By default, every "from" object added to a CollisionTraverser will test for collisions with every other CollisionNode in the scene graph, and will not test for collisions with visible geometry. For simple applications, this is sufficient, but often you will need more control.
This control is provided with the collide masks. Every CollisionNode has two collide masks: a "from" mask, which is used when the CollisionNode is acting as a "from" object (i.e. it has been added to a CollisionTraverser), and an "into" mask, which is used when the node is acting as an "into" object (i.e. it is in the scene graph, and a from object is considering it for collisions).
In addition, visible geometry nodes--that is, GeomNodes--also have an "into" mask, so that visible geometry can serve as an "into" object also. (However, only a CollisionNode can serve as a "from" object.)
Before the solids in a "from" CollisionNode are tested for collisions with another CollisionNode or with a GeomNode, the collide masks are compared. Specifically, the "from" mask of the from object, and the "into" mask of the into object, are ANDed together. If the result is not zero--meaning the two masks have at least one bit in common--then the collision test is attempted; otherwise, the two objects are ignored.
The collide masks are represented using a BitMask32 object, which is really just a 32-bit integer with some additional methods for getting and setting particular bits.
You can only set the from collide mask on a collision node, and you must set it directly on the node itself, not on the NodePath:
However, the into collide mask may be set on the NodePath, for convenience; this recursively modifies the into collide mask for all the nodes at the given NodePath level and below.
nodePath.setCollideMask(newMask, bitsToChange, nodeType)
The parameter newMask specifies the new mask to apply. The remaining parameters are optional; if they are omitted, then every node at nodePath level and below is assigned newMask as the new into collide mask. However, if bitsToChange is specified, it represents the set of bits that are to be changed from the original; bits that are 0 in bitsToChange will not be modified at each node level. If nodeType is specified, it should be a TypeHandle that represents the type of node that will be modified, e.g.
CollisionNode.getClassType() to affect only CollisionNodes.
This sets the into collide mask of nodePath, and all children of nodePath, to the value 0x10, regardless of the value each node had before.
This replaces the lower 8 bits of nodePath and all of its children with the value 0x04, leaving the upper 24 bits of each node unchanged.
The default value for both from and into collide masks for a new CollisionNode can be retrieved by
CollisionNode.getDefaultCollideMask(), and the default into collide mask for a new GeomNode is
GeomNode.getDefaultCollideMask(). Note that you can create a CollisionNode that collides with visible geometry by doing something like this:
NodePath.getCollideMask() function returns a union of all the collide masks for itself and its children. Since the
NodePath.setCollideMask() function is called recursively on its children, the following code can have a profound effect, even though it looks like it's doing nothing:
The above code actually calculates the collide mask for its children, and sets all of its children to that same collide mask, wiping out what was there before.
If you need to have only entities with a certain collision mask to be able to collide with a model, it is helpful to open the model's egg file and see where the collisions are enabled (see Egg_Syntax). Then you would set the collide mask for only that child node, using
NodePath.find() (see NodePath). For example, to create a box into only "ralph" can collide: