Public Member Functions
|Constructs an empty render context. |
|rendering_context (bounding_volume::intersection cull_flag, mat4f &modelview)|
|Constructs and initializes a render context. |
|const mat4f &||matrix () const|
|Returns a reference to the modelview matrix. |
|void||matrix (mat4f &modelview)|
|Sets the modelview matrix. |
|Track the results of intersecting node bounds with the view volume. |
|Draw the bounding volumes or not. |
The members could be arguments to the
node::render method; but there may be many arguments and adding an argument requires changing nearly every file in the core.
node::rendershould be cheap. Secondly, while adding a new member to this class is less work than adding a new argument to every
node::rendercall, it still requires a recompile of nearly the whole core. That's because pass-by-value needs to know the exact size of what's being passed. Thirdly, you can't subclass this class. It's concrete. If you pass a subclass it will just get truncated.
Why do it this way? Because it makes writing the
node::render method easier. We don't have to maintain a stack of states, we just use the call stack instead. That means no heap allocation and deallocation, which is very cool. Also, we don't have to pop the stack at the end of the method. Alternatively, I could have used the "initialization is resource acquisition" pattern, but I felt this was cleaner.
Constructs an empty render context.
An empty context should not be passed to VrmlNode::render. This constructor is useful only for debugging and experimentation.
Constructs and initializes a render context.
|cull_flag||the cull flag argument will normally be |
|modelview||The modelview matrix. The transform can be affine, but the rendering code may take advantage of an orthogonal transform if one is passed in.|
|const openvrml::mat4f & openvrml::rendering_context::matrix||(||)||const|
Returns a reference to the modelview matrix.
|void openvrml::rendering_context::matrix||(||mat4f &||modelview||)|
Sets the modelview matrix.
A rendering_context retains a pointer to the passed matrix; it does not make a copy. All memory management is up to the caller. In practice, the passed-in array will generally be a local variable in the
|modelview||the modelview matrix. Must be at least affine, although the render code may optimize for orthogonal transforms.|
Track the results of intersecting node bounds with the view volume.
bounding_volume::inside means that all the last tested bounding volume was completely inside the view volume, so all the contained bounding volumes must also be inside and we can skip further testing.
bounding_volume::partial means that the last test indicated that the bounding volume intersected the view volume, so some of the children may be visible and we must continue testing.
bounding_volume::outside means the last test indicated the bounding volume was completely outside the view volume. However, there's normally no reason to call set with
bounding_volume::outside, since the render method returns immediately. But who knows; it might be useful some day, so it's an allowed value.
Setting the cull flag to
bounding_volume::inside in the
browser at the top of the traversal has the effect of disabling the culling tests. The behavior is undefined if the flag is not one of the allowed values.
Draw the bounding volumes or not.
If true, then the renderer may draw the bounding volumes for each primitive. Or maybe not, so you shouldn't depend on this behavior. It does look kinda cool though.