hot take: non-text representations of stuff that's fundamentally directed cyclic graphs (e.g. digital logic) are Bad bc you can't diff them
Conversation
Replying to
there doesn't exist a way that allows you to render a difference of graphs embedded in 2d that scales well
2
6
Replying to
but is there a way to render 95% of real world differences well?
1
Replying to
not really. i mean, 2d embedding itself *already* doesn't scale, but it sort of passes if you suffer just enough
1
1
Replying to
briefly and it gave me a few ideas about how to do higher order programming in something labview-like
1
Replying to
you can sort of do higher-order in graph repro, but you're working against your medium
1
1
Replying to
having to do a 2d embedding artificially limits the range of abstractions you can use and i don't find that compelling
2
2
Replying to
my (untested) approach to that is: have several planes, turn your diagram "on the side" to work in another plane
2
Replying to
these would be "semantic" planes, not wiring layers
also there's UML diagramming the same thing in different visual languages