用于数据帧中多列操作的矢量化解决方案

Question

i probably will have a large dataframe which has the following first row:

   BUCHDAT     y  y1   y2    y3     y4    y5     y6     y7
7 2017-02-26 577 30.0 622.0 1785.0 2633.0 422.0 10497.0 364.0 

Now i want to replace the columns 'y' till 'y7' with a formula:

df['y'] = df['y'] - df['y1']

Is there any vectorization solution for this? So I want to apply this formula in every column, for the next column the formula should be:

df['y1'] = df['y1']- df['y2']

you have any idea how to do it?

Answer 1

Use DataFrame.sub with DataFrame.shift :

df1 = df.iloc[:, 1:].astype(float)
df.iloc[:, 1:] = df1.sub(df1.shift(-1, axis=1))
print (df)
      BUCHDAT      y     y1      y2     y3      y4       y5       y6  y7
7  2017-02-26  547.0 -592.0 -1163.0 -848.0  2211.0 -10075.0  10133.0 NaN

Answer 2

Here's one working with the underlying numpy arrays for a good performance:

df.iloc[:,1:-1] = df.values[:,1:-1] - df.values[:,2:]

print(df)

    BUCHDAT      y     y1      y2     y3      y4       y5       y6     y7
7  2017-02-26  547.0 -592.0 -1163.0 -848.0  2211.0 -10075.0  10133.0  364.0