Re: vacuum_truncate configuration parameter and isset_offset

В письме от понедельник, 24 марта 2025 г. 22:11:19 MSK пользователь Robert
Haas написал:

> I think that the answer here might be that Nikolay doesn't like this
> because it interacts badly with his "New [relation] options engine,"
There is no problem with adding isset_offset into "New [relation] options
engine", not a bit deal. But it will break all internal logic of reloption.
Both current version and my patch

> And if I'm wrong in thinking that, then I would like a detailed
> technical argument explaining exactly why.

You can't have both isset_offset flag and default_value. Either one or another.
You can't have them both, it will make whole design inconsistent. We already
have default_value. Even for boolean.
If you want custom behaviour when option is not set, you set unreachable
default value. This is the way it is done for _all_ options that has custom
behavior for unset option. Even those that a technically boolean. Why
vacuum_truncate should do another way? Either do it the way it have been done
before, or suggest total redisign for all oprions that has custom unset
behaviour via unreachable defaults. This will be consistent.

--
Nikolay Shaplov aka Nataraj
Fuzzing Engineer at Postgres Professional
Matrix IM: @dhyan:nataraj.su