Mutating table and row state - this was unexpected

Question

So, I learned in a class about 3 years ago to expect the following
SQL> create table p (pk number primary key);
Table created.
SQL> create table c (fk number references p(pk));
Table created.
SQL> create or replace trigger t_insrt
  2  before insert on p
  3  for each row
  4  begin
  5   insert into c values (:new.pk);
  6  end;
  7  /
Trigger created.
SQL> insert into p values (1);
insert into p values (1)
ERROR at line 1:
ORA-02291: integrity constraint (FORBESC.SYS_C00169150) violated - parent key
not found
ORA-06512: at "FORBESC.T_INSRT", line 2
ORA-04088: error during execution of trigger 'FORBESC.T_INSRT'and so it led me to think that replicating ON MODIFY PARENT - MODIFY CHILD functionality wouldn't work in a BEFORE ROW trigger, but it does
SQL> drop trigger t_insrt;
Trigger dropped.
SQL> create or replace trigger p_updt
  2  before update on p
  3  for each row
  4  begin
  5   update c
  6   set fk = :new.pk
  7   where fk = :old.pk;
  8  end;
  9  /
Trigger created.
SQL> insert into p values (1);
1 row created.
SQL> insert into c values (1);
1 row created.
SQL> select * from c;
        FK
         1
SQL> update p
  2  set pk = 2
  3  where pk = 1;
1 row updated.
SQL> select * from c;
        FK
         2Why would the first scenario fail while the second succeeds? The update seems prone to a parent record also not existing, at least not by the BEFORE ROW trigger.
---=Chuck

Answer 1

< mutating table and row state >
BTW, you don't seem to have run into the mutating table error though 2 other threads today are also about it. You have a constraint violation, a different thing entirely.
I believe the second scenario works because you're neatly avoiding the error of the first. The error "ORA-02291: integrity constraint (FORBESC.SYS_C00169150) violated" means that on insert Oracle is looking up the value you're trying to insert, not finding it, and raising an error. With the before trigger you are taking the assigned value from the insert, updating the parent to it, so that on actual insert when the check happens the value is there due to the update.
I'm not convinced this is a good idea because any on-the-fly approach to data entry needs to be examined carefully.