我有一些使用pivot_table的数据，如下所示：
# Sample data
import pandas as pd
df = pd.DataFrame({
'building': ['ab1', 'ab1', 'ab1', 'ab1', 'ab1', 'ab2', 'ab2', 'ab2', 'ab2'],
'year': [2017, 2017, 2017, 2018, 2018, 2017, 2017, 2018, 2018],
'unit': ['blah', 'blah', 'crap', 'blah', 'crap', 'blah', 'crap', 'blah', 'crap'],
'value1': [5.1, 5.3, 4.5, 5.7, 4.7, 3.8, 4.9, 6.0, 5.3],
'value2': [100, 102, 110, 105, 106, 112, 107, 111, 102]
})
df
building year unit value1 value2
0 ab1 2017 blah 5.1 100
1 ab1 2017 blah 5.3 102
2 ab1 2017 crap 4.5 110
3 ab1 2018 blah 5.7 105
4 ab1 2018 crap 4.7 106
5 ab2 2017 blah 3.8 112
6 ab2 2017 crap 4.9 107
7 ab2 2018 blah 6.0 111
8 ab2 2018 crap 5.3 102
# Use pivot_table to cast out unit
df2 = pd.pivot_table(df, index=['building', 'year'], values=['value1', 'value2'], columns='unit').reset_index()
df2
building year value1 value2
unit blah crap blah crap
0 ab1 2017 5.2 4.5 101 110
1 ab1 2018 5.7 4.7 105 106
2 ab2 2017 3.8 4.9 112 107
3 ab2 2018 6.0 5.3 111 102
但是，df2是一种奇怪的非数据帧格式。我怎样才能得到像下面这样的东西？
building year blah_value1 crap_value1 blah_value2 crap_value2
0 ab1 2017 5.2 4.5 101 110
1 ab1 2018 5.7 4.7 105 106
2 ab2 2017 3.8 4.9 112 107
3 ab2 2018 6.0 5.3 111 102