Metapath random walk not considering edge type?

[Wang-Yu-Qing]

In the walk function:

def walk(args):
    walk_length, start, schema = args
    # Simulate a random walk starting from start node.
    rand = random.Random()

    if schema:
        schema_items = schema.split('-')
        assert schema_items[0] == schema_items[-1]

    walk = [start]
    while len(walk) < walk_length:
        cur = walk[-1]
        candidates = []
        for node in G[cur]:
            if schema == '' or node_type[node] == schema_items[len(walk) % (len(schema_items) - 1)]:
                candidates.append(node)
        if candidates:
            walk.append(rand.choice(candidates))
        else:
            break
    return [str(node) for node in walk]

The walk doesn't consider edge types which is required in the paper's section 4.3 description.

To be specific, given a view r of the network

[Wang-Yu-Qing]

I found that the walk is done upon sub-graph of one specific edge type.

[viviqi]

Actually I have the same opinion with you, schema based random walk sample walks on one specific layer, which means relation type is fixed when sample one random walk sequence.