Thank you for the reply.
The idea is to save a value beyond the limits of the main function of a python generator, and Global Variable seems to achieve it.
Saving the value when saving the scene isn't important in this case, as the purpose is performance. On load it would re-evaluate the generator anyway.
So I'm trying to understand when global variables fail, and we need to use a plugin id instead. Is the only benefit of the plugin id that it can save the data with the file?

Hi all,
I'm trying to understand some basic concepts on good programming practices in c4d.
When we want to save a persistent data I was advised to use a plugin id that is attached to my python object.
What is the drawback of using global variables instead?
In my tests global variables seem to work, so what is the plugin id offering that the global variables don't?

My example case is manually storing a matrix array in a variable so I can compare it and see if it changes, since the GetDirty fucntions seem to be problematic for that particular object.

Update:

def main():
    obj = op.GetDown().GetCache().GetClone()
    print obj
    return obj

getting the cache before cloning does seem to work, I couldn't see the object as it was off screen before.
Still it opens some questions such as why the clone didn't initially work; I assume I needed to insert it into the doc, which I had tried without much success (crashed cinema when I tried to evaluate the doc).

In any case, the problem seems solved for now.

Hi all,
using a python generator as a source for another python generator can be helpful for prototyping. But I'm having trouble getting the cache of the child generator.

GetCache always returns me "None", and using current state to object also returns "None".

This is a code sample:

def main():
    obj = op.GetDown().GetClone()
    print obj.GetCache()
    return obj

I tried to see if it has children or siblings with GetDown or GetNext but I get:
"AttributeError: 'NoneType' object has no attribute 'GetDown'"

Code with current state to object:

def currentState(obj) :
    return c4d.utils.SendModelingCommand(c4d.MCOMMAND_CURRENTSTATETOOBJECT,[obj],c4d.MODELINGCOMMANDMODE_ALL,c4d.BaseContainer(),doc)[0]

def main():
    obj = op.GetDown().GetClone()
    collapsed = currentState(obj)
    print collapsed.GetChildren() //returns []
    return collapsed

In the docs there is a recursion function for the cache but I'm not sure how to use it.

I'm trying to understand what I'm doing wrong, are there suggestions on how to achieve this?
Here is a sample file:
python_under_python.c4d

Wow much obliged for all the replies everyone, it is really helpful!
I think evaluating in a tempdocument is what I was looking for. I have another question about best practices:

In zipit's example the original cube is removed from the temp document after use. Is there a reason to do that? Does it help with performance and/or memory management?

Hi all,
I'm trying to see if I can access the polygon data from generators that I create within the Python Generator.
E.g. the default python generator creates a cube, is it possible to print out the polygon count of that cube?

GetCache() returns None as there doesn't seem to be any cache yet.
What is the designated way to do this? Is CurrentStateToObject an optimal way?

Thanks a lot, I'll try this out.
Unfortunately iterating through the matrices one-by-one makes it much more costly to just check for the dirty state, so I'm not sure if is practical to use this solution.
It would be more viable if the built-in Get Dirty worked, maybe that's a bug or a mistake from my side.

Hi,
yes I noticed that this happens and your suggestion is more fool-proof.
But the issue is that moData.GetDirty(c4d.MDDIRTY_ALL) doesn't increase when affecting the matrix using an Effector, or when adjusting its target object.

Thanks a lot for the explanation!
I replaced the plugin id with a user data and it seems to work ok. Still trying to figure out the matrix on the other thread though.

Here is the adapted code:

def main():
    ID_PYGEN_DIRTY_CACHE = op[c4d.ID_USERDATA,3]
    linked_object = op[c4d.ID_USERDATA, 2]
    # Get the last cached dirty count and the current dirty count.
    lst_dcount = ID_PYGEN_DIRTY_CACHE
    cur_dcount = linked_object.GetDirty(c4d.DIRTYFLAGS_ALL)

    if linked_object is None:
        return c4d.BaseList2D(c4d.Onull)
    # Return the linked object if its dirty count exceeds our cached dirty
    # count or there is no cached dirty count.
    if lst_dcount is None or lst_dcount < cur_dcount or op.GetCache() is None:
        # Cache the current dirty count.
        op[c4d.ID_USERDATA,3] = cur_dcount
        clone = linked_object.GetClone(c4d.COPYFLAGS_NO_HIERARCHY)
        clone.SetMg(c4d.Matrix())
        print "Built cache"
        return clone
    # Otherwise return the cache.
    else:
        print "Returned cache"
        return op.GetCache()

And the file:
python_dirtyState_0001.c4d

Thanks a lot. I didn't notice that GetDirty gives a dirty count instead of a boolean.
I adapted your code (very helpful!) and I used a user-data instead of a plugin id, it seems to work for normal objects, but with the Matrix I still have trouble...

Changing the Matrix parameters does change the GetDirty, but when using an Effector that changes the matrices, even though I can read these matrices from the MoData, the GetDirty doesn't increase the count, at least with the "MDDIRTY_ALL" flag.

In the documentation it says for the "MDDIRTY_DATA":
"Data in the arrays changed, must be manually set."
But I'm not sure what it means by "manually set. I assume it refers to when manually setting the array data? What happens when using an effector or changing a target object?

This is the current code:

def main():
    dirty_storage = op[c4d.ID_USERDATA,1]
    moData = mo.GeGetMoData(op.GetObject())
    # Get the last cached dirty count and the current dirty count.
    lst_dcount = dirty_storage
    cur_dcount = moData.GetDirty(c4d.MDDIRTY_ALL)
    print "dirty count: " + str(cur_dcount)
    
    # count or there is no cached dirty count.
    if lst_dcount < cur_dcount:
        op[c4d.ID_USERDATA,1] = cur_dcount
        print "dirty"
    else:
        print "not dirty"

And the file:
matrix_dirty_0001.c4d