Pandas groupby "nominal" column (one hot list)

I'm attempting to group a dataframe by class where the class is represented as a one hot encoding. I build the one-hot encoding as a list in the column. However when I attempt to do a groupby it raises and error that:

> TypeError: unhashable type: 'list'

Here is a minimal reproducible example:

import pandas

data = pandas.DataFrame({"first": [0, 1, 2, 3, 4],
                         "class": [1, 0, 0, 0, 1]})

class_groups = [data for group, data in data.groupby("class")]
print(class_groups)

data = pandas.DataFrame({"first": [0, 1, 2, 3, 4],
                         "class": [[0, 1], [1, 0], [1, 0], [0, 1], [1, 0]]})

class_groups = [data for group, data in data.groupby("class")]
print(class_groups)

The first is an ordinal example which works well and then a similar nominal format which throws an error. Maybe there's another way I really should be formatting this that would be easier to groupby. But it does need to be a one-hot encoding.

You need to use non-mutable objects in the grouping Series. One option is to convert to tuple:

class_groups = [data for group, data in data.groupby(data["class"].apply(tuple))]

Output:

[   first   class
0      0  [0, 1]
3      3  [0, 1],
    first   class
1      1  [1, 0]
2      2  [1, 0]
4      4  [1, 0]]

You can do:

[data for group, data in data.groupby(data['class'].map(tuple))]

#output

[   first   class
 0      0  [0, 1]
 3      3  [0, 1],
    first   class
 1      1  [1, 0]
 2      2  [1, 0]
 4      4  [1, 0]]

Edit:

With your data you can get both class groups working by if/else:

if isinstance(data['class'][0], list):
   class_groups = [data for group, data in data.groupby(data['class'].map(tuple))] 
else:
   class_groups = [data for group, data in data.groupby("class")]

Case 1:

import pandas

data = pandas.DataFrame({"first": [0, 1, 2, 3, 4],
                         "class": [1, 0, 0, 0, 1]})
print(class_groups)

[   first  class
 1      1      0
 2      2      0
 3      3      0,
    first  class
 0      0      1
 4      4      1]

Case 2:

data = pandas.DataFrame({"first": [0, 1, 2, 3, 4],
                         "class": [[0, 1], [1, 0], [1, 0], [0, 1], [1, 0]]})

print(class_groups)

[   first   class
 0      0  [0, 1]
 3      3  [0, 1],
    first   class
 1      1  [1, 0]
 2      2  [1, 0]
 4      4  [1, 0]]