The way `dpath.util.get` is implemented means that it requires all parts of the tree to be iterable. So if you have a node which is a dictionary, but isn't iterable, it does not allow you to find the children under that node. This is a consequence of using `fold` and `walk` to implement `get`. This is also not efficient if the node is a map (that is, it still iterates over the keys when it could do a direct lookup). To make `get` more efficient and also allow non-iterable maps we need use `kvs` only when the related segment is a glob (that is, only iterate over them if it is strictly necessary). This means we can't use fold or walk anymore. There probably is additional fallout in the other methods like `view` and `search`. Unfortunately this will require some significant changes. Originally reported here: https://github.com/akesterson/dpath-python/pull/122/files#diff-f50eb92b27ecc897863ec0f0ef4f0f65R52-R66