MIME-Version: 1.0
References: <20241031113832.GA3409@wolff.to>
In-Reply-To: <20241031113832.GA3409@wolff.to>
From: Dominique Devienne <ddevienne@gmail.com>
Date: Thu, 31 Oct 2024 13:15:34 +0100
Message-ID: <CAFCRh-9dwBkb73x+ckFCn-SDy=n8Fpg3XcGfFAT+MWOm+jp-Tw@mail.gmail.com>
Subject: Re: COPY documentation with regard to references constraints
To: Bruno Wolff III <bruno@wolff.to>, pgsql-general@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: <https://www.postgresql.org/message-id/CAFCRh-9dwBkb73x%2BckFCn-SDy%3Dn8Fpg3XcGfFAT%2BMWOm%2Bjp-Tw%40mail.gmail.com>
Precedence: bulk

On Thu, Oct 31, 2024 at 1:04=E2=80=AFPM Bruno Wolff III <bruno@wolff.to> wr=
ote:
> I was unable to find any documentation clarifying if using COPY to load a
> table with rows referencing others rows in the same table, whether I
> had to worry about ordering of the input.
> What I found (in 17) is that even if the references constraint wasn't
> set as deferrable, I could have rows for children before their parents
> in the COPY data and things still worked.

AFAIK, DEFERRABLE is for temporary violations crossing statement boundaries=
,
to postpone the enforcement of FKs at COMMIT time of the transaction.

While a single COPY is just one statement, so whether a temporary violation
for self-referential FKs would occur would be implementation-dependent,
and thus "neither logical" nor predictable.

Applies to other statement types too.

But of course, more authoritative answers will come I'm sure. --DD