I have a pandas DataFrame with a combination of numpy.ndarray and categorical variables.
import pandas as pd
import numpy as np
df = pd.DataFrame({
'i': [0, 1, 2],
'a': [np.array([]), np.array([]), 'A']
})
df['a'][0]
array([], dtype=float64)
I'd like to replace the arrays with np.nan.
I tried: df['a'].replace(np.array([]), np.nan) but it didn't work.
If you want to replace empty arrays by nan, you can convert the column a as boolean mask:
df['a'] = df['a'].where(df['a'].astype(bool), np.nan)
print(df)
# Output
i a
0 0 NaN
1 1 NaN
2 2 A
One way could be checking with isinstance:
# on one column
df['a'] = df['a'].apply(lambda x: np.nan if isinstance(x, np.ndarray) else x)
#whole df
df = df.applymap(lambda x: np.nan if isinstance(x, np.ndarray) else x)
