Question

I am trying to find out which edges from a graph are bidirectional. Each row is an edge. For each starting node A, I am searching each corresponding end node B if they have node A as an ending point:

for ending_point_B in nodeA:
    nodeA in ending_points_of_B

Disregard for now repeated entries in df[‘S’]. How can I optimize this search? I suspect something along the lines of groupby. This way takes too much time for my real graph.

Thank you

from pandas import *

def missing_node(node):
    set1 = set(df[df.E == node].S.values)
    set2 = set(df.E[df.S == node].values)
    return list(set1.difference(set2))

x = [1,1,2,2,3]
y = [2,3,1,3,1]

df = DataFrame([x,y]).T
df.columns = ['S','E'] #Start & End

df['Missing'] = df.S.apply(missing_node)

df:

   S  E Missing
0  1  2      []
1  1  3      []
2  2  1      []
3  2  3      []
4  3  1     [2]

Answer 1

Pandas is great, but not sure you need it here. Something like the following should give you the links that aren’t bidirectional:

x = [1,1,2,2,3]
y = [2,3,1,3,1]

fwd = set( zip(x,y) )
rev = set( zip(y,x) )
print ' not bi: ', fwd.difference(rev)

This returns:

not bi: set([(2, 3)])