A few hours ago, when I was dealing with my thesis, I've come up with a Python function that constructs a tour for a given TSP instance using the nearest neighbor heuristic:

```
def nearestneighbor(size, dists):
tour = [0]*size
candy = [True]*size
candy[0] = False
for i in range(1, size):
prev = tour[i - 1]
tour[i] = min([(dists[prev][c], c)
for c in range(size) if candy[c] and c != prev])[1]
candy[tour[i]] = False
return tour
```

Not too bad...
Then I dig into old Python code that I wrote last year, the times when I was quitting C++ and beginning to use Python; and find this piece of code that consists of a class named `Nearestneighbor`

doing (almost) the same thing as the function above:

```
class Nearestneighbor:
def __init__(self, dists):
self.__dists = dists
self.dimension = dists.shape[0]
def construct(self, startcity = None):
if startcity is None:
self.startcity = rnd.randint(self.dimension)
else: self.startcity = startcity
self.tour = [self.startcity]
cities = range(self.dimension)
del cities[self.startcity]
while cities:
ndist = sys.maxint # nearest :) distance
nindex = 0 # nearest city's index
pcity = self.tour[-1] # previous city
ndist = self.__dists[pcity,cities[nindex]]
for i in range(1, len(cities)):
dist = self.__dists[pcity,cities[i]]
if dist < ndist:
nindex = i
ndist = dist
self.tour.append(cities[nindex])
del cities[nindex]
```

Well, this is definitely not my favorite piece of Python code ;-)