Pointer invoke method pitfall?

Question

I am writing a implementation about 2-3-4 tree. node structure is as below

type Node struct {
    items  []int
    childs []*Node
    parent *Node
}

I am confusing about the code below. In my opinon, the two part is doing the same thing. However, one of these is wrong.

• right code

cur = cur.parent
cur._insertNode(upTo, rn)
upTo, rn = cur._splitNode()

• wrong code

cur.parent._insertNode(upTo, rn)
upTo, rn = cur.parent._splitNode()
cur = cur.parent

Can anyone tells me what is the difference?

What I expect is explaintion about the question. Is it a Go pointer method pitfall? Or compiler bug?

Answer 1

Let C be the node that cur originally points to, let A be the node that is originally the parent of C, and suppose that the call to _insertNode inserts a new node B between A and C; so, we go from this:

A
|
C

(plus other nodes, not relevant to my point) to this:

A
|
B
|
C

(plus other nodes, still not relevant to my point).

The important thing to note is that before the call to _insertNode, the parent of C is A; and after the call to _insertNode, the parent of C is B.

With that in mind, here's your "right code", plus comments explaining what it's doing:

// initially, cur = C

// set cur = A:
cur = cur.parent

// insert B between A and C:
cur._insertNode(upTo, rn)

// cur is still A

// split A:
upTo, rn = cur._splitNode()

And here's your "wrong code", plus comments explaining what it's doing:

// initially, cur = C

// insert B between A and C:
cur.parent._insertNode(upTo, rn)

// cur.parent is now B

// split B:
upTo, rn = cur.parent._splitNode()

// set cur = B:
cur = cur.parent

Do you see?