1+ """Base class for all components."""
2+ import copy
3+ from abc import ABC , abstractmethod
4+
5+ import cloudpickle
6+
7+ from checkmates .exceptions import MethodPropertyNotFoundError
8+ from checkmates .pipelines .component_base_meta import ComponentBaseMeta
9+ from checkmates .utils import (
10+ _downcast_nullable_X ,
11+ _downcast_nullable_y ,
12+ classproperty ,
13+ infer_feature_types ,
14+ log_subtitle ,
15+ safe_repr ,
16+ )
17+ from checkmates .utils .logger import get_logger
18+
19+
20+ class ComponentBase (ABC , metaclass = ComponentBaseMeta ):
21+ """Base class for all components.
22+
23+ Args:
24+ parameters (dict): Dictionary of parameters for the component. Defaults to None.
25+ component_obj (obj): Third-party objects useful in component implementation. Defaults to None.
26+ random_seed (int): Seed for the random number generator. Defaults to 0.
27+ """
28+
29+ _default_parameters = None
30+ _can_be_used_for_fast_partial_dependence = True
31+ # Referring to the pandas nullable dtypes; not just woodwork logical types
32+ _integer_nullable_incompatibilities = []
33+ _boolean_nullable_incompatibilities = []
34+ is_multiseries = False
35+
36+ def __init__ (self , parameters = None , component_obj = None , random_seed = 0 , ** kwargs ):
37+ """Base class for all components.
38+
39+ Args:
40+ parameters (dict): Dictionary of parameters for the component. Defaults to None.
41+ component_obj (obj): Third-party objects useful in component implementation. Defaults to None.
42+ random_seed (int): Seed for the random number generator. Defaults to 0.
43+ kwargs (Any): Any keyword arguments to pass into the component.
44+ """
45+ self .random_seed = random_seed
46+ self ._component_obj = component_obj
47+ self ._parameters = parameters or {}
48+ self ._is_fitted = False
49+
50+ @property
51+ @classmethod
52+ @abstractmethod
53+ def name (cls ):
54+ """Returns string name of this component."""
55+
56+ @property
57+ @classmethod
58+ @abstractmethod
59+ def modifies_features (cls ):
60+ """Returns whether this component modifies (subsets or transforms) the features variable during transform.
61+
62+ For Estimator objects, this attribute determines if the return
63+ value from `predict` or `predict_proba` should be used as
64+ features or targets.
65+ """
66+
67+ @property
68+ @classmethod
69+ @abstractmethod
70+ def modifies_target (cls ):
71+ """Returns whether this component modifies (subsets or transforms) the target variable during transform.
72+
73+ For Estimator objects, this attribute determines if the return
74+ value from `predict` or `predict_proba` should be used as
75+ features or targets.
76+ """
77+
78+ @property
79+ @classmethod
80+ @abstractmethod
81+ def training_only (cls ):
82+ """Returns whether or not this component should be evaluated during training-time only, or during both training and prediction time."""
83+
84+ @classproperty
85+ def needs_fitting (self ):
86+ """Returns boolean determining if component needs fitting before calling predict, predict_proba, transform, or feature_importances.
87+
88+ This can be overridden to False for components that do not need to be fit or whose fit methods do nothing.
89+
90+ Returns:
91+ True.
92+ """
93+ return True
94+
95+ @property
96+ def parameters (self ):
97+ """Returns the parameters which were used to initialize the component."""
98+ return copy .copy (self ._parameters )
99+
100+ @classproperty
101+ def default_parameters (cls ):
102+ """Returns the default parameters for this component.
103+
104+ Our convention is that Component.default_parameters == Component().parameters.
105+
106+ Returns:
107+ dict: Default parameters for this component.
108+ """
109+ if cls ._default_parameters is None :
110+ cls ._default_parameters = cls ().parameters
111+
112+ return cls ._default_parameters
113+
114+ @classproperty
115+ def _supported_by_list_API (cls ):
116+ return not cls .modifies_target
117+
118+ def _handle_partial_dependence_fast_mode (
119+ self ,
120+ pipeline_parameters ,
121+ X = None ,
122+ target = None ,
123+ ):
124+ """Determines whether or not a component can be used with partial dependence's fast mode.
125+
126+ Args:
127+ pipeline_parameters (dict): Pipeline parameters that will be used to create the pipelines
128+ used in partial dependence fast mode.
129+ X (pd.DataFrame, optional): Holdout data being used for partial dependence calculations.
130+ target (str, optional): The target whose values we are trying to predict.
131+ """
132+ if self ._can_be_used_for_fast_partial_dependence :
133+ return pipeline_parameters
134+
135+ raise TypeError (
136+ f"Component { self .name } ,
137+ )
138+
139+ def clone (self ):
140+ """Constructs a new component with the same parameters and random state.
141+
142+ Returns:
143+ A new instance of this component with identical parameters and random state.
144+ """
145+ return self .__class__ (** self .parameters , random_seed = self .random_seed )
146+
147+ def fit (self , X , y = None ):
148+ """Fits component to data.
149+
150+ Args:
151+ X (pd.DataFrame): The input training data of shape [n_samples, n_features]
152+ y (pd.Series, optional): The target training data of length [n_samples]
153+
154+ Returns:
155+ self
156+
157+ Raises:
158+ MethodPropertyNotFoundError: If component does not have a fit method or a component_obj that implements fit.
159+ """
160+ X = infer_feature_types (X )
161+ if y is not None :
162+ y = infer_feature_types (y )
163+ try :
164+ self ._component_obj .fit (X , y )
165+ return self
166+ except AttributeError :
167+ raise MethodPropertyNotFoundError (
168+ "Component requires a fit method or a component_obj that implements fit" ,
169+ )
170+
171+ def describe (self , print_name = False , return_dict = False ):
172+ """Describe a component and its parameters.
173+
174+ Args:
175+ print_name(bool, optional): whether to print name of component
176+ return_dict(bool, optional): whether to return description as dictionary in the format {"name": name, "parameters": parameters}
177+
178+ Returns:
179+ None or dict: Returns dictionary if return_dict is True, else None.
180+ """
181+ logger = get_logger (f"{ __name__ } )
182+ if print_name :
183+ title = self .name
184+ log_subtitle (logger , title )
185+ for parameter in self .parameters :
186+ parameter_str = ("\t * {} : {}" ).format (
187+ parameter ,
188+ self .parameters [parameter ],
189+ )
190+ logger .info (parameter_str )
191+ if return_dict :
192+ component_dict = {"name" : self .name }
193+ component_dict .update ({"parameters" : self .parameters })
194+ return component_dict
195+
196+ def save (self , file_path , pickle_protocol = cloudpickle .DEFAULT_PROTOCOL ):
197+ """Saves component at file path.
198+
199+ Args:
200+ file_path (str): Location to save file.
201+ pickle_protocol (int): The pickle data stream format.
202+ """
203+ with open (file_path , "wb" ) as f :
204+ cloudpickle .dump (self , f , protocol = pickle_protocol )
205+
206+ @staticmethod
207+ def load (file_path ):
208+ """Loads component at file path.
209+
210+ Args:
211+ file_path (str): Location to load file.
212+
213+ Returns:
214+ ComponentBase object
215+ """
216+ with open (file_path , "rb" ) as f :
217+ return cloudpickle .load (f )
218+
219+ def __eq__ (self , other ):
220+ """Check for equality."""
221+ if not isinstance (other , self .__class__ ):
222+ return False
223+ random_seed_eq = self .random_seed == other .random_seed
224+ if not random_seed_eq :
225+ return False
226+ attributes_to_check = ["_parameters" , "_is_fitted" ]
227+ for attribute in attributes_to_check :
228+ if getattr (self , attribute ) != getattr (other , attribute ):
229+ return False
230+ return True
231+
232+ def __str__ (self ):
233+ """String representation of a component."""
234+ return self .name
235+
236+ def __repr__ (self ):
237+ """String representation of a component."""
238+ parameters_repr = ", " .join (
239+ [f"{ key } { safe_repr (value )} for key , value in self .parameters .items ()],
240+ )
241+ return f"{ (type (self ).__name__ )} { parameters_repr }
242+
243+ def update_parameters (self , update_dict , reset_fit = True ):
244+ """Updates the parameter dictionary of the component.
245+
246+ Args:
247+ update_dict (dict): A dict of parameters to update.
248+ reset_fit (bool, optional): If True, will set `_is_fitted` to False.
249+ """
250+ self ._parameters .update (update_dict )
251+ if reset_fit :
252+ self ._is_fitted = False
253+
254+ def _handle_nullable_types (self , X = None , y = None ):
255+ """Transforms X and y to remove any incompatible nullable types according to a component's needs.
256+
257+ Args:
258+ X (pd.DataFrame, optional): Input data to a component of shape [n_samples, n_features].
259+ May contain nullable types.
260+ y (pd.Series, optional): The target of length [n_samples]. May contain nullable types.
261+
262+ Returns:
263+ X, y with any incompatible nullable types downcasted to compatible equivalents.
264+ """
265+ X_bool_incompatible = "X" in self ._boolean_nullable_incompatibilities
266+ X_int_incompatible = "X" in self ._integer_nullable_incompatibilities
267+ if X is not None and (X_bool_incompatible or X_int_incompatible ):
268+ X = _downcast_nullable_X (
269+ X ,
270+ handle_boolean_nullable = X_bool_incompatible ,
271+ handle_integer_nullable = X_int_incompatible ,
272+ )
273+
274+ y_bool_incompatible = "y" in self ._boolean_nullable_incompatibilities
275+ y_int_incompatible = "y" in self ._integer_nullable_incompatibilities
276+ if y is not None and (y_bool_incompatible or y_int_incompatible ):
277+ y = _downcast_nullable_y (
278+ y ,
279+ handle_boolean_nullable = y_bool_incompatible ,
280+ handle_integer_nullable = y_int_incompatible ,
281+ )
282+
283+ return X , y
0 commit comments