Extending atom is really powerful and a number of live coding environments do this (e.g: atom-hydra). With Visor I opted to create the GUI from scratch to have more control. Trade-offs both ways I’m sure.
These are really fantastic!!
What is it about class wrapped sketches that make them better than bare ones?
While I haven’t live coded with Pry, I believe Visor works pretty much the same (but more constrained) as the lines of code are automatically instance eval’d on persistent layer objects (you can pretty much think of each layer as a sketch, each has it’s own state and draw method). This guarantees state retention for seamless live coding performances. Does the Pry method preserve state?
I’m interested as to why you are skeptical about REPLs in live coding! I would be keen to create another thread for the liveness discussion. And yes, Visor has since evolved to merge the REPL and Draw editors together. So instead of having your draw code in a seperate editor you just wrap it in a def draw. All the code is evaluated against the same layer object anyway so it made sense to combine the two. Regardless, the editor is still acting like a REPL.