rmanoka

Great! Closing this as completed.

I may be out of depth here, but the data in the R-Tree is only at the leaf-nodes right? So what does a post-order traversal mean for a R-Tree? Also,...

Got it. The issue is that the R-Tree itself has no user-data available at the intermediate nodes, so all the iterations have no concept of post-order. I suppose you want...

Also, note that the [root node](https://docs.rs/rstar/0.9.2/rstar/struct.RTree.html#method.root), and then [children](https://docs.rs/rstar/0.9.2/rstar/struct.ParentNode.html#method.children) of intermediate nodes can be accessed via public methods, so we can implement the union-algorithm without using the iterators.

The `SelectionFunction` already has a method `should_unpack_parent` that tells when it starts unpacking a parent node. That along with some nifty usage of a vec/stack should let us do it....

Thanks for fleshing out the details @adamreichold ! Looks quite on track; like the easy rayon support. Just a few minor comments / points to discuss: 1. The `init` parameter...

> However, while I don't know the internals of `RTree` I'm guessing this approach traverses the tree twice when it's only necessary to traverse it once. Am I right that...

One approach is to split the concern into two: 1. Implement the std [`Fromiterator`]. This allows conversion via the `collect()` method: ``` rust let _: FeatureCollection = into_iter_of_features.collect(); ``` This...

Thanks for providing more context @michaelkirk . The `FromIterator` impl. is definitely useful; in your snippet (3) above, it should help remove the double iteration, and hopefully also the extra...

Error enum and serde_jso are at loggerheads. Iiuc, Serde json will expect all our semantic errors to be String s.