Currently, the parser builds a tree whose root node is a or , depending on whether the document defines a directed graph or not.
The first child is just a text with the name of the graph, and his siblings are, either or , in the order in which they appear in the document.
Each contains a set of children whose names are the name of the 's attributes, and themselves containing just a text node with the value of each one of the attributes.
On the other hand, a is labeled with the origin of the relationship, and would be a tree node composed of

• −−
  or
  −>
  , depending on whether the document defines a directed graph or not, and
• the name of the
  node
  which appears as the target of the relationship.