On 2020/07/30 3:33, Justin Pryzby wrote:
> On Wed, Jul 29, 2020 at 03:21:19PM +0900, Michael Paquier wrote:
>> On Wed, Jul 29, 2020 at 01:27:07PM +0900, Masahiko Sawada wrote:
>>> Good catch. The patch looks good to me.
>>
>> While this patch is logically correct. I think that we should try to
>> not increase more the number of queries used to scan pg_class
>> depending on a list of relkinds. For example, did you notice that
>> your new Query_for_list_of_vacuumables becomes the same query as
>> Query_for_list_of_indexables? You can make your patch more simple.
>
> I didn't notice. There's an argument for keeping them separate, but as long as
> there's a #define in between, this is fine, too.
>
> On Wed, Jul 29, 2020 at 08:05:57PM +0900, Michael Paquier wrote:
>> On Wed, Jul 29, 2020 at 06:41:16PM +0900, Masahiko Sawada wrote:
>>> whereas Query_for_list_of_vacuumables should search for:
>>>
>>> RELKIND_RELATION
>>> RELKIND_PARTITIONED_TABLE
>>> RELKIND_MATVIEW
>>> RELKIND_TOASTVALUE
>>
>> FWIW, I don't think that we should make toast relations suggested to
>> the user at all for any command. This comes down to the same point
>> that we don't have pg_toast in search_path, and going down to this
>> level of operations is an expert-level mode, not something we should
>> recommend to the average user in psql IMO.
>
> Right. Tom's response to that suggestion a couple years ago I thought was
> pretty funny (I picture Dr. Claw at his desk using psql tab completion being
> presented with a list of pg_toast.pg_toast_NNNNNN OIDs: "which TOAST table
> should I vacuum next..")
>
> https://www.postgresql.org/message-id/14255.1536781029@sss.pgh.pa.us
> |I don't actually think that's a good idea. It's more likely to clutter
> |peoples' completion lists than offer anything they want. Even if someone
> |actually does want to vacuum a toast table, they are not likely to be
> |entering its name via tab completion; they're going to have identified
> |which table they want via some query, and then they'll be doing something
> |like copy-and-paste out of a query result.
Isn't it better to add the comment explaining why toast tables are
excluded from the tab completion for vacuum while they are vacuumable?
The patch looks good to me except that.
Regards,
--
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION