Ryven – Flow-based visual scripting for Python
ryven.org
ryven.org
Some problems in no particular order: * The presentation logic and execution logic of nodes are mixed together in the same python class. * Information density. Nodes and strings just can't fit enough on the screen. * Independent editor which doesn't integrate with other code editing tools. All the niceties of an established text editor or IDE have to be given up. * In my opinion 'get var' and 'set var' leak too much of the text based code paradigm into the flow paradigm. * It only outputs python, which already has a low barrier to entry. Outputting C or Rust or LLVM IR would make the product much more interesting.
However, after all the criticism, keep going. There is a future hidden somewhere inside this type of coding, it just hasn't been discovered yet.
I assume I'm missing something given its popularity but I personally am in no rush to use this style of development again.
To me, the problem being solved by low code tools should be focused on formatting and syntax. That's what always trips up non devs. They just don't get why capitalization, spacing and special characters always need to be so perfectly placed, with so many arbitrary rules, like functions use commas to separate parameters, but for(), which looks exactly the same, uses semicolons. Why? Reasons. New coders hate it.
TouchDevelop by Microsoft Research really nailed this problem years ago. Turn the commands and variables text into clickable objects so new coders can focus on the logic and program flow - still using "text" like normal coding - but without worrying about the syntax. Too bad it died.
There is a market for solving programming problems in various specific domains without actually programming.
Codeless programming has less value if a) you don't have domain knowledge b) you already know how to code.
If you a) have domain knowledge and b) don't know how to code then a codeless platform allows one to create models that create added value.
Excel, LabView and material editors in various CG packages, Blender included are targeted for domain experts.
I could maybe see myself using this. Maybe drawing a low res plot of each dataset as it works through the flow for visual checks.
Its an interesting take. Good stuff.
Using nodes to manipulate images hasn't been a problem for a few decades. They need to show more general cases, how to iterate, how to debug, how to manage state, how to manage errors, etc.