将两个 pandas 数据帧与两个条件结合起来

Question

There are two pandas dataframes I have which I would like to combine with checking of two conditionals.

Dataframe1:

import pandas as pd 
data = [['Z085', '2020-08', 1.33], ['Z086', '2020-08', 1.83], ['Z086', '2020-09', 1.39]] 
df1 = pd.DataFrame(data, columns = ['SN', 'Date', 'Value']) 

Dataframe2:

data = [['Z085', '2020-08', 0.34], ['Z085', '2020-09', 0.83], ['Z086', '2020-09', 0.29]] 
df2 = pd.DataFrame(data, columns = ['SN', 'Date', 'ValueX']) 
df2 

I would like to merge or append or join them in order to get the folowing dataframe: The values ("Value" and "ValueX") are being add if both "SN" and "Date" are equal.

I am not sure, if a new dataframe is required or to map the df2 to the df1.

This is what i have tried:

df1['ValueX'] = df1[('Date', 'SN')].map(df2_mean.set_index('Date', 'SN')['ValueX'])

With one conditional (for example: Date) it works ok, but i am not able to set up two conditionals.

Answer 1

This is simply a merge() operation. Don't call the columns "conditionals", just say "merge on the columns SN, Date".

However pandas (v1.1.4) has a bug (its default is to use reversed ie 'ascending') key order when doing the sort) so you can't rely on it; note below it gets sorted by 'Date' then 'SN', ie wrong-way-around:

>>> dfnew_bad = df1.merge(df2, on=['SN','Date'], how='outer')

     SN     Date  Value  ValueX
0  Z085  2020-08   1.33    0.34
1  Z086  2020-08   1.83     NaN
2  Z086  2020-09   1.39    0.29
3  Z085  2020-09    NaN    0.83

So in your case to get the correct order by SN then Date :

dfnew_good = df1.merge(df2, on=['SN','Date'], how='outer', sort=False).sort_values(['SN', 'Date'])
     SN     Date  Value  ValueX
0  Z085  2020-08   1.33    0.34
3  Z085  2020-09    NaN    0.83
1  Z086  2020-08   1.83     NaN
2  Z086  2020-09   1.39    0.29

Note that there's a flag .sort_values(ascending=True) but not pd.merge() You could also workaround by doing pd.merge(..., sort=False) then dfnew_workaround.sort_index(..., inplace=True)

Answer 2

Method 1: merge :

df_new = df1.merge(df2, on=['SN','Date'],how='outer', sort=True)
print(df_new)

Method 2: join :

df_new = df1.join(df2.set_index(['SN','Date']), on=['SN','Date'],how='outer', sort=True)
print(df_new)

In this case, one more possible way would be to use pd.concat :

df_new = pd.concat([df1.set_index(['SN','Date']),df2.set_index(['SN','Date'])],axis=1).reset_index()

Output in either case :

     SN     Date  Value  ValueX
0  Z085  2020-08   1.33    0.34
3  Z085  2020-09    NaN    0.83
1  Z086  2020-08   1.83     NaN
2  Z086  2020-09   1.39    0.29