熊猫根据另一列将条件应用于列值

Question

I have a df that looks like the below:

        date   2017-10-07  2017-10-08
id                                                                 
1    2017-09-07      46.0         5.0   
2    2017-10-07       1.0         0.0   
3    2017-11-07     123.0         0.0   
4    2017-10-07       0.0         0.0   
5    2017-10-08       0.0         0.0

and want to be able to lookup the date value from the 'date' column, and if it matches the column name of the columns with date titles, change that specific value in the df.

For instance, in the df above, the output would look like:

        date   2017-10-07  2017-10-08
id                                                                 
1    2017-09-07      46.0         5.0   
2    2017-10-07       1.0         0.0   
3    2017-11-07     123.0         0.0   
4    2017-10-07   "CHANGED"       0.0   
5    2017-10-08       0.0      "CHANGED"

I can do this in a for loop over each column and row, but it is time consuming and I know there has to be a better way.

Thanks in advance!

Answer 1

You can use numpy broadcasting, perform assignment, and reassign the result back.

v = df.values[:, 1:]
v[df.date.values[:, None] == df.columns[1:].values] = 'CHANGED'
df.iloc[:, 1:] = v

df

          date 2017-10-07 2017-10-08
id                                  
1   2017-09-07         46          5
2   2017-10-07    CHANGED          0
3   2017-11-07        123          0
4   2017-10-07    CHANGED          0
5   2017-10-08          0    CHANGED

Answer 2

By using stack and unstack

df1=df.reset_index().melt(['id','date'])
df1.loc[df1.date==df1.variable,'value']='changed'
df1.set_index(['id','variable','date']).unstack(-2)
Out[189]: 
                   value           
variable      2017-10-07 2017-10-08
id date                            
1  2017-09-07         46          5
2  2017-10-07    changed          0
3  2017-11-07        123          0
4  2017-10-07    changed          0
5  2017-10-08          0    changed