How to generate permutations without repetition

Question

I have a table that looks like below

Loc	ID	filter	P1
A	ABC1	GHY	55.6
A	DFT1	FGH	67.8
B	HJH5	GHY	67
C	HKL	BHY	78
B	GTY	FGH	60

I want the output as below. Basically, I want the records with the same Filter to be one row

Filter	ID	Loc	P1	m_ID	m_Loc	m_p1	total
GHY	ABC1	A	55.6	HJH5	B	67	122.6
FGH	DFT1	A	67.8	GTY	B	60	127.8

Is this achievable using itertools i python. If yes can someone please suggest how can we do it?

Answer 1

Here's a solution using lead and row_number that I think is a little nicer.

select filter
      ,id
      ,loc 
      ,p1
      ,m_id
      ,m_loc
      ,m_p1

from 
      (with t2 as 
      (select row_number () over( partition by filter order by filter desc) as rn
      ,*
       from t)
              select rn,filter, id, loc, p1
             ,lead(id) over( partition by filter order by filter)  as m_id
             ,lead(loc) over( partition by filter order by filter) as m_loc
             ,lead(p1) over( partition by filter order by filter)  as m_p1
              from t2) t
where rn=1  

filter	id	loc	p1	m_id	m_loc	m_p1
BHY	HKL	C	78	null	null	null
FGH	DFT1	A	67.8	GTY	B	60
GHY	ABC1	A	55.6	HJH5	B	67

Fiddle

There should be a better solution to this question, but here is a solution that's based on what you did. I used left join to not lose filters that only appear once and then I used group by to consolidate the results.

select t1.filter
      ,max(t1.id)  as id
      ,max(t1.loc) as loc
      ,max(t1.p1)  as p1
      ,min(t2.id)  as m_id
      ,min(t2.loc) as m_loc
      ,min(t2.p1)  as m_p1
      
from t as t1 left join t as t2 on t2.filter = t1.filter and t2.id <> (t1.id)
group by t1.filter

filter	id	loc	p1	m_id	m_loc	m_p1
BHY	HKL	C	78	null	null	null
FGH	GTY	B	67.8	DFT1	A	60
GHY	HJH5	B	67	ABC1	A	55.6

Fiddle

Answer 2

If the usage of pandas is possible, you can achive a flexible solutiion with the following:

Definition of the data:

df=pd.DataFrame({'Loc': {0: 'A', 1: 'A', 2: 'B ', 3: 'C', 4: 'B'},
 'ID': {0: 'ABC1', 1: 'DFT1', 2: 'HJH5', 3: 'HKL', 4: 'GTY'},
 'filter': {0: 'GHY', 1: 'FGH', 2: 'GHY', 3: 'BHY', 4: 'FGH'},
 'P1': {0: 55.6, 1: 67.8, 2: 67.0, 3: 78.0, 4: 60.0}}) 

Creation of the repetive columns:

cols=["{}_{}".format(N, c) for N in range(0,df.groupby('filter').count()['ID'].max()) for c in df.columns]

Here, I first find the maximum required repitions by looking for the max occurences of each filter df.groupby('filter').count()['ID'].max() . The remaining code is just formating by adding a leading number.

Creation of new dataframe with filter as index and the generated columns cols as columns

df_new=pd.DataFrame(index=set(df['filter']), columns=cols)

Now we have to fill in the data:

for fil in df_new.index:
    values=[val for row in df[df['filter']==fil].values for val in row]
    df_new.loc[fil,df_new.columns[:len(values)]]=values

Here two things are done: First, the selected values based on the filter name fil are flattend by [val for row in df[df['filter']==fil].values for val in row] . Then, these values are filled into the dataframe starting at the left.

The result is as expected:

    0_Loc   0_ID    0_filter    0_P1    1_Loc   1_ID    1_filter    1_P1
GHY     A   ABC1    GHY     55.6    B   HJH5    GHY     67.0
BHY     C   HKL     BHY     78.0    NaN     NaN     NaN     NaN
FGH     A   DFT1    FGH     67.8    B   GTY     FGH     60.0