#146 Refactor anomaly detection logic in series with NaN to improve clarity.

Simplified and clarified logical operations by introducing a mask for readability and maintainability. Added test cases to ensure proper handling of NaN values during anomaly detection, covering both interior and edge cases.

Author: Tim55667757

tests/test_TradeRoutines.py

@@ -604,6 +604,22 @@ def test_HampelFilterPositive(self):
                 # Serves as a stability test with noise.
                 [False, False, False, False, False, False],
             ),
+            (
+                {
+                    "series": pd.Series([1, 2, np.nan, 2, 1]),
+                    "window": 3, "sigma": 3, "scaleFactor": 1.4826,
+                },
+                # The presence of NaN in the middle should not cause anomalies. The method skips NaNs properly.
+                [False, False, False, False, False],
+            ),
+            (
+                {
+                    "series": pd.Series([np.nan, 1, 1, 11111111, 1, 1, np.nan]),
+                    "window": 1, "sigma": 1.1, "scaleFactor": 1.4826,
+                },
+                # The presence of NaN at the edges should not cause anomalies. The method skips NaNs properly.
+                [False, False, False, True, False, False, False],
+            ),
         ]
 
         for test in testData:

tksbrokerapi/TradeRoutines.py

@@ -714,7 +714,8 @@ def HampelFilter(series: Union[list, pd.Series], window: int = 5, sigma: float =
         # Step 3: Detect anomalies for valid central values:
         new = pd.Series(False, index=series.index)
         threshold = sigma * scaleFactor * rollingMAD
-        new[rollingMAD.notna()] = delta[rollingMAD.notna()] > threshold[rollingMAD.notna()]
+        mask = rollingMAD.notna()
+        new.loc[mask] = (delta[mask] > threshold[mask])
 
         # Step 4: Handle boundary values manually (first and last `window` points):
         for i in list(range(window)) + list(range(len(series) - window, len(series))):