrites

Robert Haas <robertmhaas@gmail.com>

From: Robert Haas <robertmhaas@gmail.com>

To: Noah Misch <noah@leadboat.com>

Cc: pgsql-hackers@postgresql.org

Date: 2011-02-06T17:13:05Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Change ALTER TABLE SET WITHOUT OIDS to rewrite the whole table to physically
- 6d1e36185208 8.4.0 cited

On Sun, Feb 6, 2011 at 8:58 AM, Noah Misch <noah@leadboat.com> wrote:
> On Sun, Feb 06, 2011 at 08:40:44AM -0500, Noah Misch wrote:
>> On Sun, Feb 06, 2011 at 07:54:52AM -0500, Robert Haas wrote:
>> > Yeeeeeeah, that's actually a little ugly.   It's actually a domain
>> > over a composite type, not a composite type proper, IIUC. Better
>> > ideas?
>>
>> There are no domains over composite types.  get_rels_with_domain() finds all
>> relations having columns of the (scalar) domain type.  It then calls
>> find_composite_type_dependencies to identify uses of the composite types
>> discovered in the previous step.
>>
>> Honestly, RELKIND_COMPOSITE_TYPE is a reasonable choice despite the technical
>> mismatch.  One more-correct approach would be to have two arguments, a catalog
>> OID (pg_type or pg_class, currently) and a relkind, 0 when the catalog OID !=
>> pg_class.  Might be an improvement, albeit a minor one.
>
> Scratch that.  How about classid and objid arguments, passing them to
> getObjectionDescription() internally?  We already do something very similar in
> ATExecAlterColumnType for a related case.

That's not quite so good for translators, I think.

Another option is that we could just say "relation" (table, foreign
table, etc...) or "type".  We use the word relation as a more generic
version of table in a few other places.

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