fix(rag): handle indexed empty row in Ragged.to_numpy

python/seqpro/rag/_core.py

@@ -1786,6 +1786,7 @@ def to_numpy(
                 raise ValueError("cannot convert a jagged Ragged to a dense array")
             packed = self if self.is_base else self.to_packed()
             row_len = int(lengths.flat[0]) if lengths.size else 0
+            n_rows = lengths.size
         else:
             # trust the caller: infer row_len from total // n_rows, no uniformity
             # scan. numpy's reshape still rejects a total-size mismatch for free.
@@ -1799,8 +1800,11 @@ def to_numpy(
             total = packed._rl.data.shape[0]
             row_len = total // n_rows if n_rows else 0
         leading = packed.shape[: packed.rag_dim]
+        # When fully indexed to a single ragged axis, `leading` is empty; use the
+        # explicit row count instead of -1 so an empty row (row_len == 0) reshapes
+        # cleanly (numpy cannot infer -1 against a 0 dimension). See #67.
         return packed._rl.data.reshape(  # pyrefly: ignore[no-matching-overload]
-            *(leading or (-1,)), row_len, *packed._rl.data.shape[1:]
+            *(leading or (n_rows,)), row_len, *packed._rl.data.shape[1:]
         )
 
     def __array__(self, dtype: Any = None) -> NDArray[Any]:

tests/test_ragged_core.py

@@ -330,6 +330,19 @@ def test_to_numpy_jagged_raises():
         rag.to_numpy()
 
 
+@pytest.mark.parametrize("validate", [True, False])
+def test_to_numpy_indexed_empty_row(validate):
+    # Regression for #67: an empty row (length 0) obtained by indexing a
+    # multi-dimensional Ragged (is_base=False) must convert to an empty dense
+    # array instead of raising on the final reshape.
+    data = np.array([5, 7], np.int32)
+    rag = Ragged.from_lengths(data, np.array([[[2], [0]]]))
+    empty = rag[0, 1, 0]
+    assert not empty.is_base
+    out = empty.to_numpy(validate=validate)
+    assert out.shape == (1, 0)
+
+
 def test_ingest_from_ak_numeric():
     arr = ak.Array([[1, 2, 3], [4, 5]])
     rag = Ragged(arr)