how to subtract string type columns values from another column in pandas

Use list comprehension with replace and split:

df['col2'] = [a.replace(b, '').strip() for a, b in zip(df['col2'], df['col3'])]
print (df)
  col1   col2   col3
0    A  berry  black
1    B  apple  green
2    C   wine    red

If order is not important convert splitted values to sets and subtract:

df['col2'] = [' '.join(set(a.split())-set([b])) for a, b in zip(df['col2'], df['col3'])]
print (df)
  col1   col2   col3
0    A  berry  black
1    B  apple  green
2    C   wine    red

Or use generator with if condition and join:

df['col2'] = [' '.join(c for c in a.split() if c != b) for a, b in zip(df['col2'], df['col3'])]

Performance:

pic

This was the setup used to generate the perfplot above:

def calculation(val):
    return val[0].replace(val[1],'').strip()


def regex(df):
    df.col2=df.col2.replace(regex=r'(?i)'+ df.col3,value="")
    return df

def lambda_f(df):
    df["col2"] = df.apply(lambda x: x["col2"].replace(x["col3"], "").strip(), axis=1)
    return df

def apply(df):
    df['col2'] = df[['col2','col3']].apply(calculation, axis=1)
    return df

def list_comp1(df):
    df['col2'] = [a.replace(b, '').strip() for a, b in zip(df['col2'], df['col3'])]
    return df

def list_comp2(df):
    df['col2'] = [' '.join(set(a.split())-set([b])) for a, b in zip(df['col2'], df['col3'])]
    return df

def list_comp3(df):
    df['col2'] = [' '.join(c for c in a.split() if c != b) for a, b in zip(df['col2'], df['col3'])]
    return df


def make_df(n):
    d = {'col1': {0: 'A', 1: 'B', 2: 'C'}, 'col2': {0: 'black berry', 1: 'green apple', 2: 'red wine'}, 'col3': {0: 'black', 1: 'green', 2: 'red'}}
    df = pd.DataFrame(d)
    df = pd.concat([df] * n * 100, ignore_index=True)
    return df

perfplot.show(
    setup=make_df,
    kernels=[regex, lambda_f, apply, list_comp1,list_comp2,list_comp3],
    n_range=[2**k for k in range(2, 10)],
    logx=True,
    logy=True,
    equality_check=False,  # rows may appear in different order
    xlabel='len(df)')

We can use the apply method:

def calculation(val):
    return val[0].replace(val[1],'').strip()

df['col4'] = df[['col2','col3']].apply(calculation, axis=1)

df:
  col1         col2   col3   col4
0    A  black berry  black  berry
1    B  green apple  green  apple
2    C    red wine     red   wine

Single line solution:

df["col2"] = df.apply(lambda x: x["col2"].replace(x["col3"], "").strip(), axis=1)