[FIX] Another color palette when too many colors needed

Orange/widgets/evaluate/owcalibrationplot.py

@@ -5,6 +5,8 @@
 """
 from collections import namedtuple
 
+import numpy as np
+
 from AnyQt.QtWidgets import QListWidget
 
 import pyqtgraph as pg
@@ -13,7 +15,6 @@
 from Orange.widgets import widget, gui, settings
 from Orange.widgets.evaluate.utils import check_results_adequacy
 from Orange.widgets.utils import colorpalette, colorbrewer
-from Orange.widgets.io import FileFormat
 from Orange.widgets.widget import Input
 from Orange.canvas import report
 
@@ -122,8 +123,10 @@ def _initialize(self, results):
             names = ["#{}".format(i + 1) for i in range(N)]
 
         self.classifier_names = names
-        self.colors = colorpalette.ColorPaletteGenerator(
-            N, colorbrewer.colorSchemes["qualitative"]["Dark2"])
+        scheme = colorbrewer.colorSchemes["qualitative"]["Dark2"]
+        if N > len(scheme):
+            scheme = colorpalette.DefaultRGBColors
+        self.colors = colorpalette.ColorPaletteGenerator(N, scheme)
 
         for i in range(N):
             item = self.classifiers_list_box.item(i)
@@ -138,17 +141,17 @@ def plot_curve(self, clf_idx, target):
 
         ytrue = self.results.actual == target
         probs = self.results.probabilities[clf_idx, :, target]
-        sortind = numpy.argsort(probs)
+        sortind = np.argsort(probs)
         probs = probs[sortind]
         ytrue = ytrue[sortind]
         if probs.size:
             xmin, xmax = probs.min(), probs.max()
-            x = numpy.linspace(xmin, xmax, 100)
+            x = np.linspace(xmin, xmax, 100)
             f = gaussian_smoother(probs, ytrue, sigma=0.15 * (xmax - xmin))
             observed = f(x)
         else:
-            x = numpy.array([])
-            observed = numpy.array([])
+            x = np.array([])
+            observed = np.array([])
 
         curve = Curve(x, observed)
         curve_item = pg.PlotDataItem(
@@ -159,13 +162,13 @@ def plot_curve(self, clf_idx, target):
         )
 
         rh = 0.025
-        rug_x = numpy.c_[probs, probs]
+        rug_x = np.c_[probs, probs]
         rug_x_true = rug_x[ytrue].ravel()
         rug_x_false = rug_x[~ytrue].ravel()
 
-        rug_y_true = numpy.ones_like(rug_x_true)
+        rug_y_true = np.ones_like(rug_x_true)
         rug_y_true[1::2] = 1 - rh
-        rug_y_false = numpy.zeros_like(rug_x_false)
+        rug_y_false = np.zeros_like(rug_x_false)
         rug_y_false[1::2] = rh
 
         rug1 = pg.PlotDataItem(
@@ -210,24 +213,21 @@ def send_report(self):
         self.report_caption(caption)
 
 
-import numpy
-
-
 def gaussian_smoother(x, y, sigma=1.0):
-    x = numpy.asarray(x)
-    y = numpy.asarray(y)
+    x = np.asarray(x)
+    y = np.asarray(y)
 
     gamma = 1. / (2 * sigma ** 2)
-    a = 1. / (sigma * numpy.sqrt(2 * numpy.pi))
+    a = 1. / (sigma * np.sqrt(2 * np.pi))
 
     if x.shape != y.shape:
         raise ValueError
 
     def smoother(xs):
-        W = a * numpy.exp(-gamma * ((xs - x) ** 2))
-        return numpy.average(y, weights=W)
+        W = a * np.exp(-gamma * ((xs - x) ** 2))
+        return np.average(y, weights=W)
 
-    return numpy.vectorize(smoother, otypes=[numpy.float])
+    return np.vectorize(smoother, otypes=[np.float])
 
 
 def main():
@@ -248,7 +248,7 @@ def main():
          LogisticRegressionLearner(penalty="l1"),
          SVMLearner(probability=True),
          NuSVMLearner(probability=True)
-         ],
+        ],
         store_data=True
     )
     results.learner_names = ["LR l2", "LR l1", "SVM", "Nu SVM"]

Orange/widgets/evaluate/owliftcurve.py

@@ -5,7 +5,7 @@
 """
 from collections import namedtuple
 
-import numpy
+import numpy as np
 import sklearn.metrics as skl_metrics
 
 from AnyQt import QtWidgets
@@ -19,7 +19,6 @@
 from Orange.widgets.evaluate.utils import check_results_adequacy
 from Orange.widgets.utils import colorpalette, colorbrewer
 from Orange.widgets.evaluate.owrocanalysis import convex_hull
-from Orange.widgets.io import FileFormat
 from Orange.widgets.widget import Input
 from Orange.canvas import report
 
@@ -37,7 +36,7 @@
 LiftCurve.is_valid = property(lambda self: self.points.is_valid)
 
 
-def LiftCurve_from_results(results, clf_index, target):
+def liftCurve_from_results(results, clf_index, target):
     x, y, thresholds = lift_curve_from_results(results, target, clf_index)
 
     points = CurvePoints(x, y, thresholds)
@@ -155,8 +154,10 @@ def _initialize(self, results):
         if names is None:
             names = ["#{}".format(i + 1) for i in range(N)]
 
-        self.colors = colorpalette.ColorPaletteGenerator(
-            N, colorbrewer.colorSchemes["qualitative"]["Dark2"])
+        scheme = colorbrewer.colorSchemes["qualitative"]["Dark2"]
+        if N > len(scheme):
+            scheme = colorpalette.DefaultRGBColors
+        self.colors = colorpalette.ColorPaletteGenerator(N, scheme)
 
         self.classifier_names = names
         self.selected_classifiers = list(range(N))
@@ -168,7 +169,7 @@ def _initialize(self, results):
 
     def plot_curves(self, target, clf_idx):
         if (target, clf_idx) not in self._curve_data:
-            curve = LiftCurve_from_results(self.results, clf_idx, target)
+            curve = liftCurve_from_results(self.results, clf_idx, target)
             color = self.colors[clf_idx]
             pen = QPen(color, 1)
             pen.setCosmetic(True)
@@ -182,7 +183,7 @@ def plot_curves(self, target, clf_idx):
             )
             hull_item = pg.PlotDataItem(
                 curve.hull[0], curve.hull[1],
-                pen=pen,  antialias=True
+                pen=pen, antialias=True
             )
             self._curve_data[target, clf_idx] = \
                 PlotCurve(curve, item, hull_item)
@@ -243,12 +244,12 @@ def lift_curve_from_results(results, target, clf_idx, subset=slice(0, -1)):
 
 
 def lift_curve(ytrue, ypred, target=1):
-    P = numpy.sum(ytrue == target)
+    P = np.sum(ytrue == target)
     N = ytrue.size - P
 
     if P == 0 or N == 0:
         # Undefined TP and FP rate
-        return numpy.array([]), numpy.array([]), numpy.array([])
+        return np.array([]), np.array([]), np.array([])
 
     fpr, tpr, thresholds = skl_metrics.roc_curve(ytrue, ypred, target)
     rpp = fpr * (N / (P + N)) + tpr * (P / (P + N))
@@ -273,7 +274,7 @@ def main():
          LogisticRegressionLearner(penalty="l1"),
          SVMLearner(probability=True),
          NuSVMLearner(probability=True)
-         ],
+        ],
         store_data=True
     )
     results.learner_names = ["LR l2", "LR l1", "SVM", "Nu SVM"]

Orange/widgets/evaluate/tests/base.py

@@ -0,0 +1,27 @@
+from Orange import classification, evaluation
+from Orange.data import Table
+
+
+class EvaluateTest:
+
+    def test_many_evaluation_results(self):
+        """
+        Now works with more than 9 evaluation results.
+        GH-2394 (ROC Analysis)
+        GH-2522 (Lift Curve, Calibration Plot)
+        """
+        data = Table("iris")
+        learners = [
+            classification.MajorityLearner(),
+            classification.LogisticRegressionLearner(),
+            classification.TreeLearner(),
+            classification.SVMLearner(),
+            classification.KNNLearner(),
+            classification.CN2Learner(),
+            classification.SGDClassificationLearner(),
+            classification.RandomForestLearner(),
+            classification.NaiveBayesLearner(),
+            classification.SGDClassificationLearner()
+        ]
+        res = evaluation.CrossValidation(data, learners, k=2, store_data=True)
+        self.send_signal("Evaluation Results", res)

Orange/widgets/evaluate/tests/test_owcalibrationplot.py

@@ -6,11 +6,12 @@
 import Orange.evaluation
 import Orange.classification
 
-from Orange.widgets.tests.base import WidgetTest
+from Orange.widgets.evaluate.tests.base import EvaluateTest
 from Orange.widgets.evaluate.owcalibrationplot import OWCalibrationPlot
+from Orange.widgets.tests.base import WidgetTest
 
 
-class TestOWCalibrationPlot(WidgetTest):
+class TestOWCalibrationPlot(WidgetTest, EvaluateTest):
     @classmethod
     def setUpClass(cls):
         super().setUpClass()

Orange/widgets/evaluate/tests/test_owliftcurve.py

@@ -6,12 +6,13 @@
 import Orange.evaluation
 import Orange.classification
 
+from Orange.widgets.evaluate.tests.base import EvaluateTest
 from Orange.widgets.tests.base import WidgetTest
 from Orange.widgets.tests.utils import simulate
 from Orange.widgets.evaluate.owliftcurve import OWLiftCurve
 
 
-class TestOWLiftCurve(WidgetTest):
+class TestOWLiftCurve(WidgetTest, EvaluateTest):
     @classmethod
     def setUpClass(cls):
         super().setUpClass()

Orange/widgets/evaluate/tests/test_owrocanalysis.py

@@ -1,13 +1,16 @@
+# pylint: disable=protected-access
+
 import unittest
 import copy
-import numpy
+import numpy as np
 
 import Orange.data
 import Orange.evaluation
 import Orange.classification
 
 from Orange.widgets.evaluate import owrocanalysis
 from Orange.widgets.evaluate.owrocanalysis import OWROCAnalysis
+from Orange.widgets.evaluate.tests.base import EvaluateTest
 from Orange.widgets.tests.base import WidgetTest
 
 
@@ -32,7 +35,7 @@ def test_ROCData_from_results(self):
 
         # fixed random seed because otherwise it could happen that data sample
         # contained only instances of two classes (and the test then fails)
-        data = data[numpy.random.RandomState(0).choice(len(data), size=20)]
+        data = data[np.random.RandomState(0).choice(len(data), size=20)]
         res = Orange.evaluation.LeaveOneOut(data, learners)
 
         for i, _ in enumerate(learners):
@@ -62,7 +65,7 @@ def test_ROCData_from_results(self):
                 self.assertFalse(rocdata.avg_threshold.is_valid)
 
 
-class TestOWROCAnalysis(WidgetTest):
+class TestOWROCAnalysis(WidgetTest, EvaluateTest):
 
     @classmethod
     def setUpClass(cls):
@@ -109,10 +112,10 @@ def test_basic(self):
     def test_empty_input(self):
         res = Orange.evaluation.Results(
             data=self.lenses[:0], nmethods=2, store_data=True)
-        res.row_indices = numpy.array([], dtype=int)
-        res.actual = numpy.array([])
-        res.predicted = numpy.zeros((2, 0))
-        res.probabilities = numpy.zeros((2, 0, 3))
+        res.row_indices = np.array([], dtype=int)
+        res.actual = np.array([])
+        res.predicted = np.zeros((2, 0))
+        res.probabilities = np.zeros((2, 0, 3))
 
         self.send_signal(self.widget.Inputs.evaluation_results, res)
         self.widget.roc_averaging = OWROCAnalysis.Merge
@@ -124,10 +127,10 @@ def test_empty_input(self):
         self.widget.roc_averaging = OWROCAnalysis.NoAveraging
         self.widget._replot()
 
-        res.row_indices = numpy.array([1], dtype=int)
-        res.actual = numpy.array([0.0])
-        res.predicted = numpy.zeros((2, 1))
-        res.probabilities = numpy.zeros((2, 1, 3))
+        res.row_indices = np.array([1], dtype=int)
+        res.actual = np.array([0.0])
+        res.predicted = np.zeros((2, 1))
+        res.probabilities = np.zeros((2, 1, 3))
 
         self.send_signal(self.widget.Inputs.evaluation_results, res)
         self.widget.roc_averaging = OWROCAnalysis.Merge
@@ -145,31 +148,11 @@ def test_nan_input(self):
         res.predicted = res.predicted.copy()
         res.probabilities = res.probabilities.copy()
 
-        res.actual[0] = numpy.nan
-        res.predicted[:, 1] = numpy.nan
-        res.probabilities[0, 1, :] = numpy.nan
+        res.actual[0] = np.nan
+        res.predicted[:, 1] = np.nan
+        res.probabilities[0, 1, :] = np.nan
 
         self.send_signal(self.widget.Inputs.evaluation_results, res)
         self.assertTrue(self.widget.Error.invalid_results.is_shown())
         self.send_signal(self.widget.Inputs.evaluation_results, None)
         self.assertFalse(self.widget.Error.invalid_results.is_shown())
-    def test_many_evaluation_results(self):
-        """
-        Now works with more than 9 evaluation results.
-        GH-2394
-        """
-        data = Orange.data.Table("iris")
-        learners = [
-            Orange.classification.MajorityLearner(),
-            Orange.classification.LogisticRegressionLearner(),
-            Orange.classification.TreeLearner(),
-            Orange.classification.SVMLearner(),
-            Orange.classification.KNNLearner(),
-            Orange.classification.CN2Learner(),
-            Orange.classification.SGDClassificationLearner(),
-            Orange.classification.RandomForestLearner(),
-            Orange.classification.NaiveBayesLearner(),
-            Orange.classification.SGDClassificationLearner()
-        ]
-        res = Orange.evaluation.CrossValidation(data, learners, k=2, store_data=True)
-        self.send_signal("Evaluation Results", res)