0

Is there a data-structure/model for storing a value at an arbitrary URI-based key, and then if null, backing down to a less specific path/domain? i.e.

SET example.com "hello"
SET a.example.com/foo "world"
GET example.com => "hello"
GET example.com/foo => "hello"
GET a.example.com/foo/bar => "world"

Value is simply a serialized JSON object; I don't need to do any list operations on it.

Currently, I'm using node.js/restify backed by redis (although I am open to other datastores). I realize I could have a flat key-value store, and loop through all subpaths/domains, but that feels inefficient with a dozen potentially empty calls to the datastore.

Allan Rempel
  • 73
  • 3

1 Answers1

0

You can do a binary search on a failed key lookup to find the most specific matching URI. For example, you've got a depth-8 URI; start by checking for an exact match, if this fails then check the depth-4 URI; if this fails then check the depth-2 URI, else check the depth-6 URI; assume in this case that the depth-4 and depth-6 lookups succeeded, next you'll do a depth-7 lookup, if this succeeds then return the depth-7 value, else return the depth-6 value.

As an alternative, via Google I found a trie implementation for leveldb that might do the trick, but in general there don't appear to be very many database trie implementations.

Zim-Zam O'Pootertoot
  • 17,888
  • 4
  • 41
  • 69
  • I can't use binary search to skip different depths of URIs. For example, if depth-4 and depth-7 are set, a lookup on a depth-8 URI should return the value for depth-7, not skip up to depth-4. – Allan Rempel Nov 29 '14 at 19:38
  • @AllanRempel In the example I gave you return the value of the depth-6 URI even though the depth-4 URI is also set; you don't halt the search at the first non-null value, you halt it when you've searched to a depth of log(n) at which point you'll have the longest matching prefix key – Zim-Zam O'Pootertoot Nov 29 '14 at 20:17