this is actually progress
Conversation
still very much in the "how do i want the underlying data structure to be" stage of things so i'm just drawing straight lines at the y-values of the (one-dimensional i guess) centroids of the nodes
1
this is also actually progress
read image description
ALT
3
1
what i'm doing here is just plunking down straight lines where i want the series of sankey tubes to go, this time i just made the stroke width proportional to the flow (the heights are still not quite right so i'mma fix those then replace the lines with béziers)
1
getting cloooserrrrr
almost time to switch to béziers
read image description
ALT
2
ah screw it let's just do the damn curves already
as you can see the alignment on the sink side is not right
anyway by this time tomorrow i should have a nice slick animatable sankey
read image description
ALT
2
i'm still very much in the "still deciding how this is gonna work" phase here; edge crossing minimization goes under the "boring graph theory i can implement any time" heading
1
Replying to
It may not be. In general if you have multiple sources AND sinks there’s no reason to expect the graph to be embeddable in 2d... Nash embedding theorem applies. But maybe sankeys satisfy tighter constraints and it can work.
At least that’s my rusty memory. All graphs are embeddable without edge crossings in 3d but not 2d
1
1
Replying to
yeah they have to be planar in 2d but sankeys have kinda graphs within graphs cause the flows are shown side by side; generally only really simple weighted dags* can be represented as sankey diagrams without looking ridiculous
(* you see stuff with cycles but it looks funny)
1
1
Show replies

