from collections import OrderedDict

import numpy as np
from AnyQt.QtCore import Qt

from Orange.classification.rules import (
    WeightedRelativeAccuracyEvaluator, LaplaceAccuracyEvaluator,
    EntropyEvaluator, _RuleClassifier, _RuleLearner, get_dist
)
from Orange.data import Table
from Orange.widgets import gui
from Orange.widgets.settings import Setting
from Orange.widgets.utils.owlearnerwidget import OWBaseLearner
from Orange.widgets.utils.widgetpreview import WidgetPreview


class CustomRuleClassifier(_RuleClassifier):
    """
    Custom rule induction classifier. Instances are classifier following
    either an unordered set of rules or a decision list.
    """
    def __init__(self, domain, rule_list, params):
        super().__init__(domain, rule_list)
        assert params is not None

        self.rule_ordering = params["Rule ordering"]
        self.covering_algorithm = params["Covering algorithm"]
        self.params = params

    def predict(self, X):
        if (self.rule_ordering == "ordered" and
                self.covering_algorithm == "exclusive"):
            return self.ordered_predict(X)

        if (self.rule_ordering == "unordered" or
                self.covering_algorithm == "weighted"):
            return self.unordered_predict(X)


class CustomRuleLearner(_RuleLearner):
    """
    Custom CN2 inducer that construct either a list of ordered rules or
    a set of unordered rules. Returns a CustomRuleClassifier if called
    with data.

    See Also
    --------
    For more information about function calls and the algorithm, refer
    to the base rule induction learner.
    """
    name = 'Custom rule inducer'
    __returns__ = CustomRuleClassifier

    def __init__(self, preprocessors, base_rules, params):
        super().__init__(preprocessors, base_rules)
        self.progress_advance_callback = None
        assert params is not None
        self.params = params

        # top-level control procedure (rule ordering)
        self.rule_ordering = params["Rule ordering"]

        # top-level control procedure (covering algorithm)
        self.covering_algorithm = params["Covering algorithm"]
        if self.covering_algorithm == "exclusive":
            self.cover_and_remove = self.exclusive_cover_and_remove
        elif self.covering_algorithm == "weighted":
            self.gamma = params["Gamma"]
            self.cover_and_remove = self.weighted_cover_and_remove

        # bottom-level search procedure (search algorithm)
        self.rule_finder.search_algorithm.beam_width = params["Beam width"]

        # bottom-level search procedure (search strategy)
        self.rule_finder.search_strategy.constrain_continuous = True
        self.rule_finder.search_strategy.restrict_equality = params["Restrict to equality"]

        # bottom-level search procedure (search heuristics)
        evaluation_measure = params["Evaluation measure"]
        if evaluation_measure == "entropy":
            evaluator = EntropyEvaluator()
        elif evaluation_measure == "laplace":
            evaluator = LaplaceAccuracyEvaluator()
        elif evaluation_measure == "wracc":
            evaluator = WeightedRelativeAccuracyEvaluator()
        self.rule_finder.quality_evaluator = evaluator

        # bottom-level search procedure (over-fitting avoidance heuristics)
        min_rule_cov = params["Minimum rule coverage"]
        max_rule_length = params["Maximum rule length"]
        self.rule_finder.general_validator.min_covered_examples = min_rule_cov
        self.rule_finder.general_validator.max_rule_length = max_rule_length

        # bottom-level search procedure (over-fitting avoidance heuristics)
        default_alpha = params["Default alpha"]
        parent_alpha = params["Parent alpha"]
        self.rule_finder.significance_validator.default_alpha = default_alpha
        self.rule_finder.significance_validator.parent_alpha = parent_alpha

    def set_progress_advance_callback(self, f):
        """
        Assign callback to update the corresponding widget's progress
        bar after each generated rule. Callback is used to ensure that
        the progress bar is always accessed correctly (additional
        widgets may however use the generated learner).
        """
        self.progress_advance_callback = f

    def clear_progress_advance_callback(self):
        """
        Make sure to clear the callback function immediately after the
        classifier is trained.
        """
        self.progress_advance_callback = None

    def find_rules_and_measure_progress(self, X, Y, W, target_class,
                                        base_rules, domain, progress_amount):
        """
        The top-level control procedure of the separate-and-conquer
        algorithm. For given data and target class (may be None), return
        a list of rules which all must strictly adhere to the
        requirements of rule finder's validators.

        To induce decision lists (ordered rules), set target class to
        None. To induce rule sets (unordered rules), learn rules for
        each class individually, in regard to the original learning
        data.

        Parameters
        ----------
        X, Y, W : ndarray
            Learning data.
        target_class : int
            Index of the class to model.
        base_rules : list of Rule
            An optional list of initial rules to constrain the search.
        domain : Orange.data.domain.Domain
            Data domain, used to calculate class distributions.
        progress_amount: int, percentage
            Part of the learning algorithm covered by this function
            call.

        Returns
        -------
        rule_list : list of Rule
            Induced rules.
        """
        initial_class_dist = get_dist(Y, W, domain)
        rule_list = []

        # while data allows, continuously find new rules,
        # break the loop if min. requirements cannot be met,
        # after finding a rule, remove the instances covered
        while not self.data_stopping(X, Y, W, target_class, domain):

            # remember the distribution to correctly update progress
            temp_class_dist = get_dist(Y, W, domain)

            # generate a new rule that has not been seen before
            new_rule = self.rule_finder(X, Y, W, target_class, base_rules,
                                        domain, initial_class_dist, rule_list)

            # None when no new, unique rules that pass
            # the general requirements can be found
            if new_rule is None or self.rule_stopping(new_rule):
                break

            # exclusive or weighted
            X, Y, W = self.cover_and_remove(X, Y, W, new_rule)
            rule_list.append(new_rule)

            # update progress
            if self.progress_advance_callback is not None:
                progress = (((temp_class_dist[target_class] -
                              get_dist(Y, W, domain)[target_class])
                             / initial_class_dist[target_class]
                             * progress_amount) if target_class is not None else
                            ((temp_class_dist - get_dist(Y, W, domain)).sum()
                             / initial_class_dist.sum() * progress_amount))
                self.progress_advance_callback(progress)

        return rule_list

    def fit_storage(self, data):
        rule_list = []
        X, Y, W = data.X, data.Y, data.W if data.has_weights() else None
        Y = Y.astype(dtype=int)
        if self.rule_ordering == "ordered":
            rule_list = self.find_rules_and_measure_progress(
                X, Y, np.copy(W) if W is not None else None, None,
                self.base_rules, data.domain, progress_amount=1)
            # add the default rule, if required
            if (not rule_list or rule_list and rule_list[-1].length > 0 or
                    self.covering_algorithm == "weighted"):
                rule_list.append(
                    self.generate_default_rule(X, Y, W, data.domain))

        elif self.rule_ordering == "unordered":
            for curr_class in range(len(data.domain.class_var.values)):
                rule_list.extend(self.find_rules_and_measure_progress(
                    X, Y, np.copy(W) if W is not None else None,
                    curr_class, self.base_rules, data.domain,
                    progress_amount=1/len(data.domain.class_var.values)))
            # add the default rule
            rule_list.append(self.generate_default_rule(X, Y, W, data.domain))

        return CustomRuleClassifier(domain=data.domain, rule_list=rule_list,
                                    params=self.params)


class OWRuleLearner(OWBaseLearner):
    name = "CN2 Rule Induction"
    description = "Induce rules from data using CN2 algorithm."
    icon = "icons/CN2RuleInduction.svg"
    replaces = [
        "Orange.widgets.classify.owrules.OWRuleLearner",
    ]
    priority = 19
    keywords = "cn2 rule induction"

    LEARNER = CustomRuleLearner
    supports_sparse = False

    storage_orders = ["ordered", "unordered"]
    storage_covers = ["exclusive", "weighted"]
    storage_measures = ["entropy", "laplace", "wracc"]

    # default parameter values
    rule_ordering = Setting(0)
    covering_algorithm = Setting(0)
    gamma = Setting(0.7)
    evaluation_measure = Setting(0)
    restrict_equality = Setting(False)
    beam_width = Setting(5)
    min_covered_examples = Setting(1)
    max_rule_length = Setting(5)
    default_alpha = Setting(1.0)
    parent_alpha = Setting(1.0)
    checked_default_alpha = Setting(False)
    checked_parent_alpha = Setting(False)

    # actual widget elements
    base_rules = None
    gamma_spin = None

    def add_main_layout(self):
        # top-level control procedure
        top_box = gui.hBox(widget=self.controlArea, box=None)

        rule_ordering_box = gui.hBox(widget=top_box, box="Rule ordering")
        rule_ordering_rbs = gui.radioButtons(
            widget=rule_ordering_box, master=self, value="rule_ordering",
            callback=self.settings_changed, btnLabels=("Ordered", "Unordered"))
        rule_ordering_rbs.layout().setSpacing(7)

        covering_algorithm_box = gui.hBox(
            widget=top_box, box="Covering algorithm")
        covering_algorithm_rbs = gui.radioButtons(
            widget=covering_algorithm_box, master=self,
            value="covering_algorithm",
            callback=self.settings_changed,
            btnLabels=("Exclusive", "Weighted"))
        covering_algorithm_rbs.layout().setSpacing(7)

        insert_gamma_box = gui.vBox(widget=covering_algorithm_box, box=None)
        gui.separator(insert_gamma_box, 0, 14)
        self.gamma_spin = gui.doubleSpin(
            widget=insert_gamma_box, master=self, value="gamma", minv=0.0,
            maxv=1.0, step=0.01, label="γ:", orientation=Qt.Horizontal,
            callback=self.settings_changed, alignment=Qt.AlignRight,
            enabled=self.covering_algorithm == 1)

        # bottom-level search procedure (search bias)
        middle_box = gui.vBox(widget=self.controlArea, box="Rule search")

        gui.comboBox(
            widget=middle_box, master=self, value="evaluation_measure",
            label="Evaluation measure:", orientation=Qt.Horizontal,
            items=("Entropy", "Laplace accuracy", "WRAcc"),
            callback=self.settings_changed, contentsLength=3)

        gui.spin(
            widget=middle_box, master=self, value="beam_width", minv=1,
            maxv=100, step=1, label="Beam width:", orientation=Qt.Horizontal,
            callback=self.settings_changed, alignment=Qt.AlignRight,
            controlWidth=80)

        # bottom-level search procedure (over-fitting avoidance bias)
        bottom_box = gui.vBox(widget=self.controlArea, box="Rule filtering")

        gui.spin(
            widget=bottom_box, master=self, value="min_covered_examples", minv=1,
            maxv=10000, step=1, label="Minimum rule coverage:",
            orientation=Qt.Horizontal, callback=self.settings_changed,
            alignment=Qt.AlignRight, controlWidth=80)

        gui.spin(
            widget=bottom_box, master=self, value="max_rule_length",
            minv=1, maxv=100, step=1, label="Maximum rule length:",
            orientation=Qt.Horizontal, callback=self.settings_changed,
            alignment=Qt.AlignRight, controlWidth=80)

        gui.doubleSpin(
            widget=bottom_box, master=self, value="default_alpha", minv=0.0,
            maxv=1.0, step=0.01, label="Statistical significance (default α):",
            orientation=Qt.Horizontal, callback=self.settings_changed,
            alignment=Qt.AlignRight, controlWidth=80,
            checked="checked_default_alpha")

        gui.doubleSpin(
            widget=bottom_box, master=self, value="parent_alpha", minv=0.0,
            maxv=1.0, step=0.01, label="Relative significance (parent α):",
            orientation=Qt.Horizontal, callback=self.settings_changed,
            alignment=Qt.AlignRight, controlWidth=80,
            checked="checked_parent_alpha")

        gui.checkBox(
            widget=bottom_box, master=self, value="restrict_equality",
            label="Restrict operator for categorical values to equality",
            callback=self.settings_changed,
        )

    def settings_changed(self, *args, **kwargs):
        self.gamma_spin.setDisabled(self.covering_algorithm == 0)
        super().settings_changed(*args, **kwargs)

    def update_model(self):
        """
        Reimplemented from OWBaseLearner.
        """
        self.Error.out_of_memory.clear()
        self.model = None
        if self.check_data():
            try:
                self.model = self.learner(self.data)
            except MemoryError:
                self.Error.out_of_memory()
            else:
                self.model.name = self.effective_learner_name()
                self.model.instances = self.data
                self.valid_data = True
        self.Outputs.model.send(self.model)

    def create_learner(self):
        """
        Reimplemented from OWBaseLearner.
        """
        return self.LEARNER(
            preprocessors=self.preprocessors,
            base_rules=self.base_rules,
            params=self.get_learner_parameters()
        )

    def get_learner_parameters(self):
        return OrderedDict([
            ("Rule ordering", self.storage_orders[self.rule_ordering]),
            ("Covering algorithm", self.storage_covers[self.covering_algorithm]),
            ("Gamma", self.gamma),
            ("Evaluation measure", self.storage_measures[self.evaluation_measure]),
            ("Restrict to equality", self.restrict_equality),
            ("Beam width", self.beam_width),
            ("Minimum rule coverage", self.min_covered_examples),
            ("Maximum rule length", self.max_rule_length),
            ("Default alpha", (1.0 if not self.checked_default_alpha
                               else self.default_alpha)),
            ("Parent alpha", (1.0 if not self.checked_parent_alpha
                              else self.parent_alpha))
        ])


if __name__ == "__main__":  # pragma: no cover
    WidgetPreview(OWRuleLearner).run(Table("iris"))