On Tue, Oct 28, 2025 at 09:08:36AM +0800, Chao Li wrote:
>> On Oct 28, 2025, at 04:29, Nathan Bossart <nathandbossart@gmail.com> wrote:
>> Here is what I have staged for commit. I ended up simplifying the patch a
>> bit. In particular, I thought better of the question mark business. It
>> looks like we ordinarily just skip values that can't be found, and an empty
>> search_path will appear as "" (two double-quotes), so you can still
>> distinguish empty versus not-available.
>
> +1, I like this idea.
Actually, I take it back. The following command to empty the search_path
will cause %S to be replaced with nothing, in which case you can't
distinguish empty versus not-reported.
SELECT pg_catalog.set_config('search_path', '', false);
So, I've added the question mark back.
--
nathan