[MNT] Fix warnings in widget tests

Orange/classification/sgd.py

@@ -15,7 +15,7 @@ class SGDClassificationLearner(SklLearner):
 
     def __init__(self, loss='hinge', penalty='l2', alpha=0.0001,
                  l1_ratio=0.15, fit_intercept=True, max_iter=5,
-                 tol=None, shuffle=True, epsilon=0.1, random_state=None,
+                 tol=1e-3, shuffle=True, epsilon=0.1, random_state=None,
                  learning_rate='invscaling', eta0=0.01, power_t=0.25,
                  warm_start=False, average=False, preprocessors=None):
         super().__init__(preprocessors=preprocessors)

Orange/data/table.py

@@ -983,7 +983,7 @@ def collect(attr):
                 vstack(collect("metas")),
                 merge1d(collect("W"))
             )
-            conc.ids = np.hstack(map(operator.attrgetter("ids"), tables))
+            conc.ids = np.hstack([table.ids for table in tables])
             names = [table.name for table in tables if table.name != "untitled"]
             if names:
                 conc.name = names[0]

Orange/distance/base.py

@@ -232,14 +232,15 @@ def __call__(self, e1, e2=None):
 
         x1 = _orange_to_numpy(e1)
         x2 = _orange_to_numpy(e2)
-        dist = self.compute_distances(x1, x2)
-        if self.impute and np.isnan(dist).any():
-            dist = np.nan_to_num(dist)
-        if isinstance(e1, (Table, RowInstance)):
-            dist = DistMatrix(dist, e1, e2, self.axis)
-        else:
-            dist = DistMatrix(dist)
-        return dist
+        with np.errstate(invalid="ignore"):  # nans are handled below
+            dist = self.compute_distances(x1, x2)
+            if self.impute and np.isnan(dist).any():
+                dist = np.nan_to_num(dist)
+            if isinstance(e1, (Table, RowInstance)):
+                dist = DistMatrix(dist, e1, e2, self.axis)
+            else:
+                dist = DistMatrix(dist)
+            return dist
 
     def compute_distances(self, x1, x2):
         """

Orange/regression/linear.py

@@ -84,7 +84,7 @@ class SGDRegressionLearner(LinearRegressionLearner):
     preprocessors = SklLearner.preprocessors + [Normalize()]
 
     def __init__(self, loss='squared_loss', penalty='l2', alpha=0.0001,
-                 l1_ratio=0.15, fit_intercept=True, max_iter=5, tol=None,
+                 l1_ratio=0.15, fit_intercept=True, max_iter=5, tol=1e-3,
                  shuffle=True, epsilon=0.1, n_jobs=1, random_state=None,
                  learning_rate='invscaling', eta0=0.01, power_t=0.25,
                  class_weight=None, warm_start=False, average=False,

Orange/statistics/util.py

@@ -4,6 +4,7 @@
 
 It also patches bottleneck to contain these functions.
 """
+import warnings
 from warnings import warn
 
 import bottleneck as bn
@@ -353,7 +354,10 @@ def weighted_mean():
 
 def _nan_min_max(x, func, axis=0):
     if not sp.issparse(x):
-        return func(x, axis=axis)
+        with warnings.catch_warnings():
+            warnings.filterwarnings("ignore", ".*All-NaN slice encountered.*",
+                                    RuntimeWarning)
+            return func(x, axis=axis)
     if axis is None:
         extreme = func(x.data, axis=axis) if x.nnz else float('nan')
         if sparse_has_implicit_zeros(x):
@@ -423,6 +427,9 @@ def nanmean(x, axis=None):
     """ Equivalent of np.nanmean that supports sparse or dense matrices. """
     def nanmean_sparse(x):
         n_values = np.prod(x.shape) - np.sum(np.isnan(x.data))
+        if not n_values:
+            warnings.warn(RuntimeWarning, "Mean of empty slice")
+            return np.nan
         return np.nansum(x.data) / n_values
 
     return _apply_func(x, np.nanmean, nanmean_sparse, axis=axis)

Orange/tests/test_classification.py

@@ -7,7 +7,11 @@
 import unittest
 
 import traceback
+import warnings
+
 import numpy as np
+from sklearn.exceptions import ConvergenceWarning
+
 from Orange.base import SklLearner
 
 import Orange.classification
@@ -269,6 +273,9 @@ class LearnerAccessibility(unittest.TestCase):
 
     def setUp(self):
         Variable._clear_all_caches()
+        # Convergence warnings are irrelevant for these tests
+        warnings.filterwarnings("ignore", ".*", ConvergenceWarning)
+
 
     def all_learners(self):
         classification_modules = pkgutil.walk_packages(

Orange/tests/test_neural_network.py

@@ -2,6 +2,9 @@
 # pylint: disable=missing-docstring
 
 import unittest
+import warnings
+
+from sklearn.exceptions import ConvergenceWarning
 
 from Orange.data import Table
 from Orange.classification import NNClassificationLearner
@@ -17,6 +20,11 @@ def setUpClass(cls):
         cls.housing = Table('housing')
         cls.learner = NNLearner()
 
+    def setUp(self):
+        # Convergence warnings are irrelevant for these tests
+        warnings.filterwarnings("ignore", ".*", ConvergenceWarning)
+        super().setUp()
+
     def test_NN_classification(self):
         results = CrossValidation(self.iris, [NNClassificationLearner()], k=3)
         ca = CA(results)

Orange/tests/test_sgd.py

@@ -2,8 +2,10 @@
 # pylint: disable=missing-docstring
 
 import unittest
+import warnings
 
 import numpy as np
+from sklearn.exceptions import ConvergenceWarning
 
 from Orange.data import Table
 from Orange.classification import SGDClassificationLearner
@@ -12,6 +14,11 @@
 
 
 class TestSGDRegressionLearner(unittest.TestCase):
+    def setUp(self):
+        # Convergence warnings are irrelevant for these tests
+        warnings.filterwarnings("ignore", ".*", ConvergenceWarning)
+        super().setUp()
+
     def test_SGDRegression(self):
         nrows, ncols = 500, 5
         X = np.random.rand(nrows, ncols)
@@ -32,6 +39,11 @@ class TestSGDClassificationLearner(unittest.TestCase):
     def setUpClass(cls):
         cls.iris = Table('iris')
 
+    def setUp(self):
+        # Convergence warnings are irrelevant for these tests
+        warnings.filterwarnings("ignore", ".*", ConvergenceWarning)
+        super().setUp()
+
     def test_SGDClassification(self):
         sgd = SGDClassificationLearner()
         res = CrossValidation(self.iris, [sgd], k=3)

Orange/tests/test_svm.py

@@ -2,8 +2,10 @@
 # pylint: disable=missing-docstring
 
 import unittest
+import warnings
 
 import numpy as np
+from sklearn.exceptions import ConvergenceWarning
 
 from Orange.classification import (SVMLearner, LinearSVMLearner,
                                    NuSVMLearner, OneClassSVMLearner)
@@ -24,6 +26,8 @@ def test_SVM(self):
         self.assertGreater(CA(res)[0], 0.9)
 
     def test_LinearSVM(self):
+        # This warning is irrelevant here
+        warnings.filterwarnings("ignore", ".*", ConvergenceWarning)
         learn = LinearSVMLearner()
         res = CrossValidation(self.data, [learn], k=2)
         self.assertGreater(CA(res)[0], 0.8)

Orange/widgets/data/owdatasampler.py

@@ -378,7 +378,7 @@ def __init__(self, n=0, stratified=False, replace=False,
     def __call__(self, table):
         if self.replace:
             rgen = np.random.RandomState(self.random_state)
-            sample = rgen.random_integers(0, len(table) - 1, self.n)
+            sample = rgen.randint(0, len(table), self.n)
             o = np.ones(len(table))
             o[sample] = 0
             others = np.nonzero(o)[0]

Orange/widgets/data/owfeaturestatistics.py

@@ -556,9 +556,9 @@ def data(self, index, role):
         elif output in (np.inf, -np.inf):
             output = '%s∞' % ['', '-'][output < 0]
         elif isinstance(output, int):
-            output = locale.format('%d', output, grouping=True)
+            output = locale.format_string('%d', output, grouping=True)
         elif isinstance(output, float):
-            output = locale.format('%.2f', output, grouping=True)
+            output = locale.format_string('%.2f', output, grouping=True)
 
         return output
 

Orange/widgets/data/owfile.py

@@ -201,7 +201,7 @@ def __init__(self):
         url_combo.activated.connect(self._url_set)
 
         box = gui.vBox(self.controlArea, "Info")
-        self.info = gui.widgetLabel(box, 'No data loaded.')
+        self.infolabel = gui.widgetLabel(box, 'No data loaded.')
         self.warnings = gui.widgetLabel(box, '')
 
         box = gui.widgetBox(self.controlArea, "Columns (Double click to edit)")
@@ -304,7 +304,7 @@ def load_data(self):
             self.data = None
             self.sheet_box.hide()
             self.Outputs.data.send(None)
-            self.info.setText("No data.")
+            self.infolabel.setText("No data.")
 
     def _try_load(self):
         # pylint: disable=broad-except
@@ -335,7 +335,7 @@ def _try_load(self):
             if warnings:
                 self.Warning.load_warning(warnings[-1].message.args[0])
 
-        self.info.setText(self._describe(data))
+        self.infolabel.setText(self._describe(data))
 
         self.loaded_file = self.last_path()
         add_origin(data, self.loaded_file)

Orange/widgets/data/owpaintdata.py

@@ -795,7 +795,7 @@ def __init__(self):
         self.class_model.rowsInserted.connect(self._class_count_changed)
         self.class_model.rowsRemoved.connect(self._class_count_changed)
 
-        if not self.data:
+        if self.data is None or not len(self.data):
             self.data = []
             self.__buffer = np.zeros((0, 3))
         elif isinstance(self.data, np.ndarray):

Orange/widgets/data/owrank.py

@@ -5,7 +5,7 @@
 Rank (score) features for prediction.
 
 """
-
+import warnings
 from collections import namedtuple, OrderedDict
 import logging
 from functools import partial
@@ -143,7 +143,10 @@ def headerData(self, section, orientation, role=Qt.DisplayRole):
     def setExtremesFrom(self, column, values):
         """Set extremes for columnn's ratio bars from values"""
         try:
-            vmin = np.nanmin(values)
+            with warnings.catch_warnings():
+                warnings.filterwarnings(
+                    "ignore", ".*All-NaN slice encountered.*", RuntimeWarning)
+                vmin = np.nanmin(values)
             if np.isnan(vmin):
                 raise TypeError
         except TypeError:
@@ -368,31 +371,30 @@ def set_learner(self, scorer, id):
 
     @memoize_method()
     def get_method_scores(self, method):
+        # These errors often happen, but they result in nans, which
+        # are handled correctly by the widget
         estimator = method.scorer()
         data = self.data
         try:
             scores = np.asarray(estimator(data))
         except ValueError:
-            log.warning("Scorer %s wasn't able to compute all scores at once",
-                        method.name)
             try:
                 scores = np.array([estimator(data, attr)
                                    for attr in data.domain.attributes])
             except ValueError:
-                log.error(
-                    "Scorer %s wasn't able to compute scores at all",
-                    method.name)
+                log.error("%s doesn't work on this data", method.name)
                 scores = np.full(len(data.domain.attributes), np.nan)
+            else:
+                log.warning("%s had to be computed separately for each "
+                            "variable", method.name)
         return scores
 
     @memoize_method()
     def get_scorer_scores(self, scorer):
         try:
             scores = scorer.scorer.score_data(self.data).T
         except ValueError:
-            log.error(
-                "Scorer %s wasn't able to compute scores at all",
-                scorer.name)
+            log.error("%s doesn't work on this data", scorer.name)
             scores = np.full((len(self.data.domain.attributes), 1), np.nan)
 
         labels = ((scorer.shortname,)

Orange/widgets/data/tests/test_owconcatenate.py

@@ -79,12 +79,12 @@ def get_source():
         # test adding source
         self.widget.controls.append_source_column.toggle()
         source = get_source()
-        self.assertEquals(source.name, "Source ID")
+        self.assertEqual(source.name, "Source ID")
         # test name changing
         self.widget.controls.source_attr_name.setText("Source")
         self.widget.controls.source_attr_name.callback()
         source = get_source()
-        self.assertEquals(source.name, "Source")
+        self.assertEqual(source.name, "Source")
         # test source_column role
         places = ["class_vars", "attributes", "metas"]
         for i, place in enumerate(places):

Orange/widgets/data/tests/test_owdatasampler.py

@@ -64,7 +64,7 @@ def select_sampling_type(self, sampling_type):
     def test_no_intersection_in_outputs(self):
         """ Check whether outputs intersect and whether length of outputs sums
         to length of original data """
-        self.send_signal("Data", self.zoo)
+        self.send_signal("Data", self.iris)
         w = self.widget
         sampling_types = [w.FixedProportion, w.FixedSize, w.CrossValidation]
 
@@ -78,7 +78,7 @@ def test_no_intersection_in_outputs(self):
 
                     sample = self.get_output("Data Sample")
                     other = self.get_output("Remaining Data")
-                    self.assertEqual(len(self.zoo), len(sample) + len(other))
+                    self.assertEqual(len(self.iris), len(sample) + len(other))
                     self.assertNoIntersection(sample, other)
 
     def test_bigger_size_with_replacement(self):
@@ -114,5 +114,4 @@ def set_fixed_sample_size(self, sample_size, with_replacement=False):
         return self.widget.sampleSizeSpin.value()
 
     def assertNoIntersection(self, sample, other):
-        for inst in sample:
-            self.assertNotIn(inst, other)
+        self.assertFalse(bool(set(sample.ids) & set(other.ids)))

Orange/widgets/data/tests/test_owfeaturestatistics.py

@@ -1,4 +1,5 @@
 import datetime
+import warnings
 from collections import namedtuple
 from functools import wraps, partial
 from itertools import chain
@@ -17,7 +18,6 @@
 
 VarDataPair = namedtuple('VarDataPair', ['variable', 'data'])
 
-
 # Continuous variable variations
 continuous_full = VarDataPair(
     ContinuousVariable('continuous_full'),
@@ -192,6 +192,9 @@ def setUp(self):
         self.widget = self.create_widget(
             OWFeatureStatistics, stored_settings={'auto_commit': False}
         )
+        # scipy.sparse uses matrix; this filter can be removed when it stops
+        warnings.filterwarnings(
+            "ignore", ".*the matrix subclass.*", PendingDeprecationWarning)
 
     def force_render_table(self):
         """Some fields e.g. histograms are only initialized when they actually

Orange/widgets/data/tests/test_owfile.py

@@ -170,7 +170,7 @@ def test_file_not_found(self):
         self.assertEqual(file_name, path.basename(self.widget.last_path()))
         self.assertTrue(self.widget.Error.file_not_found.is_shown())
         self.assertIsNone(self.get_output(self.widget.Outputs.data))
-        self.assertEqual(self.widget.info.text(), "No data.")
+        self.assertEqual(self.widget.infolabel.text(), "No data.")
 
         # Open a sample dataset
         self.open_dataset("iris")
@@ -285,8 +285,10 @@ def test_with_warnings(self):
         self.assertTrue(self.widget.Warning.load_warning.is_shown())
 
     def test_fail(self):
-        with named_file("name\nc\n\nstring", suffix=".tab") as fn:
+        with named_file("name\nc\n\nstring", suffix=".tab") as fn, \
+                patch("Orange.widgets.data.owfile.log.exception") as log:
             self.open_dataset(fn)
+            log.assert_called()
         self.assertTrue(self.widget.Error.unknown.is_shown())
 
     def test_read_format(self):