Question

For a certain piece of code, I needed to find a way to recognise certain aliases. Thing is, it is not known beforehand what those aliases are.

These are my requirements:

• If A and B are aliases, and B and C are aliases, A and C should be aliases as well.
• Two sets of aliases should be merged when they are connected in any way.
• In each set of aliases, one should be the primary alias.

I use the following solution, making use of what boils down to a dictionary of sets:

class Alias(object):
    def __init__(self, initial):
        self._set = {initial}
        self.initial = initial
    def add(self, alias):
        self._set.add(alias)
    def merge(self, other):
        self._set.update(other._set)
    def __iter__(self):
        return iter(self._set)

class AliasDict(object):
    def __init__(self):
        self._dict = {}
    def add(self, one, other):
        if one in self._dict:
            if other in self._dict: #merge!
                self._dict[one].merge(self._dict[other])
                for k in self._dict[other]:
                    self._dict[k] = self._dict[one]
            else:
                self._dict[one].add(other)
        elif other in self._dict:
            self._dict[other].add(one)
        else:
            self._dict[one] = self._dict[other] = Alias(one)
            self._dict[one].add(other)
    def get(self, n):
        return self._dict.get(n)
    def __contains__(self, s):
        return s in self._dict

Could this be improved upon? For example, by making use of a class in the standard library (I’ve searched, but I might have missed something useful.)

Answer 1

This is something you can map on a graph so I’d do:

from networkx import Graph
from networkx.algorithms.components.connected import connected_components

# see aliases as the edges between nodes in a graph
aliases = [('A', 'B'), ('B', 'C'), ('D','E')]

g = Graph( aliases )

# connected components are alias groups
print connected_components(g) # [['A', 'C', 'B'], ['E', 'D']]

You don’t specify which alias should be the primary one, so you might as well pick the first one from these lists.

networkx module