Drop support for python 2 and python 3.5 (#291)

* Drop Python2 support

* Fix test failures and pyflakes error

* fix pep8 issue

* Stop testing py3.5

* Update README

Author: perimosocordiae

.travis.yml

@@ -7,23 +7,12 @@ env:
     - SKGGM_VERSION=a0ed406586c4364ea3297a658f415e13b5cbdaf8
 matrix:
   include:
-  - name: "Pytest python 2.7 with skggm"
-    python: "2.7"
+  - name: "Pytest python 3.6 without skggm"
+    python: "3.6"
     before_install:
       - sudo apt-get install liblapack-dev
       - pip install --upgrade pip pytest
       - pip install wheel cython numpy scipy codecov pytest-cov scikit-learn
-      - pip install git+https://github.com/skggm/skggm.git@${SKGGM_VERSION};
-    script:
-      - pytest test --cov;
-    after_success:
-      - bash <(curl -s https://codecov.io/bash)
-  - name: "Pytest python 3.4 without skggm"
-    python: "3.4"
-    before_install:
-      - sudo apt-get install liblapack-dev
-      - pip install --upgrade pip "pytest<5"
-      - pip install wheel cython numpy scipy codecov pytest-cov scikit-learn
     script:
       - pytest test --cov;
     after_success:

README.rst

@@ -19,7 +19,7 @@ metric-learn contains efficient Python implementations of several popular superv
 
 **Dependencies**
 
--  Python 2.7+, 3.4+
+-  Python 3.6+
 -  numpy, scipy, scikit-learn>=0.20.3
 
 **Optional dependencies**

metric_learn/__init__.py

@@ -1,5 +1,3 @@
-from __future__ import absolute_import
-
 from .constraints import Constraints
 from .covariance import Covariance
 from .itml import ITML, ITML_Supervised

metric_learn/_util.py

@@ -1,5 +1,4 @@
 import numpy as np
-import six
 from numpy.linalg import LinAlgError
 from sklearn.datasets import make_spd_matrix
 from sklearn.decomposition import PCA
@@ -283,7 +282,7 @@ def make_name(estimator):
   if a string is given
   """
   if estimator is not None:
-    if isinstance(estimator, six.string_types):
+    if isinstance(estimator, str):
       estimator_name = estimator
     else:
       estimator_name = estimator.__class__.__name__

metric_learn/base_metric.py

@@ -8,12 +8,11 @@
 from sklearn.metrics import roc_auc_score, roc_curve, precision_recall_curve
 import numpy as np
 from abc import ABCMeta, abstractmethod
-import six
 from ._util import ArrayIndexer, check_input, validate_vector
 import warnings
 
 
-class BaseMetricLearner(six.with_metaclass(ABCMeta, BaseEstimator)):
+class BaseMetricLearner(BaseEstimator, metaclass=ABCMeta):
   """
   Base class for all metric-learners.
 
@@ -145,7 +144,7 @@ def get_metric(self):
     """
 
 
-class MetricTransformer(six.with_metaclass(ABCMeta)):
+class MetricTransformer(metaclass=ABCMeta):
 
   @abstractmethod
   def transform(self, X):
@@ -163,8 +162,8 @@ def transform(self, X):
     """
 
 
-class MahalanobisMixin(six.with_metaclass(ABCMeta, BaseMetricLearner,
-                                          MetricTransformer)):
+class MahalanobisMixin(BaseMetricLearner, MetricTransformer,
+                       metaclass=ABCMeta):
   r"""Mahalanobis metric learning algorithms.
 
   Algorithm that learns a Mahalanobis (pseudo) distance :math:`d_M(x, x')`,

metric_learn/constraints.py

@@ -4,7 +4,6 @@
 """
 import numpy as np
 import warnings
-from six.moves import xrange
 from sklearn.utils import check_random_state
 from sklearn.neighbors import NearestNeighbors
 
@@ -245,7 +244,7 @@ def chunks(self, num_chunks=100, chunk_size=2, random_state=None):
     chunks = -np.ones_like(self.partial_labels, dtype=int)
     uniq, lookup = np.unique(self.partial_labels, return_inverse=True)
     unknown_uniq = np.where(uniq < 0)[0]
-    all_inds = [set(np.where(lookup == c)[0]) for c in xrange(len(uniq))
+    all_inds = [set(np.where(lookup == c)[0]) for c in range(len(uniq))
                 if c not in unknown_uniq]
     max_chunks = int(np.sum([len(s) // chunk_size for s in all_inds]))
     if max_chunks < num_chunks:

metric_learn/covariance.py

@@ -2,7 +2,6 @@
 Covariance metric (baseline method)
 """
 
-from __future__ import absolute_import
 import numpy as np
 import scipy
 from sklearn.base import TransformerMixin

metric_learn/itml.py

@@ -2,10 +2,8 @@
 Information Theoretic Metric Learning (ITML)
 """
 
-from __future__ import print_function, absolute_import
 import warnings
 import numpy as np
-from six.moves import xrange
 from sklearn.exceptions import ChangedBehaviorWarning
 from sklearn.metrics import pairwise_distances
 from sklearn.utils.validation import check_array
@@ -69,7 +67,7 @@ def _fit(self, pairs, y, bounds=None):
     pos_vv = pos_pairs[:, 0, :] - pos_pairs[:, 1, :]
     neg_vv = neg_pairs[:, 0, :] - neg_pairs[:, 1, :]
 
-    for it in xrange(self.max_iter):
+    for it in range(self.max_iter):
       # update positives
       for i, v in enumerate(pos_vv):
         wtw = v.dot(A).dot(v)  # scalar

metric_learn/lfda.py

@@ -1,11 +1,9 @@
 """
 Local Fisher Discriminant Analysis (LFDA)
 """
-from __future__ import division, absolute_import
 import numpy as np
 import scipy
 import warnings
-from six.moves import xrange
 from sklearn.metrics import pairwise_distances
 from sklearn.base import TransformerMixin
 
@@ -127,7 +125,7 @@ def fit(self, X, y):
     tSb = np.zeros((d, d))
     tSw = np.zeros((d, d))
 
-    for c in xrange(num_classes):
+    for c in range(num_classes):
       Xc = X[y == c]
       nc = Xc.shape[0]
 

metric_learn/lmnn.py

@@ -1,11 +1,9 @@
 """
 Large Margin Nearest Neighbor Metric learning (LMNN)
 """
-from __future__ import print_function, absolute_import
 import numpy as np
 import warnings
 from collections import Counter
-from six.moves import xrange
 from sklearn.exceptions import ChangedBehaviorWarning
 from sklearn.metrics import euclidean_distances
 from sklearn.base import TransformerMixin
@@ -229,7 +227,7 @@ def fit(self, X, y):
             "| learning rate")
 
     # main loop
-    for it in xrange(2, self.max_iter):
+    for it in range(2, self.max_iter):
       # then at each iteration, we try to find a value of L that has better
       # objective than the previous L, following the gradient:
       while True:
@@ -293,7 +291,7 @@ def _loss_grad(self, X, L, dfG, k, reg, target_neighbors, label_inds):
     # compute the gradient
     total_active = 0
     df = np.zeros((X.shape[1], X.shape[1]))
-    for nn_idx in reversed(xrange(k)):  # note: reverse not useful here
+    for nn_idx in reversed(range(k)):  # note: reverse not useful here
       act1 = g0 < g1[:, nn_idx]
       act2 = g0 < g2[:, nn_idx]
       total_active += act1.sum() + act2.sum()

metric_learn/lsml.py

@@ -2,11 +2,9 @@
 Metric Learning from Relative Comparisons by Minimizing Squared Residual (LSML)
 """
 
-from __future__ import print_function, absolute_import, division
 import warnings
 import numpy as np
 import scipy.linalg
-from six.moves import xrange
 from sklearn.base import TransformerMixin
 from sklearn.exceptions import ChangedBehaviorWarning
 
@@ -66,7 +64,7 @@ def _fit(self, quadruplets, weights=None):
     s_best = self._total_loss(M, vab, vcd, prior_inv)
     if self.verbose:
       print('initial loss', s_best)
-    for it in xrange(1, self.max_iter + 1):
+    for it in range(1, self.max_iter + 1):
       grad = self._gradient(M, vab, vcd, prior_inv)
       grad_norm = scipy.linalg.norm(grad)
       if grad_norm < self.tol:

metric_learn/mlkr.py

@@ -1,7 +1,6 @@
 """
 Metric Learning for Kernel Regression (MLKR)
 """
-from __future__ import division, print_function
 import time
 import sys
 import warnings

metric_learn/mmc.py

@@ -1,8 +1,6 @@
 """Mahalanobis Metric for Clustering (MMC)"""
-from __future__ import print_function, absolute_import, division
 import warnings
 import numpy as np
-from six.moves import xrange
 from sklearn.base import TransformerMixin
 from sklearn.utils.validation import assert_all_finite
 from sklearn.exceptions import ChangedBehaviorWarning
@@ -110,12 +108,12 @@ def _fit_full(self, pairs, y):
 
     A_old = A.copy()
 
-    for cycle in xrange(self.max_iter):
+    for cycle in range(self.max_iter):
 
       # projection of constraints C1 and C2
       satisfy = False
 
-      for it in xrange(self.max_proj):
+      for it in range(self.max_proj):
 
         # First constraint:
         # f(A) = \sum_{i,j \in S} d_ij' A d_ij <= t              (1)

metric_learn/nca.py

@@ -2,7 +2,6 @@
 Neighborhood Components Analysis (NCA)
 """
 
-from __future__ import absolute_import
 import warnings
 import time
 import sys

metric_learn/rca.py

@@ -2,10 +2,8 @@
 Relative Components Analysis (RCA)
 """
 
-from __future__ import absolute_import
 import numpy as np
 import warnings
-from six.moves import xrange
 from sklearn.base import TransformerMixin
 from sklearn.exceptions import ChangedBehaviorWarning
 
@@ -22,7 +20,7 @@ def _chunk_mean_centering(data, chunks):
   # mean on it
   chunk_data = data[chunk_mask].astype(float, copy=False)
   chunk_labels = chunks[chunk_mask]
-  for c in xrange(num_chunks):
+  for c in range(num_chunks):
     mask = chunk_labels == c
     chunk_data[mask] -= chunk_data[mask].mean(axis=0)
 

metric_learn/sdml.py

@@ -2,7 +2,6 @@
 Sparse High-Dimensional Metric Learning (SDML)
 """
 
-from __future__ import absolute_import
 import warnings
 import numpy as np
 from sklearn.base import TransformerMixin

setup.py

@@ -29,7 +29,7 @@
       classifiers=[
           'Development Status :: 4 - Beta',
           'License :: OSI Approved :: MIT License',
-          'Programming Language :: Python',
+          'Programming Language :: Python :: 3',
           'Operating System :: OS Independent',
           'Intended Audience :: Science/Research',
           'Topic :: Scientific/Engineering'
@@ -39,7 +39,6 @@
           'numpy',
           'scipy',
           'scikit-learn',
-          'six'
       ],
       extras_require=dict(
           docs=['sphinx', 'shinx_rtd_theme', 'numpydoc'],

test/metric_learn_test.py

@@ -4,7 +4,6 @@
 import numpy as np
 import scipy
 from scipy.optimize import check_grad, approx_fprime
-from six.moves import xrange
 from sklearn.metrics import pairwise_distances, euclidean_distances
 from sklearn.datasets import (load_iris, make_classification, make_regression,
                               make_spd_matrix)
@@ -32,7 +31,7 @@
 def class_separation(X, labels):
   unique_labels, label_inds = np.unique(labels, return_inverse=True)
   ratio = 0
-  for li in xrange(len(unique_labels)):
+  for li in range(len(unique_labels)):
     Xc = X[label_inds == li]
     Xnc = X[label_inds != li]
     ratio += pairwise_distances(Xc).mean() / pairwise_distances(Xc, Xnc).mean()
@@ -385,15 +384,15 @@ def loss_fn(L, X, y, target_neighbors, reg):
        for j in target_neighbors[i]:
          loss += (1 - reg) * np.sum((Lx[i] - Lx[j]) ** 2)
          grad += (1 - reg) * np.outer(Lx[i] - Lx[j], X[i] - X[j])
-         for l in range(X.shape[0]):
-           if y[i] != y[l]:
+         for k in range(X.shape[0]):
+           if y[i] != y[k]:
              hin, active = hinge(1 + np.sum((Lx[i] - Lx[j])**2) -
-                                 np.sum((Lx[i] - Lx[l])**2))
+                                 np.sum((Lx[i] - Lx[k])**2))
              total_active += active
              if active:
                loss += reg * hin
                grad += (reg * (np.outer(Lx[i] - Lx[j], X[i] - X[j]) -
-                               np.outer(Lx[i] - Lx[l], X[i] - X[l])))
+                               np.outer(Lx[i] - Lx[k], X[i] - X[k])))
      grad = 2 * grad
      return grad, loss, total_active
 
@@ -521,7 +520,7 @@ def test_toy_ex_lmnn(X, y, loss):
   # storage
   a1 = [None] * k
   a2 = [None] * k
-  for nn_idx in xrange(k):
+  for nn_idx in range(k):
     a1[nn_idx] = np.array([])
     a2[nn_idx] = np.array([])