From 5108fdb87bc93e456e0b516fb36f9de2ad17368e Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 27 Feb 2026 04:17:31 +0000
Subject: [PATCH 1/4] Initial plan


From aaef2fdedf49da62b377e451c2bddc99b1d14f6c Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 27 Feb 2026 04:24:22 +0000
Subject: [PATCH 2/4] Fix umap_embedding edge case: return early when fewer
 than 2 non-constant rows exist

Co-authored-by: lappalainenj <34949352+lappalainenj@users.noreply.github.com>
---
 flyvis/analysis/clustering.py | 22 +++++++++++------
 tests/test_clustering.py      | 46 +++++++++++++++++++++++++++++++++++
 2 files changed, 61 insertions(+), 7 deletions(-)
 create mode 100644 tests/test_clustering.py

diff --git a/flyvis/analysis/clustering.py b/flyvis/analysis/clustering.py
index e0011cc..208c467 100644
--- a/flyvis/analysis/clustering.py
+++ b/flyvis/analysis/clustering.py
@@ -456,9 +456,11 @@ def umap_embedding(
 
     Returns:
         A tuple containing:
-        - embedding: The UMAP embedding.
-        - mask: Boolean mask for valid samples.
-        - reducer: The fitted UMAP object.
+        - embedding: The UMAP embedding (n_samples, n_components). May be NaN
+          if insufficient data.
+        - mask: Boolean mask (length n_samples) showing which rows had nonzero
+          variance and were connected.
+        - reducer: The fitted UMAP object or None if insufficient data.
 
     Raises:
         ValueError: If n_components is too large relative to sample size.
@@ -481,10 +483,16 @@ def umap_embedding(
         X = X.reshape(X.shape[0], -1)
         logging.info("reshaped X from %s to %s", shape, X.shape)
 
-    embedding = np.ones([X.shape[0], n_components]) * np.nan
-    # umap doesn't like contant rows
+    n_samples = X.shape[0]
+    embedding = np.ones([n_samples, n_components]) * np.nan
+    # umap doesn't like constant rows
     mask = ~np.isclose(X.std(axis=1), 0)
-    X = X[mask]
+    X_nonconst = X[mask]
+
+    # If fewer than 2 rows remain, skip UMAP and return embedding of NaNs.
+    if X_nonconst.shape[0] < 2:
+        return embedding, mask, None
+
     reducer = UMAP(
         n_neighbors=n_neighbors,
         min_dist=min_dist,
@@ -495,7 +503,7 @@ def umap_embedding(
         n_epochs=n_epochs,
         **kwargs,
     )
-    _embedding = reducer.fit_transform(X)
+    _embedding = reducer.fit_transform(X_nonconst)
 
     # gaussian mixture doesn't like nans through disconnected vertices in umap
     connected_vertices_mask = ~disconnected_vertices(reducer)
diff --git a/tests/test_clustering.py b/tests/test_clustering.py
new file mode 100644
index 0000000..0b33143
--- /dev/null
+++ b/tests/test_clustering.py
@@ -0,0 +1,46 @@
+import numpy as np
+import pytest
+
+from flyvis.analysis.clustering import umap_embedding
+
+
+def test_umap_embedding_single_nonzero_variance_row():
+    """Test that umap_embedding handles the edge case where only one row has
+    nonzero variance (all others are constant). UMAP should not be fitted and
+    the function should return NaN embedding with None reducer."""
+    rng = np.random.default_rng(0)
+    # One row with variance, four constant rows
+    X = np.zeros((5, 10))
+    X[2] = rng.random(10)
+
+    embedding, mask, reducer = umap_embedding(X)
+
+    assert reducer is None
+    assert np.all(np.isnan(embedding))
+    # Only the one non-constant row should be True in the mask
+    expected_mask = np.array([False, False, True, False, False])
+    np.testing.assert_array_equal(mask, expected_mask)
+
+
+def test_umap_embedding_all_zero_variance_rows():
+    """Test that umap_embedding handles all-constant rows gracefully."""
+    X = np.ones((5, 10))
+
+    embedding, mask, reducer = umap_embedding(X)
+
+    assert reducer is None
+    assert np.all(np.isnan(embedding))
+    assert not np.any(mask)
+
+
+def test_umap_embedding_returns_none_reducer_when_insufficient_data():
+    """Test that reducer is None when fewer than 2 rows have nonzero variance."""
+    X = np.zeros((4, 8))
+    # Only one non-constant row
+    X[0] = np.arange(8, dtype=float)
+
+    embedding, mask, reducer = umap_embedding(X)
+
+    assert reducer is None
+    assert embedding.shape == (4, 2)
+    assert np.all(np.isnan(embedding))

From 0e59f8eac33c138ab80817add4098aa57af8927d Mon Sep 17 00:00:00 2001
From: Janne Lappalainen <34949352+lappalainenj@users.noreply.github.com>
Date: Fri, 27 Feb 2026 12:25:25 -0800
Subject: [PATCH 3/4] Update tests/test_clustering.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 tests/test_clustering.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tests/test_clustering.py b/tests/test_clustering.py
index 0b33143..051b6d1 100644
--- a/tests/test_clustering.py
+++ b/tests/test_clustering.py
@@ -1,5 +1,4 @@
 import numpy as np
-import pytest
 
 from flyvis.analysis.clustering import umap_embedding
 

From 512596e59a7b7b75564f2a9e0e4d4a675f7e0e77 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Fri, 27 Feb 2026 20:30:16 +0000
Subject: [PATCH 4/4] Fix return type annotation: use Optional[UMAP] and
 clarify mask semantics in docstring

Co-authored-by: lappalainenj <34949352+lappalainenj@users.noreply.github.com>
---
 flyvis/analysis/clustering.py | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/flyvis/analysis/clustering.py b/flyvis/analysis/clustering.py
index 208c467..c7fdcbe 100644
--- a/flyvis/analysis/clustering.py
+++ b/flyvis/analysis/clustering.py
@@ -439,7 +439,7 @@ def umap_embedding(
     metric: str = "correlation",
     n_epochs: int = 1500,
     **kwargs,
-) -> Tuple[np.ndarray, np.ndarray, UMAP]:
+) -> Tuple[np.ndarray, np.ndarray, Optional[UMAP]]:
     """
     Perform UMAP embedding on input data.
 
@@ -458,9 +458,12 @@ def umap_embedding(
         A tuple containing:
         - embedding: The UMAP embedding (n_samples, n_components). May be NaN
           if insufficient data.
-        - mask: Boolean mask (length n_samples) showing which rows had nonzero
-          variance and were connected.
-        - reducer: The fitted UMAP object or None if insufficient data.
+        - mask: Boolean mask (length n_samples). When reducer is not None,
+          True indicates rows with nonzero variance that were also connected
+          in the UMAP graph. When reducer is None (insufficient data), True
+          indicates only rows with nonzero variance.
+        - reducer: The fitted UMAP object or None if fewer than 2 rows had
+          nonzero variance.
 
     Raises:
         ValueError: If n_components is too large relative to sample size.