Feedback on new subgraph functionality in the VFX Graph

Hi,

I tried out the recent subgraph branch ( vfx/2019.2/feature/subgraphs ) as I've been desperate for this feature to streamline my own experiments.

In general it seems great but I've got one concern - the workflow around taking an existing complex graph and converting it to a subgraph is fairly suboptimal because of one factor.

If you select a group of nodes that have multiple inputs coming from the same (non-included) node then it creates new inputs for each connection. So if there's a node called "color" that feeds into 4 nodes in the new subgraph you end up with 4 color inputs. A similar problem happens for outputs.

This makes refactoring complex graphs quite time consuming as you need to delete the duplicates and remember how they were wired up. I resorted to keeping a copy of the original graph open in a separate instance of the Unity Editor so I could refer to it.

I can't imagine a scenario where this is the preferred behaviour. Is there a reason it was chosen or is it a case of "it was easier to implement this way"?

 

Second this observation, without this current behaviour the system would be an absolute dream, instead its a bit fiddly right now having to keep everything in my mind so I can reconnect as meant to be.

 

Hi.

I'm talking about another eariler node having multiple outputs.

For example:



Everything to the right of the blue line belongs in a subgraph.

I expected a single output from each of the two "Sample Texture2D" nodes. I.e. The resulting subgraph would have two Vector4 inputs wired up exactly as the originals were.

What actually happened is that I get 8 inputs - 2 Vector4's for the pink lines and 6 float inputs for the green lines.

 

OK. So I presume a really common workflow is "take a complex graph and split it into subgraphs".

The example I posted above is probably a moderately realistic real world example. As mentioned I had similar problems with both inputs and outputs.

The current functionality creates a lot of work when I want to refactor a complex graph. Am I doing something wrong in my original design? Is the example I posted unusual?

If so then it would be useful to get an idea of how I can improve my workflow.

If not - then surely there's a flaw in the current design?

I guess another way of putting this is "how have you handled this yourselves when you're using the new functionality of something moderately complex?" Did you hit similar problems?

 

@JulienF_Unity Thanks. I just wanted to draw attention to this particular flaw as I felt it was pretty central to most common workflows (but hey - doesn't everyone thing that about their issues?)

 

Hi - just evolved to this topic. Is there a way to expand subgraphs or is this operation by design un-expandable? Could be cool to save frequently used groups as presets without having to manually create .vfx libraries of gadgets.