Sum of list values in a df, new column, values are objects-CodePudding

I have a df made of values from a dictionary. I can get rid of [], ',' and split it all in different cols (one col per number). But can't make the transfer to float, tried different ways. I'm looking for new cols with sum of values for each []. Used (df1['all'][0]) and got the numbers like in column 'all'.

import pandas as pd
data = {'all':['[0.75, 0.34, 0.91, 0.12, 0.5],[0.54, 0.65, 0.5, 0.79, 0.91],[0.81, 0.77, 0.82, 0.66, 0.38],[0.0, 0.78, 0.87, 0.81, 0.67],[0.0, 0.0, 0.56, 0.44, 0.0]']
        }
 
df1 = pd.DataFrame(data)
 
print(df1)

Preferred output

    all                all1    all2   all3    all4    all5
0  [0.75, 0.34.....    2.62    3.39   3.44    3.13    1.0

CodePudding user response：

Try:

from ast import literal_eval

df_out = pd.DataFrame(
    [
        {"all": orig} | {f"all{i}": sum(l) for i, l in enumerate(row, 1)}
        for orig, row in zip(df1["all"], df1["all"].apply(literal_eval))
    ]
)
print(df_out)

Prints:

                                                                                                                                                    all  all1  all2  all3  all4  all5
0  [0.75, 0.34, 0.91, 0.12, 0.5],[0.54, 0.65, 0.5, 0.79, 0.91],[0.81, 0.77, 0.82, 0.66, 0.38],[0.0, 0.78, 0.87, 0.81, 0.67],[0.0, 0.0, 0.56, 0.44, 0.0]  2.62  3.39  3.44  3.13   1.0

CodePudding user response：

from ast import literal_eval

pd.concat(
    [
        df1['all'], 
        pd.DataFrame(df1['all'].apply(literal_eval)[0])
            .sum(axis=1).to_frame().T
    ],
    axis=1
).rename(columns = lambda x: 'all'   str(x 1) if x != 'all' else x)

                                                 all  all1  all2  all3  all4  all5
0  [0.75, 0.34, 0.91, 0.12, 0.5],[0.54, 0.65, 0.5...  2.62  3.39  3.44  3.13   1.0