PY3 xrange, np.divide, string.uppsercase, None comparison

Fixes scikit-learn#2108.

Author: justinvf

sklearn/decomposition/fastica_.py

@@ -14,6 +14,7 @@
 
 from ..base import BaseEstimator, TransformerMixin
 from ..externals import six
+from ..externals.six import moves
 from ..utils import array2d, as_float_array, check_random_state
 
 __all__ = ['fastica', 'FastICA']
@@ -63,7 +64,7 @@ def _ica_def(X, tol, g, fun_args, max_iter, w_init):
         w = w_init[j, :].copy()
         w /= np.sqrt((w ** 2).sum())
 
-        for _ in xrange(max_iter):
+        for _ in moves.xrange(max_iter):
             wtx = np.dot(w.T, X)
             gwtx, g_wtx = g(wtx, fun_args)
 
@@ -93,7 +94,7 @@ def _ica_par(X, tol, g, fun_args, max_iter, w_init):
 
     W = _sym_decorrelation(w_init)
 
-    for _ in xrange(max_iter):
+    for _ in moves.xrange(max_iter):
         wtx = np.dot(W, X)
         gwtx, g_wtx = g(wtx, fun_args)
 

sklearn/decomposition/tests/test_fastica.py

@@ -15,6 +15,7 @@
 
 from sklearn.decomposition import FastICA, fastica, PCA
 from sklearn.decomposition.fastica_ import _gs_decorrelation
+from sklearn.externals.six import moves
 
 
 def center_and_norm(x, axis=-1):
@@ -124,7 +125,7 @@ def g_test(x):
         ica = FastICA(fun=fn, algorithm=algo, random_state=0)
         assert_raises(ValueError, ica.fit, m.T)
 
-    assert_raises(TypeError, FastICA(fun=xrange(10)).fit, m.T)
+    assert_raises(TypeError, FastICA(fun=moves.xrange(10)).fit, m.T)
 
 
 def test_fastica_nowhiten():

sklearn/feature_extraction/text.py

@@ -594,12 +594,12 @@ def __init__(self, input='content', charset='utf-8',
         if max_df < 0 or min_df < 0:
             raise ValueError("negative value for max_df of min_df")
         self.max_features = max_features
-        if not any((
-                isinstance(max_features, numbers.Integral),
-                max_features is None,
-                max_features > 0)):
-            raise ValueError(
-                "max_features is neither a positive integer nor None")
+        if max_features is not None:
+            if (not isinstance(max_features, numbers.Integral) or
+                max_features <= 0):
+                raise ValueError(
+                    "max_features=%r, neither a positive integer nor None"
+                    % max_features)
         self.ngram_range = ngram_range
         if vocabulary is not None:
             if not isinstance(vocabulary, Mapping):

sklearn/feature_selection/tests/test_base.py

@@ -1,5 +1,3 @@
-import string
-
 import numpy as np
 from scipy import sparse as sp
 
@@ -34,7 +32,7 @@ def _get_support_mask(self):
 Xinv = X.copy()
 Xinv[:, 1::2] = 0
 y = [0, 1]
-feature_names = list(string.uppercase[:10])
+feature_names = list('ABCDEFGHIJ')
 feature_names_t = feature_names[::2]
 feature_names_inv = np.array(feature_names)
 feature_names_inv[1::2] = ''

sklearn/metrics/tests/test_metrics.py

@@ -519,7 +519,8 @@ def test(y_true, y_pred):
         assert_array_almost_equal(mcc, 0.57, decimal=2)
 
     test(y_true, y_pred)
-    test(map(str, y_true), map(str, y_pred))
+    test([str(y) for y in y_true],
+         [str(y) for y in y_pred])
 
 
 def test_matthews_corrcoef_nan():
@@ -639,7 +640,9 @@ def test(y_true, y_pred, string_type=False):
                                 [4, 24, 3]])
 
     test(y_true, y_pred)
-    test(map(str, y_true), map(str, y_pred), string_type=True)
+    test(list(str(y) for y in y_true),
+         list(str(y) for y in y_pred),
+         string_type=True)
 
 
 def test_confusion_matrix_multiclass_subset_labels():
@@ -820,7 +823,7 @@ def test_losses():
     assert_equal(accuracy_score(y_true, y_pred),
                  1 - zero_one_loss(y_true, y_pred))
 
-    with warnings.catch_warnings(True):
+    with warnings.catch_warnings(record=True):
     # Throw deprecated warning
         assert_equal(zero_one_score(y_true, y_pred),
                      1 - zero_one_loss(y_true, y_pred))
@@ -1001,7 +1004,7 @@ def test_hinge_loss_binary():
     pred_decision = np.array([-8.5, 0.5, 1.5, -0.3])
     assert_equal(hinge_loss(y_true, pred_decision), 1.2 / 4)
 
-    with warnings.catch_warnings(True):
+    with warnings.catch_warnings(record=True):
         # Test deprecated pos_label
         assert_equal(
             hinge_loss(-y_true, pred_decision),
@@ -1011,7 +1014,7 @@ def test_hinge_loss_binary():
     pred_decision = np.array([-8.5, 0.5, 1.5, -0.3])
 
     assert_equal(hinge_loss(y_true, pred_decision), 1.2 / 4)
-    with warnings.catch_warnings(True):
+    with warnings.catch_warnings(record=True):
         # Test deprecated pos_label
         assert_equal(hinge_loss(y_true, pred_decision, pos_label=2,
                                 neg_label=0), 1.2 / 4)

sklearn/utils/fixes.py

@@ -213,8 +213,13 @@ def safe_copy(X):
     safe_copy = np.copy
 
 try:
-    np.divide(1, 1, dtype=np.float)
+    if not np.allclose(np.divide(.4, 1),
+                       np.divide(.4, 1, dtype=np.float),
+                       .4):
+        raise TypeError('Divide not working with dtype kwarg: '
+                        'https://github.com/numpy/numpy/issues/3484')
     divide = np.divide
+
 except TypeError:
     # Compat for old versions of np.divide that do not provide support for
     # the dtype args

sklearn/utils/tests/test_fixes.py

@@ -7,7 +7,7 @@
 from nose.tools import assert_equal
 from numpy.testing import assert_array_equal
 
-from ..fixes import _in1d, _copysign
+from ..fixes import _in1d, _copysign, divide
 
 
 def test_in1d():
@@ -16,6 +16,10 @@ def test_in1d():
     assert_equal(_in1d(a, b).sum(), 5)
 
 
+def test_divide():
+    assert_equal(divide(.6, 1), .600000000000)
+
+
 def test_copysign():
     a = np.array([-1, 1, -1])
     b = np.array([1, -1, 1])