src/skimage2/feature/_peaks.py (1)

117-119: Consider handling 1D coordinate input explicitly.

When coords is a 1D array (single point like [1, 2, 3]), np.atleast_2d converts it to shape (1, 3) which is correct. However, the initial assignment output = coords on line 117 means if len(coords) == 0 returns the original input type (could be a list []), but non-empty inputs return an ndarray. This inconsistency is minor but could surprise callers.

♻️ Optional: ensure consistent return type

-    output = coords
-    if len(coords):
+    if len(coords) == 0:
+        return np.zeros((0, 1), dtype=float)
+    else:
         coords = np.atleast_2d(coords)

tests/skimage2/feature/test_peaks.py (4)

19-35: Consider using a fixed random seed for reproducibility.

The test uses np.random.randn without setting a seed, which makes test results non-deterministic. If a test fails, it may be difficult to reproduce. Consider using np.random.default_rng(seed) for reproducible tests.

♻️ Proposed fix

 `@pytest.mark.parametrize`("p", [1, 2, np.inf])
 `@pytest.mark.parametrize`("size", [30, 50, None])
 def test_ensure_spacing_trivial(p, size):
+    rng = np.random.default_rng(42)
     # --- Empty input
     assert _ensure_spacing([], p_norm=p) == []

     # --- A unique point
-    coord = np.random.randn(1, 2)
+    coord = rng.standard_normal((1, 2))
     assert np.array_equal(coord, _ensure_spacing(coord, p_norm=p, min_split_size=size))

     # --- Verified spacing
-    coord = np.random.randn(100, 2)
+    coord = rng.standard_normal((100, 2))

---

48-60: Consider setting a random seed for reproducibility.

Same concern as test_ensure_spacing_trivial - using np.random.randn without a seed makes the test non-deterministic.

Also, minor grammar nit: Line 53 comment says "between the point" but should be "between the points" (plural).

♻️ Proposed fix

 `@pytest.mark.parametrize`("p", [1, 2, np.inf])
 `@pytest.mark.parametrize`("size", [50, 100, None])
 def test_ensure_spacing_batch_processing(p, size):
-    coord = np.random.randn(100, 2)
+    rng = np.random.default_rng(123)
+    coord = rng.standard_normal((100, 2))

-    # --- Consider the average distance between the point as spacing
+    # --- Consider the average distance between the points as spacing
     spacing = np.median(pdist(coord, metric=minkowski, p=p))

---

63-81: Timing-based test may still be flaky on resource-constrained CI.

While the relaxed assertion (dur1 < 1.33 * dur2) helps, timing-based tests remain inherently sensitive to system load. Consider either:

1. Using pytest.mark.slow to optionally skip in CI
2. Increasing the tolerance further if flakiness persists
3. Using a fixed seed for np.random.randint for reproducibility

♻️ Optional: add slow marker and fixed seed

+@pytest.mark.slow
 def test_ensure_spacing_max_batch_size():
     """Small batches are slow, large batches -> large allocations -> also slow.

     https://github.com/scikit-image/scikit-image/pull/6035#discussion_r751518691
     """
-    coords = np.random.randint(low=0, high=1848, size=(40000, 2))
+    rng = np.random.default_rng(0)
+    coords = rng.integers(low=0, high=1848, size=(40000, 2))

---

84-93: Same recommendations: fixed seed and grammar fix.

Consider using a fixed random seed for reproducibility, and update the comment on line 89 to "between the points" (plural).

♻️ Proposed fix

 `@pytest.mark.parametrize`("p", [1, 2, np.inf])
 `@pytest.mark.parametrize`("size", [30, 50, None])
 def test_ensure_spacing_p_norm(p, size):
-    coord = np.random.randn(100, 2)
+    rng = np.random.default_rng(456)
+    coord = rng.standard_normal((100, 2))

-    # --- Consider the average distance between the point as spacing
+    # --- Consider the average distance between the points as spacing
     spacing = np.median(pdist(coord, metric=minkowski, p=p))

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro