Making Plugs Dirty And Use Of Compute

[dannybe2]

Hi,

I'm having a few problems with creating a custom mpxLocatorNode, which I think is mainly down to my misunderstanding of how it works.

I have a custom node overriding an mpxLocatorNode with a load of openGL drawing code in it. I want to create a half dozen of these in the scene and connect them up in the hypergraph. Now, if I manually set an attr on the first node, I would like that to trickle through the dg and update all my nodes in the scene using my compute within each node, which then makes all the openGL views change. But this doesn't seem to want to happen unless I explicitly getAttr the attribute that is on the end node far downstream.

How do I trigger an update of all the nodes when setting an attribute on my first node?

Thanks,
Dan

[joojaa]

Nodes do NOT update because upstream nodes change values. They update because downstream nodes need that value in order for them to operate. Thus if your nodes do not have right attribute affects then they will never update. So only those values that affect something maya cosiders important has any way of updating.

[dannybe2]

Right, exactly. So how do I tell Maya that it is important to update the nodes by purely viewing their openGL?

Is there a method I can override such as its boundingBox that will get called on drawing which I can use to make the attributes of the node dirty forcing the whole dg to update?

[joojaa]

Well you define a Draw function draw function needs to update every time (otheriwse nothing is drawn). And then the compute needs to be defined too and that compute node should work basically exactly the way the example cvColorNode that ships with maya devkit ie it should update your internals. Locators are automatically end of the line nodes. That is and only if their attributes affect some of the nodes outputs! Because a affect newrok must be in place for maya to know what to update.

So pay attention to the cvColor::initialize function!

[dannybe2]

Ok, I realise I was approaching this in the wrong way. I have now moved all my logic from the getInternalValue / setInternalValue into the compute function and instead of referencing the internal values when drawing, I pull the information I need from the plugs of the node and that forces the update (though I could probably get away with doing this for just the outputs).

However, my attributeAffects seem to be more complicated than need be, so I'll just describe a psuedo example that illustrates this:

I have a node with three inputs I1, I2, I3 and three outputs A, B, C. The outputs can be accessed as an attribute of the node in maya of course. These three attributes are dependent on each other like this:

A -> B -> C (meaning if you want C, A must be computed first, then B and finally C)

So the ideal way to set up my attributeAffects would be like this:

CODE
attributeAffects(I1, A)
attributeAffects(I2, A)
attributeAffects(I3, A)

attributeAffects(A, B)
attributeAffects(B, C)

This doesn't seem to work though, I have to explicitly associate all of the inputs with all of the outputs, meaning that I have a add a load of new code in order to add an extra input for example. Is there a way of avoiding this?

PS. The devkit is quite extensive and I have used it before in the past, but is there a description of what each plug-in does and the kind of solutions it is demonstrating? Unless I become much more familiar with the library, there doesn't seem to be a simple way of finding the particular plug-in that solves a specific problem.

[joojaa]

Well attribute affects are a function of how complex things you do thats just life of how maya works. Since the plugs work both ways it must be defined as the conduit goes both ways.

Sorry not much time to THINK very deep right now.

Anyway it does not really matter maya does cache solutions so if you mark something dirty and you do not update it then it does not matter much.

Also the idea of compute is to checks what OUTPUT was accessed and calculate the data that you need to do for that output so if c is accessed then calculate A also otherwise leave it be. Since when you calcualte you markit clean there will be no extra call for A calculate when c is asked for. If the other way then just chek to see if A is dirty then compute it otherwise use the data you have clean (because its allready computed.)

QUOTE(dan_b @ 09/21/08, 05:53 PM) [snapback]292063[/snapback]
PS. The devkit is quite extensive and I have used it before in the past, but is there a description of what each plug-in does and the kind of solutions it is demonstrating?

Yes in the manual:

Developer Resources -> API Guide -> Example Plug-ins under that you have mainly:
Overview of example plug-ins
Example C++ plug-in descriptions

But theres a few other categories there also

[dannybe2]

That's great.

Thanks very much for your help.