MAINT Add parameter validation to HashingVectorizer. (scikit-learn#24181)

Co-authored-by: jeremie du boisberranger <[email protected]>

Author: stefmolin

sklearn/feature_extraction/text.py

@@ -15,7 +15,6 @@
 from collections import defaultdict
 from collections.abc import Mapping
 from functools import partial
-import numbers
 from numbers import Integral, Real
 from operator import itemgetter
 import re
@@ -631,7 +630,7 @@ class HashingVectorizer(TransformerMixin, _VectorizerMixin, BaseEstimator):
         'strict', meaning that a UnicodeDecodeError will be raised. Other
         values are 'ignore' and 'replace'.
 
-    strip_accents : {'ascii', 'unicode'}, default=None
+    strip_accents : {'ascii', 'unicode'} or callable, default=None
         Remove accents and perform other character normalization
         during the preprocessing step.
         'ascii' is a fast method that only works on characters that have
@@ -664,7 +663,7 @@ class HashingVectorizer(TransformerMixin, _VectorizerMixin, BaseEstimator):
         will be removed from the resulting tokens.
         Only applies if ``analyzer == 'word'``.
 
-    token_pattern : str, default=r"(?u)\\b\\w\\w+\\b"
+    token_pattern : str or None, default=r"(?u)\\b\\w\\w+\\b"
         Regular expression denoting what constitutes a "token", only used
         if ``analyzer == 'word'``. The default regexp selects tokens of 2
         or more alphanumeric characters (punctuation is completely ignored
@@ -740,6 +739,25 @@ class HashingVectorizer(TransformerMixin, _VectorizerMixin, BaseEstimator):
     (4, 16)
     """
 
+    _parameter_constraints: dict = {
+        "input": [StrOptions({"filename", "file", "content"})],
+        "encoding": [str],
+        "decode_error": [StrOptions({"strict", "ignore", "replace"})],
+        "strip_accents": [StrOptions({"ascii", "unicode"}), None, callable],
+        "lowercase": ["boolean"],
+        "preprocessor": [callable, None],
+        "tokenizer": [callable, None],
+        "stop_words": [StrOptions({"english"}), list, None],
+        "token_pattern": [str, None],
+        "ngram_range": [tuple],
+        "analyzer": [StrOptions({"word", "char", "char_wb"}), callable],
+        "n_features": [Interval(Integral, 1, np.iinfo(np.int32).max, closed="left")],
+        "binary": ["boolean"],
+        "norm": [StrOptions({"l1", "l2"}), None],
+        "alternate_sign": ["boolean"],
+        "dtype": "no_validation",  # delegate to numpy
+    }
+
     def __init__(
         self,
         *,
@@ -796,6 +814,8 @@ def partial_fit(self, X, y=None):
         self : object
             HashingVectorizer instance.
         """
+        # TODO: only validate during the first call
+        self._validate_params()
         return self
 
     def fit(self, X, y=None):
@@ -814,6 +834,8 @@ def fit(self, X, y=None):
         self : object
             HashingVectorizer instance.
         """
+        self._validate_params()
+
         # triggers a parameter validation
         if isinstance(X, str):
             raise ValueError(
@@ -935,7 +957,7 @@ class CountVectorizer(_VectorizerMixin, BaseEstimator):
         Remove accents and perform other character normalization
         during the preprocessing step.
         'ascii' is a fast method that only works on characters that have
-        an direct ASCII mapping.
+        a direct ASCII mapping.
         'unicode' is a slightly slower method that works on any characters.
         None (default) does nothing.
 
@@ -1359,12 +1381,8 @@ def fit_transform(self, raw_documents, y=None):
 
         if not self.fixed_vocabulary_:
             n_doc = X.shape[0]
-            max_doc_count = (
-                max_df if isinstance(max_df, numbers.Integral) else max_df * n_doc
-            )
-            min_doc_count = (
-                min_df if isinstance(min_df, numbers.Integral) else min_df * n_doc
-            )
+            max_doc_count = max_df if isinstance(max_df, Integral) else max_df * n_doc
+            min_doc_count = min_df if isinstance(min_df, Integral) else min_df * n_doc
             if max_doc_count < min_doc_count:
                 raise ValueError("max_df corresponds to < documents than min_df")
             if max_features is not None:
@@ -1771,11 +1789,11 @@ class TfidfVectorizer(CountVectorizer):
         'strict', meaning that a UnicodeDecodeError will be raised. Other
         values are 'ignore' and 'replace'.
 
-    strip_accents : {'ascii', 'unicode'}, default=None
+    strip_accents : {'ascii', 'unicode'} or callable, default=None
         Remove accents and perform other character normalization
         during the preprocessing step.
         'ascii' is a fast method that only works on characters that have
-        an direct ASCII mapping.
+        a direct ASCII mapping.
         'unicode' is a slightly slower method that works on any characters.
         None (default) does nothing.
 

sklearn/tests/test_common.py

@@ -465,7 +465,6 @@ def test_estimators_do_not_raise_errors_in_init_or_set_params(Estimator):
 
 PARAM_VALIDATION_ESTIMATORS_TO_IGNORE = [
     "DictionaryLearning",
-    "HashingVectorizer",
     "MiniBatchDictionaryLearning",
     "MultiTaskElasticNet",
     "MultiTaskLasso",