scikit-learn · ogrisel · Jan 2, 2020 · Dec 2, 2019 · Dec 4, 2019 · Dec 4, 2019
diff --git a/.circleci/config.yml b/.circleci/config.yml
@@ -101,7 +101,7 @@ jobs:
 
   pypy3:
     docker:
-      - image: pypy:3.6-7.1.1
+      - image: pypy:3.6-7.2.0
     steps:
       - restore_cache:
           keys:

diff --git a/azure-pipelines.yml b/azure-pipelines.yml
@@ -12,8 +12,14 @@ jobs:
     - bash: conda create --name flake8_env --yes flake8
       displayName: Install flake8
     - bash: |
-        source activate flake8_env
-        ./build_tools/circle/linting.sh
+        if [[ $BUILD_SOURCEVERSIONMESSAGE =~ \[lint\ skip\] ]]; then
+          # skip linting
+          echo "Skipping linting"
+          exit 0
+        else
+          source activate flake8_env
+          ./build_tools/circle/linting.sh
+        fi
       displayName: Run linting
 
 
@@ -65,10 +71,9 @@ jobs:
         JOBLIB_VERSION: '0.12.3'
         COVERAGE: 'true'
       # Linux environment to test the latest available dependencies and MKL.
-      # It runs tests requiring pandas and PyAMG.
+      # It runs tests requiring lightgbm, pandas and PyAMG.
       pylatest_pip_openblas_pandas:
         DISTRIB: 'conda-pip-latest'
-        # FIXME: pinned until SciPy wheels are available for Pyhon 3.8
         PYTHON_VERSION: '3.8'
         PYTEST_VERSION: '4.6.2'
         COVERAGE: 'true'

diff --git a/benchmarks/bench_plot_randomized_svd.py b/benchmarks/bench_plot_randomized_svd.py
@@ -104,7 +104,7 @@
 # in case the reconstructed (dense) matrix is too large
 MAX_MEMORY = np.int(2e9)
 
-# The following datasets can be dowloaded manually from:
+# The following datasets can be downloaded manually from:
 # CIFAR 10: https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
 # SVHN: http://ufldl.stanford.edu/housenumbers/train_32x32.mat
 CIFAR_FOLDER = "./cifar-10-batches-py/"

diff --git a/benchmarks/bench_text_vectorizers.py b/benchmarks/bench_text_vectorizers.py
@@ -32,7 +32,7 @@ def f():
 text = fetch_20newsgroups(subset='train').data[:1000]
 
 print("="*80 + '\n#' + "    Text vectorizers benchmark" + '\n' + '='*80 + '\n')
-print("Using a subset of the 20 newsrgoups dataset ({} documents)."
+print("Using a subset of the 20 newsgroups dataset ({} documents)."
       .format(len(text)))
 print("This benchmarks runs in ~1 min ...")
 

diff --git a/build_tools/azure/install.sh b/build_tools/azure/install.sh
@@ -13,7 +13,7 @@ make_conda() {
 version_ge() {
     # The two version numbers are separated with a new line is piped to sort
     # -rV. The -V activates for version number sorting and -r sorts in
-    # decending order. If the first argument is the top element of the sort, it
+    # descending order. If the first argument is the top element of the sort, it
     # is greater than or equal to the second argument.
     test "$(printf "${1}\n${2}" | sort -rV | head -n 1)" == "$1"
 }
@@ -92,6 +92,8 @@ elif [[ "$DISTRIB" == "conda-pip-latest" ]]; then
     python -m pip install numpy scipy cython joblib
     python -m pip install pytest==$PYTEST_VERSION pytest-cov pytest-xdist
     python -m pip install pandas matplotlib pyamg
+    # do not install dependencies for lightgbm since it requires scikit-learn
+    python -m pip install lightgbm --no-deps
 fi
 
 if [[ "$COVERAGE" == "true" ]]; then

diff --git a/build_tools/circle/build_test_pypy.sh b/build_tools/circle/build_test_pypy.sh
@@ -18,11 +18,14 @@ source pypy-env/bin/activate
 python --version
 which python
 
-# XXX: numpy version pinning can be reverted once PyPy
-#      compatibility is resolved for numpy v1.6.x. For instance,
-#      when PyPy3 >6.0 is released (see numpy/numpy#12740)
-pip install --extra-index https://antocuni.github.io/pypy-wheels/ubuntu numpy Cython pytest
-pip install scipy sphinx numpydoc docutils joblib pillow
+pip install -U pip
+
+# pins versions to install wheel from https://antocuni.github.io/pypy-wheels/manylinux2010
+pip install --extra-index-url https://antocuni.github.io/pypy-wheels/manylinux2010 numpy==1.18.0 scipy==1.3.2
+
+# Install Cython directly
+pip install https://antocuni.github.io/pypy-wheels/ubuntu/Cython/Cython-0.29.14-py3-none-any.whl
+pip install sphinx numpydoc docutils joblib pillow pytest
 
 ccache -M 512M
 export CCACHE_COMPRESS=1

diff --git a/conftest.py b/conftest.py
@@ -37,7 +37,7 @@ def pytest_collection_modifyitems(config, items):
         skip_marker = pytest.mark.skip(
             reason='FeatureHasher is not compatible with PyPy')
         for item in items:
-            if item.name.endswith(('hashing.FeatureHasher',
+            if item.name.endswith(('_hash.FeatureHasher',
                                    'text.HashingVectorizer')):
                 item.add_marker(skip_marker)
 

diff --git a/doc/Makefile b/doc/Makefile
@@ -2,7 +2,7 @@
 #
 
 # You can set these variables from the command line.
-SPHINXOPTS    =
+SPHINXOPTS    = -j auto
 SPHINXBUILD  ?= sphinx-build
 PAPER         =
 BUILDDIR      = _build

diff --git a/doc/conf.py b/doc/conf.py
@@ -51,6 +51,7 @@
 if os.environ.get('NO_MATHJAX'):
     extensions.append('sphinx.ext.imgmath')
     imgmath_image_format = 'svg'
+    mathjax_path = ''
 else:
     extensions.append('sphinx.ext.mathjax')
     mathjax_path = ('https://cdn.jsdelivr.net/npm/mathjax@3/es5/'
@@ -304,7 +305,9 @@ def __call__(self, directory):
         'branch': binder_branch,
         'dependencies': './binder/requirements.txt',
         'use_jupyter_lab': True
-    }
+    },
+    # avoid generating too many cross links
+    'inspect_global_variables': False,
 }
 
 

diff --git a/doc/developers/advanced_installation.rst b/doc/developers/advanced_installation.rst
@@ -49,7 +49,7 @@ feature, code or documentation improvement).
    If you plan on submitting a pull-request, you should clone from your fork
    instead.
 
-#. Install a compiler with OpenMP_ support for your platform. See intructions
+#. Install a compiler with OpenMP_ support for your platform. See instructions
    for :ref:`compiler_windows`, :ref:`compiler_macos`, :ref:`compiler_linux`
    and :ref:`compiler_freebsd`.
 
@@ -339,7 +339,7 @@ architecture (e.g. ARM), you can install the system versions::
 
 On Red Hat and clones (e.g. CentOS), install the dependencies using::
 
-    sudo yum -y install gcc gcc-c++ python-devel numpy scipy
+    sudo yum -y install gcc gcc-c++ python3-devel numpy scipy
 
 Linux compilers from conda-forge
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

diff --git a/doc/developers/contributing.rst b/doc/developers/contributing.rst
@@ -251,7 +251,7 @@ modifying code and submitting a PR:
    to record your changes in Git, then push the changes to your GitHub
    account with::
 
-       $ git push -u origin my-feature
+       $ git push -u origin my_feature
 
 10. Follow `these
     <https://help.github.com/articles/creating-a-pull-request-from-a-fork>`_
@@ -377,7 +377,7 @@ complies with the following rules before marking a PR as ``[MRG]``. The
    methods available in scikit-learn.
 
 10. New features often need to be illustrated with narrative documentation in
-    the user guide, with small code snipets. If relevant, please also add
+    the user guide, with small code snippets. If relevant, please also add
     references in the literature, with PDF links when possible.
 
 11. The user guide should also include expected time and space complexity
@@ -437,6 +437,7 @@ message, the following actions are taken.
     ---------------------- -------------------
     [scipy-dev]            Add a Travis build with our dependencies (numpy, scipy, etc ...) development builds
     [ci skip]              CI is skipped completely
+    [lint skip]            Azure pipeline skips linting
     [doc skip]             Docs are not built
     [doc quick]            Docs built, but excludes example gallery plots
     [doc build]            Docs built including example gallery plots

diff --git a/doc/developers/maintainer.rst b/doc/developers/maintainer.rst
@@ -62,7 +62,7 @@ Making a release
 2. On the branch for releasing, update the version number in
    sklearn/__init__.py, the ``__version__`` variable by removing ``dev*`` only
    when ready to release.
-   On master, increment the verson in the same place (when branching for
+   On master, increment the version in the same place (when branching for
    release).
 
 3. Create the tag and push it::

diff --git a/doc/developers/tips.rst b/doc/developers/tips.rst
@@ -181,10 +181,10 @@ Issue/Comment: Linking to comments
 
         Please use links to comments, which make it a lot easier to see what you are referring to, rather than just linking to the issue. See [this](https://stackoverflow.com/questions/25163598/how-do-i-reference-a-specific-issue-comment-on-github) for more details.
 
-PR-NEW: Better description
+PR-NEW: Better description and title
     ::
 
-        Thanks for the pull request! Please make the title of the PR descriptive so that we can easily recall the issue it is resolving. You should state what issue (or PR) it fixes/resolves in the description (see [here](http://scikit-learn.org/dev/developers/contributing.html#contributing-pull-requests)).
+        Thanks for the pull request! Please make the title of the PR more descriptive. The title will become the commit message when this is merged. You should state what issue (or PR) it fixes/resolves in the description using the syntax described [here](http://scikit-learn.org/dev/developers/contributing.html#contributing-pull-requests).
 
 PR-NEW: Fix #
     ::

diff --git a/doc/glossary.rst b/doc/glossary.rst
@@ -1161,7 +1161,7 @@ Methods
 
             TODO: `This gist
             <https://gist.github.com/jnothman/4807b1b0266613c20ba4d1f88d0f8cf5>`_
-            higlights the use of the different formats for multilabel.
+            highlights the use of the different formats for multilabel.
         multioutput classification
             A list of 2d arrays, corresponding to each multiclass decision
             function.

diff --git a/doc/install.rst b/doc/install.rst
@@ -126,7 +126,7 @@ If you have not installed NumPy or SciPy yet, you can also install these using
 conda or pip. When using pip, please ensure that *binary wheels* are used,
 and NumPy and SciPy are not recompiled from source, which can happen when using
 particular configurations of operating system and hardware (such as Linux on
-a Raspberry Pi). 
+a Raspberry Pi).
 
 If you must install scikit-learn and its dependencies with pip, you can install
 it as ``scikit-learn[alldeps]``.
@@ -255,3 +255,38 @@ WinPython for Windows
 
 The `WinPython <https://winpython.github.io/>`_ project distributes
 scikit-learn as an additional plugin.
+
+
+Troubleshooting
+===============
+
+.. _windows_longpath:
+
+Error caused by file path length limit on Windows
+-------------------------------------------------
+
+It can happen that pip fails to install packages when reaching the default path
+size limit of Windows if Python is installed in a nested location such as the
+`AppData` folder structure under the user home directory, for instance::
+
+    C:\Users\username>C:\Users\username\AppData\Local\Microsoft\WindowsApps\python.exe -m pip install scikit-learn
+    Collecting scikit-learn
+    ...
+    Installing collected packages: scikit-learn
+    ERROR: Could not install packages due to an EnvironmentError: [Errno 2] No such file or directory: 'C:\\Users\\username\\AppData\\Local\\Packages\\PythonSoftwareFoundation.Python.3.7_qbz5n2kfra8p0\\LocalCache\\local-packages\\Python37\\site-packages\\sklearn\\datasets\\tests\\data\\openml\\292\\api-v1-json-data-list-data_name-australian-limit-2-data_version-1-status-deactivated.json.gz'
+
+In this case it is possible to lift that limit in the Windows registry by
+using the ``regedit`` tool:
+
+#. Type "regedit" in the Windows start menu to launch ``regedit``.
+
+#. Go to the
+   ``Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem``
+   key.
+
+#. Edit the value of the ``LongPathsEnabled`` property of that key and set
+   it to 1.
+
+#. Reinstall scikit-learn (ignoring the previous broken installation)::
+
+       pip install --exists-action=i scikit-learn
diff --git a/doc/modules/clustering.rst b/doc/modules/clustering.rst
@@ -775,7 +775,7 @@ core sample, and is at least ``eps`` in distance from any core sample, is
 considered an outlier by the algorithm.
 
 While the parameter ``min_samples`` primarily controls how tolerant the
-algorithm is towards noise (on noisy and large data sets it may be desiable
+algorithm is towards noise (on noisy and large data sets it may be desirable
 to increase this parameter), the parameter ``eps`` is *crucial to choose
 appropriately* for the data set and distance function and usually cannot be
 left at the default value. It controls the local neighborhood of the points.

diff --git a/doc/modules/model_evaluation.rst b/doc/modules/model_evaluation.rst
@@ -1348,8 +1348,8 @@ the one-vs-rest algorithm computes the average of the ROC AUC scores for each
 class against all other classes. In both cases, the predicted labels are
 provided in an array with values from 0 to ``n_classes``, and the scores
 correspond to the probability estimates that a sample belongs to a particular
-class. The OvO and OvR algorithms supports weighting uniformly
-(``average='macro'``) and weighting by the prevalence (``average='weighted'``).
+class. The OvO and OvR algorithms support weighting uniformly
+(``average='macro'``) and by prevalence (``average='weighted'``).
 
 **One-vs-one Algorithm**: Computes the average AUC of all possible pairwise
 combinations of classes. [HT2001]_ defines a multiclass AUC metric weighted
@@ -1380,10 +1380,10 @@ the keyword argument ``multiclass`` to ``'ovo'`` and ``average`` to
 ``'weighted'``. The ``'weighted'`` option returns a prevalence-weighted average
 as described in [FC2009]_.
 
-**One-vs-rest Algorithm**: Computes the AUC of each class against the rest.
-The algorithm is functionally the same as the multilabel case. To enable this
-algorithm set the keyword argument ``multiclass`` to ``'ovr'``. Similar to
-OvO, OvR supports two types of averaging: ``'macro'`` [F2006]_ and
+**One-vs-rest Algorithm**: Computes the AUC of each class against the rest
+[PD2000]_. The algorithm is functionally the same as the multilabel case. To
+enable this algorithm set the keyword argument ``multiclass`` to ``'ovr'``.
+Like OvO, OvR supports two types of averaging: ``'macro'`` [F2006]_ and
 ``'weighted'`` [F2001]_.
 
 In applications where a high false positive rate is not tolerable the parameter
@@ -1422,6 +1422,10 @@ to the given limit.
        <https://www.math.ucdavis.edu/~saito/data/roc/ferri-class-perf-metrics.pdf>`_
        Pattern Recognition Letters. 30. 27-38.
 
+    .. [PD2000] Provost, F., Domingos, P. (2000). Well-trained PETs: Improving
+       probability estimation trees (Section 6.2), CeDER Working Paper #IS-00-04,
+       Stern School of Business, New York University.
+
     .. [F2006] Fawcett, T., 2006. `An introduction to ROC analysis.
        <http://www.sciencedirect.com/science/article/pii/S016786550500303X>`_
        Pattern Recognition Letters, 27(8), pp. 861-874.
-Original file line number
+Diff line change
@@ Expand Up / @@ -101,7 +101,7 @@ jobs: @@
       pypy3:
         docker:
-          - image: pypy:3.6-7.1.1
+          - image: pypy:3.6-7.2.0
         steps:
           - restore_cache:
               keys:
@@ Expand Down @@