Convert a list of dictionaries into a set of dictionaries

Question

How can i make a set of dictionaries from one list of dictionaries?

Example:

import copy

v1 = {'k01': 'v01', 'k02': {'k03': 'v03', 'k04': {'k05': 'v05'}}}
v2 = {'k11': 'v11', 'k12': {'k13': 'v13', 'k14': {'k15': 'v15'}}}

data = []
N = 5
for i in range(N):
    data.append(copy.deepcopy(v1))
    data.append(copy.deepcopy(v2))

print data

How would you create a set of dictionaries from the list data?

NS: One dictionary is equal to another when they are structurally the same. That means, they got exactly the same keys and same values (recursively)

Answer 1

A cheap workaround would be to serialize your dicts, for example:

import json

dset = set()

d1 = {'a':1, 'b':{'c':2}}
d2 = {'b':{'c':2}, 'a':1} # the same according to your definition
d3 = {'x': 42}

dset.add(json.dumps(d1, sort_keys=True))
dset.add(json.dumps(d2, sort_keys=True))
dset.add(json.dumps(d3, sort_keys=True))

for p in dset:
    print json.loads(p) 

In the long run it would make sense to wrap the whole thing in a class like SetOfDicts.

Answer 2

Dictionaries are mutable and therefore not hashable in python.

You could either create a dict-subclass with a __hash__ method. Make sure that the hash of a dictionary does not change while it is in the set (that probably means that you cannot allow modifying the members). See http://code.activestate.com/recipes/414283-frozen-dictionaries/ for an example implementation of frozendicts.

If you can define a sort order on your (frozen) dictionaries, you could alternatively use a data structure based on a binary tree instead of a set. This boils down to the bisect solution provided in the link below.

See also https://stackoverflow.com/a/18824158/5069869 for an explanation why sets without hash do not make sense.