get_weight() inconsistent changing of indexing bug

[koleckar]

• TSPLIB 95 version: 0.7.1
• Python version: 3.6
• Operating System: Win

graph = tsplib95.load('data/gr17.tsp')
print(graph.get_weight(17, 1))  # IndexError: (17, 1) is out of bonuds

graph = tsplib95.load('data/gr120.tsp')
print(graph.get_weight(120, 1))  # ok


graph = tsplib95.load('data/gr17.tsp')
print(graph.get_weight(16, 0))  # ok

graph = tsplib95.load('data/gr120.tsp')
print(graph.get_weight(119, 0))  # IndexError: (118, -1) is out of bonuds

So inside get_weight() for some graphs (with exactly same edge_weight_type/format etc..) there is NONCONSISTENCY
of what the function does with indexing.
Eg.: for gr120.tsp instance there is implicit -1 for index passed into get_weight() but for gr17/24/48 there is not.

The subtracitng 1 issue I found for 'berlin52.tsp', 'pa561.tsp' and 'gr120'.

Btw there is typo in "raise IndexError(f'({i}, {j}) is out of bonuds" as well.

Also for some problems (eg: si_xxx.tsp) there is flipped 'i' and 'j' in "raise IndexError(f'({i}, {j}) is out of bonuds"

graph = tsplib95.load('data/si535.tsp')
print(graph.get_weight(535, 1)) # IndexError: (1, 535) is out of bonuds

Quite annoying bug tbh.

[George-Ogden]

@koleckar one solution is to use the get_nodes() function:

n = len(list(problem.get_nodes()))
adjacency_matrix = np.zeros((n, n), dtype=np.float64)
for i, u in enumerate(problem.get_nodes()):
    for j, v in enumerate(problem.get_nodes()):
        if i != j:
            adjacency_matrix[i][j] = problem.get_weight(u, v)