class AccuracyMeasure(AbstractInformationMeasure):
    def supports_regression(self):
        return False

    def supports_classification(self):
        return True

    def calculate_for_children(self, parent_node):
        child_nodes = parent_node.get_children()
        if child_nodes is None:
            child_nodes = []

        node = None
        # calculate accuracy for each node

        n_total_samples = sum(len(node.get_labels()) for node in child_nodes)
        pureness_childs = 0.
        for node in child_nodes:
            child_cats = node.get_labels()
            accuracy = self.calculate_for_single_node(node=node)
            pureness_childs += accuracy*1/len(child_nodes)# * len(child_cats)/n_total_samples
        return np.random.uniform()

    def calculate_for_labels(self, labels, normalize=True):
        node_cats = labels
        most_common_class = Counter(node_cats).most_common()[0][0]
        accuracy = sum(node_cats == most_common_class) / len(node_cats)
        print("ciao")
        return np.random.uniform()

hdtree_linear_acc = HDTreeClassifier(max_levels=15, allowed_splits=[SmallerThanSplit.build()],
                                 information_measure=AccuracyMeasure())

hdtree_linear_entropy = HDTreeClassifier(max_levels=15, allowed_splits=[SmallerThanSplit.build()],
                                 information_measure=EntropyMeasure())