[GraphBolt] fix type_per_edge when convert DGLGraph to CSCSamplingGraph (#6314)

python/dgl/graphbolt/impl/csc_sampling_graph.py

@@ -729,7 +729,10 @@ def from_dglgraph(g: DGLGraph, is_homogeneous=False) -> CSCSamplingGraph:
     ntype_count.insert(0, 0)
     node_type_offset = torch.cumsum(torch.LongTensor(ntype_count), 0)
 
-    type_per_edge = None if is_homogeneous else homo_g.edata[ETYPE]
+    # Sort edge type according to columns as `csc` is used.
+    _, dst = homo_g.edges()
+    _, dst_indices = torch.sort(dst)
+    type_per_edge = None if is_homogeneous else homo_g.edata[ETYPE][dst_indices]
 
     return CSCSamplingGraph(
         torch.ops.graphbolt.from_csc(

tests/python/pytorch/graphbolt/impl/test_csc_sampling_graph.py

@@ -1048,12 +1048,12 @@ def test_multiprocessing_with_shared_memory():
 )
 def test_from_dglgraph_homogeneous():
     dgl_g = dgl.rand_graph(1000, 10 * 1000)
-    gb_g = gb.from_dglgraph(dgl_g)
+    gb_g = gb.from_dglgraph(dgl_g, is_homogeneous=True)
 
     assert gb_g.num_nodes == dgl_g.num_nodes()
     assert gb_g.num_edges == dgl_g.num_edges()
     assert torch.equal(gb_g.node_type_offset, torch.tensor([0, 1000]))
-    assert torch.all(gb_g.type_per_edge == 0)
+    assert gb_g.type_per_edge is None
     assert gb_g.metadata.node_type_to_id == {"_N": 0}
     assert gb_g.metadata.edge_type_to_id == {"_N:_E:_N": 0}
 
@@ -1063,46 +1063,27 @@ def test_from_dglgraph_homogeneous():
     reason="Graph on GPU is not supported yet.",
 )
 def test_from_dglgraph_heterogeneous():
-    def create_random_hetero():
-        num_nodes = {"n1": 1000, "n2": 1010, "n3": 1020}
-        etypes = [
-            ("n1", "r12", "n2"),
-            ("n2", "r21", "n1"),
-            ("n1", "r13", "n3"),
-            ("n2", "r23", "n3"),
-        ]
-        edges = {}
-        for etype in etypes:
-            src_ntype, _, dst_ntype = etype
-            arr = spsp.random(
-                num_nodes[src_ntype],
-                num_nodes[dst_ntype],
-                density=0.001,
-                format="coo",
-                random_state=100,
+    dgl_g = dgl.heterograph(
+        {
+            ("author", "writes", "paper"): ([1, 2, 3, 4, 5], [1, 2, 3, 4, 5]),
+            ("paper", "cites", "paper"): ([2, 3, 4, 5, 6], [1, 2, 3, 4, 5]),
+        }
     )
-            edges[etype] = (arr.row, arr.col)
-        return dgl.heterograph(edges, num_nodes)
-
-    dgl_g = create_random_hetero()
-    gb_g = gb.from_dglgraph(dgl_g)
+    gb_g = gb.from_dglgraph(dgl_g, is_homogeneous=False)
 
     assert gb_g.num_nodes == dgl_g.num_nodes()
     assert gb_g.num_edges == dgl_g.num_edges()
+    assert torch.equal(gb_g.node_type_offset, torch.tensor([0, 6, 13]))
     assert torch.equal(
-        gb_g.node_type_offset, torch.tensor([0, 1000, 2010, 3030])
+        gb_g.type_per_edge, torch.tensor([0, 1, 0, 1, 0, 1, 0, 1, 0, 1])
     )
-    assert torch.all(gb_g.type_per_edge[:-1] <= gb_g.type_per_edge[1:])
     assert gb_g.metadata.node_type_to_id == {
-        "n1": 0,
-        "n2": 1,
-        "n3": 2,
+        "author": 0,
+        "paper": 1,
     }
     assert gb_g.metadata.edge_type_to_id == {
-        "n1:r12:n2": 0,
-        "n1:r13:n3": 1,
-        "n2:r21:n1": 2,
-        "n2:r23:n3": 3,
+        "author:writes:paper": 0,
+        "paper:cites:paper": 1,
     }