Hydra uses named buffers in code. This is great for feedback effects like src(o0). This works because src(o0) gives you the previous frame of that buffer. That delay brings the feedback fun that we love Hydra for.
In a node graph, connecting an output back into itself doesn't create that behaviour; it would represent the current frame, not the previous one.
One option is to replicate Hydra's model directly to nodes, having 'output nodes' which users can select elsewhere src sources elsewhere. But this feels like mapping a solution made for the (effectively) 1D world of code, to the 2D world of nodes.
A more intuitive approach in a node editor is to support a 'delay edge'. A delay edge says "use the output of this node from the previous frame, not the current one'.
Visually it can be just a different edge style. Behind the scenes, it can just use out(o0) and src(o0). But for a user, I think it's clearer and better matches the flow of a graph.
This would give us feedback without exposing buffers to the user or needing to think about the different o0, o1 etc. It'll all just still be wires.
Hydra uses named buffers in code. This is great for feedback effects like
src(o0). This works becausesrc(o0)gives you the previous frame of that buffer. That delay brings the feedback fun that we love Hydra for.In a node graph, connecting an output back into itself doesn't create that behaviour; it would represent the current frame, not the previous one.
One option is to replicate Hydra's model directly to nodes, having 'output nodes' which users can select elsewhere
srcsources elsewhere. But this feels like mapping a solution made for the (effectively) 1D world of code, to the 2D world of nodes.A more intuitive approach in a node editor is to support a 'delay edge'. A delay edge says "use the output of this node from the previous frame, not the current one'.
Visually it can be just a different edge style. Behind the scenes, it can just use
out(o0)andsrc(o0). But for a user, I think it's clearer and better matches the flow of a graph.This would give us feedback without exposing buffers to the user or needing to think about the different
o0,o1etc. It'll all just still be wires.