MAINT: Mayavi / NumPy dep (mne-tools#7412)

* MAINT: Mayavi / NumPy dep

* FIX: Restore tests for pre

* MAINT: Bump NumPy

* FIX: Bump reqs to ones we can test

* FIX: Bump again

* DOC: Correct README [ci skip]

* FIX: Nest import

* FIX: Def

* FIX: Missed one

.travis.yml

@@ -33,7 +33,7 @@ matrix:
         # Old dependencies
         - os: linux
           env: PYTHON_VERSION=3.5
-               CONDA_DEPENDENCIES="numpy=1.12 scipy=0.18 matplotlib=2.0 pandas=0.19 scikit-learn=0.18"
+               CONDA_DEPENDENCIES="numpy=1.13 scipy=1.0 matplotlib=2.1 pandas=0.21 scikit-learn=0.19"
                CONDA_CHANNELS="conda-forge"
 
         # Minimal (runs with and without testing data)
@@ -116,8 +116,6 @@ script:
     # need the --pre wheels to skip slow ones.
     - if [ "${TRAVIS_OS_NAME}" == "osx" ]; then
         CONDITION='not slowtest';
-      elif [ -z "$CONDA_ENV" ] && [ -z "$CONDA_DEPENDENCIES" ]; then
-        CONDITION='not slowtest';
       else
         CONDITION='not ultraslowtest';
       fi;

README.rst

@@ -81,22 +81,22 @@ Dependencies
 The minimum required dependencies to run MNE-Python are:
 
 - Python >= 3.5
-- NumPy >= 1.12.1
-- SciPy >= 0.18.1
+- NumPy >= 1.13.3
+- SciPy >= 1.0.0
 
 For full functionality, some functions require:
 
-- Matplotlib >= 2.0.2
+- Matplotlib >= 2.1
 - Mayavi >= 4.6
 - PySurfer >= 0.8
-- Scikit-learn >= 0.18.2
+- Scikit-learn >= 0.19.1
 - Numba >= 0.40
 - NiBabel >= 2.1.0
-- Pandas >= 0.19.2
+- Pandas >= 0.21
 - Picard >= 0.3
 - CuPy >= 4.0 (for NVIDIA CUDA acceleration)
 - DIPY >= 0.10.1
-- PyVista >= 0.20.1
+- PyVista >= 0.23.1
 
 Contributing to MNE-Python
 ^^^^^^^^^^^^^^^^^^^^^^^^^^

doc/changes/latest.inc

@@ -282,6 +282,14 @@ Bug
 API
 ~~~
 
+- Bumped minimum requirements to fall 2017 versions by `Eric Larson`_:
+
+  - NumPy 1.13.3
+  - SciPy 1.0.0
+  - matplotlib 2.1
+  - scikit-learn 1.19.1 (optional requirement)
+  - pandas 0.21 (optional requirement)
+
 - :meth:`mne.Epochs.plot` now accepts an ``event_id`` parameter (useful in tandem with ``event_colors`` for specifying event colors by name) by `Daniel McCloy`_.
 
 - New time conversion options for methods :meth:`mne.io.Raw.to_data_frame`, :meth:`mne.Epochs.to_data_frame`, :meth:`mne.Evoked.to_data_frame`, and :meth:`mne.SourceEstimate.to_data_frame`, by `Daniel McCloy`_.

mne/conftest.py

@@ -82,6 +82,7 @@ def pytest_configure(config):
     ignore:.*TraitTuple.*trait.*handler.*deprecated.*:DeprecationWarning
     ignore:.*rich_compare.*metadata.*deprecated.*:DeprecationWarning
     ignore:.*In future, it will be an error for 'np.bool_'.*:DeprecationWarning
+    ignore:.*Converting `np\.character` to a dtype is deprecated.*:DeprecationWarning
     always:.*get_data.* is deprecated in favor of.*:DeprecationWarning
     """  # noqa: E501
     for warning_line in warning_lines.split('\n'):

mne/cov.py

@@ -964,7 +964,7 @@ def _compute_covariance_auto(data, method, info, method_params, cv,
         estimator_cov_info = list()
         msg = 'Estimating covariance using %s'
 
-        ok_sklearn = check_version('sklearn', '0.15')
+        ok_sklearn = check_version('sklearn')
         if not ok_sklearn and (len(method) != 1 or method[0] != 'empirical'):
             raise ValueError('scikit-learn is not installed, `method` must be '
                              '`empirical`, got %s' % (method,))

mne/decoding/base.py

@@ -225,35 +225,17 @@ def _set_cv(cv, estimator=None, X=None, y=None):
     else:
         est_is_classifier = is_classifier(estimator)
     # Setup CV
-    if check_version('sklearn', '0.18'):
-        from sklearn import model_selection as models
-        from sklearn.model_selection import (check_cv, StratifiedKFold, KFold)
-        if isinstance(cv, (int, np.int)):
-            XFold = StratifiedKFold if est_is_classifier else KFold
-            cv = XFold(n_splits=cv)
-        elif isinstance(cv, str):
-            if not hasattr(models, cv):
-                raise ValueError('Unknown cross-validation')
-            cv = getattr(models, cv)
-            cv = cv()
-        cv = check_cv(cv=cv, y=y, classifier=est_is_classifier)
-    else:
-        from sklearn import cross_validation as models
-        from sklearn.cross_validation import (check_cv, StratifiedKFold, KFold)
-        if isinstance(cv, (int, np.int)):
-            if est_is_classifier:
-                cv = StratifiedKFold(y=y, n_folds=cv)
-            else:
-                cv = KFold(n=len(y), n_folds=cv)
-        elif isinstance(cv, str):
-            if not hasattr(models, cv):
-                raise ValueError('Unknown cross-validation')
-            cv = getattr(models, cv)
-            if cv.__name__ not in ['KFold', 'LeaveOneOut']:
-                raise NotImplementedError('CV cannot be defined with str for'
-                                          ' sklearn < .017.')
-            cv = cv(len(y))
-        cv = check_cv(cv=cv, X=X, y=y, classifier=est_is_classifier)
+    from sklearn import model_selection as models
+    from sklearn.model_selection import (check_cv, StratifiedKFold, KFold)
+    if isinstance(cv, (int, np.int)):
+        XFold = StratifiedKFold if est_is_classifier else KFold
+        cv = XFold(n_splits=cv)
+    elif isinstance(cv, str):
+        if not hasattr(models, cv):
+            raise ValueError('Unknown cross-validation')
+        cv = getattr(models, cv)
+        cv = cv()
+    cv = check_cv(cv=cv, y=y, classifier=est_is_classifier)
 
     # Extract train and test set to retrieve them at predict time
     if hasattr(cv, 'split'):

mne/decoding/tests/test_base.py

@@ -9,7 +9,7 @@
 import pytest
 
 from mne.fixes import is_regressor, is_classifier
-from mne.utils import requires_version, check_version
+from mne.utils import requires_sklearn, check_version
 from mne.decoding.base import (_get_inverse_funcs, LinearModel, get_coef,
                                cross_val_multiscore)
 from mne.decoding.search_light import SlidingEstimator
@@ -55,7 +55,7 @@ def _make_data(n_samples=1000, n_features=5, n_targets=3):
     return X, Y, A
 
 
-@requires_version('sklearn', '0.17')
+@requires_sklearn
 def test_get_coef():
     """Test getting linear coefficients (filters/patterns) from estimators."""
     from sklearn.base import TransformerMixin, BaseEstimator
@@ -206,7 +206,7 @@ def inverse_transform(self, X):
     lm.fit(X, Y, sample_weight=np.ones(len(Y)))
 
 
-@requires_version('sklearn', '0.15')
+@requires_sklearn
 def test_linearmodel():
     """Test LinearModel class for computing filters and patterns."""
     # check categorical target fit in standard linear model
@@ -260,7 +260,7 @@ def test_linearmodel():
         clf.fit(X, wrong_y)
 
 
-@requires_version('sklearn', '0.18')
+@requires_sklearn
 def test_cross_val_multiscore():
     """Test cross_val_multiscore for computing scores on decoding over time."""
     from sklearn.model_selection import KFold, StratifiedKFold, cross_val_score

mne/decoding/tests/test_ems.py

@@ -8,7 +8,7 @@
 import pytest
 
 from mne import io, Epochs, read_events, pick_types
-from mne.utils import requires_version, check_version, run_tests_if_main
+from mne.utils import requires_sklearn, run_tests_if_main
 from mne.decoding import compute_ems, EMS
 
 data_dir = op.join(op.dirname(__file__), '..', '..', 'io', 'tests', 'data')
@@ -21,9 +21,10 @@
 event_id = dict(aud_l=1, vis_l=3)
 
 
-@requires_version('sklearn', '0.15')
+@requires_sklearn
 def test_ems():
     """Test event-matched spatial filters."""
+    from sklearn.model_selection import StratifiedKFold
     raw = io.read_raw_fif(raw_fname, preload=False)
 
     # create unequal number of events
@@ -58,12 +59,7 @@ def test_ems():
     # test compute_ems cv
     epochs = epochs['aud_r', 'vis_l']
     epochs.equalize_event_counts(epochs.event_id)
-    if check_version('sklearn', '0.18'):
-        from sklearn.model_selection import StratifiedKFold
-        cv = StratifiedKFold(n_splits=3)
-    else:
-        from sklearn.cross_validation import StratifiedKFold
-        cv = StratifiedKFold(epochs.events[:, 2])
+    cv = StratifiedKFold(n_splits=3)
     compute_ems(epochs, cv=cv)
     compute_ems(epochs, cv=2)
     pytest.raises(ValueError, compute_ems, epochs, cv='foo')

mne/decoding/tests/test_receptive_field.py

@@ -10,7 +10,7 @@
 
 from mne import io, pick_types
 from mne.fixes import einsum, rfft, irfft
-from mne.utils import requires_version, requires_sklearn, run_tests_if_main
+from mne.utils import requires_sklearn, run_tests_if_main
 from mne.decoding import ReceptiveField, TimeDelayingRidge
 from mne.decoding.receptive_field import (_delay_time_series, _SCORERS,
                                           _times_to_delays, _delays_to_slice)
@@ -537,7 +537,6 @@ def test_inverse_coef():
 
 
 @requires_sklearn
-@requires_version('scipy', '1.0')
 def test_linalg_warning():
     """Test that warnings are issued when no regularization is applied."""
     from sklearn.linear_model import Ridge

mne/decoding/tests/test_search_light.py

@@ -6,7 +6,7 @@
 from numpy.testing import assert_array_equal, assert_equal
 import pytest
 
-from mne.utils import requires_version
+from mne.utils import requires_sklearn
 from mne.fixes import _get_args
 from mne.decoding.search_light import SlidingEstimator, GeneralizingEstimator
 from mne.decoding.transformer import Vectorizer
@@ -24,7 +24,7 @@ def make_data():
     return X, y
 
 
-@requires_version('sklearn', '0.17')
+@requires_sklearn
 def test_search_light():
     """Test SlidingEstimator."""
     from sklearn.linear_model import Ridge, LogisticRegression
@@ -171,7 +171,7 @@ def transform(self, X):
         assert (isinstance(pipe.estimators_[0], BaggingClassifier))
 
 
-@requires_version('sklearn', '0.17')
+@requires_sklearn
 def test_generalization_light():
     """Test GeneralizingEstimator."""
     from sklearn.pipeline import make_pipeline
@@ -258,7 +258,7 @@ def test_generalization_light():
     assert_array_equal(y_preds[0], y_preds[1])
 
 
-@requires_version('sklearn', '0.19')  # 0.18 does not raise when it should
+@requires_sklearn
 def test_cross_val_predict():
     """Test cross_val_predict with predict_proba."""
     from sklearn.linear_model import LinearRegression

mne/decoding/tests/test_transformer.py

@@ -14,7 +14,7 @@
 from mne.decoding import (Scaler, FilterEstimator, PSDEstimator, Vectorizer,
                           UnsupervisedSpatialFilter, TemporalFilter)
 from mne.defaults import DEFAULTS
-from mne.utils import requires_version, run_tests_if_main, check_version
+from mne.utils import requires_sklearn, run_tests_if_main, check_version
 
 tmin, tmax = -0.2, 0.5
 event_id = dict(aud_l=1, vis_l=3)
@@ -42,11 +42,9 @@ def test_scaler():
     infos = (epochs.info, epochs.info, None, None)
     epochs_data_t = epochs_data.transpose([1, 0, 2])
     for method, info in zip(methods, infos):
-        if method == 'median' and not check_version('sklearn', '0.17'):
-            pytest.raises(ValueError, Scaler, info, method)
-            continue
-        if method == 'mean' and not check_version('sklearn', ''):
-            pytest.raises(ImportError, Scaler, info, method)
+        if method in ('mean', 'median') and not check_version('sklearn'):
+            with pytest.raises(ImportError, match='No module'):
+                Scaler(info, method)
             continue
         scaler = Scaler(info, method)
         X = scaler.fit_transform(epochs_data, y)
@@ -172,7 +170,7 @@ def test_vectorizer():
                   np.random.rand(102, 12, 12))
 
 
-@requires_version('sklearn', '0.16')
+@requires_sklearn
 def test_unsupervised_spatial_filter():
     """Test unsupervised spatial filter."""
     from sklearn.decomposition import PCA

mne/decoding/transformer.py

@@ -13,7 +13,7 @@
 from .. import pick_types
 from ..filter import filter_data, _triage_filter_params
 from ..time_frequency.psd import psd_array_multitaper
-from ..utils import check_version, fill_doc, _check_option
+from ..utils import fill_doc, _check_option
 from ..io.pick import (pick_info, _pick_data_channels, _picks_by_type,
                        _picks_to_idx)
 from ..cov import _check_scalings_user
@@ -125,9 +125,6 @@ def __init__(self, info=None, scalings=None, with_mean=True,
             from sklearn.preprocessing import StandardScaler
             self._scaler = StandardScaler(self.with_mean, self.with_std)
         else:  # scalings == 'median':
-            if not check_version('sklearn', '0.17'):
-                raise ValueError("median requires version 0.17 of "
-                                 "sklearn library")
             from sklearn.preprocessing import RobustScaler
             self._scaler = RobustScaler(self.with_mean, self.with_std)
 

mne/filter.py

@@ -9,7 +9,7 @@
 from .io.pick import _picks_to_idx
 from .cuda import (_setup_cuda_fft_multiply_repeated, _fft_multiply_repeated,
                    _setup_cuda_fft_resample, _fft_resample, _smart_pad)
-from .fixes import minimum_phase, _sosfreqz, irfft, ifftshift, fftfreq
+from .fixes import irfft, ifftshift, fftfreq
 from .parallel import parallel_func, check_n_jobs
 from .time_frequency.multitaper import _mt_spectra, _compute_mt_params
 from .utils import (logger, verbose, sum_squared, check_version, warn, _pl,
@@ -363,6 +363,7 @@ def _construct_fir_filter(sfreq, freq, gain, filter_length, phase, fir_window,
     else:
         assert fir_design == 'firwin'
         fir_design = partial(_firwin_design, sfreq=sfreq)
+    from scipy.signal import minimum_phase
 
     # issue a warning if attenuation is less than this
     min_att_db = 12 if phase == 'minimum' else 20
@@ -624,7 +625,7 @@ def construct_iir_filter(iir_params, f_pass=None, f_stop=None, sfreq=None,
     For more information, see the tutorials
     :ref:`disc-filtering` and :ref:`tut-filter-resample`.
     """  # noqa: E501
-    from scipy.signal import iirfilter, iirdesign, freqz
+    from scipy.signal import iirfilter, iirdesign, freqz, sosfreqz
     known_filters = ('bessel', 'butter', 'butterworth', 'cauer', 'cheby1',
                      'cheby2', 'chebyshev1', 'chebyshev2', 'chebyshevi',
                      'chebyshevii', 'ellip', 'elliptic')
@@ -691,7 +692,7 @@ def construct_iir_filter(iir_params, f_pass=None, f_stop=None, sfreq=None,
     # get the gains at the cutoff frequencies
     if Wp is not None:
         if output == 'sos':
-            cutoffs = _sosfreqz(system, worN=Wp * np.pi)[1]
+            cutoffs = sosfreqz(system, worN=Wp * np.pi)[1]
         else:
             cutoffs = freqz(system[0], system[1], worN=Wp * np.pi)[1]
         # 2 * 20 here because we do forward-backward filtering
@@ -1821,15 +1822,13 @@ def savgol_filter(self, h_freq, verbose=None):
         >>> evoked.savgol_filter(10.)  # low-pass at around 10 Hz # doctest:+SKIP
         >>> evoked.plot()  # doctest:+SKIP
         """  # noqa: E501
+        from scipy.signal import savgol_filter
         _check_preload(self, 'inst.savgol_filter')
         h_freq = float(h_freq)
         if h_freq >= self.info['sfreq'] / 2.:
             raise ValueError('h_freq must be less than half the sample rate')
 
         # savitzky-golay filtering
-        if not check_version('scipy', '0.14'):
-            raise RuntimeError('scipy >= 0.14 must be installed for savgol')
-        from scipy.signal import savgol_filter
         window_length = (int(np.round(self.info['sfreq'] /
                                       h_freq)) // 2) * 2 + 1
         logger.info('Using savgol length %d' % window_length)