使用列表、元组或字典作为单个条目值从字典创建 Pandas DataFrame

Question

I have a dictionary

data = { 'x' : 1,
         'y' : [1,2,3],
         'z' : (4,5,6),
         'w' : {1:2, 3:4}
       }

I'd like to construct a Pandas DataFrame such that the list and tuple do not get broadcasted:

df = pd.DataFrame(some_transformation(data), index=['a'])

to get

df = 
      x         y         z          w
a     1   (1,2,3)   (4,5,6)  (1,2,3,4)

Or some sort of flattening and/or string-fy of the list/tuple/dict. What is the easiest / most efficient way of doing so, without having to go down the exact data structure of each dictionary entry?

Answer 1

without going down the exact data structure, I think the easiest way to achieve what you want is:

      data={k:str(v) for k,v in data}

Above statement will make all values as string type. Now you can convert the data dictionary to a dataframe by using below line:

    df=pd.DataFrame(data, index=[0])

This will get you the output in below form:

           w        x          y          z
    0 {1: 2, 3: 4}  1      [1, 2, 3]  (4, 5, 6)

Now for your desired output: (you can use other efficent methods as well for string replacement in dataframe)

      for acol in df.columns:
            a[acol]=a[acol].values[0].strip('[{()}]')
            a[acol]=a[acol].values[0].replace(':', ',')

Output looks like

                 w         x        y          z

            1, 2, 3, 4     1    1, 2, 3     4, 5, 6

Answer 2

You cannot apply one transformation to lists/tuples and dictionaries. They have very different properties. You can flatten all dictionaries and then create a pd.Series out of the updated dictionary.

for key in data:
    if isinstance(data[key],dict):
        data[key] = list(data[key].keys())+list(data[key].values())
pd.Series(data)
#w    [1, 3, 2, 4]
#x               1
#y       [1, 2, 3]
#z       (4, 5, 6)
#dtype: object

Convert it further into a DataFrame, if you want:

df = pd.DataFrame(pd.Series(data)).T
#              w  x          y          z
#0  [1, 3, 2, 4]  1  [1, 2, 3]  (4, 5, 6)

You can handle lists in the same spirit (convert them to tuples).

Answer 3

This is one way.

def transformer(data):
    for k, v in data.items():
        if isinstance(v, list):
            data[k] = [tuple(v)]
        elif isinstance(v, dict):
            data[k] = [tuple(chain(*(v.items())))]
        else:
            data[k] = [v]
    return data

df = pd.DataFrame(transformer(data), index=['a'])

#               w  x          y          z
# a  (1, 2, 3, 4)  1  (1, 2, 3)  (4, 5, 6)

Answer 4

You can use set_value to assign those elements to the df and then transform dict and list to tuples.

df=pd.DataFrame(columns=data.keys())
[df.set_value(0,k,v) for k,v in data.items()]
df = df.applymap(lambda x: sum([[k,v] for k,v in x.items()],[]) if isinstance(x,dict) else x)
df = df.applymap(lambda x: tuple(x) if isinstance(x,list) else x)
Out[716]: 
   x          y          z             w
0  1  (1, 2, 3)  (4, 5, 6)  (1, 2, 3, 4)