On 2026-04-05 Su 12:35 PM, Jelte Fennema-Nio wrote:
> On Sun, 5 Apr 2026 at 17:06, Andrew Dunstan <andrew@dunslane.net> wrote:
>> Pushed. I have moved the remaining get_*_ddl items to PG20-1
> +1 on having this feature in general. But I'm not sure I understand
> why it needs the whole bespoke string-based option parsing in the
> first commit. Why not use named arguments for this, i.e. have the
> usage syntax be:
>
> SELECT * FROM pg_get_role_ddl('regress_role_ddl_test3', pretty => true);
>
> Instead of the current:
>
> SELECT * FROM pg_get_role_ddl('regress_role_ddl_test3', 'pretty', 'true');
There was quite a deal of discussion around this mechanism. See Euler's
review at [1] and follow-up at [2] for the original discussion of the
VARIADIC option-parsing design and the use cases it was meant to
address. I'm prepared to revisit it is there's a strong consensus on the
point.
cheers
andrew
[1]
https://www.postgresql.org/message-id/4e60bcae-8222-4e1f-8e5b-d73b59c93304%40app.fastmail.com
[2]
https://www.postgresql.org/message-id/4c695e76-5ab7-449f-8060-76518dd41468%40app.fastmail.com
--
Andrew Dunstan
EDB: https://www.enterprisedb.com
