enum CacheEnabled { OFF, ON, AUTO }

virtual SoType getTypeId (void) const

SoSeparator (void)

SoSeparator (const int nchildren)

virtual void doAction (SoAction *action)

virtual void GLRender (SoGLRenderAction *action)

virtual void GLRenderBelowPath (SoGLRenderAction *action)

virtual void GLRenderInPath (SoGLRenderAction *action)

virtual void GLRenderOffPath (SoGLRenderAction *action)

virtual void callback (SoCallbackAction *action)

virtual void getBoundingBox (SoGetBoundingBoxAction *action)

virtual void getMatrix (SoGetMatrixAction *action)

virtual void handleEvent (SoHandleEventAction *action)

virtual void rayPick (SoRayPickAction *action)

virtual void search (SoSearchAction *action)

virtual void getPrimitiveCount (SoGetPrimitiveCountAction *action)

virtual void audioRender (SoAudioRenderAction *action)

virtual void notify (SoNotList *nl)

virtual SbBool affectsState (void) const

static SoType getClassTypeId (void)

static void initClass (void)

static void setNumRenderCaches (const int howmany)

static int getNumRenderCaches (void)

SoSFEnum renderCaching

SoSFEnum boundingBoxCaching

SoSFEnum renderCulling

SoSFEnum pickCulling

virtual const SoFieldData * getFieldData (void) const

virtual ~SoSeparator ()

virtual SbBool cullTest (SoState *state)

virtual SbBool cullTest (SoGLRenderAction *action, int &cullresults)

virtual SbBool readInstance (SoInput *in, unsigned short flags)

static const SoFieldData ** getFieldDataPtr (void)

Enumeration of flags for the fields deciding which optimalizations to do in SoSeparator nodes. There are two types of settings available: caching policies or culling policies. See doumentation of fields.

Enumerator:

OFF

No caching.

ON

Always try to cache state.

AUTO

Use heuristics to try to figure out the optimal caching policy.

Default constructor.

Constructor.

The argument should be the approximate number of children which is expected to be inserted below this node. The number need not be exact, as it is only used as a hint for better memory resource allocation.

Destructor. Frees resources used to implement caches.

This static method returns the SoType object associated with objects of this class.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation, SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

Returns the type identification of an object derived from a class inheriting SoBase. This is used for run-time type checking and 'downward' casting.

Usage example:

  void foo(SoNode * node)
  {
    if (node->getTypeId() == SoFile::getClassTypeId()) {
      SoFile * filenode = (SoFile *)node;  // safe downward cast, knows the type
    }
    else if (node->getTypeId().isOfType(SoGroup::getClassTypeId())) {
      SoGroup * group = (SoGroup *)node;  // safe downward cast, knows the type
    }
  }

For application programmers wanting to extend the library with new nodes, engines, nodekits, draggers or others: this method needs to be overridden in all subclasses. This is typically done as part of setting up the full type system for extension classes, which is usually accomplished by using the pre-defined macros available through for instance Inventor/nodes/SoSubNode.h (SO_NODE_INIT_CLASS and SO_NODE_CONSTRUCTOR for node classes), Inventor/engines/SoSubEngine.h (for engine classes) and so on.

For more information on writing Coin extensions, see the class documentation of the toplevel superclasses for the various class groups.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation, SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

This API member is considered internal to the library, as it is not likely to be of interest to the application programmer.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation, SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

Returns a pointer to the class-wide field data storage object for this instance. If no fields are present, returns NULL.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation, SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

Sets up initialization for data common to all instances of this class, like submitting necessary information to the Coin type system.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation, SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

This function performs the typical operation of a node for any action.

Reimplemented from SoGroup.

Action method for the SoGLRenderAction.

This is called during rendering traversals. Nodes influencing the rendering state in any way or who wants to throw geometry primitives at OpenGL overrides this method.

Reimplemented from SoGroup.

Reimplemented in SoAnnotation.

SGI Open Inventor v2.1 obsoleted support for SoGLRenderAction::addMethod(). Instead, GLRender() might be called directly, and to optimize traversal, the SoSeparator node calls GLRenderBelowPath whenever the path code is BELOW_PATH or NO_PATH (path code is guaranteed not to change). To be compatible with SGI's Inventor (and thereby also TGS') we have chosen to follow their implementation in this respect.

SoSeparator::GLRenderBelowPath() do not traverse its children using SoChildList::traverse(), but calls GLRenderBelowPath() directly for all its children.

Reimplemented from SoNode.

Reimplemented in SoAnnotation, SoExtSelection, and SoLocateHighlight.

Implements the SoAction::IN_PATH traversal method for the rendering action.

Reimplemented from SoNode.

Reimplemented in SoAnnotation, and SoLocateHighlight.

Implements the SoAction::OFF_PATH traversal method for the rendering action.

Reimplemented from SoNode.

Reimplemented in SoAnnotation.

Action method for SoCallbackAction.

Simply updates the state according to how the node behaves for the render action, so the application programmer can use the SoCallbackAction for extracting information about the scene graph.

Reimplemented from SoGroup.

Action method for the SoGetBoundingBoxAction.

Calculates bounding box and center coordinates for node and modifies the values of the action to encompass the bounding box for this node and to shift the center point for the scene more towards the one for this node.

Nodes influencing how geometry nodes calculates their bounding box also overrides this method to change the relevant state variables.

Reimplemented from SoGroup.

Action method for SoGetMatrixAction.

Updates action by accumulating with the transformation matrix of this node (if any).

Reimplemented from SoGroup.

Action method for SoHandleEventAction.

Inspects the event data from action, and processes it if it is something which this node should react to.

Nodes influencing relevant state variables for how event handling is done also overrides this method.

Reimplemented from SoGroup.

Reimplemented in SoExtSelection, SoLocateHighlight, SoSelection, and SoWWWAnchor.

Action method for SoRayPickAction.

Checks the ray specification of the action and tests for intersection with the data of the node.

Nodes influencing relevant state variables for how picking is done also overrides this method.

Reimplemented from SoNode.

Action method for SoSearchAction.

Compares the search criteria from the action to see if this node is a match. Searching is done by matching up all criteria set up in the SoSearchAction -- if any of the requested criteria is a miss, the search is not deemed successful for the node.

See also: SoSearchAction

Reimplemented from SoGroup.

Action method for the SoGetPrimitiveCountAction.

Calculates the number of triangle, line segment and point primitives for the node and adds these to the counters of the action.

Nodes influencing how geometry nodes calculates their primitive count also overrides this method to change the relevant state variables.

Reimplemented from SoGroup.

Action method for SoAudioRenderAction.

Does common processing for SoAudioRenderAction action instances.

Reimplemented from SoGroup.

Notifies all auditors for this instance when changes are made.

Reimplemented from SoNode.

Set the maximum number of caches that SoSeparator nodes may allocate for render caching.

This is a global value which will be used for all SoSeparator nodes, but the value indicate the maximum number per SoSeparator node.

More caches might give better performance, but will use more memory. The built-in default value is 2.

The value can also be changed globally by setting the host system's environment variable IV_SEPARATOR_MAX_CACHES to the wanted number. This is primarily meant as an aid during debugging, to make it easy to turn off rendercaching completely (by setting 'IV_SEPARATOR_MAX_CACHES=0') without having to change any application code.

Returns maximum number of caches SoSeparator nodes are allowed to use for render caching.

See also: setNumRenderCaches()

Returns TRUE if the node could have any effect on the state during traversal.

If it returns FALSE, no data in the traversal-state will change from the pre-traversal state to the post-traversal state. The SoSeparator node will for instance return FALSE, as it pushes and pops the state before and after traversal of its children. All SoShape nodes will also return FALSE, as just pushing out geometry data to the rendering engine won't affect the actual rendering state.

The default method returns TRUE, on a 'better safe than sorry' philosophy.

Reimplemented from SoNode.

Internal method which do view frustum culling. For now, view frustum culling is performed if the renderCulling field is AUTO or ON, and the bounding box cache is valid.

Returns TRUE if this separator is outside view frustum, FALSE if inside.

This is an internal Open Inventor method. We've implemented view frustum culling in a different manner. Let us know if you need this function, and we'll consider implementing it.

This method is mainly intended for internal use during file import operations.

It reads a definition of an instance from the input stream in. The input stream state points to the start of a serialized / persistant representation of an instance of this class type.

TRUE or FALSE is returned, depending on if the instantiation and configuration of the new object of this class type went ok or not. The import process should be robust and handle corrupted input streams by returning FALSE.

flags is used internally during binary import when reading user extension nodes, group nodes or engines.

Reimplemented from SoGroup.

Policy for caching of rendering instructions for faster execution. This will typically use the OpenGL displaylist mechanism.

Default value is SoSeparator::AUTO.

If you know that some parts of your scene will never change, rendering might happen faster if you explicitly set this field to SoSeparator::ON. If you on the other hand know that parts of the scene will change a lot (like for every redraw), it will be beneficial to set this field to SoSeparator::OFF for the top-level separator node of this (sub)graph.

Usually the default setting of AUTO will handle any scene very well. The advantages that can be had from setting SoSeparator::renderCaching to ON are:

•

If you positively know that the geometry under the SoSeparator is static, you get the cache set up right away.

Otherwise, the code in Coin will do a bit of testing and decide by some heuristics whether or not to enable it. That will make the rendering be a tiny bit slower right after start-up than with renderCaching set to ON.

(The slow-down should hardly be noticable, though, so we don't advice application programmers to do this.)

•

For many of the shape nodes that can contain many basic primitives, like e.g. SoFaceSet, SoIndexedFaceSet, SoLineSet, etc etc, there is an internal threshold for how many primitives a node can contain before we don't do caching when SoSeparator::renderCaching is set to AUTO.

The reason we do this is because OpenGL renderlists can potentially suck up a lot of memory resources on the graphics card.

But if you know that it will be advantageous on your particular platform, you can override this by setting SoSeparator::renderCaching equal to ON.

(We don't advice application programmers to do this either. A better solution in these cases would simply be to get in touch with SIM and describe the platform and the problem, and we could integrate a proper fix into Coin.)

There are good reasons for setting renderCaching to OFF, like when you know the geometry will be changing a lot. Still, Coin should work fairly well without even this optimization. (If renderCaching is AUTO over a sub-graph with changing geometry or other cache smashing nodes, the caching heuristics will stop the SoSeparator node from trying to make caches -- at least after a few tries has been made and failed.)

The short story about how auto-caching works is as follows:

•

For vertex-based shapes with fewer than 100 triangles and where the geometry is detected to be fairly static, caching is enabled.

•

For shapes with more than 1000 trangles, it is disabled, to avoid spending too much of the on-board graphics card's memory resources.

•

For shapes with between 100 and 1000 shapes, displaylist caching will be turned on if our heuristics decides that the geometry can be considered static.

The maximum threshold (of 1000) is higher when doing remote rendering (as when rendering from one X11-based system to another).

Disabling the displaylist caching takes precedence over enabling, so if you have an SoSeparator with a shape with more than 1000 triangles and a shape with fewer than 100 triangles, caching will be disabled for the SoSeparator.

It's possible to tune the limits using some environment variables:

•

COIN_AUTOCACHE_LOCAL_MIN can be used to change the enable-caching limit, while COIN_AUTOCACHE_LOCAL_MAX controls the disable-caching limit.

•

The corresponding variables for remote rendering are COIN_AUTOCACHE_REMOTE_MIN and COIN_AUTOCACHE_REMOTE_MAX.

Policy for caching bounding box calculations. Default value is SoSeparator::AUTO.

See also documentation for SoSeparator::renderCaching.

Policy for doing viewport culling during rendering traversals. Default value is SoSeparator::AUTO.

When the render culling is turned off for Coin, it will be left to be done for the underlying immediate mode rendering library. This will often be faster than doing culling from within Coin, so be careful to monitor the change in execution speed if setting this field to SoSeparator::ON.

See also documentation for SoSeparator::renderCaching.

Policy for doing viewport culling during pick traversals. Default value is SoSeparator::AUTO.

See documentation for SoSeparator::renderCulling.