Major testing improvements and numerous bug fixes

.flake8

@@ -10,8 +10,11 @@ per-file-ignores =
     __init__.py: F401, F403
     obsidian/dash/*: F401, F403
 
-    # Often creating variables but not accessing them in testing
-    obsidian/tests/*: F841
+    # Often importing and creating unaccessed objects during testing
+    obsidian/tests/*: F401, F841
+
+    # No good way around comparing types for recursive state-dict comparison
+    obsidian/tests/utils.py: E721
 
 exclude =
     projects/

CHANGELOG.md

@@ -1,7 +1,16 @@
 # Changelog
 
-## [0.7.13]
+## [0.8.0]
+### Added
+- Major improvements to testing and numerous small bug fixes to improve code robustness
+- Code coverage > 90%
+- New method for asserting equivalence of state_dicts during serialization
 
+### Modified
+- Objective PyTests separated
+- Constraint PyTests separated
+
+## [0.7.13]
 ### Added
 - Campaign.Explainer now added to PyTests
 - Docstrings and typing to Explainer methods

obsidian/__init__.py

@@ -1,5 +1,5 @@
 """obsidian: Automated experiment design and black-box optimization"""
-__version__ = '0.7.12'
+__version__ = '0.8.0'
 
 # Import key objects
 from obsidian.campaign import Campaign

obsidian/campaign/campaign.py

@@ -98,7 +98,7 @@ def add_data(self, df: pd.DataFrame):
 
     def clear_data(self):
         """Clears campaign data"""
-        self.data = None
+        self.data = pd.DataFrame()
 
     @property
     def optimizer(self) -> Optimizer:
@@ -189,12 +189,11 @@ def y(self) -> pd.Series | pd.DataFrame:
         """
         Experimental response data
 
-        Raises:
-            ValueError: If no target(s) are specified.
         """
-        if not self.target:
-            raise ValueError('No target(s) specified')
-        return self.data[self.y_names]
+        if not self.data.empty:
+            return self.data[self.y_names]
+        else:
+            return None
 
     @property
     def response_max(self) -> float | pd.Series:
@@ -296,10 +295,7 @@ def load_state(cls,
         new_campaign.data = pd.DataFrame(obj_dict['data'])
         new_campaign.data.index = new_campaign.data.index.astype('int')
 
-        try:
-            new_campaign.iter = new_campaign.data['Iteration'].astype('int').max()
-        except KeyError:
-            new_campaign.iter = 0
+        new_campaign.iter = new_campaign.data['Iteration'].astype('int').max()
 
         return new_campaign
 

obsidian/campaign/explainer.py

@@ -2,6 +2,7 @@
 
 from obsidian.parameters import Param_Continuous, ParamSpace
 from obsidian.optimizer import Optimizer
+from obsidian.exceptions import UnfitError
 
 import shap
 from shap import KernelExplainer, Explanation
@@ -36,7 +37,7 @@ def __init__(self,
                  X_space: ParamSpace | None = None) -> None:
 
         if not optimizer.is_fit:
-            raise ValueError('Surrogate model in optimizer is not fit to data. ')
+            raise UnfitError('Surrogate model in optimizer is not fit to data. ')
 
         self.set_optimizer(optimizer)
         self.X_space = optimizer.X_space if X_space is None else X_space
@@ -117,7 +118,7 @@ def pred_func(X):
     def shap_summary(self) -> Figure:
         """SHAP Summary Plot (Beeswarm)"""
         if not self.shap:
-            raise ValueError('shap explainer is not fit.')
+            raise UnfitError('SHAP explainer is not fit.')
 
         fig = plt.figure()
         shap.summary_plot(self.shap['values'], self.shap['X_sample'],
@@ -129,7 +130,7 @@ def shap_summary(self) -> Figure:
     def shap_summary_bar(self) -> Figure:
         """SHAP Summary Plot (Bar Plot / Importance)"""
         if not self.shap:
-            raise ValueError('shap explainer is not fit.')
+            raise UnfitError('SHAP explainer is not fit.')
 
         fig = plt.figure()
         shap.plots.bar(self.shap['explanation'],
@@ -159,7 +160,7 @@ def shap_pdp_ice(self,
 
         """
         if not self.shap:
-            raise ValueError('shap explainer is not fit.')
+            raise UnfitError('SHAP explainer is not fit.')
 
         fig, ax = partial_dependence(
                 ind=ind,
@@ -193,7 +194,7 @@ def shap_single_point(self,
 
         """
         if not self.shap:
-            raise ValueError('shap explainer is not fit.')
+            raise UnfitError('SHAP explainer is not fit.')
 
         if isinstance(X_new, pd.Series):
             X_new = X_new.copy().to_frame().T

obsidian/parameters/base.py

@@ -12,8 +12,9 @@ def __init__(self,
                  name: str):
         self.name = name
 
+    @abstractmethod
     def __repr__(self):
-        return f"{self.__class__.__name__}(name={self.name})"
+        pass  # pragma: no cover
 
     @abstractmethod
     def _validate_value(self,

obsidian/parameters/discrete.py

@@ -97,7 +97,7 @@ def __init__(self,
             self.categories = categories.split(',')
         else:
             self.categories = categories
-        for c in categories:
+        for c in self.categories:
             self._validate_value(c)
 
     def __repr__(self):
@@ -183,12 +183,14 @@ def range(self):
 
     def __init__(self,
                  name,
-                 categories: int | float | list[int | float]):
-        if not isinstance(categories, (int, float, list)):
+                 categories: list[int | float]):
+
+        if not isinstance(categories, list):
             raise TypeError('Categories must be a number or list of numbers')
-        if isinstance(categories, list):
-            if not all(isinstance(x, (int, float)) for x in categories):
-                raise TypeError('Categories must be a list of numbers')
+
+        self.categories = categories
+        for c in self.categories:
+            self._validate_value(c)
 
         self.name = name
         self.categories = categories if isinstance(categories, list) else [categories]

obsidian/parameters/targets.py

@@ -64,26 +64,31 @@ def transform_f(self,
         if not (isinstance(f, (pd.Series, pd.DataFrame, np.ndarray, list, float, int))
                 or torch.is_tensor(f)):
             raise TypeError('f being transformed must be numeric or array-like')
-
-        if not fit:
-            if not hasattr(self, 'f_transform_func'):
-                raise UnfitError('Transform function is being called without being fit first.')
-
+
+        # Convert everything to numpy except Tensors
         if isinstance(f, (float, int)):
-            f = [f]
-
+            f = np.array([f])
+        if isinstance(f, (list)):
+            f = np.array(f)
         if isinstance(f, (pd.Series, pd.DataFrame)):
             f = f.values
 
         if not torch.is_tensor(f):
+            # Check that types are valid, then convert to Tensor
+            if not all(np.issubdtype(f_i.dtype, np.number) for f_i in f.flatten()):
+                raise TypeError('Each element of f being transformed must be numeric')
             f = torch.tensor(f)
 
+        if not fit:
+            if not hasattr(self, 'f_transform_func'):
+                raise UnfitError('Transform function is being called without being fit first.')
+
         if f.ndim == 1:
             f = f.reshape(-1, 1)
 
         if inverse:
-            f_obj = self.f_transform_func.inverse(f)
-            return pd.Series(f_obj.flatten(), name=self.name) * self.multiplier
+            f_obj = self.f_transform_func.inverse(f * self.multiplier)
+            return pd.Series(f_obj.flatten(), name=self.name)
         else:
             if fit:
                 self.f_transform_func = f_transform_dict[self.f_transform]()

obsidian/parameters/transforms.py

@@ -127,7 +127,7 @@ def forward(self,
             X_s = self.params['scale']*(X - self.params['loc'])
             valid_range = (X_s >= 0).all() and (X_s <= 1).all()
             if not valid_range:
-                warnings.warn('Invalid range provided for logit scaler, proceeding with min-max fit')
+                warnings.warn('Invalid range provided for logit scaler, proceeding with min-max fit', UserWarning)
                 self._fit_minmax(X)
                 return self.forward(X)
         else:

obsidian/tests/param_configs.py

@@ -1,6 +1,9 @@
+"""Preset parameter configurations for unit testing"""
+
 from obsidian.parameters import Param_Continuous, Param_Ordinal, Param_Categorical, \
     Param_Observational, Param_Discrete_Numeric, ParamSpace
 
+# Set up ap master list of parameter spaces for testing
 params = [
         Param_Continuous('Parameter 1', 0, 10),
         Param_Continuous('Parameter 2', -20, 0),
@@ -15,14 +18,25 @@
         Param_Ordinal('Parameter 11', ['N'])
     ]
 
+# Subset some default selections
 default = [params[i] for i in [0, 1, 2, 6]]  # 2 continuous, 1 static, 1 categorical
+X_sp_default = ParamSpace(params=default)
+
+# Numeric
 cont_small = [params[i] for i in [0, 1, 2]]  # continuous including edge cases
 numeric = [params[i] for i in [0, 1, 2, 3, 4, 5]]  # numeric including edge cases
+X_sp_cont_small = ParamSpace(params=cont_small)
+X_sp_numeric = ParamSpace(params=numeric)
 
+# Nominal
 cat_small = [params[i] for i in [6, 7, 8]]  # categorical including edge cases
 disc_small = [params[i] for i in [6, 9]]  # 1 categorical, 1 ordinal
 disc_large = [params[i] for i in [6, 7, 8, 9, 10]]  # discrete including edge cases
+X_sp_cat_small = ParamSpace(params=cat_small)
+X_sp_disc_small = ParamSpace(params=disc_small)
+X_sp_disc_large = ParamSpace(params=disc_large)
 
+# Set up a range of continuous parameters
 params_cont_large = [
         Param_Continuous('Parameter 1', 0, 10),
         Param_Continuous('Parameter 2', 0, 10),
@@ -37,15 +51,8 @@
         Param_Continuous('Parameter 11', 0, 10),
         Param_Continuous('Parameter 12', 0, 10),
 ]
-
-X_sp_default = ParamSpace(params=default)
-X_sp_cont_small = ParamSpace(params=cont_small)
 X_sp_cont_large = ParamSpace(params=params_cont_large)
-X_sp_numeric = ParamSpace(params=numeric)
-X_sp_cat_small = ParamSpace(params=cat_small)
-X_sp_disc_small = ParamSpace(params=disc_small)
-X_sp_disc_large = ParamSpace(params=disc_large)
+X_sp_cont_ndims = [ParamSpace(params_cont_large[:i]) for i in range(len(params_cont_large))]
 
+# Wrap everything for iteration during testing
 test_X_space = [X_sp_default, X_sp_cont_small, X_sp_numeric, X_sp_cat_small, X_sp_disc_small, X_sp_disc_large]
-
-X_sp_cont_ndims = [ParamSpace(params_cont_large[:i]) for i in range(len(params_cont_large))]