[Feature] Add tests and pytest github action (#28)

- Add tests and pytest github action
- Remove `model` property in `MQDataset` class
- Add epsilon validation logic

.github/workflows/pypi_release.yaml

@@ -18,3 +18,4 @@ jobs:
       uses: JRubics/poetry-publish@v2.0
       with:
         pypi_token: ${{ secrets.PYPI_TOKEN }}
+        poetry_install_options: "--without dev"

.github/workflows/test.yaml

@@ -0,0 +1,33 @@
+name: Tests
+
+on:
+  pull_request:
+    types: [opened, reopened, ready_for_review, review_requested]
+
+jobs:
+  run-pytest:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Set up python
+      uses: actions/setup-python@v5
+      with:
+        python-version: '3.12'
+
+    - name: Install Poetry
+      uses: snok/install-poetry@v1
+
+    - name: Install dependencies
+      run: poetry install --no-interaction --with dev
+
+    - name: Run tests
+      run: poetry run pytest --junitxml=pytest.xml --cov-report=term-missing:skip-covered --cov=tests/ > pytest-coverage.txt
+
+    - name: Pytest coverage comment
+      uses: MishaKav/pytest-coverage-comment@main
+      with:
+        pytest-coverage-path: ./pytest-coverage.txt
+        junitxml-path: ./pytest.xml

mqboost/constraints.py

@@ -1,13 +1,6 @@
 import pandas as pd
 
-from mqboost.base import (
-    FUNC_TYPE,
-    ModelName,
-    MQStr,
-    ParamsLike,
-    TypeName,
-    ValidationException,
-)
+from mqboost.base import FUNC_TYPE, ModelName, MQStr, ParamsLike, TypeName
 
 
 def set_monotone_constraints(
@@ -27,10 +20,6 @@ def set_monotone_constraints(
         ParamsLike
     """
     constraints_fucs = FUNC_TYPE.get(model_name).get(TypeName.constraints_type)
-    if MQStr.obj.value in params:
-        raise ValidationException(
-            "The parameter named 'objective' must be excluded in params"
-        )
     _params = params.copy()
     if MQStr.mono.value in _params:
         _monotone_constraints = list(_params[MQStr.mono.value])

mqboost/dataset.py

@@ -74,11 +74,6 @@ def predict_dtype(self) -> Callable:
         """Get the data type function for prediction data."""
         return self._predict_dtype
 
-    @property
-    def model(self) -> ModelName:
-        """Get the model type."""
-        return self._model
-
     @property
     def columns(self) -> pd.Index:
         """Get the column names of the input features."""

mqboost/objective.py

@@ -4,7 +4,7 @@
 import numpy as np
 
 from mqboost.base import DtrainLike, ModelName, ObjectiveName
-from mqboost.utils import delta_validate
+from mqboost.utils import delta_validate, epsilon_validate
 
 CHECK_LOSS: str = "check_loss"
 
@@ -26,19 +26,13 @@ def _hess_rho(error: np.ndarray, alpha: float) -> np.ndarray:
 
 
 # Huber loss
-def _error_delta_compare(
-    error: np.ndarray, delta: float
-) -> tuple[np.ndarray, np.ndarray]:
-    """Rerutn two boolean arrays indicating where the errors are smaller or larger than delta."""
-    _abs_error = np.abs(error)
-    return (_abs_error <= delta).astype(int), (_abs_error > delta).astype(int)
-
-
 def _grad_huber(error: np.ndarray, alpha: float, delta: float) -> np.ndarray:
     """Compute the gradient of the huber loss function."""
-    _smaller_delta, _bigger_delta = _error_delta_compare(error=error, delta=delta)
-    _grad = _grad_rho(error=error, alpha=alpha)
+    _abs_error = np.abs(error)
+    _smaller_delta = (_abs_error <= delta).astype(int)
+    _bigger_delta = (_abs_error > delta).astype(int)
     _r = _rho(error=error, alpha=alpha)
+    _grad = _grad_rho(error=error, alpha=alpha)
     return _r * _smaller_delta + _grad * _bigger_delta
 
 
@@ -174,9 +168,10 @@ def __init__(
         if objective == ObjectiveName.check:
             self._fobj = partial(check_loss_grad_hess, alphas=alphas)
         elif objective == ObjectiveName.huber:
-            _delta = delta_validate(delta=delta)
-            self._fobj = partial(huber_loss_grad_hess, alphas=alphas, delta=_delta)
+            delta_validate(delta=delta)
+            self._fobj = partial(huber_loss_grad_hess, alphas=alphas, delta=delta)
         elif objective == ObjectiveName.approx:
+            epsilon_validate(epsilon=epsilon)
             self._fobj = partial(approx_loss_grad_hess, alphas=alphas, epsilon=epsilon)
 
         if model == ModelName.lightgbm:

mqboost/optimize.py

@@ -19,7 +19,7 @@
 from mqboost.constraints import set_monotone_constraints
 from mqboost.dataset import MQDataset
 from mqboost.objective import MQObjective
-from mqboost.utils import delta_validate
+from mqboost.utils import delta_validate, epsilon_validate, params_validate
 
 __all__ = ["MQOptimizer"]
 
@@ -93,9 +93,12 @@ def __init__(
         epsilon: float = 1e-5,
     ) -> None:
         """Initialize the MQOptimizer."""
+        delta_validate(delta=delta)
+        epsilon_validate(epsilon=epsilon)
+
         self._model = ModelName.get(model)
         self._objective = ObjectiveName.get(objective)
-        self._delta = delta_validate(delta)
+        self._delta = delta
         self._epsilon = epsilon
         self._get_params = _GET_PARAMS_FUNC.get(self._model)
 
@@ -179,6 +182,7 @@ def __optuna_objective(
     ) -> float:
         """Objective function for Optuna to minimize."""
         params = get_params_func(trial=trial)
+        params_validate(params=params)
         params = set_monotone_constraints(
             params=params,
             columns=self._dataset.columns,

mqboost/regressor.py

@@ -6,6 +6,7 @@
 from mqboost.constraints import set_monotone_constraints
 from mqboost.dataset import MQDataset
 from mqboost.objective import MQObjective
+from mqboost.utils import params_validate
 
 __all__ = ["MQRegressor"]
 
@@ -46,6 +47,7 @@ def __init__(
         epsilon: float = 1e-5,
     ) -> None:
         """Initialize the MQRegressor."""
+        params_validate(params=params)
         self._params = params
         self._model = ModelName.get(model)
         self._objective = ObjectiveName.get(objective)

mqboost/utils.py

@@ -4,7 +4,14 @@
 import numpy as np
 import pandas as pd
 
-from mqboost.base import AlphaLike, ValidationException, XdataLike, YdataLike
+from mqboost.base import (
+    AlphaLike,
+    MQStr,
+    ParamsLike,
+    ValidationException,
+    XdataLike,
+    YdataLike,
+)
 
 
 def alpha_validate(
@@ -49,7 +56,7 @@ def prepare_x(
     _alpha_repeat_count_list = [list(repeat(alpha, len(x))) for alpha in alphas]
     _alpha_repeat_list = list(chain.from_iterable(_alpha_repeat_count_list))
 
-    _repeated_x = pd.concat([x] * len(alphas), axis=0)
+    _repeated_x = pd.concat([x] * len(alphas), axis=0).reset_index(drop=True)
     _repeated_x = _repeated_x.assign(
         _tau=_alpha_repeat_list,
     )
@@ -64,7 +71,7 @@ def prepare_y(
     return np.concatenate(list(repeat(y, len(alphas))))
 
 
-def delta_validate(delta: float) -> float:
+def delta_validate(delta: float) -> None:
     """Validates the delta parameter ensuring it is a positive float and less than or equal to 0.05."""
     _delta_upper_bound: float = 0.05
 
@@ -77,4 +84,18 @@ def delta_validate(delta: float) -> float:
     if delta > _delta_upper_bound:
         warnings.warn("Delta should be 0.05 or less.")
 
-    return delta
+
+def epsilon_validate(epsilon: float) -> None:
+    if not isinstance(epsilon, float):
+        raise ValidationException("Epsilon is not float type")
+
+    if epsilon <= 0:
+        raise ValidationException("Epsilon must be positive")
+
+
+def params_validate(params: ParamsLike) -> None:
+    """Validates the model parameter ensuring its key dosen't contain 'objective'."""
+    if MQStr.obj.value in params:
+        raise ValidationException(
+            "The parameter named 'objective' must be excluded in params"
+        )