Re: vacuum_truncate configuration parameter and isset_offset

В письме от среда, 26 марта 2025 г. 18:41:13 MSK пользователь Robert Haas
написал:

> What I'm upset about is that it feels to me like Nikolay is trying to
> win the argument by yelling.

I do not think it is fair. When votes are against you (and I did not ask
anybody else's help for it) it turned out, that I am yelling, and thus voting
is not valid.

I do not demand enum, you can create new reloption data type that can be true
false and unset. I even would agree with isset_offset, it you manage to make
code consistent. I doubt anybody manages do it, but may be I am wrong. You did
not even try, you just insist on keeping code inconsistent, because nobody
care. Nikolay cares, but he is yelling while caring, so let's set him aside.

The only thing I am asking for: please keep code consistent. Better keep
reloption core the way it was, but if it is not possible, then keep it
consistent then.

Enum is the cheapest way to achieve the goal Nathan want, and keep code
consistent. Cheapest, not the best.

When the dust settles down I will try to invent boolean-based data type, that
would not be ugly as this enum, and even may be suitable for
vacuum_index_cleanup option. I see there is a demand for this. But this will
need carefull thinking and development. And starting from ugly Enum with
boolean values would be much easy then from boolean with isset_offset.

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