One more time about COUNT(*)

Question

We have had related discussions, but looks like I missed something.
Anyway I do not have a good explanation on the following:
large table has unique index on field x
SELECT COUNT(*) takes some time.
If make field x as primary key, it takes much less to execute (about 1/4).
(9i)
link will be good.
This is not "urgent"

Answer 1

You could run the below to demonstrate the situation "SomeoneElse" just described:
drop table t;
create table t as select rownum rn, rpad('x',200) cola from dual connect by rownum <= 10000;
create unique index i on t(rn);
explain plan for select count(*) from t;
select * from table(dbms_xplan.display);
alter table t modify rn number not null;
explain plan for select count(*) from t;
select * from table(dbms_xplan.display);