[x3d-public] review of various X3D statement implementations forX3DOM

[Andreas Plesch]

Yeah, this is currently what is recommended for x3dom. However, this
is not very satisfying since DEF/USE is popular and the issue keeps
coming up.

Overall, the issue is really x3dom specific due to the way it does
picking. Most browser would start from the TouchSensor, find all
influenced geometries and then generate events for the TouchSensor
analysed. In this case, USE and DEF nodes among the influenced
geometries do not need to be distinguished. x3dom reverts the process.
It starts from the geometries by using the GPU to render a special
picking image with id information (which is very efficient especially
if you have multiple TouchSensors), which then any TouchSensor (or dom
event dispatcher) uses to determine if events need to generated. Since
all geometries are rendered there is no way to distinguish between
influenced geometries and other geometries if the ids are the same.
The fix equivalent to the first picking method would be to render a
special picking image with only influenced geometries, separately for
each Touchsensor or dom event listener which seems wasteful. Another
fix would be to use an actually unique id which distinguishes a USE
node from its DEF node.

-Andreas

On Wed, Dec 20, 2017 at 2:21 PM, John Carlson <yottzumm at gmail.com> wrote:
> Andreas, I think if you want to distinguish USE nodes with picking, you
> would use a separate node.
>
>
>
> John
>
>
>
> Sent from Mail for Windows 10
>
>
>
> From: Andreas Plesch
> Sent: Wednesday, December 20, 2017 2:18 PM
> To: Michalis Kamburelis
> Cc: X3D Graphics public mailing list
> Subject: Re: [x3d-public] review of various X3D statement implementations
> forX3DOM
>
>
>
> Hi Michalis,
>
>
>
> yes, that is how DEF/USE should be implemented. The point I am trying
>
> to understand is if there is absolutely no distinction between the
>
> original DEF node and a USE node which references the DEF node, how
>
> can the USE node be identified, for example during picking ? The USE
>
> node still needs to have at least a distinct ID, right ?
>
>
>
> A conceptual scene graph structure may look like this:
>
>
>
> group1
>
>   node 1: DEF "original"
>
> group2
>
>   node 2: USE "original"
>
>
>
> There are still two nodes. node 2 is not a copy of node 1 but is still
>
> a separate node which happens to reference node 1.
>
>
>
> Or
>
>
>
> group1
>
> node 1: DEF "orginal"
>
> group2
>
> node 1 (inserted a second time during scene graph construction)
>
>
>
> Closer to spec. but makes it hard (impossible?) to distinguish the two
>
> node 1 instances from each other.
>
>
>
> So I was thinking something like an additional ID may have to exist:
>
>
>
> 1. group1
>
> 2.  node 1: DEF "orginal"
>
> 3. group2
>
> 4.  node 1 (inserted during scene graph construction)
>
>
>
> ?
>
>
>
> -Andreas
>
>
>
>
>
>
>
>
>
> On Wed, Dec 20, 2017 at 11:03 AM, Michalis Kamburelis
>
> <michalis.kambi at gmail.com> wrote:
>
>> 2017-12-20 15:51 GMT+01:00 Andreas Plesch <andreasplesch at gmail.com>:
>
>>> (DEF/USE in x3dom: There are a several x3dom issues about picking a
>
>>> USE Shape. Since x3dom literally places the referenced DEF Shape in
>
>>> the scene graph, the picking actually reports the node of the
>
>>> referenced DEF Shape when the USE Shape is picked. This is because
>
>>> x3dom actually does not keep a distinct representation of the USE node
>
>>> at all. Except for the fact that the DEF node now has two parents.
>
>>> This may be too literal an interpretation of the DEF/USE concept, and
>
>>> I wonder if in the spec. there is a section which explains how the
>
>>> node which has a USE reference is still a distinct node. I suspect
>
>>> that keeping such a distinct node around is how most x3d browsers work
>
>>> ?
>
>>
>
>> This DEF / USE (in x3dom) treatment is required by the X3D
>
>> specification, and it (should) be the same in all VRML / X3D browsers.
>
>> I know it is the same in view3dscene and Castle Game Engine. When we
>
>> encounter a "USE Xxx" then internally we *do not* make a copy of the
>
>> "Xxx" node, instead we make an additional link to the "Xxx" node that
>
>> in effect has 2 parents. This implementation is most natural and
>
>> practically required by the X3D specification:
>
>>
>
>>
>> http://www.web3d.org/documents/specifications/19775-1/V3.2/Part01/concepts.html#DEFL_USESemantics
>
>>
>
>> """
>
>> The USE statement does not create a copy of the node. Instead, the
>
>> same node is inserted into the scene graph a second time, resulting in
>
>> the node having multiple parents (see 4.3.5 Transformation hierarchy,
>
>> for restrictions on self-referential nodes).
>
>> """
>
>>
>
>> And that's a very good concept, in my opinion: it allows to use DEF /
>
>> USE mechanism for various optimizations. Since you have only one copy
>
>> of the node, you should also be able to have only one copy of various
>
>> internal things (like OpenGL VBO data) associated e.g. with Shape that
>
>> is USEd multiple times. This is really powerful, you can construct a
>
>> large scene by USEing the same shapes multiple times, and it will
>
>> still render fast and not eat too much memory:)
>
>>
>
>> In contrast, expanding the PROTOtypes makes a copy on the contents.
>
>>
>
>> Regards,
>
>> Michalis
>
>
>
>
>
>
>
> --
>
> Andreas Plesch
>
> Waltham, MA 02453
>
>
>
> _______________________________________________
>
> x3d-public mailing list
>
> x3d-public at web3d.org
>
> http://web3d.org/mailman/listinfo/x3d-public_web3d.org
>
>



-- 
Andreas Plesch
Waltham, MA 02453