11import re
2- import six
32from typing import (
4- Any , Iterable , Iterator , Tuple , Sized , List , Optional , Dict ,
5- Union , Callable , Pattern
6- )
3+ Any , Iterable , Iterator , Sized , Optional , Union , Callable , Pattern )
74
85import numpy as np
96import scipy .sparse as sp
@@ -14,15 +11,14 @@ class FeatureNames(Sized, Iterable):
1411 A list-like object with feature names. It allows
1512 feature names for unknown features to be generated using
1613 a provided template, and to avoid making copies of large objects
17- in get_feature_names .
14+ in get_feature_names_out .
1815 """
1916 def __init__ (self ,
2017 feature_names = None ,
21- bias_name = None , # type: str
22- unkn_template = None , # type: str
23- n_features = None , # type: int
18+ bias_name : Optional [ str ] = None ,
19+ unkn_template : Optional [ str ] = None ,
20+ n_features : Optional [ int ] = None ,
2421 ):
25- # type: (...) -> None
2622 if not (feature_names is not None or
2723 (unkn_template is not None and n_features )):
2824 raise ValueError (
@@ -39,20 +35,17 @@ def __init__(self,
3935 'unkn_template should be set for sparse features' )
4036 self .feature_names = feature_names
4137 self .unkn_template = unkn_template
42- self .n_features = n_features or len (feature_names ) # type: int
38+ self .n_features : int = n_features or len (feature_names )
4339 self .bias_name = bias_name
4440
45- def __repr__ (self ):
46- # type: () -> str
41+ def __repr__ (self ) -> str :
4742 return '<FeatureNames: {} features {} bias>' .format (
4843 self .n_features , 'with' if self .has_bias else 'without' )
4944
50- def __len__ (self ):
51- # type: () -> int
45+ def __len__ (self ) -> int :
5246 return self .n_features + int (self .has_bias )
5347
54- def __iter__ (self ):
55- # type: () -> Iterator[str]
48+ def __iter__ (self ) -> Iterator [str ]:
5649 return (self [i ] for i in range (len (self )))
5750
5851 def __getitem__ (self , idx ):
@@ -69,10 +62,10 @@ def __getitem__(self, idx):
6962 return self .unkn_template % idx
7063 raise IndexError ('Feature index out of range' )
7164
72- def _slice (self , aslice ):
73- # type: (slice) -> Any
65+ def _slice (self , aslice : slice ):
7466 if isinstance (self .feature_names , (list , np .ndarray )):
7567 # Fast path without going through __getitem__
68+ lst : Union [list , np .ndarray ]
7669 if self .has_bias :
7770 lst = list (self .feature_names )
7871 lst .append (self .bias_name )
@@ -84,29 +77,26 @@ def _slice(self, aslice):
8477 return [self [idx ] for idx in indices ]
8578
8679 @property
87- def has_bias (self ):
88- # type: () -> bool
80+ def has_bias (self ) -> bool :
8981 return self .bias_name is not None
9082
9183 @property
92- def bias_idx (self ):
93- # type: () -> Optional[int]
84+ def bias_idx (self ) -> Optional [int ]:
9485 if self .has_bias :
9586 return self .n_features
9687 return None
9788
98- def filtered (self , feature_filter , x = None ):
99- # type: (Callable, Any) -> Tuple[FeatureNames, List[int]]
89+ def filtered (self , feature_filter : Callable , x = None ) -> tuple ['FeatureNames' , list [int ]]:
10090 """ Return feature names filtered by a regular expression
10191 ``feature_re``, and indices of filtered elements.
10292 """
10393 indices = []
10494 filtered_feature_names = []
105- indexed_names = None # type : Optional[Iterable[Tuple [int, Any]]]
95+ indexed_names : Optional [Iterable [tuple [int , Any ]]] = None
10696 if isinstance (self .feature_names , (np .ndarray , list )):
10797 indexed_names = enumerate (self .feature_names )
10898 elif isinstance (self .feature_names , dict ):
109- indexed_names = six . iteritems ( self .feature_names )
99+ indexed_names = self .feature_names . items ( )
110100 elif self .feature_names is None :
111101 indexed_names = []
112102 assert indexed_names is not None
@@ -116,8 +106,7 @@ def filtered(self, feature_filter, x=None):
116106 assert x .shape [0 ] == 1
117107 flt = lambda nm , i : feature_filter (nm , x [0 , i ])
118108 else :
119- # FIXME: mypy warns about x[i] because it thinks x can be None
120- flt = lambda nm , i : feature_filter (nm , x [i ]) # type: ignore
109+ flt = lambda nm , i : feature_filter (nm , x [i ])
121110 else :
122111 flt = lambda nm , i : feature_filter (nm )
123112
@@ -141,10 +130,9 @@ def filtered(self, feature_filter, x=None):
141130
142131 def handle_filter (self ,
143132 feature_filter ,
144- feature_re , # type: Pattern[str]
145- x = None , # type: Any
146- ):
147- # type: (...) -> Tuple[FeatureNames, Union[List[int], None]]
133+ feature_re : Pattern [str ],
134+ x = None ,
135+ ) -> tuple ['FeatureNames' , Union [list [int ], None ]]:
148136 if feature_re is not None and feature_filter :
149137 raise ValueError ('pass either feature_filter or feature_re' )
150138 if feature_re is not None :
@@ -158,8 +146,7 @@ def handle_filter(self,
158146 else :
159147 return self , None
160148
161- def add_feature (self , feature ):
162- # type: (Any) -> int
149+ def add_feature (self , feature ) -> int :
163150 """ Add a new feature name, return it's index.
164151 """
165152 # A copy of self.feature_names is always made, because it might be
@@ -179,8 +166,7 @@ def add_feature(self, feature):
179166 return idx
180167
181168
182- def _all_feature_names (name ):
183- # type: (Union[str, bytes, List[Dict]]) -> List[str]
169+ def _all_feature_names (name : Union [str , bytes , list [dict ]]) -> list [str ]:
184170 """ All feature names for a feature: usually just the feature itself,
185171 but can be several features for unhashed features with collisions.
186172 """
0 commit comments