Knockout does incremental updates as well as memoization. It instruments the evaluation of computed observables to determine the dependency relationships. I haven't delved into Ractive.js closely, but it sounds like it does something similar. How is what they do different from what you're describing?
Re-evaluation is easy enough when your computation is pure. But once you throw in some imperative effects, like adding a name binding to a symbol table, all of the declarative data binding frameworks fall down very quickly.
There is a lot of research in this area, concurrent revisions at MSR or self-adjusting computations at CMU to name a couple. Also, if you've seen Bret Victor's videos and wonder how any of those could be implemented for real, then this is where we need to go. It will be a fun trip though.
You're awesome. I've been playing around with a dependency network in JS recently, and I'd just been puzzling over exactly that question: how best to handle the case where a computation has side-effects. So I'm now busy googling concurrent revisions and self-adjusting computations. Got any further references to throw my way? Is there a standard literature on this stuff?
Actually it's a problem I'm working on right now, but I have nothing to share yet. Some sort of restricted programming model is needed, in my current system effects must be reversible and commutative, the last one being a major restriction but workable for most programs (e.g. set add is typically commutative, list add is not but can be hacked, dictionary set to the same key is not but can restricted by set once semantics).
