Fixes before 0.5.0

README.md

@@ -24,7 +24,7 @@ for more information.
 
 ## Installation
 
-The preferred way to install `dtaianomaly` is via PyPi. See the [documentation](https://dtaianomaly.readthedocs.io/en/stable/index.html) 
+The preferred way to install `dtaianomaly` is via PyPi. See the [documentation](https://dtaianomaly.readthedocs.io/en/stable/getting_started/installation.html) 
 for more options.
 ```
 pip install dtaianomaly

docs/additional_information/changelog.rst

@@ -42,12 +42,14 @@ Changed
 - Replaced dependency on ``tslearn>=0.6.3`` by ``sktime[clustering]``, which includes the ``tslearn``
   dependency. Before, ``tslearn`` was only used for ``KShapeAnomalyDetector`` to do the clustering. This
   capability is also offered by ``sktime`` through a direct interface to ``tslearn``.
+- Added dependency ``tslearn>=0.6.3`` as it is not installed for Python >= 3.12 via ``sktime``.
 
 Fixed
 ^^^^^
 
 - np.datetime64 are now also valid values in the ``utils.is_valid_array_like``.
   This was especially necessary for time indexes in a dataset.
+- Added missing tests and removed temporary code to test the doctests.
 
 [0.4.2] - 2025-07-03
 --------------------

docs/getting_started/installation.rst

@@ -29,10 +29,12 @@ Currently, following subsets are available:
 - ``docs``: Dependencies for generating the documentation.
 - ``notebooks``: Dependencies for using jupyter notebooks.
 - ``lint``: Dependencies for linting the code.
-- ``chronos``: Install autogluon.timeseries, necessary for running Chronos.
-- ``moment``: Install momentfm, necessary for running Chronos. **Warning:** Not included when
-  installing ``dtaianomaly[all]`` due to dependency conflicts!
-- ``time_moe``: Install transformers==4.40.1, necessary for running MOMENT.
+- ``chronos``: Install autogluon.timeseries, necessary for running Chronos. **Warning:** Not
+  included when installing ``dtaianomaly[all]``!
+- ``moment``: Install momentfm, necessary for running MOMENT. **Warning:** Not included when
+  installing ``dtaianomaly[all]``!
+- ``time_moe``: Install transformers==4.40.1, necessary for running Time-MoE. **Warning:** Not
+  included when  installing ``dtaianomaly[all]``!
 - ``in_time_ad``: Dependencies for running the demonstrator.
 
 To install version ``X.Y.Z``, use the following command:

dtaianomaly/anomaly_detection/_BaseNeuralForecastingDetector.py

@@ -198,11 +198,6 @@ def _evaluate_batch(self, batch: list[torch.Tensor]) -> torch.Tensor:
                 torch.abs(forecast - future), dim=tuple(range(1, forecast.ndim))
             )
 
-        # Raise an error if invalid metric is given
-        raise ValueError(
-            f"Unknown error_metric '{self.error_metric}'. Valid options are ['mean-squared-error', 'mean-absolute-error']"
-        )
-
     def _evaluate(self, data_loader: torch.utils.data.DataLoader) -> np.array:
         decision_scores = super()._evaluate(data_loader)
         return np.concatenate(

dtaianomaly/anomaly_detection/_Chronos.py

@@ -94,11 +94,11 @@ class Chronos(BaseDetector):
 
     Examples
     --------
-    >>> from dtaianomaly.anomaly_detection import Chronos
-    >>> from dtaianomaly.data import demonstration_time_series
-    >>> x, y = demonstration_time_series()
-    >>> chronos = Chronos(10).fit(x)
-    >>> chronos.decision_function(x)  # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
+    >>> from dtaianomaly.anomaly_detection import Chronos  # doctest: +SKIP
+    >>> from dtaianomaly.data import demonstration_time_series  # doctest: +SKIP
+    >>> x, y = demonstration_time_series()  # doctest: +SKIP
+    >>> chronos = Chronos(10).fit(x)  # doctest: +SKIP
+    >>> chronos.decision_function(x)  # doctest: +SKIP
     array([0.00027719, 0.00027719, 0.00027719, ..., 0.00058781, 0.02628242,
            0.00010728]...)
     """

dtaianomaly/anomaly_detection/_HybridKNearestNeighbors.py

@@ -121,7 +121,7 @@ class HybridKNearestNeighbors(BaseDetector):
     >>> hybrid_knn = HybridKNearestNeighbors(64, seed=0).fit(x)
     >>> hybrid_knn.decision_function(x)  # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
     array([0.12284644, 0.38202247, 0.43220974, ..., 0.83470662, 0.81722846,
-           0.85243446])
+           0.85243446]...)
     """
 
     window_size: WINDOW_SIZE_TYPE

dtaianomaly/anomaly_detection/_KShapeAnomalyDetector.py

@@ -76,7 +76,7 @@ class KShapeAnomalyDetector(BaseDetector):
     >>> kshape = KShapeAnomalyDetector(window_size=50).fit(x)
     >>> kshape.decision_function(x)  # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
     array([7.07106781, 7.07106781, 7.07106781, ..., 7.07106781, 7.07106781,
-           7.07106781])
+           7.07106781]...)
     """
 
     window_size: WINDOW_SIZE_TYPE
@@ -203,9 +203,3 @@ def _ncc_c(x: np.array, y: np.array) -> np.array:
     cc = np.fft.ifft(np.fft.fft(x, fft_size) * np.conj(np.fft.fft(y, fft_size)))
     cc = np.concatenate((cc[-(x.shape[0] - 1) :], cc[: x.shape[0]]))
     return np.real(cc) / den
-
-
-if __name__ == "__main__":
-    import doctest
-
-    doctest.testmod()

dtaianomaly/anomaly_detection/_MatrixProfileDetector.py

@@ -186,26 +186,3 @@ def is_fitted(self) -> bool:
                 for attr in self.__annotations__
                 if attr.endswith("_") and attr != "X_reference_"
             )
-
-    def requires_fitting(self) -> bool:
-        """
-        Check whether this object requires fitting.
-
-        Check whether any of the attributes of this object ends with an
-        underscore ('_'), which indicates that the attribute is set when
-        the object is fitted. Note that this method does not check whether
-        the object is fitted, i.e., whether the attributes have been set.
-
-        Returns
-        -------
-        bool
-            True if and only if this object has attributes that end with '_'.
-        """
-        if self.novelty:
-            return super().requires_fitting()
-        else:
-            return any(
-                attr.endswith("_")
-                for attr in self._all_annotations()
-                if attr != "X_reference_"
-            )

dtaianomaly/anomaly_detection/_SpectralResidual.py

@@ -112,9 +112,3 @@ def _decision_function(self, X: np.ndarray) -> np.array:
         saliency_map = np.sqrt(inverse_fourier.real**2 + inverse_fourier.imag**2)
 
         return saliency_map
-
-
-if __name__ == "__main__":
-    import doctest
-
-    doctest.testmod()

dtaianomaly/anomaly_detection/_TimeMoE.py

@@ -74,11 +74,11 @@ class TimeMoE(BaseDetector):
 
     Examples
     --------
-    >>> from dtaianomaly.anomaly_detection import TimeMoE
-    >>> from dtaianomaly.data import demonstration_time_series
-    >>> x, y = demonstration_time_series()
-    >>> time_moe = TimeMoE(10).fit(x)
-    >>> time_moe.decision_function(x)  # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE +SKIP
+    >>> from dtaianomaly.anomaly_detection import TimeMoE  # doctest: +SKIP
+    >>> from dtaianomaly.data import demonstration_time_series  # doctest: +SKIP
+    >>> x, y = demonstration_time_series()  # doctest: +SKIP
+    >>> time_moe = TimeMoE(10).fit(x)  # doctest: +SKIP
+    >>> time_moe.decision_function(x)  # doctest: +SKIP
     array([6.34949149e-05, 6.34949149e-05, 6.34949149e-05, ...,
            6.34949149e-05, 6.34949149e-05, 6.34949149e-05]...)
     """

dtaianomaly/anomaly_detection/baselines/_AlwaysAnomalous.py

@@ -20,7 +20,7 @@ class AlwaysAnomalous(BaseDetector):
     >>> x, y = demonstration_time_series()
     >>> baseline = AlwaysAnomalous().fit(x)
     >>> baseline.decision_function(x)
-    array([1., 1., 1., ..., 1., 1., 1.])
+    array([1., 1., 1., ..., 1., 1., 1.]...)
     """
 
     def __init__(self):

dtaianomaly/anomaly_detection/baselines/_AlwaysNormal.py

@@ -20,7 +20,7 @@ class AlwaysNormal(BaseDetector):
     >>> x, y = demonstration_time_series()
     >>> baseline = AlwaysNormal().fit(x)
     >>> baseline.decision_function(x)
-    array([0., 0., 0., ..., 0., 0., 0.])
+    array([0., 0., 0., ..., 0., 0., 0.]...)
     """
 
     def __init__(self):

dtaianomaly/anomaly_detection/baselines/_MovingWindowVariance.py

@@ -41,7 +41,7 @@ class MovingWindowVariance(BaseDetector):
     >>> baseline = MovingWindowVariance(16).fit(x)
     >>> baseline.decision_function(x)
     array([0.06820711, 0.07130246, 0.07286874, ..., 0.01125165, 0.00984333,
-           0.00986772])
+           0.00986772]...)
     """
 
     window_size: WINDOW_SIZE_TYPE

dtaianomaly/anomaly_detection/baselines/_RandomDetector.py

@@ -26,7 +26,7 @@ class RandomDetector(BaseDetector):
     >>> baseline = RandomDetector(seed=0).fit(x)
     >>> baseline.decision_function(x)
     array([0.63696169, 0.26978671, 0.04097352, ..., 0.70724404, 0.90315986,
-           0.8944909 ])
+           0.8944909 ]...)
     """
 
     seed: int | None

dtaianomaly/anomaly_detection/baselines/_SquaredDifference.py

@@ -30,7 +30,7 @@ class SquaredDifference(BaseDetector):
     >>> baseline = SquaredDifference().fit(x)
     >>> baseline.decision_function(x)
     array([0.00779346, 0.00779346, 0.00260361, ..., 0.00286662, 0.05578398,
-           0.02683475])
+           0.02683475]...)
     """
 
     square_errors: bool
@@ -54,19 +54,3 @@ def _decision_function(self, X: np.ndarray) -> np.array:
         if self.square_errors:
             decision_scores = np.square(decision_scores)
         return decision_scores
-
-
-def main():
-    from dtaianomaly.data import demonstration_time_series
-    from dtaianomaly.visualization import plot_anomaly_scores
-
-    x, y = demonstration_time_series()
-    x = x.reshape(-1, 1)
-
-    baseline = SquaredDifference().fit(x)
-    y_pred = baseline.decision_function(x)
-    plot_anomaly_scores(x, y, y_pred, figsize=(20, 5)).show()
-
-
-if __name__ == "__main__":
-    main()

dtaianomaly/pipeline/_Pipeline.py

@@ -35,7 +35,7 @@ class Pipeline(BaseDetector):
     >>> pipeline = Pipeline(StandardScaler(), IsolationForest(16))
     >>> pipeline.fit(X).decision_function(X)
     array([-0.01080726, -0.01053199, -0.00883758, ..., -0.05298726,
-           -0.05898066, -0.05713733])
+           -0.05898066, -0.05713733]...)
     """
 
     preprocessor: Preprocessor

dtaianomaly/windowing/_sliding_window.py

@@ -59,9 +59,3 @@ def sliding_window(X: np.ndarray, window_size: int, stride: int) -> np.ndarray:
     ]
     windows.append(X[-window_size:].ravel())
     return np.array(windows)
-
-
-if __name__ == "__main__":
-    import doctest
-
-    doctest.testmod()

pyproject.toml

@@ -25,7 +25,8 @@ dependencies = [
     "matplotlib>=3.7",
     "statsmodels>=0.6",
     "pyod>=2.0.0",
-    "sktime[clustering]",
+    "tslearn>=0.6.3",
+    "sktime",
     "toml",
     "torch>=1.8.0",
 ]
@@ -73,8 +74,6 @@ all = [  # All the optional dependencies
     "isort",
     'tqdm',
     "sphinxcontrib-bibtex",
-    "autogluon.timeseries>=1.3.1",
-    "transformers==4.40.1",
     'tqdm',
     "streamlit",
     "plotly"

tests/anomaly_detection/test_Chronos.py

@@ -8,72 +8,114 @@
 from dtaianomaly.anomaly_detection._Chronos import MODEL_PATHS
 
 
+def setup():
+    # create dummy parent package and submodule
+    autogluon = types.ModuleType("autogluon")
+    timeseries = types.ModuleType("autogluon.timeseries")
+    autogluon.timeseries = timeseries
+
+    # register both in sys.modules
+    sys.modules["autogluon"] = autogluon
+    sys.modules["autogluon.timeseries"] = timeseries
+
+
+def cleanup():
+    del sys.modules["autogluon"]
+    del sys.modules["autogluon.timeseries"]
+
+
 class TestChronos:
 
     def test_supervision(self):
+        setup()
         detector = Chronos(1)
         assert detector.supervision == Supervision.UNSUPERVISED
+        cleanup()
 
     def test_str(self):
+        setup()
         assert str(Chronos(5)) == "Chronos(window_size=5)"
         assert str(Chronos("fft")) == "Chronos(window_size='fft')"
         assert str(Chronos(15, "large")) == "Chronos(window_size=15,model_path='large')"
         assert str(Chronos(25, batch_size=3)) == "Chronos(window_size=25,batch_size=3)"
+        cleanup()
 
     @pytest.mark.parametrize("model_path", MODEL_PATHS)
     def test_model_path_valid(self, model_path):
+        setup()
         detector = Chronos(window_size="fft", model_path=model_path)
         assert detector.model_path == model_path
+        cleanup()
 
     @pytest.mark.parametrize("model_path", [0, True, None, ["a", "list"]])
     def test_model_path_invalid_type(self, model_path):
+        setup()
         with pytest.raises(TypeError):
             Chronos(window_size="fft", model_path=model_path)
+        cleanup()
 
     @pytest.mark.parametrize("model_path", ["invalid"])
     def test_model_path_invalid_value(self, model_path):
+        setup()
         with pytest.raises(ValueError):
             Chronos(window_size="fft", model_path=model_path)
+        cleanup()
 
     @pytest.mark.parametrize("batch_size", [8, 16, 32])
     def test_batch_size_valid(self, batch_size):
+        setup()
         detector = Chronos(window_size="fft", batch_size=batch_size)
         assert detector.batch_size == batch_size
+        cleanup()
 
     @pytest.mark.parametrize("batch_size", ["8", 8.0])
     def test_batch_size_invalid_type(self, batch_size):
+        setup()
         with pytest.raises(TypeError):
             Chronos(window_size="fft", batch_size=batch_size)
+        cleanup()
 
     @pytest.mark.parametrize("batch_size", [0, -8])
     def test_batch_size_invalid_value(self, batch_size):
+        setup()
         with pytest.raises(ValueError):
             Chronos(window_size="fft", batch_size=batch_size)
+        cleanup()
 
     @pytest.mark.parametrize("forecast_horizon", [32, 16, 8])
     def test_forecast_horizon_valid(self, forecast_horizon):
+        setup()
         detector = Chronos(window_size=16, forecast_horizon=forecast_horizon)
         assert detector.forecast_horizon == forecast_horizon
+        cleanup()
 
     @pytest.mark.parametrize("forecast_horizon", ["32", 16.0, True])
     def test_forecast_horizon_invalid_type(self, forecast_horizon):
+        setup()
         with pytest.raises(TypeError):
             Chronos(window_size=16, forecast_horizon=forecast_horizon)
+        cleanup()
 
     @pytest.mark.parametrize("forecast_horizon", [0, -1, -16])
     def test_forecast_horizon_invalid_value(self, forecast_horizon):
+        setup()
         with pytest.raises(ValueError):
             Chronos(window_size=16, forecast_horizon=forecast_horizon)
+        cleanup()
 
     @pytest.mark.parametrize("do_fine_tuning", [True, False])
     def test_do_fine_tuning_valid(self, do_fine_tuning):
+        setup()
         detector = Chronos(window_size="fft", do_fine_tuning=do_fine_tuning)
         assert detector.do_fine_tuning == do_fine_tuning
+        cleanup()
 
     @pytest.mark.parametrize("do_fine_tuning", [5, 1.0, "invalid"])
     def test_do_fine_tuning_invalid_type(self, do_fine_tuning):
+        setup()
         with pytest.raises(TypeError):
             Chronos(window_size="fft", do_fine_tuning=do_fine_tuning)
+        cleanup()
 
     def test_raises_if_autogluon_missing(self, monkeypatch):
         # simulate ImportError when trying to import autogluon.timeseries
@@ -93,11 +135,6 @@ def fake_import(name, *args, **kwargs):
             Chronos(15)
 
     def test_no_error_if_autogluon_available(self, monkeypatch):
-        # simulate a dummy autogluon.timeseries module
-        sys.modules["autogluon.timeseries"] = types.ModuleType("autogluon.timeseries")
-
-        # should NOT raise
+        setup()
         Chronos(15)  # just runs, no exception
-
-        # cleanup (avoid side effects on other tests)
-        del sys.modules["autogluon.timeseries"]
+        cleanup()