The Scene Graph

The Scene Graph: a Tree of Nodes

Many simple 3D engines maintain a list of 3D models to render every frame. In these simple engines, one must allocate a 3D model (or load it from disk), and then insert it into the list of models to render. The model is not “visible” to the renderer until it is inserted into the list.

Panda3D is slightly more sophisticated. Instead of maintaining a list of objects to render, it maintains a tree of objects to render. An object is not visible to the renderer until it is inserted into the tree.

The tree consists of objects of class PandaNode. This is actually a superclass for a number of other classes: ModelNode, GeomNode, LightNode, and so forth. Throughout this manual, it is common for us to refer to objects of these classes as simply nodes, and the tree that is being made up by these nodes is known as the scene graph.

There may be many scene graphs in an application. In fact, any tree of nodes in existence is technically a scene graph all on its own. But for the purposes of rendering 3D models, we usually talk about the standard 3D scene graph, at the root of which is a node called render.

Panda3D also creates a scene graph for rendering 2D objects. This is used for putting text, images and GUI elements on top of the 3D scene. This scene graph is positioned under a root called render2d, but this will be explained further in a later section.

What you Need to Know about the Hierarchical Scene Graph

Here are the most important things you need to know about the hierarchical arrangement of the scene graph:

  1. You control where objects go in the tree. When you insert an object into the tree, you specify where to insert it. You can move branches of the tree around. You can make the tree as deep or as shallow as you like.

  2. Positions of objects are specified relative to their parent in the tree. For example, if you have a 3D model of a hat, you might want to specify that it always stays five units above a 3D model of a certain person’s head. Insert the hat as a child of the head, and set the position of the hat to (0,0,5).

  3. When models are arranged in a tree, any rendering attributes you assign to a node will propagate to its children. For example, if you specify that a given node should be rendered with depth fog, then its children will also be rendered with depth fog, unless you explicitly override at the child level.

  4. Panda3D generates bounding boxes for each node in the tree. A good organizational hierarchy can speed frustum and occlusion culling. If the bounding box of an entire branch is outside the frustum, there is no need to examine the children.

Beginners usually choose to make their tree completely flat–everything is inserted immediately beneath the root. This is actually a very good initial design. Eventually, you will find a reason to want to add a little more depth to the hierarchy. But it is wise not to get complicated until you have a clear, specific reason to do so.


Most manipulations of the scene graph are performed using the NodePath class. This is a very small object containing a pointer to a node, plus some administrative information. For now, you can ignore the administrative information; it will be explained in a later section of the manual. It is the intent of the Panda3D designers that you should think of a NodePath as a handle to a node. Any function that creates a node returns a NodePath that refers to the newly-created node.

A NodePath isn’t exactly a pointer to a node; it’s a “handle” to a node. Conceptually, this is almost a distinction without a difference. However, there are certain API functions that expect you to pass in a NodePath, and there are other API functions that expect you to pass in a node pointer. Because of this, although there is little conceptual difference between them, you still need to know that both exist.

You can convert a NodePath into a “regular” pointer at any time by calling nodePath.node(). However, there is no unambiguous way to convert back. That’s important: sometimes you need a NodePath, sometimes you need a node pointer. Because of this, it is recommended that you store NodePaths, not node pointers. When you pass parameters, you should probably pass NodePaths, not node pointers. The callee can always convert the NodePath to a node pointer if it needs to.

NodePath methods and Node methods

There are many methods that you can invoke on NodePaths, which are appropriate for nodes of any type. Specialized node types, like LODNode and Camera (for instance), provide additional methods that are available only for nodes of that type, which you must invoke on the node itself. Here are some assorted examples:

myNodePath.set_pos(x, y, z);

myNodePath.node()->add_switch(1000, 100);
myNodePath.node()->set_center(LPoint3(0, 5, 0));

myNodePath.node()->set_lens(new PerspectiveLens());

Always remember: when you invoke a method of NodePath, you are actually performing an operation on the node to which it points.

In the example above, we call node-methods by first converting the NodePath into a node, and then immediately calling the node-method. This is the recommended style.

Table of Contents