> david.g.johns...@gmail.com wrote: > > The convention on these lists is to inline or bottom post (and to trim the > reply to just the pertinent parts).
Just for completeness, I expected this test to run without error. (I tried it in PG Version 14.1). create table t(k int primary key, v text not null); insert into t(k, v) values (1, 'one'), (2, 'two'); select k, v from t order by k; start transaction; delete from t where k = 1; insert into t(k, v) values(1, 'new one'); commit; select k, v from t order by k; Indeed it did run without error. And I saw the results that I expected.
