scikit-learn · jnothman · Feb 28, 2019 · Dec 27, 2018 · Dec 27, 2018 · Feb 27, 2019
diff --git a/sklearn/__check_build/_check_build.pyx b/sklearn/__check_build/_check_build.pyx
@@ -1,2 +1,4 @@
+# cython: language_level=3
+
 def check_build():
-    return
+    return
diff --git a/sklearn/_isotonic.pyx b/sklearn/_isotonic.pyx
@@ -3,16 +3,15 @@
 # Uses the pool adjacent violators algorithm (PAVA), with the
 # enhancement of searching for the longest decreasing subsequence to
 # pool at each step.
+#
+# cython: language_level=3, boundscheck=False, wraparound=False, cdivision=True
 
 import numpy as np
 cimport numpy as np
 cimport cython
 from cython cimport floating
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def _inplace_contiguous_isotonic_regression(floating[::1] y, floating[::1] w):
     cdef:
         Py_ssize_t n = y.shape[0], i, k
@@ -64,9 +63,6 @@ def _inplace_contiguous_isotonic_regression(floating[::1] y, floating[::1] w):
             i = k
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def _make_unique(np.ndarray[dtype=floating] X,
                  np.ndarray[dtype=floating] y,
                  np.ndarray[dtype=floating] sample_weights):

diff --git a/sklearn/cluster/_dbscan_inner.pyx b/sklearn/cluster/_dbscan_inner.pyx
@@ -1,6 +1,8 @@
 # Fast inner loop for DBSCAN.
 # Author: Lars Buitinck
 # License: 3-clause BSD
+#
+# cython: language_level=3, boundscheck=False, wraparound=False
 
 cimport cython
 from libcpp.vector cimport vector
@@ -14,8 +16,6 @@ cdef inline void push(vector[np.npy_intp] &stack, np.npy_intp i) except +:
     stack.push_back(i)
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
 def dbscan_inner(np.ndarray[np.uint8_t, ndim=1, mode='c'] is_core,
                  np.ndarray[object, ndim=1] neighborhoods,
                  np.ndarray[np.npy_intp, ndim=1, mode='c'] labels):

diff --git a/sklearn/cluster/_hierarchical.pyx b/sklearn/cluster/_hierarchical.pyx
@@ -1,4 +1,6 @@
 # Author: Gael Varoquaux <gael.varoquaux@normalesup.org>
+#
+# cython: language_level=3
 
 import numpy as np
 cimport numpy as np

diff --git a/sklearn/cluster/_k_means.pyx b/sklearn/cluster/_k_means.pyx
@@ -7,6 +7,8 @@
 #         Lars Buitinck
 #
 # License: BSD 3 clause
+#
+# cython: language_level=3, boundscheck=False, wraparound=False, cdivision=True
 
 from libc.math cimport sqrt
 import numpy as np
@@ -25,9 +27,6 @@ ctypedef np.int32_t INT
 np.import_array()
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 cpdef DOUBLE _assign_labels_array(np.ndarray[floating, ndim=2] X,
                                   np.ndarray[floating, ndim=1] sample_weight,
                                   np.ndarray[floating, ndim=1] x_squared_norms,
@@ -94,9 +93,6 @@ cpdef DOUBLE _assign_labels_array(np.ndarray[floating, ndim=2] X,
     return inertia
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 cpdef DOUBLE _assign_labels_csr(X, np.ndarray[floating, ndim=1] sample_weight,
                                 np.ndarray[DOUBLE, ndim=1] x_squared_norms,
                                 np.ndarray[floating, ndim=2] centers,
@@ -158,9 +154,6 @@ cpdef DOUBLE _assign_labels_csr(X, np.ndarray[floating, ndim=1] sample_weight,
     return inertia
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def _mini_batch_update_csr(X, np.ndarray[floating, ndim=1] sample_weight,
                            np.ndarray[DOUBLE, ndim=1] x_squared_norms,
                            np.ndarray[floating, ndim=2] centers,
@@ -262,9 +255,6 @@ def _mini_batch_update_csr(X, np.ndarray[floating, ndim=1] sample_weight,
     return squared_diff
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def _centers_dense(np.ndarray[floating, ndim=2] X,
         np.ndarray[floating, ndim=1] sample_weight,
         np.ndarray[INT, ndim=1] labels, int n_clusters,
@@ -332,9 +322,6 @@ def _centers_dense(np.ndarray[floating, ndim=2] X,
     return centers
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def _centers_sparse(X, np.ndarray[floating, ndim=1] sample_weight,
         np.ndarray[INT, ndim=1] labels, n_clusters,
         np.ndarray[floating, ndim=1] distances):

diff --git a/sklearn/cluster/_k_means_elkan.pyx b/sklearn/cluster/_k_means_elkan.pyx
@@ -1,7 +1,7 @@
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
-# cython: profile=True
+# cython: language_level=3
 #
 # Author: Andreas Mueller
 #

diff --git a/sklearn/datasets/_svmlight_format.pyx b/sklearn/datasets/_svmlight_format.pyx
@@ -4,6 +4,8 @@
 #          Lars Buitinck
 #          Olivier Grisel <olivier.grisel@ensta.org>
 # License: BSD 3 clause
+#
+# cython: language_level=2, boundscheck=False, wraparound=False
 
 import array
 from cpython cimport array
@@ -21,8 +23,6 @@ cdef bytes COMMA = u','.encode('ascii')
 cdef bytes COLON = u':'.encode('ascii')
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
 def _load_svmlight_file(f, dtype, bint multilabel, bint zero_based,
                         bint query_id, long long offset, long long length):
     cdef array.array data, indices, indptr

diff --git a/sklearn/decomposition/_online_lda.pyx b/sklearn/decomposition/_online_lda.pyx
@@ -1,3 +1,6 @@
+#
+# cython: language_level=3, boundscheck=False, wraparound=False
+
 cimport cython
 cimport numpy as np
 import numpy as np
@@ -8,8 +11,6 @@ from libc.math cimport exp, fabs, log
 from numpy.math cimport EULER
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
 def mean_change(np.ndarray[ndim=1, dtype=np.float64_t] arr_1,
                 np.ndarray[ndim=1, dtype=np.float64_t] arr_2):
     """Calculate the mean difference between two arrays.
@@ -29,8 +30,6 @@ def mean_change(np.ndarray[ndim=1, dtype=np.float64_t] arr_1,
     return total / size
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
 def _dirichlet_expectation_1d(np.ndarray[ndim=1, dtype=np.float64_t] doc_topic,
                               double doc_topic_prior,
                               np.ndarray[ndim=1, dtype=np.float64_t] out):
@@ -59,8 +58,6 @@ def _dirichlet_expectation_1d(np.ndarray[ndim=1, dtype=np.float64_t] doc_topic,
         out[i] = exp(psi(doc_topic[i]) - psi_total)
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
 def _dirichlet_expectation_2d(np.ndarray[ndim=2, dtype=np.float64_t] arr):
     """Dirichlet expectation for multiple samples:
     E[log(theta)] for theta ~ Dir(arr).

diff --git a/sklearn/decomposition/cdnmf_fast.pyx b/sklearn/decomposition/cdnmf_fast.pyx
@@ -1,6 +1,7 @@
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
+# cython: language_level=3
 
 # Author: Mathieu Blondel, Tom Dupre la Tour
 # License: BSD 3 clause

diff --git a/sklearn/ensemble/_gradient_boosting.pyx b/sklearn/ensemble/_gradient_boosting.pyx
@@ -1,6 +1,7 @@
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
+# cython: language_level=3
 #
 # Author: Peter Prettenhofer
 #

diff --git a/sklearn/feature_extraction/_hashing.pyx b/sklearn/feature_extraction/_hashing.pyx
@@ -1,5 +1,7 @@
 # Author: Lars Buitinck
 # License: BSD 3 clause
+#
+# cython: language_level=3, boundscheck=False, cdivision=True
 
 import sys
 import array
@@ -15,8 +17,6 @@ from sklearn.utils.fixes import sp_version
 np.import_array()
 
 
-@cython.boundscheck(False)
-@cython.cdivision(True)
 def transform(raw_X, Py_ssize_t n_features, dtype, bint alternate_sign=1):
     """Guts of FeatureHasher.transform.
 

diff --git a/sklearn/linear_model/cd_fast.pyx b/sklearn/linear_model/cd_fast.pyx
@@ -5,6 +5,8 @@
 #         Manoj Kumar <manojkumarsivaraj334@gmail.com>
 #
 # License: BSD 3 clause
+#
+# cython: language_level=3, boundscheck=False, wraparound=False, cdivision=True
 
 from libc.math cimport fabs
 cimport numpy as np
@@ -100,9 +102,6 @@ cdef floating diff_abs_max(int n, floating* a, floating* b) nogil:
     return m
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def enet_coordinate_descent(floating[::1] w,
                             floating alpha, floating beta,
                             floating[::1, :] X,
@@ -258,9 +257,6 @@ def enet_coordinate_descent(floating[::1] w,
     return w, gap, tol, n_iter + 1
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def sparse_enet_coordinate_descent(floating [::1] w,
                             floating alpha, floating beta,
                             np.ndarray[floating, ndim=1, mode='c'] X_data,
@@ -475,9 +471,6 @@ def sparse_enet_coordinate_descent(floating [::1] w,
     return w, gap, tol, n_iter + 1
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def enet_coordinate_descent_gram(floating[::1] w,
                                  floating alpha, floating beta,
                                  np.ndarray[floating, ndim=2, mode='c'] Q,
@@ -630,9 +623,6 @@ def enet_coordinate_descent_gram(floating[::1] w,
     return np.asarray(w), gap, tol, n_iter + 1
 
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
-@cython.cdivision(True)
 def enet_coordinate_descent_multi_task(floating[::1, :] W, floating l1_reg,
                                        floating l2_reg,
                                        np.ndarray[floating, ndim=2, mode='fortran'] X,

diff --git a/sklearn/linear_model/sag_fast.pyx.tp b/sklearn/linear_model/sag_fast.pyx.tp
@@ -29,16 +29,23 @@ def get_dispatch(dtypes):
 
 }}
 
-
-"""
-SAG and SAGA implementation
-WARNING: Do not edit .pyx file directly, it is generated from .pyx.tp
-"""
 #------------------------------------------------------------------------------
 
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
+# cython: language_level=3
+#
+# Authors: Danny Sullivan <dbsullivan23@gmail.com>
+#          Tom Dupre la Tour <tom.dupre-la-tour@m4x.org>
+#          Arthur Mensch <arthur.mensch@m4x.org
+#
+# License: BSD 3 clause
+
+"""
+SAG and SAGA implementation
+WARNING: Do not edit .pyx file directly, it is generated from .pyx.tp
+"""
 
 cimport numpy as np
 import numpy as np

diff --git a/sklearn/linear_model/sgd_fast.pxd b/sklearn/linear_model/sgd_fast.pxd
@@ -1,5 +1,7 @@
-"""Helper to load LossFunction from sgd_fast.pyx to sag_fast.pyx"""
 # License: BSD 3 clause
+#
+# cython: language_level=3
+"""Helper to load LossFunction from sgd_fast.pyx to sag_fast.pyx"""
 
 cdef class LossFunction:
     cdef double loss(self, double p, double y) nogil

diff --git a/sklearn/linear_model/sgd_fast.pyx b/sklearn/linear_model/sgd_fast.pyx
@@ -1,6 +1,7 @@
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
+# cython: language_level=3
 #
 # Author: Peter Prettenhofer <peter.prettenhofer@gmail.com>
 #         Mathieu Blondel (partial_fit support)

diff --git a/sklearn/manifold/_barnes_hut_tsne.pyx b/sklearn/manifold/_barnes_hut_tsne.pyx
@@ -1,6 +1,8 @@
 # cython: boundscheck=False
 # cython: wraparound=False
 # cython: cdivision=True
+# cython: language_level=3
+#
 # Author: Christopher Moody <chrisemoody@gmail.com>
 # Author: Nick Travers <nickt@squareup.com>
 # Implementation by Chris Moody & Nick Travers

diff --git a/sklearn/manifold/_utils.pyx b/sklearn/manifold/_utils.pyx
@@ -1,3 +1,5 @@
+# cython: language_level=3, boundscheck=False
+
 from libc cimport math
 cimport cython
 import numpy as np
@@ -10,7 +12,6 @@ cdef extern from "numpy/npy_math.h":
 cdef float EPSILON_DBL = 1e-8
 cdef float PERPLEXITY_TOLERANCE = 1e-5
 
-@cython.boundscheck(False)
 cpdef np.ndarray[np.float32_t, ndim=2] _binary_search_perplexity(
         np.ndarray[np.float32_t, ndim=2] affinities,
         np.ndarray[np.int64_t, ndim=2] neighbors,

diff --git a/sklearn/metrics/cluster/expected_mutual_info_fast.pyx b/sklearn/metrics/cluster/expected_mutual_info_fast.pyx
@@ -1,6 +1,8 @@
 # cython: cdivision=True
 # cython: boundscheck=False
 # cython: wraparound=False
+# cython: language_level=3
+#
 # Authors: Robert Layton <robertlayton@gmail.com>
 #           Corey Lynch <coreylynch9@gmail.com>
 # License: BSD 3 clause
@@ -15,8 +17,7 @@ from sklearn.utils.lgamma cimport lgamma
 np.import_array()
 ctypedef np.float64_t DOUBLE
 
-@cython.boundscheck(False)
-@cython.wraparound(False)
+
 def expected_mutual_information(contingency, int n_samples):
     """Calculate the expected mutual information for two labelings."""
     cdef int R, C

diff --git a/sklearn/metrics/pairwise_fast.pyx b/sklearn/metrics/pairwise_fast.pyx
@@ -1,7 +1,8 @@
 #cython: boundscheck=False
 #cython: cdivision=True
 #cython: wraparound=False
-
+# cython: language_level=3
+#
 # Author: Andreas Mueller <amueller@ais.uni-bonn.de>
 #         Lars Buitinck
 #

diff --git a/sklearn/neighbors/ball_tree.pyx b/sklearn/neighbors/ball_tree.pyx
@@ -2,6 +2,7 @@
 #cython: boundscheck=False
 #cython: wraparound=False
 #cython: cdivision=True
+# cython: language_level=3
 
 # Author: Jake Vanderplas <vanderplas@astro.washington.edu>
 # License: BSD 3 clause

diff --git a/sklearn/neighbors/binary_tree.pxi b/sklearn/neighbors/binary_tree.pxi
@@ -1,4 +1,6 @@
 #!python
+# cython: language_level=3
+
 
 # KD Tree and Ball Tree
 # =====================
@@ -152,10 +154,10 @@ import numpy as np
 import warnings
 from ..utils import check_array
 
-from typedefs cimport DTYPE_t, ITYPE_t, DITYPE_t
-from typedefs import DTYPE, ITYPE
+from .typedefs cimport DTYPE_t, ITYPE_t, DITYPE_t
+from .typedefs import DTYPE, ITYPE
 
-from dist_metrics cimport (DistanceMetric, euclidean_dist, euclidean_rdist,
+from .dist_metrics cimport (DistanceMetric, euclidean_dist, euclidean_rdist,
                            euclidean_dist_to_rdist, euclidean_rdist_to_dist)
 
 cdef extern from "numpy/arrayobject.h":
@@ -999,7 +1001,7 @@ def newObj(obj):
 
 ######################################################################
 # define the reverse mapping of VALID_METRICS
-from dist_metrics import get_valid_metric_ids
+from .dist_metrics import get_valid_metric_ids
 VALID_METRIC_IDS = get_valid_metric_ids(VALID_METRICS)