I think what diotalevi is trying to acheive is the reverse of B's next method

$ perl584 -MO=Concise -e'print "a", $b+1;'
9  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 1 -e:1) v ->3
8     <@> print vK ->9
3        <0> pushmark s ->4
4        <$> const(PV "a") s ->5
7        <2> add[t1] sK/2 ->8
-           <1> ex-rv2sv sK/1 ->6
5              <$> gvsv(*b) s ->6
6           <$> const(IV 1) s ->7

$ perl584 -MO=Concise,-exec -e'print "a", $b+1;'
1  <0> enter 
2  <;> nextstate(main 1 -e:1) v
3  <0> pushmark s
4  <$> const(PV "a") s
5  <$> gvsv(*b) s
6  <$> const(IV 1) s
7  <2> add[t1] sK/2
8  <@> print vK
9  <@> leave[1 ref] vKP/REFC
[download]

Internally Perl doesn't store pointers to previous siblings, not does it store a pointer to the parent. So my initial reaction is that the OP's approach is correct. However, since the B modules provide a parent() method when internally there is no such pointer, it's posssible that B:;* can do other clever stuff too. I've never looked at it myself.

Dave.

That just goes to show how little I really know about B. I thought if you had stored a child op that you could use next on it to get the next sibling. Hence my comment about reversing it to create a previous method. Looks like I'll need to reread that stuff again. :)