Why does .validate() raise error when data.num_nodes == data.edge_index.max() for hetero_data objects?

[aaronwtr]

Going into hetero_data.py we see that

                if (num_src_nodes is not None
                        and store.edge_index[0].max() >= num_src_nodes):
                    status = False
                    warn_or_raise(
                        f"'edge_index' of edge type {edge_type} contains "
                        f"larger source indices than the number of nodes "
                        f"({num_src_nodes}) of this node type in '{cls_name}' "
                        f"(found {int(store.edge_index[0].max())})",
                        raise_on_error)

where store.edge_index[0].max() >= num_src_nodes), is the crucial part causing the error in my code because, for me, these two are equal. This also logically makes sense to me: for every node, I have an index in edge_index pointing to that node. This intuitively makes sense to me.

We can observe similar behaviour in message_passing.py:

        elif isinstance(edge_index, Tensor):
            try:
                index = edge_index[dim]
                return src.index_select(self.node_dim, index)
            except (IndexError, RuntimeError) as e:
                if index.min() < 0 or index.max() >= src.size(self.node_dim):
                    raise IndexError(
                        f"Encountered an index error. Please ensure that all "
                        f"indices in 'edge_index' point to valid indices in "
                        f"the interval [0, {src.size(self.node_dim) - 1}] "
                        f"(got interval "
                        f"[{int(index.min())}, {int(index.max())}])")
                else:
                    raise e

Again here the greater / equal sign breaks my code if I remove the greater than or equal sign from hetero_data.py. Also not in the text we subtract 1 from the node_dim? I suppose it is there for a reason, although I can not grasp what this reason is. Anyone willing to share any insight on this matter would be greatly appreciated.

[akihironitta]

For example, when you have three nodes, it expects the index to be [0, 1, 2].