replace celle within a dataframe according to matching element in another df in pandas-CodePudding

I have two dataframes such as

Tab1 :

ORFs_values      Groups       SP1  SP2   SP3
SP_ORF1          Group1       1    1     0
SP_ORF1          Group2       0    0     0
SP_ORF1          Group3       0    1     0
SP_ORF1          Group4       1    1     1
SP_ORF1          Group5       1    1     1

and another:

Tab2

Groups         SP_names   SP_names2
Group1         SP1        SP1_A
Group1         SP2        SP2_A
Group3         SP2        SP2_BL
Group4         SP1        SP1_BI
Group4         SP2        SP2_OP
Group4         SP3        SP3_I
Group5         SP3        SP3_0 
Group5         SP3        SP3_P
Group5         SP3        SP3_K

And I would like to replace within the Tab1 all cells with 1, and with their corresponding SP_names2 for each matching Groups and SP_names.

For instance, in the first row of Tab1, the SP_names2 value for SP1 in Group1 is SP1_A, so I replace 1 by SP1_A and so on...

When there is duplicated as in the Group5 example, I need to add all the duplicates SP_names2 separated by a pipe?

I should then get the following output:

New_tab1:

ORFs_values      Groups       SP1      SP2       SP3
SP_ORF1          Group1       SP1_A    SP2_A     0
SP_ORF1          Group2       0         0        0
SP_ORF1          Group3       0         SP2_BL   0
SP_ORF1          Group4       SP1_BI    SP2_OP   SP3_I
SP_ORF1          Group5       0         0        SP3_0|SP3_P|SP3_K

So far I got this try :

df = df.set_index('Groups')
(df.mask(df.eq(1),
         df2.set_index(['Groups','SP_names'])['SP_names2'].unstack())
 .reset_index())

But it does not take into account the multiple SP_names2 elements separated by a pipe...

Here are the two df in dict format if it can help:

Tab1

{'ORFs_values': {0: 'SP_ORF1', 1: 'SP_ORF1', 2: 'SP_ORF1', 3: 'SP_ORF1'}, 'Groups': {0: 'Group1', 1: 'Group2', 2: 'Group3', 3: 'Group4'}, 'SP1': {0: 1, 1: 0, 2: 0, 3: 1}, 'SP2': {0: 1, 1: 0, 2: 1, 3: 1}, 'SP3': {0: 0, 1: 0, 2: 0, 3: 1}}

Tab2

{'Groups': {0: 'Group1', 1: 'Group1', 2: 'Group3', 3: 'Group4', 4: 'Group4', 5: 'Group4', 6: 'Group5', 7: 'Group5', 8: 'Group5'}, 'SP_names': {0: 'SP1', 1: 'SP2', 2: 'SP2', 3: 'SP1', 4: 'SP2', 5: 'SP3', 6: 'SP3', 7: 'SP3', 8: 'SP3'}, 'SP_names2': {0: 'SP1_A', 1: 'SP2_A', 2: 'SP2_BL', 3: 'SP1_BI', 4: 'SP2_OP', 5: 'SP3_I', 6: 'SP3_0 ', 7: 'SP3_P', 8: 'SP3_K '}}

CodePudding user response：

Use DataFrame.pivot_table with aggregate function join:

df2['SP_names2'] = df2['SP_names2'].str.strip()
df2 = df2.pivot_table(index='Groups', 
                      columns='SP_names', 
                      values='SP_names2', 
                      aggfunc='|'.join, 
                      fill_value=0)
print (df2)
SP_names     SP1     SP2                SP3
Groups                                     
Group1     SP1_A   SP2_A                  0
Group3         0  SP2_BL                  0
Group4    SP1_BI  SP2_OP              SP3_I
Group5         0       0  SP3_0|SP3_P|SP3_K

If non missing values are same like 1 in df1 is possible use DataFrame.join:

df = df1[['ORFs_values','Groups']].join(df2, on='Groups').fillna(0)
print (df)
  ORFs_values  Groups     SP1     SP2    SP3
0     SP_ORF1  Group1   SP1_A   SP2_A      0
1     SP_ORF1  Group2       0       0      0
2     SP_ORF1  Group3       0  SP2_BL      0
3     SP_ORF1  Group4  SP1_BI  SP2_OP  SP3_I

If need replace 1 by df2:

df = df.set_index('Groups')
df = df.mask(df.eq(1), df2).reset_index().reindex(df1.columns, axis=1)
print (df)
  ORFs_values  Groups     SP1     SP2    SP3
0     SP_ORF1  Group1   SP1_A   SP2_A      0
1     SP_ORF1  Group2       0       0      0
2     SP_ORF1  Group3       0  SP2_BL      0
3     SP_ORF1  Group4  SP1_BI  SP2_OP  SP3_I