Issue #46/#44 add basic test for whittaker_f

tests/test_whittaker.py

@@ -0,0 +1,37 @@
+from datetime import datetime, timedelta
+
+import numpy as np
+import numpy.testing
+
+from fusets.whittaker import whittaker_f
+
+
+def test_whittaker_f():
+    def generate_data(xs: np.array):
+        """Generate test data from array of ints (day offsets)"""
+        ts = [datetime(2022, 1, 1) + timedelta(days=int(x)) for x in xs]
+        ys = np.cos(0.35 * xs)
+        return ts, ys
+
+    n = 32
+    # Input: unevenly spaced timestamps and missing data
+    xs = np.array([(x + x // 3) for x in range(n)], dtype="int")
+    ts, ys = generate_data(xs)
+    ys_with_nan = ys.copy()
+    ys_with_nan[xs % 5 >= 2] = np.nan
+    assert 0.25 < np.isnan(ys_with_nan).mean() < 0.75
+
+    # Smooth
+    ys_smooth, ts_smooth, ys_smooth_sampled, ts_smooth_sampled = whittaker_f(x=ts, y=ys, lmbd=1, d=4)
+
+    xs_expected = np.arange(42)
+    ts_expected, ys_expected = generate_data(xs_expected)
+    assert ts_smooth == ts_expected
+    assert np.isnan(ys_smooth).sum() == 0
+    numpy.testing.assert_allclose(ys_smooth, ys_expected, atol=0.1)
+
+    xs_expected = np.arange(42)[::4]
+    ts_expected, ys_expected = generate_data(xs_expected)
+    assert ts_smooth_sampled == ts_expected
+    assert np.isnan(ys_smooth_sampled).sum() == 0
+    numpy.testing.assert_allclose(ys_smooth_sampled, ys_expected, atol=0.1)