I think need convert values to lists and then call DataFrame contructor:
df = pd.DataFrame({'list':[np.array([1,2,3]), np.array([7,8,3]), np.array([3,7,0])]})
print (df)
list
0 [1, 2, 3]
1 [7, 8, 3]
2 [3, 7, 0]
df = pd.DataFrame(df['list'].values.tolist(), index=df.index)
print (df)
0 1 2
0 1 2 3
1 7 8 3
2 3 7 0
And last join to original df:
df = df.join(pd.DataFrame(df['list'].values.tolist(), index=df.index))
print (df)
list 0 1 2
0 [1, 2, 3] 1 2 3
1 [7, 8, 3] 7 8 3
2 [3, 7, 0] 3 7 0
Another slowier solution is:
df = df.join(df['list'].apply(pd.Series))
Performance:
The plot was created with perfplot:
np.random.seed(57)
def apply(df):
df = df.join(df['list'].apply(pd.Series))
return df
def values(df):
df = df.join(pd.DataFrame(df['list'].values.tolist(), index=df.index))
return df
def make_df(n):
df = pd.DataFrame({'list': np.random.randint(10, size=(n, 10)).tolist()})
return df
perfplot.show(
setup=make_df,
kernels=[ apply, values],
n_range=[2**k for k in range(2, 17)],
logx=True,
logy=True,
equality_check=False, # rows may appear in different order
xlabel='len(df)')
