me nefcanto <sn.1361@gmail.com> writes:
> The point is, that there is already an `on_error ignore` clause there. This
> means that somewhere there is a try/catch per row. If I'm wrong, please let
> me know.
You are wrong. See
https://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=d9f7f5d32https://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=9e2d87011
and a ton of related commits.
It's possible that the specific case of unique-index violations
could be handled in a similar style. But "catch any error whatever"
is simply not going to happen here, because a subtransaction per
row is not practical.
regards, tom lane