[Bf-committers] Meeting minutes - nodal logic proposal

[Benoit Bolsee]

On Tue, 3 Nov 2009 09:43:13 -0500, Charles Wardlaw
<cwardlaw at marchentertainment.com> wrote:
> The nodal logic proposal sounds fantastic.

Thanks, it's really in the design phase at the moment. The real trick is
to find a generic way to build a node system that allows all types of
graphical programmation: imperative programming, data flow programming,
state engine, hierarchical state engine, protocol stack, etc.  It's not
a new subject, many Nodal programmation systems exist already. I've
tried to extract the essense of these systems and with the help of the
robotics experts at the KUL who have already implemented node systems
like that, I came up with this design. 

> I guess my first question is: why is it limited to the game engine?   
> People have wanted object nodes for years, and that's basically what  
> this is.  Instead of keyboard or joystick input, all you really need  
> is the ability to either run input from a time node (already 
> noted on  
> the wiki page) or to have data pulled towards end result 
> nodes rather  
> than pushed on a frame-by-frame basis.
> 

I'm getting requests to make the engine independent on the game engine.
That should be possible but I haven't considered all the implications of
that. The generic node that is at the center of the system is a
combination of C++ container and python objects. 
Note that I want to provide push data flow (in my new design the logic
input/ouput will become push input/output that will allow passing
complex data and not just pulses) and pull data flow as you describe.

> In Maya's dependency graph, for example, result nodes (nodes with no  
> further downstream dependencies) are checked to see if they're  
> "dirty".  If they are, then they call upstream sources and ask them  
> for data.  Nodes asked for data will check themselves, 
> calculate, and  
> cache their values, then send that data downstream.
> 

An important aspect of my system is that there will be no hidden
processing such as automatic detection of result nodes, etc. If you want
to evaluate a DAG, just build it and connect its result node(s) to a
push logic node to trigger it's evaluation. This node can be triggered
by a time node (for realtime applications) or by a 'start' pulse that
Blender could generate to evaluate the graph.  After that, each node of
the DAG can implement caching logic as they want, it's up to the creator
of the node. 

> If the nodal logic system were just a bit more generalized, 
> and didn't  
> rely on being in game engine mode to work, it could become a rigging  
> tool on par with Maya or XSI.  Add in the ability to enable / 
> disable  
> node traversals per object (possibly by wrapping nodes as a modifier  
> or something), and bam said the lady-- I have a much stronger 
> position  
> to argue from re: using Blender on a longer project at my studio.
> 
> How will iteration over lists of objects be handled?  Or will it?
> 

Quite simply I guess: someone just need to implement an enumerator node
that push each object reference to a graph (encapsulated in a compound
node for instance) which is looped back to the enumarator node. In other
words, a graphical representation of a for loop. You could also
implement some nodes that can operate on a single object or a list of
objects. Inside each node there is a Python class, it's up to the node
creator to code it right.

> The other question I have is on this line:
> 
> > Nodes may be distributed via the standard Blender install, or
> > through open source libraries, or as closed-source licensed 
> routines  
> > from vendors.
> 
> I doubt I'll ever fully understand how the GPL works in conjunction  
> with other software licenses, but wouldn't a closed-source node need  
> to be distributed as a C++ dynamic library, and as such not 
> be usable  
> with Blender due to a GPL violation?  Although, if Blender offered  
> some sort of script locking of compiled Python files (a la 
> MaxScript)  
> I could see people releasing paid node collections.
> 

There are requests to develop the node system under a more liberal
license than GPL. I am not against that idea but I haven't thought of
any of its implications. I don't want to make my life too complicated
before starting to code.  For example, I don't know if using Python is
already a big drawback for non-GPL applications. But basically I am in
favor of using a liberal license and will try to do it.

> On 2009-11-02, at 9:49 PM, Alex Fraser wrote:
> 
> > Hi,
> >
> > On Mon, Nov 2, 2009 at 4:48 AM, Ton Roosendaal <ton at blender.org>
> > wrote:
> >> - Benoit is working now on a new logic editing system for the GE:
> >>   
> http://wiki.blender.org/index.php/Dev:Source/GameEngine/NodalLogic
> >>   Feedback welcome, design is in progress still.
> >
> > This looks great. I've used nodal logic in another game 
> engine before, 
> > and this proposal looks pretty much spot-on. I made a 
> couple of small 
> > changes to the wiki - I hope that's a good place to leave feedback.
> >

That's fine as long as you make you changes clearly identifiable.
I've made progresses with the design and I will refresh the wiki with
more detailed informations.

/Benoit