python - Pandas Grouby with OR condition

Question

Consider the following data set.

i_id	g_id	s_id	p_id
1001	1001	1001
1002	1001	1001
1003	1001	1001	1003
1004	1004	1001	1003
1005	1004	1001	1003
1006	1004		1003
1007	1007	1007	1003
1008	1007	1007	1003
1009	1009	1007
1010	1010	1007

question from:https://stackoverflow.com/questions/65914186/pandas-grouby-with-or-condition

Answer 1

Use networkx with connected_components for dictionary created by minimal common values and then mapping to new column:

import networkx as nx

df1 = (df.melt('i_id')
        .dropna(subset=['value'])
        .groupby('i_id', as_index=False)['value'].min())
print (df1)
    i_id value
0   1001  1001
1   1002  1001
2   1003  1001
3   1004  1001
4   1005  1001
5   1006  1003
6   1007  1003
7   1008  1003
8   1009  1007
9   1010  1007
10  1011  1011
11  1012  1012
12  1013  1013
13  1014  1014

# Create the graph from the dataframe
g = nx.Graph()
g = nx.from_pandas_edgelist(df1,'i_id','value')

connected_components = nx.connected_components(g)
# print (list(connected_components))

d = {y: min(x) for x in connected_components for y in x}
df['Key'] = df['i_id'].map(d)

---

print (df)
    i_id  g_id  s_id  p_id   Key
0   1001  1001  1001   NaN  1001
1   1002  1001  1001   NaN  1001
2   1003  1001  1001  1003  1001
3   1004  1004  1001  1003  1001
4   1005  1004  1001  1003  1001
5   1006  1004   NaN  1003  1001
6   1007  1007  1007  1003  1001
7   1008  1007  1007  1003  1001
8   1009  1009  1007   NaN  1001
9   1010  1010  1007   NaN  1001
10  1011  1011  1011  1011  1011
11  1012  1012  1012  1012  1012
12  1013  1013  1013  1013  1013
13  1014  1014  1014  1014  1014