Inverting a dictionary with list values

Question

I have this index as a dict.

index = {
    'Testfil2.txt': ['nisse', 'hue', 'abe', 'pind'],
    'Testfil1.txt': ['hue', 'abe', 'tosse', 'svend']}

I need to invert the index so it will be a dict with duplicates of values merged into one key with the 2 original keys as values, like this:

inverse = {
    'nisse': ['Testfil2.txt'],
    'hue': ['Testfil2.txt', 'Testfil1.txt'],
    'abe': ['Testfil2.txt', 'Testfil1.txt'],
    'pind': ['Testfil2.txt'], 
    'tosse': ['Testfil1.txt'],
    'svend': ['Testfil1.txt']}

My textbook has this function for inverting dictionaries:

def invert_dict(d): 
    inverse = dict() 
    for key in d: 
        val = d[key] 
        if val not in inverse: 
            inverse[val] = [key] 
        else: 
            inverse[val].append(key) 
    return inverse

It works fine for simple key:value pairs, BUT, when I try that function with a dict that has lists as values such as my index, I get this error message:

Traceback (most recent call last):
  File "<pyshell#153>", line 1, in <module>
    invert_dict(index)
  File "<pyshell#150>", line 5, in invert_dict
    if val not in inverse:
TypeError: unhashable type: 'list'

The book is no help, and I suspect that I can use tuples in some way, but I am not sure how.

Answer 1

My solution for reversing a dictionary.

inverse = {}
for k,v in index.items():
    for x in v:
        inverse.setdefault(x, []).append(k)

Output:

{'nisse': ['Testfil2.txt'],
 'hue': ['Testfil2.txt', 'Testfil1.txt'],
 'abe': ['Testfil2.txt', 'Testfil1.txt'],
 'pind': ['Testfil2.txt'],
 'tosse': ['Testfil1.txt'],
 'svend': ['Testfil1.txt']}

Answer 2

I've tried around and you want to use val not in inverse but it can't be checked if a "list is in a dict". (val is a list)

For your code a simple change will do what you want:

def invert_dict(d): 
    inverse = dict() 
    for key in d: 
        # Go through the list that is saved in the dict:
        for item in d[key]:
            # Check if in the inverted dict the key exists
            if item not in inverse: 
                # If not create a new list
                inverse[item] = [key] 
            else: 
                inverse[item].append(key) 
    return inverse

Answer 3

As a nested comprehension:

inverse = { v: k for k, l in index.items() for v in l }

or, perhaps more clearly:

inverse = { 
            new_key: index_key                              #body
            for index_key, index_value in index.items()     #outer loop
                for new_key in index_value                  #inner loop
            }

which is roughly equivalent to:

new_keys    =   []
new_values  =   []

for index_key, index_value in index.items():
    for new_key in index_value:
        new_keys.append(new_key)
        new_values.append(index_key)
        
inverse     =   dict(zip(new_keys,new_values))

Answer 4

You can not use list objects as dictionary keys, since they should be hashable objects. You can loop over your items and use dict.setdefault method to create the expected result:

new = {}
for k,value in index.items():
    for v in value:
        new.setdefault(v, []).append(k)

Result:

{'nisse': ['Testfil2.txt'],
 'hue': ['Testfil2.txt', 'Testfil1.txt'],
 'abe': ['Testfil2.txt', 'Testfil1.txt'],
 'pind': ['Testfil2.txt'],
 'tosse': ['Testfil1.txt'],
 'svend': ['Testfil1.txt']}

and if you are dealing with larger datasets for refusing of calling creating an empty list at each calling the setdefault() method you can use collections.defaultdict() which will calls the missing function just when it encounter a new key.

from collections import defaultdict

new = defaultdict(list)
for k,value in index.items():
    for v in value:
        new[v].append(k)

Result:

defaultdict(<type 'list'>,
    {'nisse': ['Testfil2.txt'],
     'hue': ['Testfil2.txt', 'Testfil1.txt'],
     'abe': ['Testfil2.txt', 'Testfil1.txt'],
     'pind': ['Testfil2.txt'],
     'tosse': ['Testfil1.txt'],
     'svend': ['Testfil1.txt']})

Answer 5

Here is a variation that uses a comprehension plus set to remove duplicates.

def invert_setdict(setdict):
    inverse = {}
    vk = [(v, k) for k, vs in index.items() for v in vs]
    for k, v in vk:
       inverse.setdefault(k, set()).add(v)

    return inverse

Example

>>> index = {
... 'Testfil2.txt': ['nisse', 'hue', 'abe', 'pind'],
... 'Testfil1.txt': ['hue', 'abe', 'tosse', 'svend']}

>>> inverse = invert_setdict(index)
>>> inverse
{'nisse': {'Testfil2.txt'},
 'hue': {'Testfil1.txt', 'Testfil2.txt'},
 'abe': {'Testfil1.txt', 'Testfil2.txt'},
 'pind': {'Testfil2.txt'},
 'tosse': {'Testfil1.txt'},
 'svend': {'Testfil1.txt'}}

If you want to convert the set values to lists:

>>> inverse = {k:list(v) for k, v in inverse.items()}

Answer 6

Two liner solution using unpacking operator * and nested compression.

for k,v in old_dict.items():
    new_dict = {**new_dict,**{vi:k for vi in v}}