On 18.03.24 11:02, Aleksander Alekseev wrote:
> Hi,
>
>> Anyway, in order to move this forward, here is an updated patch where
>> the ADD CONSTRAINT ... NOT NULL behavior for domains matches the
>> idempotent behavior of tables. This uses the patch that Jian He posted.
>
> I tested the patch on Raspberry Pi 5 and Intel MacBook and also
> experimented with it. Everything seems to work properly.
>
> Personally I believe new functions such as
> validateDomainNotNullConstraint() and findDomainNotNullConstraint()
> could use a few lines of comments (accepts..., returns..., etc).
Done.
> Also
> I think that the commit message should explicitly say that supporting
> NOT VALID constraints is out of scope of this patch.
Not done. I don't know what NOT VALID has to do with this. This patch
just changes the internal catalog representation, it doesn't claim to
add or change any features. The documentation already accurately states
that NOT VALID is not supported for NOT NULL constraints.
> Except for named nitpicks v4 LGTM.
Committed, thanks.