Maikel van de Lisdonk 2021-10-31 11:07:43 Hi, finally I got around to making the video that I wanted to make for some time.. here I'll show the progress I made on the visual webassembly compiler. Basically this is a side project of a side project.. because the visual webassembly compiler uses the flowrunner-canvas project and extends it with webassembly compilation and visualizations. Enjoy! https://youtu.be/MUuYKLbfHhw
Daniel Krasner 2021-10-31 15:03:45 that’s really nice! Are you thinking of some sort of graph-like node collapse/expand navigation?
Kartik Agaram 2021-10-31 15:23:50 I had trouble following because the voice-over seemed to be unrelated to the visuals. Is that accurate? Did you record video first and overlay a narrative over it?
The visuals look amazing (I kept rewinding to pay attention to details) but I don't quite follow what's going on. Is all the code for the compiler also on screen or something like that?
(Sorry I haven't watched all the previous videos in your series.. when I make these I try not to assume what viewers have watched, and always describe some foundations for each video. Which gets increasingly challenging as you make more videos 😄)
Kartik Agaram 2021-10-31 15:35:06 I went back and rewatched https://www.youtube.com/watch?v=hxCrRkkhgiM, and that helped a lot.
Have you considered having the dynamic recompilation in the context of a text-based language? Just smush the text from all the nodes in a function into a single text area. Is that really worse than all the boxes and arrows with strange names?
Maikel van de Lisdonk 2021-10-31 15:35:30 Yes, i recorded the video and audio separately. I was already wondering if they were related enough. 2 minutes is quite challenging to put everything in😅. I'll try to do better next time☺
Maikel van de Lisdonk 2021-10-31 15:39:26 Everything that is on screen is being compiled.. so all the matrix operations are also on screen. I agree that laying them out like this is not quite making things clearer. Multiplying a matrix takes like 20 nodes.. in comparison, I've actually put the code for the inverse matrix in one node, and that really helps. Although it still has 5 nodes on the top of my head
Maikel van de Lisdonk 2021-10-31 15:43:13 My plan is also to be able to put function/sub-flows in a separate layer , although I am not yet sure how to show the different layers visually. Perhaps showing them further away and with some fog or blur applied, or in a single node with a portal like functionality. Although the current implementation is too heavy already using both 2d canvas and DOM nodes.
Kartik Agaram 2021-10-31 15:50:56 It might help to describe your goals and assumed designed constraints. Either in text or video.
There's always a trade-off in going from the density of text to a visual framing. But different projects like PANE or Apparatus have some subtly different reasons for giving up the density. So it would be helpful to understand the details of where you're coming from.
Maikel van de Lisdonk 2021-10-31 16:15:47 The project that I showed in this video is very early stage and for sure needs a lot of extra functionality to abstract away a lot of the things that are now in separate nodes. For accessing memory the addresses need to be calculated your self currently, which offcourse is not how you want to do it🙈. My other project (the flowrunner-canvas project which participates in the review-jam) operates on a higher level
Maikel van de Lisdonk 2021-10-31 16:16:43 It probably helps if I make a longer video with a presentation for the share-your-work channel to talk more in dept about my projects and goals
Kartik Agaram 2021-10-31 16:41:46 Yeah I think it's far enough along now that I and maybe others will be interested in something more in-depth. Either text or video. A sense of plans, what you want to build first vs later. What you think you will never build, etc.
