Solved Make editable an object generated by GetVirtualObjects()

Hi,

your code does look a bit confusing to me.

1. If you invoke self.SetOptimizeCache(True) in ObjectData.Init() then you do not have to handle your cache manually in ObjectData.GetVirtualObjects() (GVO).

2. Only letting your GVO logic execute if the BaseObject node attached to your ObjectData has no children and then grabing the first object of the scene as an input is really weird.

3. If you do some file I/O, you should not do it in GVO if possible. Also the following snippet is a crash waiting to happen (when your first object in the scene is a MyObject object).

preset = doc.GetFirstObject()
obj = preset.GetClone()

4. You do modify the scene graph and invoke events in GVO. You are not allowed to do that.

5. You also use GVO against its purpose since you never do return any objects constructed in your GVO. You only return the cache when your object node isn't flagged dirty. But that does not mean anything since you never return any objects to populate that cache. In most cases your GVO will return None, which will be read by c4d as that you have indicated an memory error. If you want to return nothing you have to return a null object. So there is nothing to convert in your ObjectData.

6. Think of your ObjectData as a null object. In GVO you have to return the object tree that c4d is supposed to put under that null. These children objects will be hidden from the user. When the user then invokes "current state to object" c4d will remove that null and return the hidden object tree.

Cheers
zipit

@zipit
Hi,

Here is an example of the exact method that I had used with the GVO and I never get a crash on the application.

import c4d, os
from c4d import documents, plugins, bitmaps

PLUGIN_ID = 100000000

class MyObject(c4d.plugins.ObjectData):
    
    def Init(self, op):
            self.SetOptimizeCache(True)
            return True

    def GetVirtualObjects(self, op, hierarchyhelp):

            doc = documents.GetActiveDocument()
            
            if not op.GetDown():     
                # Merge object
                preset = os.path.join(os.path.dirname(__file__), "res", "my_scene.c4d")                       
                load = c4d.documents.MergeDocument(doc, preset, c4d.SCENEFILTER_OBJECTS )
                if load is False:
                    return True
        
                # Search object // I have put the plugin id in the object name to ensure that there is no other object with same.
                preset = doc.SearchObject("my_object[100000000]")
                preset.InsertUnder(op)
           
                
                preset.ChangeNBit(c4d.NBIT_OHIDE, c4d.NBITCONTROL_SET)
                for obj in preset.GetChildren():
                    try:
                        obj.ChangeNBit(c4d.NBIT_OHIDE, c4d.NBITCONTROL_SET)
                        obj.GetDown().ChangeNBit(c4d.NBIT_OHIDE, c4d.NBITCONTROL_SET)
                        if obj.GetDown().GetDown():
                             obj.GetDown().GetDown().ChangeNBit(c4d.NBIT_OHIDE, c4d.NBITCONTROL_SET)

                        if obj.GetDown().GetNext():
                             obj.GetDown().GetNext().ChangeNBit(c4d.NBIT_OHIDE, c4d.NBITCONTROL_SET)
                    except :
                        continue
                
                # Unfold in manager
                for bit in xrange(c4d.NBIT_OM1_FOLD, c4d.NBIT_OM4_FOLD + 1):
                    op.ChangeNBit(bit, c4d.NBITCONTROL_SET)            
                
                op.SetBit(c4d.BIT_ACTIVE)

                # Update the document
                c4d.EventAdd()  
 
            # Here I add the modifications to the returned object (preset) and also to his children objects.
            obj = op.GetDown()

            return op.GetCache(hierarchyhelp)

if __name__ == "__main__":
    icon = bitmaps.BaseBitmap()
    icon.InitWith(os.path.join(os.path.dirname(__file__), 'res', 'icon.tif'))
    plugins.RegisterObjectPlugin(PLUGIN_ID, 'My Object', MyObject, '', c4d.OBJECT_GENERATOR, icon)

Note:
The parent object ("my_object[100000000]") has a python tag that contain the majority of my python code.

Because my object is too complex (it contain multiples sub-objects) and if I reload it at each change on op parameters, over a period of time when the preset object is reloading the object is disappeared from the scene.

And to make editable my object I just added a button "Make Editable" on my description resource.

Hi,

The quoted lines in point 3 are a problem because when your first object is a MyObject, you will end up with an infinite recursion, because this object will try to build its caches when it is added to the scene (either in the illegal direct scene graph modification way of yours or by the correct way by returning that MyObject as a result of your GVO). Building cashes will mean that it will try invoke its GVO which then is the start of the infinite recursion. But this is only a somewhat minor issue with your code.

There are two major problems with both your code snippets.

1. You modify the scene graph and invoke events in a threaded context (see point 4 in my last posting).

2. You 'misuse' GVO as you basically treat it like a script or CommandData plugin (see points 5 and 6 in my last posting).

Disregarding the threaded context of a method and the mentioned limitations can lead to corrupted objects and files or crashes of the main application. I was probably a bit unclear about the importance of the points in my first posting, I just did wrote them down as I red your code. The important points you should focus on are 4, 5, and 6. You will have to change your approach.

If you are unclear on how to do that, check the example plugins on github and then explain here to us what you are trying to do. Because I frankly, have no idea, what your code is supposed to do on a higher level of meaning :)

Cheers
zipit

Hi,
I'm actually developing a new version of HDRI Studio Generator Tool, is a complex Rig using UserData and Xpresso. So for the new version I want to replace all the UserData by the Description Resource and using the GVO to benefit from the c4d.plugins.ObjectData Automated Handle Interface: GetHandle()/SetHandle()/GetHandleCount() and other advantages.

If you don't mind, here is a full plugin that use a similar method:
primitive_objects.zip (I don't know if I am authorized to share zip file here)

In comparison with this plugin example my actual rig is very complex it contain hundreds of description resource and twenty Handle Count. the plugin example it's only to see the method that I use.

I hope that is clear. I down know how explain you what I want to do exactly :)

Hi,

yes, while I still have no c4d here (so I could not use your zip), this does make things a lot more understandable for me.

Here is a quick sketch on how I would go about implementing, what you want to do.

class LightRig(c4d.plugins.ObjectData):
    
    def __init__(self):
        '''
        '''
        self._light_rig_cache = {}
    
    def Init(self, op):
        '''
        '''
        self.SetOptimizeCache(True)
        return True
        
    def _load_first_object_from_file(self, path):
        ''' Returns the first object from a file path interpreted as a c4d
         document.
            
            Args:
                path (str): The file path.
            
            Raises:
                IOError: When the file path does not exists or could not be red.
                IOError: When the document does not contain any objects.
            
            Returns:
                c4d.BaseObject: The first object of the document.
        '''
        doc = c4d.documents.LoadDocument(path, c4d.SCENEFILTER_NONE, None)
        
        if doc is None:
            msg = 'Could not load file.'
            raise IOError(msg)
        
        op = doc.GetFirstObject()
        if op is None:
            msg = 'Target object not found.'
            raise IOError(msg)
            
        return op.GetClone()
        
    def _get_light_rig(self, lid):
        ''' Either loads the light rig with the given id from a file or
         returns a copy of the local cache.
            
            Args:
                lid (int): The integer identifier for the requested light rig.
                
            Raises:
                IOError: When the hardcoded file path does not exists or could
                 not be red.
                ValueError: If lid is not a legal light rig id.
            
            Returns:
                c4d.BaseObject: The requested light rig.
        '''
        # Return an cached object from our ligth rig hash table
        if lid in self._light_rig_cache:
            return self._light_rig_cache[lid].GetClone()
        
        # Or load it from a file and add it to the cache
        if lid == c4d.ID_MY_FANCY_LIGHT_RIG_1:
            rig = self._load_first_object_from_file(SOME_FRIGGING_FILE_PATH_1)
            self._light_rig_cache[lid] = rig
        elif lid == c4d.ID_MY_FANCY_LIGHT_RIG_2:
            rig = self._load_first_object_from_file(SOME_FRIGGING_FILE_PATH_2)
            self._light_rig_cache[lid] = rig
        # ...
        else:
            msg = 'Illegal light rig id: {}.'
            raise ValueError(msg.format(lid))
        return rig.GetClone()
            
    def GetVirtualObjects(self, op, hierarchyhelp):
        ''' Returns the preloaded light rig cache.
        '''
        # The more verbose form would be somehing like:
        # if op[c4d.LIGTH_RIG_TYPE] == c4d.ID_MY_FANCY_LIGHT_RIG_1:
        #     return self.get_light_rig(c4d.ID_MY_FANCY_LIGHT_RIG_1)
        # elif ...
        return self._get_light_rig(op[c4d.LIGHT_RIG_TYPE])

A few notes:

1. I wrote this on an iPad from my head, so, jeah, view it as a 'draft'.

2. Using predefined data from a document is still a rather exotic concept. You might have to preload all data, when you run into threading issues.

3. The preloading / caching of your light rigs is in general a problem, as it will increase the memory footprint of your plugin.

4. Building your rigs procedurally might be a better option performancewise.

Cheers
zipit

@zipit
Hi,
Here is short video capture to compare between the result of my method and the result of your method, in my method the object stay visible when I moving the Handle point or make change in object data instance but with your method the object is not visible when I moving the Handle point or make change in object data instance.
For this reason I'm using this method.

Thank you for your time.

Hi,

1. I am absolutely fine with it if you prefer something else.
2. I do not know what you are doing exactly in your video, but yes, "your" side does look better. But that almost has certainly nothing todo how you impement your GVO.
3. I have to point out again that there is no my method and your method. You are using GVO in a way that is not allowed by the c4d SDK. Read the provided links above on threading. They are crystal clear on that topic. Or contact the the SDK support team, they will tell you the same.
4. While things might work in your small test environment, they won't in a production environment. Consider how pissed your clients or your company will be when your plugin has corrupted their files.

Consider yourself to have been warned ;)

Cheers
zipit

@zipit
Hi,
Sorry, I said "My method" and "Your method" just to differentiate between the two methods. I don't said that my method is better I'm just searching a legal method that allow me to obtain a similar result as the method that I'm using now.
in the method that I'm using the preset is loaded in document only one time. So, for example when I move the object handle points the change is applied into the previous loaded object.

The loaded object has a python tag that contain all object controls.
here is an example of python tag content:

import c4d

def main():
    obj = op.GetObject()
    parent = obj.GetUp()
    scale = parent[c4d.PRIMITIVEOBJECTS_GLOBAL_SCALE]
    
    # Objects scale
    obj.SetAbsScale(c4d.Vector(scale))

But with second method at each change the preset object is reloaded from disc or cache with the new changes (the preset object is removed then reloaded with the new changes) and that result that the object is becomes not visible in the View during a few thousandths of a second.

I'm thinking about another alternative, is the following code cause the same problem:

def GetVirtualObjects(self, op, hierarchyhelp):
        doc = documents.GetActiveDocument()

        path = os.path.join(os.path.dirname(__file__), "res", "my_scene.c4d")
        load = c4d.documents.MergeDocument(doc, path, c4d.SCENEFILTER_OBJECTS)
        if load is False:
            return True
        
        return c4d.BaseObject(c4d.Onull)

Hi,

no need to be sorry, I am / was not offended. I was just trying to be very clear about the implications of your approach.

1. If you want to have a statically generated object, ObjectData is not the correct base class for you, as its whole purpose is to provide a dynamically generated object when you overwrite GVO.
2. Loading an object from a preset file is quite an unusal thing to do for the ObjectData type. The more common thing to do, would be to construct your rig procedurally within your GVO. Which would probably also (apart from performance benefits) alleviate the problems you have shown in the video.
3. Also noteworthy is, that a python scripting tag also calls the draw() function of your script, which you could use to draw / react to to handles with your original xpresso and user data steup.

Cheers
zipit

@zipit
Hi,
Thank you, I'm trying to rebuild all my rig inside the GVO and I think I will using C++.
is it better to use c++?

Cheers
Mustapha

Hi,

@mfersaoui said:

is it better to use c++?

I think the answer to that is highly subjective. Both SDKs and languages have their strengths. As your plugin probably is neither going to be computationally expensive, nor will be in need of any of the more advanced features only the C++ SDK does offer, I would say, Python wins this cost/gain comparison by a long shot.

But that is subjective ;)

Cheers
zipit

Hello,

just to make some things clear:

• The purpose of GetVirtualObjects() is to fill a cache with virtual objects.
• It is NOT the purpose of GetVirtualObjects() to modify the document / the scene graph.
• It is forbidden to edit the scene graph from within the scene execution pipeline e.g. from within GetVirtualObjects().
• One must only edit the scene graph in reaction to some user interaction e.g. after the user pressed some button in the UI or used a tool.
• One must never call GetActiveDocument() in a NodeData based function. The active document is the document currently displayed in the viewport; but there is no reason to think that this is the document that hosts the object.
• Consequentially, one must never call EventAdd() in a NodeData based function; especially not from within the execution pipeline.
• If you want your generator to load a given scene file, you can load that scene with LoadDocument() and just COPY the content into the cache.
• The Python API is an incomplete wrapper around the C++ API. So some functions may be missing in the Python API.

best wishes,
Sebastian

@s_bach said in Make editable an object generated by GetVirtualObjects():

GetVirtualObjects()

Hello,
Thank you, I decided to use C++ to do that, I just post new topic "MAXPORTS GraphView Flags - Description Resource" is a continuity of this topic.

@s_bach said in Make editable an object generated by GetVirtualObjects():

LoadDocument()

Hello,
I gave up the idea to merge objects from scene file, I had generate my object and all her sub-objects inside the GetVirtualObjects(). But now I'm searching how to assign materials to my sub-objects, I would imagine that is the same as objects, should I use the LoadDocument()? if yes are there examples? and then can I change these materials proprieties from the object description resource or I must to use CommandData dialog?
Thank you.

Hello,

a generator (GetVirtualObjects()) cannot create materials. You must generate materials in reaction to user-interaction (Material Manual). Then you can assign materials to your (virtual) objects using a TextureTag (TextureTag Manual).

for example, you could detect when your generator object is created from the UI by listening to the MSG_MENUPREPARE message in NodeData.Message(). Then you can create all the materials you want to use later.

best wishes,
Sebastian

@s_bach
Hello,
Thank you, As you explained, I create all materials in NodeData.Message() and then I called them in the GVO. I just want to know if is it allowed to edit the called material parameters inside the GVO or not? see example bellow:

BaseObject* MyObject::GetVirtualObjects(BaseObject* op, HierarchyHelp* hh) {
        ...
        BaseObject* container = BaseObject::Alloc(Onull);
	if (!container)
		return BaseObject::Alloc(Onull);

	container->SetName("Container");

	BaseMaterial* material = doc->SearchMaterial("my_material");
	if (material) {
		// create the texture tag
		TextureTag* const textureTag = static_cast<TextureTag*>(container->MakeTag(Ttexture));
		if (textureTag == nullptr)
			return maxon::OutOfMemoryError(MAXON_SOURCE_LOCATION);

		const Float Brightness = data->GetFloat(MYOBJECT_LIGHT_INTENSITY);
		
		// Is the following action allowed inside GVO:
		material->SetParameter(DescID(MATERIAL_LUMINANCE_BRIGHTNESS), Brightness, DESCFLAGS_SET_0);

		textureTag->SetMaterial(material);
	}
        ...
}

Best,
Mustapha

I think this is not allowed. I've just seen the list of the Forbidden Functions and "Make any changes to materials." is inside.

Hello,

as said before, it is not the purpose of GetVirtualObject() to edit the scene or to edit elements of the scene; this includes materials.