Tree lenses

[cfhammill]

I find data.tree cumbersome because of mutation. I want a library free of this inconvenience, but it does a lot of nice things I would miss. Convenient print methods, easy data.frame conversion, automatic ancestry/depth tracking. I am torn between writing lenses for data.tree and writing a simpler tree library built around a lens API. Provided I write simple conversions between the two we could potentially have the best of both worlds.

Writing lenses for data.tree would involve checking to see if the lens is pointing at a node, following the node's ancestry to the tree root, and triggering a full copy on set. It is possible I'm over-estimating the performance problems associated with this. R triggers full list copies on modify too, although I suspect R is making very smart choices about copying/GC that make it dramatically faster than copying data.trees (although I should really test).

Hacking a tree API seems easier to me than optimizing copy speed or decreasing copy frequency in data.tree but involves writing a lot more code.

Decisions, decisions.

[cfhammill]

cc: @jasonlerch