scikit-learn · dalmia · Feb 6, 2017 · Feb 6, 2017 · Feb 16, 2017 · Feb 17, 2017
diff --git a/doc/tutorial/statistical_inference/unsupervised_learning_fixture.py b/doc/tutorial/statistical_inference/unsupervised_learning_fixture.py
@@ -1,10 +1,11 @@
-"""Fixture module to skip the unsupervised_learning.rst doctest for 
-versions of SciPy earlier than 0.12.0. 
+"""Fixture module to skip the unsupervised_learning.rst doctest for
+versions of SciPy earlier than 0.12.0.
 """
 from sklearn.utils.testing import SkipTest
 from sklearn.utils.fixes import sp_version
+from sklearn.utils import parse_version
 
 def setup_module(module):
-    if sp_version < (0, 12):
+    if sp_version < parse_version('0.12'):
         raise SkipTest("Skipping because SciPy version earlier than 0.12.0 and "
                        "thus does not include the scipy.misc.face() image.")
diff --git a/sklearn/cross_decomposition/pls_.py b/sklearn/cross_decomposition/pls_.py
@@ -4,11 +4,10 @@
 
 # Author: Edouard Duchesnay <edouard.duchesnay@cea.fr>
 # License: BSD 3 clause
-from distutils.version import LooseVersion
 from sklearn.utils.extmath import svd_flip
 
 from ..base import BaseEstimator, RegressorMixin, TransformerMixin
-from ..utils import check_array, check_consistent_length
+from ..utils import check_array, check_consistent_length, parse_version
 from ..externals import six
 
 import warnings
@@ -22,7 +21,7 @@
 
 import scipy
 pinv2_args = {}
-if LooseVersion(scipy.__version__) >= LooseVersion('0.12'):
+if parse_version(scipy.__version__) >= parse_version('0.12'):
     # check_finite=False is an optimization available only in scipy >=0.12
     pinv2_args = {'check_finite': False}
 

diff --git a/sklearn/feature_extraction/tests/test_image.py b/sklearn/feature_extraction/tests/test_image.py
@@ -14,8 +14,9 @@
 from sklearn.utils.graph import connected_components
 from sklearn.utils.testing import SkipTest, assert_equal, assert_true
 from sklearn.utils.fixes import sp_version
+from sklearn.utils.version import parse_version
 
-if sp_version < (0, 12):
+if sp_version < parse_version('0.12'):
     raise SkipTest("Skipping because SciPy version earlier than 0.12.0 and "
                    "thus does not include the scipy.misc.face() image.")
 

diff --git a/sklearn/linear_model/least_angle.py b/sklearn/linear_model/least_angle.py
@@ -13,7 +13,6 @@
 from math import log
 import sys
 import warnings
-from distutils.version import LooseVersion
 
 import numpy as np
 from scipy import linalg, interpolate
@@ -28,9 +27,11 @@
 from ..externals.six.moves import xrange
 from ..externals.six import string_types
 
+from ..utils import parse_version
+
 import scipy
 solve_triangular_args = {}
-if LooseVersion(scipy.__version__) >= LooseVersion('0.12'):
+if parse_version(scipy.__version__) >= parse_version('0.12'):
     solve_triangular_args = {'check_finite': False}
 
 

diff --git a/sklearn/linear_model/omp.py b/sklearn/linear_model/omp.py
@@ -6,7 +6,6 @@
 # License: BSD 3 clause
 
 import warnings
-from distutils.version import LooseVersion
 
 import numpy as np
 from scipy import linalg
@@ -18,9 +17,11 @@
 from ..model_selection import check_cv
 from ..externals.joblib import Parallel, delayed
 
+from ..utils import parse_version
+
 import scipy
 solve_triangular_args = {}
-if LooseVersion(scipy.__version__) >= LooseVersion('0.12'):
+if parse_version(scipy.__version__) >= parse_version('0.12'):
     # check_finite=False is an optimization available only in scipy >=0.12
     solve_triangular_args = {'check_finite': False}
 

diff --git a/sklearn/linear_model/tests/test_logistic.py b/sklearn/linear_model/tests/test_logistic.py
@@ -16,6 +16,7 @@
 from sklearn.exceptions import ConvergenceWarning
 from sklearn.utils import compute_class_weight
 from sklearn.utils.fixes import sp_version
+from sklearn.utils.version import parse_version
 
 from sklearn.linear_model.logistic import (
     LogisticRegression,
@@ -899,7 +900,7 @@ def test_max_iter():
 
     solvers = ['newton-cg', 'liblinear', 'sag']
     # old scipy doesn't have maxiter
-    if sp_version >= (0, 12):
+    if sp_version >= parse_version('0.12'):
         solvers.append('lbfgs')
 
     for max_iter in range(1, 5):
@@ -969,7 +970,7 @@ def test_warm_start():
 
     solvers = ['newton-cg', 'sag']
     # old scipy doesn't have maxiter
-    if sp_version >= (0, 12):
+    if sp_version >= parse_version('0.12'):
         solvers.append('lbfgs')
 
     for warm_start in [True, False]:

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
@@ -26,6 +26,7 @@
 from sklearn.utils.testing import assert_not_equal
 from sklearn.utils.testing import ignore_warnings
 from sklearn.utils.mocking import MockDataFrame
+from sklearn.utils.version import parse_version
 
 from sklearn.metrics import accuracy_score
 from sklearn.metrics import average_precision_score
@@ -691,7 +692,7 @@ def test_classification_report_multiclass_with_unicode_label():
 
 avg / total       0.51      0.53      0.47        75
 """
-    if np_version[:3] < (1, 7, 0):
+    if np_version < parse_version('1.7.0'):
         expected_message = ("NumPy < 1.7.0 does not implement"
                             " searchsorted on unicode data correctly.")
         assert_raise_message(RuntimeError, expected_message,

diff --git a/sklearn/model_selection/_search.py b/sklearn/model_selection/_search.py
@@ -27,7 +27,7 @@
 from ..exceptions import NotFittedError
 from ..externals.joblib import Parallel, delayed
 from ..externals import six
-from ..utils import check_random_state
+from ..utils import check_random_state, parse_version
 from ..utils.fixes import sp_version
 from ..utils.fixes import rankdata
 from ..utils.fixes import MaskedArray
@@ -251,7 +251,7 @@ def __iter__(self):
                 params = dict()
                 for k, v in items:
                     if hasattr(v, "rvs"):
-                        if sp_version < (0, 16):
+                        if sp_version < parse_version('0.16'):
                             params[k] = v.rvs()
                         else:
                             params[k] = v.rvs(random_state=rnd)

diff --git a/sklearn/model_selection/tests/test_search.py b/sklearn/model_selection/tests/test_search.py
@@ -25,6 +25,7 @@
 from sklearn.utils.testing import assert_almost_equal
 from sklearn.utils.testing import ignore_warnings
 from sklearn.utils.mocking import CheckingClassifier, MockDataFrame
+from sklearn.utils.version import parse_version
 
 from scipy.stats import bernoulli, expon, uniform
 
@@ -688,7 +689,7 @@ def test_param_sampler():
                                n_iter=3, random_state=0)
     assert_equal([x for x in sampler], [x for x in sampler])
 
-    if sp_version >= (0, 16):
+    if sp_version >= parse_version('0.16'):
         param_distributions = {"C": uniform(0, 1)}
         sampler = ParameterSampler(param_distributions=param_distributions,
                                    n_iter=10, random_state=0)

diff --git a/sklearn/neighbors/tests/test_dist_metrics.py b/sklearn/neighbors/tests/test_dist_metrics.py
@@ -9,24 +9,13 @@
 from sklearn.neighbors.dist_metrics import DistanceMetric
 from sklearn.neighbors import BallTree
 from sklearn.utils.testing import SkipTest, assert_raises_regex
+from sklearn.utils.version import parse_version
 
 
 def dist_func(x1, x2, p):
     return np.sum((x1 - x2) ** p) ** (1. / p)
 
 
-def cmp_version(version1, version2):
-    version1 = tuple(map(int, version1.split('.')[:2]))
-    version2 = tuple(map(int, version2.split('.')[:2]))
-
-    if version1 < version2:
-        return -1
-    elif version1 > version2:
-        return 1
-    else:
-        return 0
-
-
 class TestMetrics:
     def __init__(self, n1=20, n2=25, d=4, zero_frac=0.5,
                  rseed=0, dtype=np.float64):
@@ -70,7 +59,8 @@ def test_cdist(self):
             yield self.check_cdist_bool, metric, D_true
 
     def check_cdist(self, metric, kwargs, D_true):
-        if metric == 'canberra' and cmp_version(scipy.__version__, '0.9') <= 0:
+        if metric == 'canberra' and \
+                parse_version(scipy.__version__) <= parse_version('0.9'):
             raise SkipTest("Canberra distance incorrect in scipy < 0.9")
         dm = DistanceMetric.get_metric(metric, **kwargs)
         D12 = dm.pairwise(self.X1, self.X2)
@@ -94,7 +84,8 @@ def test_pdist(self):
             yield self.check_pdist_bool, metric, D_true
 
     def check_pdist(self, metric, kwargs, D_true):
-        if metric == 'canberra' and cmp_version(scipy.__version__, '0.9') <= 0:
+        if metric == 'canberra' and \
+                parse_version(scipy.__version__) <= parse_version('0.9'):
             raise SkipTest("Canberra distance incorrect in scipy < 0.9")
         dm = DistanceMetric.get_metric(metric, **kwargs)
         D12 = dm.pairwise(self.X1)

diff --git a/sklearn/preprocessing/label.py b/sklearn/preprocessing/label.py
@@ -19,7 +19,7 @@
 from ..utils.fixes import sparse_min_max
 from ..utils.fixes import astype
 from ..utils.fixes import in1d
-from ..utils import column_or_1d
+from ..utils import column_or_1d, parse_version
 from ..utils.validation import check_array
 from ..utils.validation import check_is_fitted
 from ..utils.validation import _num_samples
@@ -47,7 +47,7 @@ def _check_numpy_unicode_bug(labels):
       https://github.com/numpy/numpy/pull/243
 
     """
-    if np_version[:3] < (1, 7, 0) and labels.dtype.kind == 'U':
+    if np_version < parse_version('1.7.0') and labels.dtype.kind == 'U':
         raise RuntimeError("NumPy < 1.7.0 does not implement searchsorted"
                            " on unicode data correctly. Please upgrade"
                            " NumPy to use LabelEncoder with unicode inputs.")

diff --git a/sklearn/preprocessing/tests/test_data.py b/sklearn/preprocessing/tests/test_data.py
@@ -9,7 +9,6 @@
 import numpy as np
 import numpy.linalg as la
 from scipy import sparse
-from distutils.version import LooseVersion
 
 from sklearn.utils import gen_batches
 
@@ -29,6 +28,7 @@
 from sklearn.utils.testing import assert_no_warnings
 from sklearn.utils.testing import assert_allclose
 from sklearn.utils.testing import skip_if_32bit
+from sklearn.utils.version import parse_version
 
 from sklearn.utils.sparsefuncs import mean_variance_axis
 from sklearn.preprocessing.data import _transform_selected
@@ -206,7 +206,7 @@ def test_standard_scaler_numerical_stability():
     # was empirically found to cause numerical problems with np.mean & np.std.
 
     x = np.zeros(8, dtype=np.float64) + np.log(1e-5, dtype=np.float64)
-    if LooseVersion(np.__version__) >= LooseVersion('1.9'):
+    if parse_version(np.__version__) >= parse_version('1.9'):
         # This does not raise a warning as the number of samples is too low
         # to trigger the problem in recent numpy
         x_scaled = assert_no_warnings(scale, x)

diff --git a/sklearn/utils/__init__.py b/sklearn/utils/__init__.py
@@ -13,6 +13,7 @@
                          check_random_state, column_or_1d, check_array,
                          check_consistent_length, check_X_y, indexable,
                          check_symmetric)
+from .version import parse_version
 from .class_weight import compute_class_weight, compute_sample_weight
 from ..externals.joblib import cpu_count
 from ..exceptions import DataConversionWarning
@@ -25,7 +26,7 @@
            "compute_class_weight", "compute_sample_weight",
            "column_or_1d", "safe_indexing",
            "check_consistent_length", "check_X_y", 'indexable',
-           "check_symmetric", "indices_to_mask", "deprecated"]
+           "check_symmetric", "indices_to_mask", "deprecated", "parse_version"]
 
 
 def safe_mask(X, mask):

diff --git a/sklearn/utils/arpack.py b/sklearn/utils/arpack.py
@@ -52,7 +52,7 @@
 import scipy
 import functools
 import operator
-from distutils.version import LooseVersion
+from .version import parse_version
 
 __docformat__ = "restructuredtext en"
 
@@ -276,9 +276,9 @@
 
 
 # CHECK IF BACKPORT IS ACTUALLY NEEDED
-if scipy.version.version >= LooseVersion('0.12'):
+if parse_version(scipy.version.version) >= parse_version('0.12'):
     BACKPORT_TO = None
-elif scipy.version.version >= LooseVersion('0.11'):
+elif parse_version(scipy.version.version) >= parse_version('0.11'):
     BACKPORT_TO = '0.10'
 else:
     BACKPORT_TO = '0.09'
@@ -1853,7 +1853,7 @@ def matvec_XH_X(x):
 
 
 # Redefine the backported function
-if scipy.version.version >= LooseVersion('0.12'):
+if parse_version(scipy.version.version) >= parse_version('0.12'):
     from scipy.sparse.linalg import eigs, eigsh, svds
 else:
     eigs, eigsh, svds = _eigs, _eigsh, _svds
diff --git a/sklearn/utils/extmath.py b/sklearn/utils/extmath.py
@@ -26,6 +26,7 @@
 from .sparsefuncs_fast import csr_row_norms
 from .validation import check_array
 from ..exceptions import NonBLASDotWarning
+from .version import parse_version
 
 
 def norm(x):
@@ -40,7 +41,7 @@ def norm(x):
 
 
 # Newer NumPy has a ravel that needs less copying.
-if np_version < (1, 7, 1):
+if np_version < parse_version('1.7.1'):
     _ravel = np.ravel
 else:
     _ravel = partial(np.ravel, order='K')
@@ -131,7 +132,7 @@ def _have_blas_gemm():
 
 
 # Only use fast_dot for older NumPy; newer ones have tackled the speed issue.
-if np_version < (1, 7, 2) and _have_blas_gemm():
+if np_version < parse_version('1.7.2') and _have_blas_gemm():
     def fast_dot(A, B):
         """Compute fast dot products directly calling BLAS.
 
@@ -860,7 +861,7 @@ def stable_cumsum(arr, axis=None, rtol=1e-05, atol=1e-08):
         Absolute tolerance, see ``np.allclose``
     """
     # sum is as unstable as cumsum for numpy < 1.9
-    if np_version < (1, 9):
+    if np_version < parse_version('1.9'):
         return np.cumsum(arr, axis=axis, dtype=np.float64)
 
     out = np.cumsum(arr, axis=axis, dtype=np.float64)