Fix: edge_feat_embedding crashes on empty edge_index when normalized_features=True*

devograph/datasets/datasets1.py

@@ -301,6 +301,8 @@ def normalize_array(self, array):
         return array
 
     def edge_feat_embedding(self, x, edge_index):
+        if edge_index.shape[1] == 0:
+            return np.empty((0, x.shape[1]), dtype=np.float32)
         src, trg = edge_index
         sub_x = x[src] - x[trg]
         abs_sub = np.abs(sub_x)

test/test_datasets1.py

@@ -0,0 +1,48 @@
+import sys
+import types
+from unittest.mock import MagicMock
+
+# hydra is not installed in this env; mock it before importing the module
+hydra_mock = types.ModuleType("hydra")
+hydra_utils_mock = types.ModuleType("hydra.utils")
+hydra_utils_mock.get_original_cwd = MagicMock(return_value="/tmp")
+hydra_mock.utils = hydra_utils_mock
+sys.modules.setdefault("hydra", hydra_mock)
+sys.modules.setdefault("hydra.utils", hydra_utils_mock)
+
+import numpy as np
+import torch
+
+from devograph.datasets.datasets1 import CellTrackDataset
+
+
+def make_dataset():
+    """Bypass __init__ and set only the attributes edge_feat_embedding needs."""
+    ds = object.__new__(CellTrackDataset)
+    ds.edge_feat_embed_dict = {'p': 1, 'use_normalized_x': True, 'normalized_features': True}
+    ds.which_preprocess = 'MinMax'
+    ds.separate_models = False
+    ds.normalize_cols = np.array([True, True, True])
+    return ds
+
+
+def test_edge_feat_embedding_empty_edge_index():
+    """No crash and correct shape when graph has zero edges."""
+    ds = make_dataset()
+    x = np.random.rand(5, 3).astype(np.float32)
+    edge_index = torch.empty((2, 0), dtype=torch.long)
+
+    result = ds.edge_feat_embedding(x, edge_index)
+
+    assert result.shape == (0, 3), f"Expected (0, 3), got {result.shape}"
+
+
+def test_edge_feat_embedding_nonempty_edge_index():
+    """Normal case: returns (E, F) edge features."""
+    ds = make_dataset()
+    x = np.random.rand(5, 3).astype(np.float32)
+    edge_index = torch.tensor([[0, 1, 2], [1, 2, 3]], dtype=torch.long)
+
+    result = ds.edge_feat_embedding(x, edge_index)
+
+    assert result.shape == (3, 3), f"Expected (3, 3), got {result.shape}"