Question

I have a simple nested dict to indicate a graph with nodes, vertices, and edge weights in the form dict:

{node: {vertex: weight, ... } ...}.

Here is how I create it:

with open(file) as f:
        __, __1 = next(f).split()
        for line in f:
            tail, head, weight = line.split()
                g1[tail] = {}
                g1[tail][head] = int(weight)

This code gives me the dict I want. However, when I run more code that creates a similar nested dict as well as accesses this dict, I run into an error. Here is the code:

nodes = g1.keys()
distance = {}

    for n in nodes:
        distance[n] = {}
    for k in nodes:
        distance[n][k] = graph[n][k]

Somewhere in here, this error crops up:

distance[n][k] = graph[n][k]
KeyError: '344'

indicating that I haven’t yet created a key for ‘344,’ which is what this code is trying to do. I thought by initializing each node n to a dict, I could then create an entry for distance[n][k]. I’ve tried using default dicts — the result is the same. Why?

Answer 1

for n in nodes:
    distance[n] = {}

After this for-loop completes, n is equal to the last value in nodes.

There is no guarantee that for every k in nodes, graph[n][k] exists:

for k in nodes:
    distance[n][k] = graph[n][k]

The nodes are all tails, not heads. So k is iterating over tails. Yet graph[n][k] is placing k in the position expected for a head.

The tails and the heads are not necessarily interchangeable, and the last tail, n, may not be connected to every other tail, k.

---

If you are trying to make distance a copy of graph, then use

import copy
distance = copy.deepcopy(graph)