How to handle typeErrors when doing vectorize calculations?

Question

I want to avoid crashes when performing vectorized calculations using pandas dataframes (python-3.6).

For example I have a dataframe with 2 Columns A,B. I want to create a column C that will be C = A - B. However one cell in column A is a string and this cause a TypeError. Have a look at the picture below.

Column C is the outcome that I want to achieve.

Currently I get an Type Error message:

TypeError: unsupported operand type(s) for -: 'float' and 'str'

which is expected.

Answer 1

It is possible by numpy.select , but get mixed values in output:

df = pd.DataFrame({
         'A':[7,8,9,10,5],
         'B':[1,2,3,'str',np.nan],
})

b = pd.to_numeric(df['B'], errors='coerce')
df['C'] = np.select([df['B'].isna(), b.isna()], [np.nan, 'ERROR'], default=df['A'] - b)
print (df)
    A    B      C
0   7    1    6.0
1   8    2    6.0
2   9    3    6.0
3  10  str  ERROR
4   5  NaN    nan

The best is convert to numeric by to_numeric and subtract only if need processing column later:

b = pd.to_numeric(df['B'], errors='coerce')
df['C'] = df['A'] - b
print (df)
    A    B    C
0   7    1  6.0
1   8    2  6.0
2   9    3  6.0
3  10  str  NaN
4   5  NaN  NaN