Added exception in pathological case in transition.py plus test.

Author: Eli

tests/test_transition.py

@@ -0,0 +1,72 @@
+import pytest
+import pandas as pd
+import numpy as np
+from vtools import transition_ts
+
+
+def pink_noise(size, alpha=1.0):
+    f = np.fft.rfftfreq(size)
+    f[0] = 1e-6
+    spectrum = 1 / f ** (alpha / 2.0)
+    phases = np.exp(2j * np.pi * np.random.rand(len(f)))
+    signal = np.fft.irfft(spectrum * phases, n=size)
+    return signal / np.std(signal)
+
+
+@pytest.fixture
+def test_series():
+    freq = "15min"
+    date0 = pd.date_range("2023-01-01", "2023-06-01", freq=freq)
+    date1 = pd.date_range("2023-05-20", "2023-09-01", freq=freq)
+    t0 = np.linspace(0, 1, len(date0))
+    t1 = np.linspace(0, 1, len(date1))
+    ts0 = pd.Series(1.0 + 0.5 * pink_noise(len(date0)) + 0.1 * np.sin(2 * np.pi * 3 * t0), index=date0)
+    ts1 = pd.Series(2.0 + 0.4 * pink_noise(len(date1)) + 0.15 * np.cos(2 * np.pi * 2 * t1), index=date1)
+    return ts0, ts1
+
+
+def test_natural_gap_linear(test_series):
+    ts0, ts1 = test_series
+    with pytest.raises(ValueError, match="overlap must be resolved with create_gap"):
+        transition_ts(ts0, ts1, method='linear', create_gap=None, return_type='series')
+
+
+def test_explicit_gap_linear(test_series):
+    ts0, ts1 = test_series
+    gap = ["2023-05-15", "2023-06-10"]
+    result = transition_ts(ts0, ts1, method="linear", create_gap=gap, return_type="series")
+    glue = transition_ts(ts0, ts1, method="linear", create_gap=gap, return_type="glue")
+    assert glue.index[0] == pd.Timestamp(gap[0])
+    assert glue.index[-1] == pd.Timestamp(gap[1])
+
+
+def test_explicit_gap_pchip_with_overlap(test_series):
+    ts0, ts1 = test_series
+    gap = ["2023-05-15", "2023-06-10"]
+    result = transition_ts(ts0, ts1, method="pchip", create_gap=gap, overlap=(10, 10), return_type="series")
+    assert result.index.is_monotonic_increasing
+    assert not result.index.duplicated().any()
+
+
+def test_gap_inside_natural_gap(test_series):
+    ts0, ts1 = test_series
+    gap = ["2023-05-10", "2023-05-15"]
+    transition = transition_ts(ts0, ts1, method="linear", create_gap=gap, return_type="glue")
+    diff = np.diff(transition.values)
+    assert np.std(diff) < 0.5
+
+
+def test_overlapping_series_with_gap(test_series):
+    ts0, ts1 = test_series
+    gap = ["2023-05-25", "2023-05-28"]
+    result = transition_ts(ts0, ts1, method="pchip", create_gap=gap, overlap=(12, 12), return_type="series")
+    assert result.index.is_monotonic_increasing
+    assert not result.index.duplicated().any()
+
+
+def test_glue_return_type(test_series):
+    ts0, ts1 = test_series
+    gap = ["2023-05-15", "2023-06-10"]
+    glue = transition_ts(ts0, ts1, method="linear", create_gap=gap, return_type="glue")
+    assert glue.index[0] >= pd.Timestamp(gap[0])
+    assert glue.index[-1] <= pd.Timestamp(gap[1])

vtools/functions/transition.py

@@ -1,10 +1,10 @@
-
 from scipy.interpolate import PchipInterpolator
 import pandas as pd
 import numpy as np
 
 __all__ = ["transition_ts"]
 
+
 def transition_ts(
     ts0, ts1, method="linear", create_gap=None, overlap=(0, 0), return_type="series"
 ):
@@ -37,6 +37,10 @@ def transition_ts(
             end_val = ts1.loc[trans_end:].iloc[0]
 
     else:
+        if ts0.index[-1] >= ts1.index[0]:
+            raise ValueError(
+                "ts0 ends after ts1 starts; overlap must be resolved with create_gap."
+            )
         trans_start = ts0.index[-1] + freq
         trans_end = ts1.index[0] - freq
         start_time = ts0.index[-1]
@@ -107,4 +111,4 @@ def transition_ts(
         ts1_trunc = ts1.loc[trans_end + freq :]
         return pd.concat([ts0_trunc, interpolated, ts1_trunc])
     else:
-        raise ValueError("return_type must be either 'glue' or 'series'.")
+        raise ValueError("return_type must be either 'glue' or 'series'.")