On the contrary. I believe I’ve been looking at server-side VDOM diffing since Drab.Live came out and (through talking to @OvermindDL1) I understood that controlling a dynamic UI through the server was viable.
I got pretty excited with the first texas announcement, and then the excitement died out when I understood that texas seemed to depend on parsing HTML strings with Floki (wtf?!).
I’ve never talked to you directly because I’ve been busy with other things, but I’ve been waiting for something faster to come out (parsing HTML at runtime is a big no for me)
If you use smart heuristics (like react does), you can diff a tree in linear time on the number of nodes.
This requires storing the previous tree in memoy and comparing it with the new one though, which costs memory. By diffing the dynamic data directly, you save a lot of memory.
If you restrict the data to a sensible datatype, auch as a map, you can diff it using already available algorithms, which are linear on the size of the map or list. Then, you can implement something like a mutation watcher, in which you build a dependency graph between the raw data and the VDOM and update only the parts that depend on data that has changed.
The Javascript framework VueJS does some clever things (which I don’t understand) with mutation watchers and dependency graphs, and as a result is both fast and intuitive.
@OvermindDL1 has a client-side framework written in OCaml that compiles to Javascript, which he claims to be very fast.
I think Drab.Live does something like that (and @grych is the correct person to explain it to you).
