Pandas DataFrame：减去字符串数据类型的列

Question

How can I subtract two columns that contain values of type string? No values are indicated by '---' and should lead to a '---' in the result. The result should also be of value type string.

Source

df1 = pd.DataFrame({'x': ['a', 'b', 'c'], 'y': ['5', '---', '7']})

    x   y
0   'a' '5'
1   'b' '---'
2   'c' '7'

df2 = pd.DataFrame({'x': ['a', 'b', 'c'], 'y': ['1', '2', '---']})

    x    y
0   'a'  '1'
1   'b'  '2'
2   'c'  '---'

Target

df3 = df1 - df2

    x   y
0   'a' '4'
1   'b' '---'
2   'c' '---'

Answer 1

Try with:

df1.set_index('x').apply(lambda x: pd.to_numeric(x,errors='coerce')).sub(
      df2.set_index('x').apply(lambda x: pd.to_numeric(x,errors='coerce'))).fillna('--')\
                                                                .reset_index()

---

   x   y
0  a   4
1  b  --
2  c  --

Answer 2

You could use pd.to_numeric to both replace all '---' to NaNs , and also also cast all values to floats :

df1['y'] = pd.to_numeric(df1['y'], errors='coerce')
df2['y'] = pd.to_numeric(df2['y'], errors='coerce')

The simply subtract both columns and store the result in df1 for instance:

df1['y'] = (df1['y'] - df2['y']).replace(np.nan,'---')

   x    y
0  a    4
1  b  ---
2  c  ---