ExpandingMan

I'm still a bit confused about what you're trying to achieve here (I still haven't read everything thoroughly, though I did read your responses. I suspect you are trying to...

It's still sounding to me a bit like you don't want indexed trees at all, from your description I'm still imagining that what I wrote above is a special case...

> The penny finally dropped last night that in your example the every node is itself a tree. This is essentially the difference between the indexed trees and regular trees....

This is the situation we wind up in when we upgrade a package with so many dependencies. Any help pushing through PR's to update bounds is appreciated.

This seems a bit aggressive, there is a huge class of cases in which this will make the display much worse... Perhaps our real problem here is that we use...

Small update: I just remembered that `getindex` works on `pairs(dict)` just like it does on `dict`. This has the implication for `Dict` that `getdescendant` works even on current master. That...

I think we should avoid complicating the interface. Since my realization that `pairs(dict)` can be indexed it seems very unlikely that we will change this. Certainly we would not get...

The other problem is that `values` does not return an object which has `getindex` methods for the keys. I think a lot of this goes back to it probably having...

Another perspective is that the problem is with `Base`, not AbstractTrees, and we are merely adhering to the somewhat-regrettable interface that it established. Users who want better behavior can use...

I have looked into the first of the non-trivial breaks, and made [this PR](https://github.com/fabid/BehaviorTree.jl/pull/1) to BehaviorTree.jl. It appears the authors were not really using `ShadowTree`, which support the decision to...