test(diffusion,eval): cover non-square matrix, 1D input, max_iters cap, repeated-event traces

tests/test_diffusion.py

@@ -52,5 +52,34 @@ def test_ppr_higher_alpha_concentrates_at_seed() -> None:
     seed = np.array([1.0, 0.0, 0.0])
     r_low = personalized_pagerank(p.T, seed, alpha=0.05)
     r_high = personalized_pagerank(p.T, seed, alpha=0.9)
-    # higher alpha → more mass at seed
+    # higher alpha -> more mass at seed
     assert r_high[0] > r_low[0]
+
+
+def test_ppr_non_square_matrix_raises() -> None:
+    p_t = np.zeros((2, 3))
+    with pytest.raises(ValueError):
+        personalized_pagerank(p_t, np.zeros(2))
+
+
+def test_ppr_1d_input_raises() -> None:
+    with pytest.raises(ValueError):
+        personalized_pagerank(np.zeros(3), np.zeros(3))
+
+
+def test_ppr_max_iters_cap_returns_normalized() -> None:
+    # 4-node ring that needs many iterations to converge.
+    # max_iters=1 exits before convergence but result must still be
+    # a valid L1-normalized probability vector.
+    p = np.array(
+        [
+            [0.0, 1.0, 0.0, 0.0],
+            [0.0, 0.0, 1.0, 0.0],
+            [0.0, 0.0, 0.0, 1.0],
+            [1.0, 0.0, 0.0, 0.0],
+        ]
+    )
+    seed = np.array([1.0, 0.0, 0.0, 0.0])
+    r = personalized_pagerank(p.T, seed, alpha=0.15, max_iters=1)
+    assert pytest.approx(float(r.sum()), rel=1e-6) == 1.0
+    assert float(r.min()) >= 0.0

tests/test_eval.py

@@ -83,3 +83,17 @@ def test_demo_traces_round_trip_through_eval() -> None:
     # Demo is constructed so PPR should localize *something* -
     # top-10 over a 10-node graph should include the truth often.
     assert score.top_k[10] > 0.5
+
+
+def test_extract_cases_repeated_events_in_trace() -> None:
+    cases = extract_cases([["a", "a", "a"]])
+    assert len(cases) == 2
+    assert cases[0] == LocalizationCase(prefix=["a"], next_event="a")
+    assert cases[1] == LocalizationCase(prefix=["a", "a"], next_event="a")
+
+
+def test_evaluate_ks_deduplication() -> None:
+    idx = build_index(demo_graph(), demo_events())
+    cases = extract_cases(demo_traces())
+    score = evaluate(idx, cases, ks=[1, 1, 3])
+    assert set(score.top_k.keys()) == {1, 3}