[FIXED] Count unique values in each dataframe's column and transform to dictionary

Issue

data = {'Col1': [1, 2, 2, 3, 1],
    'Col2': ['A', 'B', 'B', 'A', 'C']}
df = pd.DataFrame(data)

i want to get a dictionary like:

{'Col1': {1:2, 2:2, 3:1},
 'Col2': {'A':2, 'B':2, 'C':1}

without using any kind of loop, apply or agg method.

I have tried something like this:

count_matrix = df.stack().groupby(level=1).value_counts()
count_matrix = count_matrix.unstack(0)
count_matrix = count_matrix.to_dict()

but it does not work, because it inserts nan value to fill blanks when unstacking.

Solution

If you want to use pure pandas without explicit loops, you can go with agg, value_counts and to_dict:

df.agg(lambda x: x.value_counts().to_dict()).to_dict()

Or, overriding the loop limitation with map:

dict(map(lambda x: (x[0], x[1].value_counts().to_dict()), df.items()))

Output:

{'Col1': {1: 2, 2: 2, 3: 1}, 'Col2': {'A': 2, 'B': 2, 'C': 1}}

Answered By - mozway