Refactorize test_classification.py

tests/test_classification.py

@@ -22,109 +22,122 @@ def test_vqc_classical_should_return_value_error():
         QuanticVQC(quantum=False)
 
 
-def test_qsvm_init():
+def test_qsvm_init_quantum_wrong_token():
+    with pytest.raises(Exception):
+        q = QuanticSVM(quantum=True, q_account_token="INVALID")
+        q._init_quantum()
+
+
+@pytest.mark.parametrize('quantum', [False, True])
+def test_qsvm_init(quantum):
     """Test init of quantum classifiers"""
-    # if "classical" computation enable,
-    # no provider and backend should be defined
-    q = QuanticSVM(quantum=False)
-    q._init_quantum()
-    assert not q.quantum
-    assert not hasattr(q, "backend")
-    assert not hasattr(q, "provider")
-    # if "quantum" computation enabled, but no accountToken are provided,
-    # then "quantum" simulation will be enabled
-    # i.e., no remote quantum provider will be defined
-    q = QuanticSVM(quantum=True)
+
+    q = QuanticSVM(quantum=quantum)
     q._init_quantum()
-    assert q.quantum
-    assert hasattr(q, "_backend")
+    assert q.quantum == quantum
+    assert hasattr(q, "_backend") == quantum
+
+    # A provider is only assigned when running on a real quantum backend
     assert not hasattr(q, "_provider")
-    # if "quantum" computation enabled, and accountToken is provided,
-    # then real quantum backend is used
-    # this should raise a error as uncorrect API Token is passed
-    try:
-        q = QuanticSVM(labelsquantum=True, qAccountToken="Test")
-        assert False  # Should never reach this line
-    except Exception:
-        pass
 
 
-def test_qsvm_splitclasses(get_dataset):
+class BinaryTest:
+    def prepare(self, n_samples, n_features, quantum_instance, random):
+        self.n_classes = 2
+        self.n_samples = n_samples
+        self.n_features = n_features
+        self.quantum_instance = quantum_instance
+        self.random = random
+        self.class_len = n_samples // self.n_classes
+
+    def test(self, get_dataset):
+        # there is no __init__ method with pytest
+        n_samples, n_features, quantum_instance, random = self.get_params()
+        self.prepare(n_samples, n_features, quantum_instance, random)
+        self.samples, self.labels = get_dataset(self.n_samples,
+                                                self.n_features,
+                                                self.n_classes,
+                                                self.random)
+        self.additional_steps()
+        self.check()
+
+    def get_params(self):
+        raise NotImplementedError
+
+    def additional_steps(self):
+        raise NotImplementedError
+
+    def check(self):
+        raise NotImplementedError
+
+
+class TestQSVMSplitClasses(BinaryTest):
     """Test _split_classes method of quantum classifiers"""
-    q = QuanticSVM(quantum=False)
+    def get_params(self):
+        quantum_instance = QuanticSVM(quantum=False)
+        return 100, 9, quantum_instance, True
+
+    def additional_steps(self):
+        # As fit method is not called here, classes_ is not set.
+        # so we need to provide the classes ourselves.
+        self.quantum_instance.classes_ = range(0, self.n_classes)
+        self.x_class1, \
+            self.x_class0 = self.quantum_instance._split_classes(self.samples,
+                                                                 self.labels)
 
-    n_samples, n_features, n_classes = 100, 9, 2
-    samples, labels = get_dataset(n_samples, n_features, n_classes)
+    def check(self):
+        assert np.shape(self.x_class1) == (self.class_len, self.n_features)
+        assert np.shape(self.x_class0) == (self.class_len, self.n_features)
 
-    # As fit method is not called here, classes_ is not set.
-    # so we need to provide the classes ourselves.
-    q.classes_ = range(0, n_classes)
 
-    x_class1, x_class0 = q._split_classes(samples, labels)
-    class_len = n_samples // n_classes  # balanced set
-    assert np.shape(x_class1) == (class_len, n_features)
-    assert np.shape(x_class0) == (class_len, n_features)
+class BinaryFVT(BinaryTest):
+    def additional_steps(self):
+        self.quantum_instance.fit(self.samples, self.labels)
+        self.prediction = self.quantum_instance.predict(self.samples)
 
 
-def test_quantic_fvt_Classical(get_dataset):
+class TestClassicalSVM(BinaryFVT):
     """ Perform standard SVC test
     (canary test to assess pipeline correctness)
     """
-    # When quantum=False, it should use
-    # classical SVC implementation from SKlearn
-    q = QuanticSVM(quantum=False, verbose=False)
-    # We need to have different values for first and second classes
-    # in our samples or vector machine will not converge
-    n_samples, n_features, n_classes = 100, 9, 2
-    class_len = n_samples // n_classes  # balanced set
-    samples, labels = get_dataset(n_samples, n_features, n_classes,
-                                  random=False)
-
-    q.fit(samples, labels)
-    # This will autodefine testing sets
-    prediction = q.predict(samples)
-    # In this case, using SVM, predicting accuracy should be 100%
-    assert prediction[:class_len].all() == q.classes_[0]
-    assert prediction[class_len:].all() == q.classes_[1]
-
-
-def test_quantic_svm_fvt_simulated_quantum(get_dataset):
+    def get_params(self):
+        quantum_instance = QuanticSVM(quantum=False, verbose=False)
+        return 100, 9, quantum_instance, False
+
+    def check(self):
+        assert self.prediction[:self.class_len].all() == \
+               self.quantum_instance.classes_[0]
+        assert self.prediction[self.class_len:].all() == \
+               self.quantum_instance.classes_[1]
+
+
+class TestQuanticSVM(BinaryFVT):
     """Perform SVC on a simulated quantum computer.
     This test can also be run on a real computer by providing a qAccountToken
     To do so, you need to use your own token, by registering on:
     https://quantum-computing.ibm.com/
     Note that the "real quantum version" of this test may also take some time.
     """
-    # We will use a quantum simulator on the local machine
-    q = QuanticSVM(quantum=True, verbose=False)
-    # We need to have different values for target and non-target in our samples
-    # or vector machine will not converge
-    # To achieve testing in a reasonnable amount of time,
-    # we will lower the size of the feature and the number of trials
-    n_samples, n_features, n_classes = 10, 4, 2
-    class_len = n_samples // n_classes  # balanced set
-    samples, labels = get_dataset(n_samples, n_features, n_classes,
-                                  random=False)
-
-    q.fit(samples, labels)
-    prediction = q.predict(samples)
-    # In this case, using SVM, predicting accuracy should be 100%
-    assert prediction[:class_len].all() == q.classes_[0]
-    assert prediction[class_len:].all() == q.classes_[1]
-
-
-def test_quantic_vqc_fvt_simulated_quantum(get_dataset):
+    def get_params(self):
+        quantum_instance = QuanticSVM(quantum=True, verbose=False)
+        return 10, 4, quantum_instance, False
+
+    def check(self):
+        assert self.prediction[:self.class_len].all() == \
+               self.quantum_instance.classes_[0]
+        assert self.prediction[self.class_len:].all() == \
+               self.quantum_instance.classes_[1]
+
+
+class TestQuanticVQC(BinaryFVT):
     """Perform VQC on a simulated quantum computer"""
-    # We will use a quantum simulator on the local machine
-    # quantum parameter for VQC is always true
-    q = QuanticVQC(verbose=False)
-    # To achieve testing in a reasonnable amount of time,
-    # we will lower the size of the feature and the number of trials
-    n_samples, n_features, n_classes = 4, 4, 2
-    samples, labels = get_dataset(n_samples, n_features, n_classes)
-
-    q.fit(samples, labels)
-    prediction = q.predict(samples)
-    # Considering the inputs, this probably make no sense to test accuracy.
-    # Instead, we could consider this test as a canary test
-    assert len(prediction) == len(labels)
+    def get_params(self):
+        quantum_instance = QuanticVQC(verbose=False)
+        # To achieve testing in a reasonnable amount of time,
+        # we will lower the size of the feature and the number of trials
+        return 4, 4, quantum_instance, True
+
+    def check(self):
+        # Considering the inputs, this probably make no sense to test accuracy.
+        # Instead, we could consider this test as a canary test
+        assert len(self.prediction) == len(self.labels)