On 6/12/23 14:46, Andrew Dunstan wrote:
>
> On 2023-06-11 Su 16:15, Joel Jacobson wrote:
>> On Sun, Jun 11, 2023, at 17:03, Tomas Vondra wrote:
>>> On 6/11/23 12:26, Joel Jacobson wrote:
>>>> I think there are some good arguments that speaks in favour of including it in core:
>> ...
>>> I agree with all of that, but ...
>>>
>>> It's just past feature freeze, so the earliest release this could appear
>>> in is 17, about 15 months away.
>>>
>>> Once stuff gets added to core, it's tied to the release cycle, so no new
>>> features in between.
>>>
>>> Presumably people would like to use the extension in the release they
>>> already use, without backporting.
>>>
>>> Postgres is extensible for a reason, exactly so that we don't need to
>>> have everything in core.
>> Interesting, I've never thought about this one before:
>> What if something is deemed to be fundamental and therefore qualify for core inclusion,
>> and at the same time is suitable to be made an extension.
>> Would it be possible to ship such extension as pre-installed?
>>
>> What was the json/jsonb story, was it ever an extension before
>> being included in core?
>
>
> No, and the difficulty is that an in-core type and associated functions
> will have different oids, so migrating from one to the other would be at
> best painful.
>
> So it's a kind of now or never decision. I think extensions are
> excellent for specialized types. But I don't regard a set type in that
> light.
>
Perhaps. So you're proposing to have this as a regular built-in type?
It's hard for me to judge how popular this feature would be, but I guess
people often use arrays while they actually want set semantics ...
If we do that, I wonder if we could do something similar to arrays, with
the polymorphism and SQL grammar support. Does SQL have any concept of
sets (or arrays, for that matter) as data types?
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company