如何在pydatatable中加入具有不同键列名称的两个数据框？

Question

I have a X dataframe as,

DT_X = dt.Frame({
    
    'date':['2020-09-01','2020-09-02','2020-09-03'],
    'temp':[35.3,32.9,43.2]
    
})

Out[4]: 
   | date        temp
-- + ----------  ----
 0 | 2020-09-01  35.3
 1 | 2020-09-02  32.9
 2 | 2020-09-03  43.2

[3 rows x 2 columns]

Another dataframe Y as,

DT_Y = dt.Frame({
    
    'stop_date' : ['2020-08-01','2020-09-01','2020-09-03','2020-09-07'],
    'is_arrested':[True,False,False,True]
    
})

Out[6]: 
   | stop_date   is_arrested
-- + ----------  -----------
 0 | 2020-08-01            1
 1 | 2020-09-01            0
 2 | 2020-09-03            0
 3 | 2020-09-07            1

[4 rows x 2 columns]

Now I would like to perform JOIN operation on X and Y, for that i'm supposed to assign a key on X dataframe as,

DT_X.key='date'

Out[8]: 
date       | temp
---------- + ----
2020-09-01 | 35.3
2020-09-02 | 32.9
2020-09-03 | 43.2

[3 rows x 2 columns]

Next I'm joining X and Y as ,

DT_Y[:,:,join(DT_X)]

Here it is throwing out an error as ,

In [9]: DT_Y[:,:,join(DT_X)]                                                                                                                                                                      
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-9-a3bc1690fb98> in <module>
----> 1 DT_Y[:,:,join(DT_X)]

ValueError: Key column `date` does not exist in the left Frame

Of course date is not existed in DT_Y, it has a column name as stop_date .

How to perform join operations in this scenario ?? ie No match of column names.

Note :

An work around for this is to change the column name of DT_Y as

DT_Y.names = {'stop_date':'date'}

DT_Y[:,:,join(DT_X)]

The joined DT can be viewed as,

Out[11]: 
   | date        is_arrested  temp
-- + ----------  -----------  ----
 0 | 2020-08-01            1  NA  
 1 | 2020-09-01            0  35.3
 2 | 2020-09-03            0  43.2
 3 | 2020-09-07            1  NA  

[4 rows x 3 columns]

Here is the expected output:

Out[13]: 
   | stop_date   is_arrested  temp
-- + ----------  -----------  ----
 0 | 2020-08-01            1  NA  
 1 | 2020-09-01            0  35.3
 2 | 2020-09-03            0  43.2
 3 | 2020-09-07            1  NA  

[4 rows x 3 columns]

Answer 1

Right now, join() only supports the same column names in the both frames, please refer to documentation for more details. However, there is an open issue to improve the join functionality/API.

Meanwhile, if you prefer not to rename the columns you can do the following

DT_Y_date = DT_Y[:, {"date":f[0], "is_arrested":f[1]}]
DT_YX_joined = DT_Y_date[:, :, join(DT_X)]

Then, DT_YX_joined will have the data you are looking for

   | date        is_arrested  temp
-- + ----------  -----------  ----
 0 | 2020-08-01            1  NA  
 1 | 2020-09-01            0  35.3
 2 | 2020-09-03            0  43.2
 3 | 2020-09-07            1  NA 

You can even do a one-liner like

DT_YX_joined = DT_Y[:, {"date":f[0], "is_arrested":f[1]}][:, :, join(DT_X)]

but it may not be readable enough. Also note, that no data copies are created here, it is only the column name that changes.