import pandas as pd 

data = {'A': [1,2],
'B':[[1,1,1,2,2,4,4,4,4],[5, 4, 8, 1, 1, 1, 3, 2, 4, 2, 2, 2, 1, 1, 1]]}

df = pd.DataFrame(data)

def top_frequent(a):

    import numpy 
    k = {}
    for j in a:
        if j in k:
            k[j]  =1
        else:
            k[j] =1

    occ = []
    for key, val in k.items():
        occ.append(val)
    Z = numpy.percentile(occ, 75, interpolation='higher')
    print(Z)
    
    bucket = [[] for l in range(len(a) 1)]    
    for key, val in k.items():
        if val >= Z :
            if val != 1 : 
                bucket[val].append(key)

    res = []
    for i in reversed(range(len(bucket))):
        if bucket[i]:
            res.extend(bucket[i])

    return res

df['C'] = df.apply(top_frequent(df['B']))

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_13728/2052560572.py in <module>
     28     return res
     29 
---> 30 df['C'] = df.apply(top_frequent(df['B']))

~\AppData\Local\Temp/ipykernel_13728/2052560572.py in top_frequent(ids)
      4     k = {}
      5     for j in ids:
----> 6         if j in k:
      7             k[j]  =1
      8         else:

TypeError: unhashable type: 'list'

When I apply the function on just one row it works fine But when I apply it for all lines I get this error : TypeError: unhashable type: 'list'

CodePudding user response：

The problem is that when you pass df['B'] into top_frequent(), df['B'] is a column of list, you can view is as a list of list.

So in your for j in a:, you are getting item from outer list. For list of list, what you get is a list.

Then in k[j], you are using a list as key which is not supported by Python list. So it gives you the error TypeError: unhashable type: 'list'.

You can try

df['C'] = df['B'].apply(top_frequent)

# or

df['C'] = df.apply(lambda row: top_frequent(row['B']), axis=1)

Besides you can use a more pandas way to do this

df['C'] = df['B'].apply(lambda x: (lambda y: (y[y==y.max()].index.tolist()))(pd.Series(x).value_counts()))