Pandas计算数据帧中不同列上的所有匹配项

Question

I have a data frame similar to this one GRP HOST1 HOST2 HOST3 FILESIZE 0 0 srv39 srv45 srv47 203498176 1 1 srv102 srv36 srv38 452763956 2 1 srv101 srv36 srv45 453277268 3 1 srv101 srv34 srv45 448174741 4 1 srv36 srv49 srv50 452728577 5 2 srv100 srv47 srv48 454617541 6 2 srv100 srv45 srv49 454617541 7 2 srv38 srv49 srv47 454617541

Now what I would like to achieve is count all occurrences that I have across HOST1 HOST2 and HOST3 column grouped by the GRP column, like this

-- GRP HOST count 1 srv101 2 srv36 3 It would be perfect if I would be able to sum the value of the FILESIZE column. I was trying to shape a solution using suggestions that I have found here , but I have not been able to get the count grouped by GRP.

Any suggestion about which would be the best approach to obtain the results that I need with pandas?

Answer 1

Use melt for reshape anf then aggregate size :

df = (df.melt(id_vars='GRP', value_vars=['HOST1','HOST2','HOST3'], value_name='HOST')
        .groupby(['GRP', 'HOST'])
        .size()
        .reset_index(name='count'))
print (df)
    GRP    HOST  count
0     0   srv39      1
1     0   srv45      1
2     0   srv47      1
3     1  srv101      2
4     1  srv102      1
5     1   srv34      1
6     1   srv36      3
7     1   srv38      1
8     1   srv45      2
9     1   srv49      1
10    1   srv50      1
11    2  srv100      2
12    2   srv38      1
13    2   srv45      1
14    2   srv47      2
15    2   srv48      1
16    2   srv49      2

If want sum of column FILESIZE use agg :

df1 = (df.melt(id_vars=['GRP', 'FILESIZE'], value_vars=['HOST1','HOST2','HOST3'], value_name='HOST')
        .groupby(['GRP', 'HOST'])['FILESIZE']
        .agg(['size','sum'])
        .reset_index()
        )
print (df1)
    GRP    HOST  size         sum
0     0   srv39     1   203498176
1     0   srv45     1   203498176
2     0   srv47     1   203498176
3     1  srv101     2   901452009
4     1  srv102     1   452763956
5     1   srv34     1   448174741
6     1   srv36     3  1358769801
7     1   srv38     1   452763956
8     1   srv45     2   901452009
9     1   srv49     1   452728577
10    1   srv50     1   452728577
11    2  srv100     2   909235082
12    2   srv38     1   454617541
13    2   srv45     1   454617541
14    2   srv47     2   909235082
15    2   srv48     1   454617541
16    2   srv49     2   909235082

Answer 2

You can using stack , then follow with groupby and size

s=df.set_index('GRP')[['HOST1','HOST2','HOST3']].stack().to_frame('HOST')
s.groupby([s.index.get_level_values(level=0),s.HOST]).size()
Out[229]: 
GRP  HOST  
0    srv39     1
     srv45     1
     srv47     1
1    srv101    2
     srv102    1
     srv34     1
     srv36     3
     srv38     1
     srv45     2
     srv49     1
     srv50     1
2    srv100    2
     srv38     1
     srv45     1
     srv47     2
     srv48     1
     srv49     2
dtype: int64

If you need sum

s=df.set_index(['GRP','FILESIZE'])[['HOST1','HOST2','HOST3']].stack().to_frame('HOST').reset_index(level=1)
s.groupby([s.index.get_level_values(level=0),s.HOST.values]).FILESIZE.agg(['count','sum'])