Styjun

I have test names in a list like this in Python v2.7:

all_tests = ['test1', 'test2', 'test3', ..., 'testN']

And for each test name, I can lookup a set of properties that the test has:

series['test'] = 'prop1', 'prop2', 'prop3', ..., 'propN'

E.g.,

series['test1'] = 'yellow', 'blue', 'orange'
series['test2'] = 'blue', 'red', 'black'
series['test3'] = 'yellow', 'green', 'black'

Now, I wish to reverse the direction and create a dictionary from properties to the set of tests that have that property.

So for the above example three tests, I'd like to create this dictionary:

result = 
 'yellow': 'test1', 'test3',
 'blue': 'test1', 'test2',
 'orange': 'test1',
 'red': 'test2',
 'black': 'test2', 'test3'
 'green': 'test3'

I imagine there is a way with list comprehension to build the dictionary, but I'm unclear how to do so. I'm thinking something along these lines:

tuple( series[test], test ) for test in all_tests

But don't know how to put that into the result dictionary and have the set of dictionary values for each key, keep getting added to.

A straightforward looping solution using collections.defaultdict

from collections import defaultdict
d = defaultdict(set)
for k,v in series.items():
 for prop in v:
 d[prop].add(k)
print(d)
#defaultdict(set,
# 'black': 'test2', 'test3',
# 'blue': 'test1', 'test2',
# 'green': 'test3',
# 'orange': 'test1',
# 'red': 'test2',
# 'yellow': 'test1', 'test3')

Another approach using pandas

First convert series into a DataFrame:

import pandas as pd
df = pd.DataFrame(k: list(v) for k, v in series.items())
print(df)
# test1 test2 test3
#0 blue blue black
#1 orange black green
#2 yellow red yellow

Next melt the DataFrame, groupby the value and use set as the aggregate function:

print(pd.melt(df).groupby("value", as_index=True).agg(set))
# variable
#value 
#black test3, test2
#blue test1, test2
#green test3
#orange test1
#red test2
#yellow test1, test3

Finally to go back to a dictionary, call to_records() and apply the dict constructor:

print(dict(pd.melt(df).groupby("value", as_index=True).agg(set).to_records()))
#'black': 'test2', 'test3',
# 'blue': 'test1', 'test2',
# 'green': 'test3',
# 'orange': 'test1',
# 'red': 'test2',
# 'yellow': 'test1', 'test3'

Here's an alternative that creates a regular dictionary:

series = 
 'test1': 'yellow', 'blue', 'orange',
 'test2': 'blue', 'red', 'black',
 'test3': 'yellow', 'green', 'black',


result = 
for name, props in series.items():
 for prop in props:
 result.setdefault(prop, set()).add(name)

from pprint import pprint
pprint(result)

Output:

'black': set(['test2', 'test3']),
 'blue': set(['test1', 'test2']),
 'green': set(['test3']),
 'orange': set(['test1']),
 'red': set(['test2']),
 'yellow': set(['test1', 'test3'])

Another alternative to get same result:

from itertools import chain

properties = set(chain.from_iterable(series.values())) # All possible.
result = prop: name for name, props in series.items() if prop in props
 for prop in properties

from pprint import pprint
pprint(result)

Do you already have a set or any iterable containing all properties, e.g. all_props?

Then your reverse dict could be created with

rev_series = p: k for k, v in series.items() if p in v for p in all_props

If not:

all_props = set()
for s in series.values():
 all_props = all_props.union(s)

Build a set of all the available colors. Then, for each color create a set of the series entries that contain that color:

colors = set( color for colors in series.values() for color in colors ) 
result = color:set(test for test in series if color in series[test]) for color in colors 
print(result)
# 'green': 'test3', 'red': 'test2', 'black': 'test2', 'test3', 'yellow': 'test1', 'test3', 'orange': 'test1', 'blue': 'test1', 'test2'