Ways by node

graph.clj

(defn collect-ways-by-node [ways]
  (loop [ways ways
         accum {}]
    (if (seq ways)
      (let [{:keys [nodes id]} (first ways)
            new-accum (zipmap nodes (map vector (repeat id)))]
        (recur (rest ways) (merge-with (comp vec concat) accum new-accum)))
      accum)))

graph.py

import collections

def ways_by_node(ways):
    ret = collections.defaultdict(list)
    for way in ways:
        way_id = way['id']
        for node_id in way['nodes']:
            ret[node_id].append(way_id)
    return ret

Original post: https://plus.google.com/105985814958698329518/posts/CowT87M61WC

Problem description: a Way is an object with an id and a list of node ids, like so (in json): {'id': 'foo', 'nodes': [10, 11]}. Given a list of ways, return a mapping of node ids to a list of way ids that contain each node. Given a list of the way described previously, the return would be {10: ['foo'], 11: ['foo']}

Speculations over why the Clojure solution looks worse than the Python solution:

• I wrote terrible Clojure and someone with more than a month of experience can improve my code.
• The Clojure solution is fine and just looks worse because I'm less familiar with the language
• Immutable data structures are just clumsier to work with despite their other benefits.
• defaultdict is just super-handy and the clojure code would be easier if defaultdict existed in clojure.core. (It didn't show up in the Python stdlib until version 2.5.)