# Implementing the CutHill-McKee Algorithm

I am very much interested in the Reverse Cuthil McKee Algorithm. I have seen Fortran and C or C++ implementations of it, and I decided that it would be a nice exercise to implement it in Python. I know this algorithm is quite domain specific, but I would still be happy to see what kind of comments I get regarding:

• Correctness - I am not sure my only test case works for others, although I did some comparison to the Octave and Matlab version.
• Speed - Of course a C version would be faster. However, is there some Python improvements which can be done?
• Readability - Is this code clear enough to other peer programmers?

The code:

import numpy as np

def getDegree(Graph):
"""
find the degree of each node. That is the number
of neighbours or connections.
(number of non-zero elements) in each row minus 1.
Graph is a Cubic Matrix.
"""
degree = *Graph.shape
for row in range(Graph.shape):
degree[row] = len(np.flatnonzero(Graph[row]))-1
return degree

"""
return the adjacncy matrix for each node
"""
for i in xrange(Mat.shape):
q=np.flatnonzero(Mat[i])
q=list(q)
q.pop(q.index(i))

"""
Reverse Cuthil McKee ordering of an adjacency Matrix
"""
digar=np.array(deg)
# use np.where here to get indecies of minimums
if start not in R:
R.append(start)
for idx, item in enumerate(Q):
if item not in R:
R.append(item)
if set(Q).issubset(set(R)) and len(R) < len(deg) :
p = pivots
pivots.pop(0)
elif len(R) < len(deg):
else:
R.reverse()
return R

def test():
"""
test the RCM loop
"""
A = np.diag(np.ones(8))
print A
nzc=[,[2,5,7],[1,4],,[0,2],[1,7],,[1,5]]

for i in range(len(nzc)):
for j in nzc[i]:
A[i,j]=1
# define the Result queue
R = ["C"]*A.shape
degree = getDegree(A)
digar=np.array(degree)
pivots = list(np.where(digar == digar.min()))
inl=[]
print degree
print "solution:", Res
print "correct:", [6,3,7,5,1,2,4,0]

if __name__ == '__main__':
test()


Regarding readability, it is generally a good practice to avoid using unnecessarily shortened forms like adjcncy, as it is only two characters short of adjacency. Also using R for result_queue is discouraged in general (subjective opinion).