from django.db import models
from common.models import Schedule, Sequence, ScheduleHasSequences
from operator import attrgetter
from decimal import *
import time

DEFAULT_PLAN_START = Decimal(2457485.250000)  # April 6th 2016, 18:00:00.0 UT
DEFAULT_PLAN_END = Decimal(2457485.916667)  # April 7th 2016, 10:00:00.0 UT

PRECISION = Decimal(0.0000000001)

SIMULATION = False

TIMESTAMP_JD = 2440587.500000
MAX_OVERHEAD = 25
MAX_OVERHEAD_JD = Decimal(MAX_OVERHEAD / (24 * 60 * 60))

REJECTED_QUOTA = "Insufficient quota"
REJECTED_ROOM = "Insufficient room for this sequence"

'''
    Note : the following functions are necessary due to a too-high precision of Decimal objects
'''


def is_nearby_equal(a: Decimal, b: Decimal, precision: Decimal=PRECISION):
    '''
        Compare the two decimal, according to the given precision
    '''
    return (True if abs(b - a) < PRECISION else False)


def is_nearby_sup_or_equal(a: Decimal, b: Decimal, precision: Decimal=PRECISION):
    '''
        Compare the two decimal, according to the given precision
    '''
    if (a > b):
        return True
    return (True if abs(b - a) < PRECISION else False)


def is_nearby_less_or_equal(a: Decimal, b: Decimal, precision: Decimal=PRECISION):
    '''
        Compare the two decimal, according to the given precision
    '''
    if (a < b):
        return True
    return (True if abs(b - a) < PRECISION else False)


class Interval:
    """
    Simple class that represents an interval of time
    Julian days should be used
    """

    def __init__(self, start, end):
        self._start = Decimal(start)
        self._end = Decimal(end)
        self.duration = Decimal(end - start)

    def __str__(self):
        print("[" + str(self.start) + " - " + str(self.end) + "]")

    def _get_start(self):
        return self._start

    def _set_start(self, start):
        if start > self._end:
            raise ValueError(
                "Cannot set start (%d): must be lower than end (%d)" % (start, self._end))
        self._start = start
        self.duration = self._end - self._start

    def _get_end(self):
        return self._end

    def _set_end(self, end):
        if end < self._start:
            raise ValueError(
                "Cannot set end (%d): must be bigger than start (%d)" % (end, self._start))
        self._end = end
        self.duration = self._end - self._start

    start = property(fget=_get_start, fset=_set_start)
    end = property(fget=_get_end, fset=_set_end)


class Scheduler():
    """
   Role : create a planning for the following/current night

   Read in DB : Sequence, PyrosUser and parents
   Create in DB : Schedule
   Update in DB : Schedule, Sequence
   Delete in DB : None

   Entry point(s) :
           - make_schedule
           - simulate_schedule
    """

    """
    TODO:
        - définition de plan_start et plan_end
        - calcul de la priorité
        - calcul des quotas
        - définir l'attribut 'flag' de Schedule
        - remplissage des espaces libres

    """

    def __init__(self):
        self.schedule = Schedule.objects.create()
        # TODO: quel est le "flag" dans le schedule ??
        self.intervals = []
        self.max_overhead = MAX_OVERHEAD_JD

    def get_night_limits(self):
        '''
        determines and set plan_start and plan_end (beginning & end of the observation night)        
        '''

        # TODO: définir comment on calcule plan_start et plan_end (via quels
        # moyens)
        self.schedule.plan_start = DEFAULT_PLAN_START  # default value
        self.schedule.plan_end = DEFAULT_PLAN_END  # default value

    def set_night_limits(self, plan_start, plan_end):
        '''
        Sets given schedule start & end (in julian day)
        '''
        self.schedule.plan_start = Decimal(plan_start)
        self.schedule.plan_end = Decimal(plan_end)

    def make_schedule(self, first_schedule):
        '''
        ENTRY POINT

        Check all 'OBSERVABLE' sequences to create the most optimized planning for the following/current night

        It is assumed that all sequences that MUST and CAN be analyse have the OBSERVABLE status        

        shs means 'ScheduleHasSequences'
        self.sequences is a list of tuples (sequence, shs)
        
        :returns : The new schedule
        
        :side-effect :
            - modify sequences status and dates in DB
        '''

        global SIMULATION
        SIMULATION = False

        if first_schedule is False:
            self.copy_from_previous_schedule()
        else:
            self.schedule.plan_night_start = self.schedule.plan_start

        self.sequences = list(Sequence.objects.filter(status=Sequence.OBSERVABLE))
        shs_list = []
        for sequence in self.sequences:
            shs_list.append(ScheduleHasSequences(sequence=sequence, schedule=self.schedule))
        self.sequences = [(sequence, shs_list[index]) for index, sequence in enumerate(self.sequences)]
        self.compute_schedule()
        self.save_schedule()
        return self.schedule

    def copy_from_previous_schedule(self):
        '''
            Copy needed information from the previous schedule :
                - gets the executed sequences from the previous schedule and copy them into the new schedule
                - gets plan_start and plan_end
                - computes new plan_restart

            shs means 'ScheduleHasSequences'
        '''

        previous_sched = Schedule.objects.order_by('-created')[1]
        previous_exc_seq = previous_sched.sequences.filter(status=Sequence.EXECUTED)
        for seq in previous_exc_seq:
            shs = seq.shs
            shs.pk = None
            shs.schedule = self.schedule
            if SIMULATION == False:
                shs.save()

        self.schedule.plan_night_start = previous_sched.plan_night_start
        self.schedule.plan_end = previous_sched.plan_end

        ''' Schedule starts in MAX_OVERHEAD seconds '''
        self.schedule.plan_start = time.time() / 86400 + TIMESTAMP_JD + self.max_overhead

    def simulate_schedule(self, sequences):
        '''
        ENTRY POINT - SIMULATION

        Do the same as make_schedule but do not touch the DB

        :type sequences : list of Sequence
        :param sequences : sequences to plan

        :returns : a tuple (Schedule, list of sequences)
        '''

        global SIMULATION
        SIMULATION = True

        self.schedule.plan_night_start = self.schedule.plan_start
        self.sequences = sequences
        shs_list = []
        for sequence in self.sequences:
            shs_list.append(ScheduleHasSequences(sequence=sequence, schedule=self.schedule))
        self.sequences = [(sequence, shs_list[index]) for index, sequence in enumerate(self.sequences)]
        self.compute_schedule()
        return (self.schedule, self.sequences)

    def compute_schedule(self):
        self.intervals.append(
            Interval(self.schedule.plan_start, self.schedule.plan_end))
        self.check_sequences_validity()
        self.determine_priorities()
        self.remove_not_eligible_sequences()
        self.sort_by_jd2_and_priorities()
        self.organize_sequences()

    def check_sequences_validity(self):
        '''
        Checks come sequence attributes to validate their integrity

        :side-effect :
            - remove invalid sequences from self.sequences
            - set INVALID status for invalid sequences in DB
        '''

        ''' Note(1) '''
        for sequence, shs in list(self.sequences):
            if sequence.jd1 < 0 or sequence.jd2 < 0 or is_nearby_less_or_equal(sequence.duration, 0) or sequence.jd2 - sequence.jd1 < sequence.duration:
                self.sequences.remove((sequence, shs))
                sequence.status = Sequence.INVALID
                if SIMULATION == False:
                    sequence.save()

    def determine_priorities(self):
        '''
        Computes sequences priority according to the user, the scientific program, ...        
        '''

        # TODO: définir comment on calcule la priorité
        pass

    def remove_not_eligible_sequences(self):
        '''
        Computes overlap between [jd1; jd2] and [plan_start; plan_end]
        Removes from self.sequences all the sequences that cannot be observed between plan_start and plan_end
        Set UNPLANNABLE sequences if jd2 < plan_start

        :side-effect :
            - remove unwanted sequences from self.sequences
        '''

        ''' Note (1) '''
        for sequence, shs in list(self.sequences):
            overlap = min(self.schedule.plan_end, sequence.jd2) - \
                max(self.schedule.plan_start, sequence.jd1) - self.max_overhead
            if overlap < sequence.duration:
                if sequence.jd1 < self.schedule.plan_start:
                    """ Note (2) """
                    sequence.status = Sequence.UNPLANNABLE
                    if SIMULATION == False:
                        sequence.save()
                self.sequences.remove((sequence, shs))

    def sort_by_jd2_and_priorities(self):
        '''
        Sort by priority and jd2, priority being the main sorting parameter        
        '''

        self.sequences.sort(key=lambda x: x[0].jd2)
        self.sequences.sort(key=lambda x: x[0].priority)

    def organize_sequences(self):
        '''
        Main function of the Scheduler
        Arrange a maximum of observable sequences in the planning

        Algorithm (for each sequence) :
            - check quota (remove sequence from list if quota is too low)
            - select matching intervals
            - IF matching intervals => place sequence according to tPrefered
            - IF NO matching intervals => try to move other sequences to place this one

        :side-effect :
            - remove unwanted sequences from self.sequences
            - change status and dates of sequences in self.sequences (but not in DB yet)
        '''

        ''' Note (1) '''
        for sequence, shs in list(self.sequences):
            quota = self.determine_quota(sequence)
            if quota < sequence.duration:
                shs.status = Sequence.REJECTED
                shs.desc = REJECTED_QUOTA
                continue

            matching_intervals = self.get_matching_intervals(sequence)
            if len(matching_intervals) > 0:
                self.place_sequence(sequence, shs, matching_intervals)
                sequence_placed = True
            else:
                sequence_placed = self.try_shifting_sequences(sequence, shs)
            if sequence_placed == True:
                shs.status = Sequence.PENDING
                self.update_quota(sequence)
            else:
                shs.status = Sequence.REJECTED
                shs.desc = REJECTED_ROOM

    def determine_quota(self, sequence: Sequence) -> float:
        '''
        Determines the quota (in minutes) according to the current planning duration and the quota of the user and scientific program associated

        :returns : The quota (float)
        '''
        # TODO: définir comment on calcule le quota

        return sequence.request.pyros_user.quota  # default value

    def get_matching_intervals(self, sequence: Sequence):
        '''
        Find the intervals where the sequence could be inserted

        :returns : list of matching Intervals
        '''

        matching_intervals = []

        for interval in self.intervals:
            overlap = min(sequence.jd2, interval.end) - \
                max(sequence.jd1, interval.start) - self.max_overhead
            if overlap > sequence.duration or is_nearby_equal(overlap, sequence.duration):
                matching_intervals.append(interval)

        return matching_intervals

    def place_sequence(self, sequence: Sequence, shs: ScheduleHasSequences, matching_intervals):
        '''
        Place the sequence in the better interval, according to the t_prefered

        :type matching_intervals: list [Interval]
        :param matching_intervals: Intervals in which the sequence can be placed

        :side-effect :
            - changes self.intervals
            - change the sequence if it it placed
        '''

        if len(matching_intervals) == 0:
            raise ValueError("matching_intervals shall not be empty")

        prefered_interval = self.get_prefered_interval(
            sequence, matching_intervals)
        sequence_position_in_interval = self.get_sequence_position_in_interval(
            sequence, prefered_interval)
        self.insert_sequence_in_interval(
            sequence, shs, prefered_interval, sequence_position_in_interval)
        self.cut_interval(sequence, shs, prefered_interval)
        self.update_other_deltas(sequence, shs, prefered_interval)

    def get_prefered_interval(self, sequence: Sequence, matching_intervals) -> Interval:
        '''
        Find the better interval, according to the t_prefered (get the nearest)

        :type matching_intervals: list [Interval]
        :param matching_intervals: Intervals in which the sequence can be placed

        :returns : An Interval that fits sequence.t_prefered at most
        '''

        if len(matching_intervals) == 0:
            raise ValueError("matching_intervals shall not be empty")

        if sequence.t_prefered == 0 or len(matching_intervals) == 1:
            prefered_interval = matching_intervals[0]
        else:
            for index, interval in enumerate(matching_intervals):
                if is_nearby_less_or_equal(interval.start, sequence.t_prefered) and is_nearby_less_or_equal(sequence.t_prefered, interval.end):
                    prefered_interval = interval
                    break
                elif sequence.t_prefered < interval.start:
                    if index == 0:
                        prefered_interval = interval
                    elif interval.start + self.max_overhead - sequence.t_prefered < sequence.t_prefered - matching_intervals[index - 1].end:
                        prefered_interval = interval
                    else:
                        prefered_interval = matching_intervals[index - 1]
                    break
        return prefered_interval

    def get_sequence_position_in_interval(self, sequence: Sequence, interval: Interval) -> str:
        '''
        Determines where the sequence will be inserted in the interval, regarding sequence.t_prefered

        :returns : A string in ["START", "END", "PREFERED"] describing where the sequence will be inserted in the interval
        '''

        if is_nearby_less_or_equal(interval.start, sequence.t_prefered) and is_nearby_less_or_equal(sequence.t_prefered, interval.end):
            if is_nearby_less_or_equal(sequence.t_prefered - Decimal(0.5) * sequence.duration, interval.start):
                position_in_interval = "START"
            elif is_nearby_sup_or_equal(sequence.t_prefered + Decimal(0.5) * sequence.duration, interval.end):
                position_in_interval = "END"
            else:
                position_in_interval = "PREFERED"
        else:
            if sequence.t_prefered < interval.start:
                position_in_interval = "START"
            else:
                position_in_interval = "END"
        return position_in_interval

    def insert_sequence_in_interval(self, sequence: Sequence, shs: ScheduleHasSequences, interval: Interval, position: str):
        '''
        Inserts the sequence in the interval:
            - sets sequence.tsp and sequence.tep
            - sets sequence.deltaTL and sequence.deltaTR

        :param interval: Interval in which the sequence will be inserted
        :param position: String describing where the sequence will be inserted in the interval

        :side-effect :
            - modify sequence attributes (tsp, tep, deltaTL, deltaTR)
        '''

        if position not in ["START", "END", "PREFERED"]:
            raise ValueError(
                "position must be either 'START', 'END' or 'PREFERED'")

        if position == "START":
            shs.tsp = max(
                interval.start + self.max_overhead, sequence.jd1)
            shs.tep = shs.tsp + sequence.duration
            shs.deltaTL = 0
            shs.deltaTR = min(
                interval.end, sequence.jd2) - shs.tep
        elif position == "END":
            shs.tep = min(interval.end, sequence.jd2)
            shs.tsp = shs.tep - sequence.duration
            shs.deltaTL = shs.tsp - \
                max(interval.start + self.max_overhead, sequence.jd1)
            shs.deltaTR = 0
        else:
            shs.tsp = max(
                sequence.jd1, sequence.t_prefered - Decimal(0.5) * sequence.duration)
            if shs.tsp - interval.start < self.max_overhead:
                shs.tsp = interval.start + self.max_overhead
            shs.tep = shs.tsp + sequence.duration
            shs.deltaTL = shs.tsp - \
                max(interval.start + self.max_overhead, sequence.jd1)
            shs.deltaTR = min(
                interval.end, sequence.jd2) - shs.tep

    def cut_interval(self, sequence: Sequence, shs: ScheduleHasSequences, interval: Interval):
        '''
        Separates the interval in two parts regarding to the sequence position
        Sorts the interval list in time order

        :param interval : the interval in which the sequence was added

        :side-effect :
            - removes interval from self.intervals
            - add created intervals to self.intervals
            - sorts self.intervals
        '''

        interval_before_sequence = Interval(
            interval.start, shs.tsp - self.max_overhead)
        interval_after_sequence = Interval(shs.tep, interval.end)

        self.intervals.remove(interval)
        if interval_before_sequence.duration > 0:
            self.intervals.append(interval_before_sequence)
        if interval_after_sequence.duration > 0:
            self.intervals.append(interval_after_sequence)
        self.intervals.sort(key=lambda interval: interval.start, reverse=False)

    def update_other_deltas(self, sequence: Sequence, shs: ScheduleHasSequences, interval: Interval):
        '''
        Update deltaTL and deltaTR of sequences planned near this sequence

        :param interval: Interval in which the sequence was added

        :side-effect :
            - modify deltaTL and deltaTR of sequences before and after the interval
        '''

        for sequence_, shs_ in self.sequences:
            if shs_.status == Sequence.PENDING:
                if is_nearby_equal(shs_.tep, interval.start):
                    sequence_before_interval, shs_b_i = sequence_, shs_
                elif is_nearby_equal(shs_.tsp - self.max_overhead, interval.end):
                    sequence_after_interval, shs_a_i = sequence_, shs_

        if 'sequence_before_interval' in locals():
            shs_b_i.deltaTR = min(
                shs.tsp - self.max_overhead, sequence_before_interval.jd2) - shs_b_i.tep
        if 'sequence_after_interval' in locals():
            shs_a_i.deltaTL = shs_a_i.tsp - self.max_overhead - \
                max(shs.tep, sequence_after_interval.jd1)

    def try_shifting_sequences(self, sequence: Sequence, shs: ScheduleHasSequences) -> bool:
        '''
        Tries to find a place in the planning for the sequence, moving the other sequences

        :returns : A boolean -> True if the sequence was placed, False otherwise

        :side-effect:
            - might change some sequences' deltaTL and/or deltaTR
        '''

        potential_intervals = self.get_potential_intervals(sequence)
        potential_intervals.sort(key=attrgetter("duration"), reverse=True)
        for interval in potential_intervals:
            ''' we get the adjacent sequences '''
            for sequence_, shs_ in self.sequences:
                if shs_.status == Sequence.PENDING:
                    if is_nearby_equal(shs_.tep, interval.start):
                        sequence_before_interval, shs_b_i = sequence_, shs_
                    elif is_nearby_equal(shs_.tsp - self.max_overhead, interval.end):
                        sequence_after_interval, shs_a_i = sequence_, shs_

            available_duration = min(
                interval.end, sequence.jd2) - max(interval.start, sequence.jd1)
            missing_duration = sequence.duration - \
                available_duration + self.max_overhead
            if 'sequence_before_interval' in locals():
                possible_move_to_left = min(
                    shs_b_i.deltaTL, interval.start - sequence.jd1)
            else:
                possible_move_to_left = 0

            if 'sequence_after_interval' in locals():
                possible_move_to_right = min(
                    shs_a_i.deltaTR, sequence.jd2 - interval.end)
            else:
                possible_move_to_right = 0

            if is_nearby_sup_or_equal(possible_move_to_left, missing_duration):
                self.move_sequence(
                    sequence_before_interval, shs_b_i, missing_duration, "LEFT")
            elif is_nearby_sup_or_equal(possible_move_to_right, missing_duration):
                self.move_sequence(
                    sequence_after_interval, shs_a_i, missing_duration, "RIGHT")
            elif is_nearby_sup_or_equal(possible_move_to_left + possible_move_to_right, missing_duration):
                self.move_sequence(
                    sequence_before_interval, shs_b_i, possible_move_to_left, "LEFT")
                self.move_sequence(
                    sequence_after_interval, shs_a_i, missing_duration - possible_move_to_left, "RIGHT")
            else:
                continue

            matching_intervals = self.get_matching_intervals(sequence)

            if len(matching_intervals) != 1:
                raise ValueError(
                    "There should be one and only one matching interval after shifting")
            self.place_sequence(sequence, shs, matching_intervals)

            return True

        return False

    def get_potential_intervals(self, sequence: Sequence):
        '''
        Find the intervals where a part of the sequence could be inserted

        :returns : list of partially-matching Intervals
        '''

        potential_intervals = []

        for interval in self.intervals:
            overlap = min(sequence.jd2, interval.end) - \
                max(sequence.jd1, interval.start) - self.max_overhead
            if overlap > 0:
                potential_intervals.append(interval)

        return potential_intervals

    def move_sequence(self, sequence: Sequence, shs: ScheduleHasSequences, time_shift: Decimal, direction: str):
        '''
        Moves the sequence in the wanted direction, decreasing its deltaTL or deltaTR.

        :param sequence: sequence to be moved
        :param time_shift: amplitude of the shift
        :param direction: "LEFT" or "RIGHT"

        :side-effect :
            - modify the sequence in self.sequences
            - changes the interval before and the interval after the sequence
        '''

        if direction not in ["LEFT", "RIGHT"]:
            raise ValueError("direction must be 'LEFT' or 'RIGHT'")
        if time_shift > (shs.deltaTL if direction == "LEFT" else shs.deltaTR):
            raise ValueError("Shift value is bigger than deltaT(R/L)")

        for interval in self.intervals:
            if is_nearby_equal(interval.end, shs.tsp - self.max_overhead):
                interval_before = interval
            elif is_nearby_equal(interval.start, shs.tep):
                interval_after = interval

        if direction == "LEFT":
            interval_before.end -= time_shift
            if "interval_after" in locals():
                interval_after.start -= time_shift
            shs.tsp -= time_shift
            shs.tep -= time_shift
            shs.deltaTL -= time_shift
            shs.deltaTR += time_shift
        else:
            if "interval_before" in locals():
                interval_before.end += time_shift
            interval_after.start += time_shift
            shs.tsp += time_shift
            shs.tep += time_shift
            shs.deltaTL += time_shift
            shs.deltaTR -= time_shift
        if "interval_before" in locals() and is_nearby_equal(interval_before.duration, 0):
            self.intervals.remove(interval_before)
        if "interval_after" in locals() and is_nearby_equal(interval_after.duration, 0):
            self.intervals.remove(interval_after)

    def update_quota(self, sequence: Sequence):
        '''
        Update the quota of the user / scientific program / whatever by substracting the sequence duration to the quotas

        :side-effect:
            - Modify User quota in DB
        '''

        # TODO: faire les vrais calculs de quota
        user = sequence.request.pyros_user
        # action par défaut qui correspond au code de self.determine_quota
        user.quota -= float(sequence.duration)

        if SIMULATION == False:
            user.save()

    def save_schedule(self):
        '''
        Final function : save in the db all modifications done to sequences, and the schedule

        :side-effect :
            - change sequences status and dates in DB
            - add a schedule in the DB
        '''

        self.schedule.save()
        for sequence, shs in self.sequences:
            shs.schedule = self.schedule
            shs.save()

    def print_schedule(self):
        '''
        ONLY FOR DEBUG

        Prints the planned sequences
        '''

        sequences = Sequence.objects.filter(
            shs__status=Sequence.PENDING).order_by('shs__tsp')

        print("---- There are %d sequence(s) planned ----" % len(sequences))

        for sequence in sequences:
            print("name: %r\t, start: %d\t, end: %d\t, duration: %d\t, deltaTL: %d\t, deltaTR: %d\t"
                  % (sequence.name, sequence.shs.tsp, sequence.shs.tep, sequence.duration, sequence.shs.deltaTL, sequence.shs.deltaTR))

        print("---- There are %d free interval(s) ----" % len(self.intervals))

        for interval in self.intervals:
            print("start: %d\t, end: %d\t" % (interval.start, interval.end))

''' Notes

(1) list(self.sequences) creates a copy in order to modify self.sequences and still iterate on it without unexpected behavior
(2) UNPLANNABLE is a definitive status meaning that the sequence will never be able to be scheduled

'''