Ensure downsample works with non-contiguous arrays

holoviews/operation/downsample.py

@@ -100,6 +100,13 @@ def _lttb_inner(x, y, n_out, sampled_x, offset):
     )
 
 
+def _ensure_contiguous(x, y):
+    """
+    Ensures the arrays are contiguous in memory (required by tsdownsample).
+    """
+    return np.ascontiguousarray(x), np.ascontiguousarray(y)
+
+
 def _lttb(x, y, n_out, **kwargs):
     """
     Downsample the data using the LTTB algorithm.
@@ -115,6 +122,7 @@ def _lttb(x, y, n_out, **kwargs):
     """
     try:
         from tsdownsample import LTTBDownsampler
+        x, y = _ensure_contiguous(x, y)
         return LTTBDownsampler().downsample(x, y, n_out=n_out, **kwargs)
     except ModuleNotFoundError:
         pass
@@ -168,6 +176,7 @@ def _min_max(x, y, n_out, **kwargs):
             'The min-max downsampling algorithm requires the tsdownsample '
             'library to be installed.'
         ) from None
+    x, y = _ensure_contiguous(x, y)
     return MinMaxDownsampler().downsample(x, y, n_out=n_out, **kwargs)
 
 def _min_max_lttb(x, y, n_out, **kwargs):
@@ -178,6 +187,7 @@ def _min_max_lttb(x, y, n_out, **kwargs):
             'The minmax-lttb downsampling algorithm requires the tsdownsample '
             'library to be installed.'
         ) from None
+    x, y = _ensure_contiguous(x, y)
     return MinMaxLTTBDownsampler().downsample(x, y, n_out=n_out, **kwargs)
 
 def _m4(x, y, n_out, **kwargs):
@@ -188,6 +198,7 @@ def _m4(x, y, n_out, **kwargs):
             'The m4 downsampling algorithm requires the tsdownsample '
             'library to be installed.'
         ) from None
+    x, y = _ensure_contiguous(x, y)
     n_out = n_out - (n_out % 4)  # n_out must be a multiple of 4
     return M4Downsampler().downsample(x, y, n_out=n_out, **kwargs)
 

holoviews/tests/operation/test_downsample.py

@@ -2,7 +2,8 @@
 import pandas as pd
 import pytest
 
-import holoviews as hv
+from holoviews.core.overlay import NdOverlay, Overlay
+from holoviews.element import Curve
 from holoviews.operation.downsample import _ALGORITHMS, downsample1d
 
 try:
@@ -20,16 +21,26 @@ def test_downsample1d_multi(plottype):
     assert N > downsample1d.width
 
     if plottype == "overlay":
-        figure = hv.Overlay([hv.Curve(range(N)), hv.Curve(range(N))])
+        figure = Overlay([Curve(range(N)), Curve(range(N))])
     elif plottype == "ndoverlay":
-        figure = hv.NdOverlay({"A": hv.Curve(range(N)), "B": hv.Curve(range(N))})
+        figure = NdOverlay({"A": Curve(range(N)), "B": Curve(range(N))})
 
     figure_values = downsample1d(figure, dynamic=False).data.values()
     for n in figure_values:
         for value in n.data.values():
             assert value.size == downsample1d.width
 
 
+@pytest.mark.skipif(not tsdownsample, reason="tsdownsample not installed")
+@pytest.mark.parametrize("algorithm", algorithms)
+def test_downsample1d_non_contiguous(algorithm):
+    x = np.arange(20)
+    y = np.arange(40).reshape(1, 40)[0, ::2]
+
+    downsampled = downsample1d(Curve((x, y), datatype=['array']), dynamic=False, width=10, algorithm=algorithm)
+    assert len(downsampled)
+
+
 def test_downsample1d_shared_data():
     runs = [0]
 
@@ -42,7 +53,7 @@ def _compute_mask(self, element):
 
     N = 1000
     df = pd.DataFrame({c: range(N) for c in "xyz"})
-    figure = hv.Overlay([hv.Curve(df, kdims="x", vdims=c) for c in "yz"])
+    figure = Overlay([Curve(df, kdims="x", vdims=c) for c in "yz"])
 
     # We set x_range to trigger _compute_mask
     mocksample(figure, dynamic=False, x_range=(0, 500))
@@ -61,7 +72,7 @@ def _compute_mask(self, element):
 
     N = 1000
     df = pd.DataFrame({c: range(N) for c in "xyz"})
-    figure = hv.Overlay([hv.Curve(df, kdims="index", vdims=c) for c in "xyz"])
+    figure = Overlay([Curve(df, kdims="index", vdims=c) for c in "xyz"])
 
     # We set x_range to trigger _compute_mask
     mocksample(figure, dynamic=False, x_range=(0, 500))