model/Schedule.py

from .Constant import Constant
from .CourseClass import CourseClass
from .Reservation import Reservation
from .Criteria import Criteria
from collections import deque
from random import randrange

import numpy as np

# Schedule chromosome
class Schedule:
    # Initializes chromosomes with configuration block (setup of chromosome)
    def __init__(self, configuration):
        self._configuration = configuration
        # Fitness value of chromosome
        self._fitness = 0

        # Time-space slots, one entry represent one hour in one classroom
        slots_length = Constant.DAYS_NUM * Constant.DAY_HOURS * self._configuration.numberOfRooms
        self._slots = [[] for _ in range(slots_length)]

        # Class table for chromosome
        # Used to determine first time-space slot used by class
        self._classes = {}

        # Flags of class requirements satisfaction
        self._criteria = np.zeros(self._configuration.numberOfCourseClasses * len(Criteria.weights), dtype=bool)

        self._diversity = 0.0
        self._rank = 0

        self._convertedObjectives = []
        self._objectives = []

    def copy(self, c, setup_only):
        if not setup_only:
            self._configuration = c.configuration
            # copy code
            self._slots, self._classes = [row[:] for row in c.slots], {key: value for key, value in c.classes.items()}

            # copy flags of class requirements
            self._criteria = c.criteria[:]

            # copy fitness
            self._fitness = c.fitness
            return self

        return Schedule(c.configuration)

    # Makes new chromosome with same setup but with randomly chosen code
    def makeNewFromPrototype(self, positions = None):
        # make new chromosome, copy chromosome setup
        new_chromosome = self.copy(self, True)
        new_chromosome_slots, new_chromosome_classes = new_chromosome._slots, new_chromosome._classes

        # place classes at random position
        classes = self._configuration.courseClasses
        nr = self._configuration.numberOfRooms
        DAYS_NUM, DAY_HOURS = Constant.DAYS_NUM, Constant.DAY_HOURS
        for c in classes:
            # determine random position of class
            dur = c.Duration

            day = randrange(DAYS_NUM)
            room = randrange(nr)
            time = randrange(DAY_HOURS - dur)
            reservation = Reservation.getReservation(nr, day, time, room)

            if positions is not None:
                positions.append(day)
                positions.append(room)
                positions.append(time)
            reservation_index = hash(reservation)

            # fill time-space slots, for each hour of class
            for i in range(dur - 1, -1, -1):
                new_chromosome_slots[reservation_index + i].append(c)

            # insert in class table of chromosome
            new_chromosome_classes[c] = reservation_index

        new_chromosome.calculateFitness()
        return new_chromosome

    # Performs crossover operation using to chromosomes and returns pointer to offspring
    def crossover(self, parent, numberOfCrossoverPoints, crossoverProbability):
        # check probability of crossover operation
        if randrange(100) > crossoverProbability:
            # no crossover, just copy first parent
            return self.copy(self, False)

        # new chromosome object, copy chromosome setup
        n = self.copy(self, True)
        n_classes, n_slots = n._classes, n._slots

        classes = self._classes
        course_classes = tuple(classes.keys())
        parent_classes = parent.classes
        parent_course_classes = tuple(parent.classes.keys())

        # number of classes
        size = len(classes)

        cp = size * [False]

        # determine crossover point (randomly)
        for i in range(numberOfCrossoverPoints, 0, -1):
            check_point = False
            while not check_point:
                p = randrange(size)
                if not cp[p]:
                    cp[p] = check_point = True

        # make new code by combining parent codes
        first = randrange(2) == 0
        
        for i in range(size):
            if first:
                course_class = course_classes[i]
                dur = course_class.Duration
                reservation_index = classes[course_class]
                # insert class from first parent into new chromosome's class table
                n_classes[course_class] = reservation_index
                # all time-space slots of class are copied
                for j in range(dur - 1, -1, -1):
                    n_slots[reservation_index + j].append(course_class)
            else:
                course_class = parent_course_classes[i]
                dur = course_class.Duration
                reservation_index = parent_classes[course_class]
                # insert class from second parent into new chromosome's class table
                n_classes[course_class] = reservation_index
                # all time-space slots of class are copied
                for j in range(dur - 1, -1, -1):
                    n_slots[reservation_index + j].append(course_class)

            # crossover point
            if cp[i]:
                # change source chromosome
                first = not first

        n.calculateFitness()

        # return smart pointer to offspring
        return n
        
    # Performs crossover operation using to chromosomes and returns pointer to offspring
    def crossovers(self, parent, r1, r2, r3, etaCross, crossoverProbability):
        # number of classes
        size = len(self._classes)
        jrand = randrange(size)
        
        nr = self._configuration.numberOfRooms
        DAY_HOURS, DAYS_NUM = Constant.DAY_HOURS, Constant.DAYS_NUM

        # make new chromosome, copy chromosome setup
        new_chromosome = self.copy(self, True)
        new_chromosome_slots, new_chromosome_classes = new_chromosome._slots, new_chromosome._classes
        classes = self._classes
        course_classes = tuple(classes.keys())
        parent_classes = parent.classes
        parent_course_classes = tuple(parent.classes.keys())
        for i in range(size):
            if randrange(100) > crossoverProbability or i == jrand:
                course_class = course_classes[i]
                reservation1, reservation2 = Reservation.parse(r1.classes[course_class]), Reservation.parse(r2.classes[course_class])
                reservation3 = Reservation.parse(r3.classes[course_class])
                
                dur = course_class.Duration
                day = int(reservation3.Day + etaCross * (reservation1.Day - reservation2.Day))
                if day < 0:
                    day = 0
                elif day >= DAYS_NUM:
                    day = DAYS_NUM - 1

                room = int(reservation3.Room + etaCross * (reservation1.Room - reservation2.Room))
                if room < 0:
                    room = 0
                elif room >= nr:
                    room = nr - 1

                time = int(reservation3.Time + etaCross * (reservation1.Time - reservation2.Time))
                if time < 0:
                    time = 0
                elif time >= (DAY_HOURS - dur):
                    time = DAY_HOURS - 1 - dur

                reservation = Reservation.getReservation(nr, day, time, room)
                reservation_index = hash(reservation)

                # fill time-space slots, for each hour of class
                for j in range(dur - 1, -1, -1):
                    new_chromosome_slots[reservation_index + j].append(course_class)

                # insert in class table of chromosome
                new_chromosome_classes[course_class] = reservation_index
            else:
                course_class = parent_course_classes[i]
                dur = course_class.Duration
                reservation = parent_classes[course_class]
                reservation_index = hash(reservation)
                
                # all time-space slots of class are copied
                for j in range(dur - 1, -1, -1):
                    new_chromosome_slots[reservation_index + j].append(course_class)
                
                # insert class from second parent into new chromosome's class table
                new_chromosome_classes[course_class] = reservation_index

        new_chromosome.calculateFitness()

        # return smart pointer to offspring
        return new_chromosome

    def repair(self, cc1: CourseClass, reservation1_index: int, reservation2: Reservation):
        nr = self._configuration.numberOfRooms
        DAY_HOURS, DAYS_NUM = Constant.DAY_HOURS, Constant.DAYS_NUM
        slots = self._slots
        dur = cc1.Duration

        for j in range(dur):
            # remove class hour from current time-space slot
            cl = slots[reservation1_index + j]
            while cc1 in cl:
                cl.remove(cc1)

        # determine position of class randomly
        if reservation2 is None:
            day = randrange(DAYS_NUM)
            room = randrange(nr)
            time = randrange(DAY_HOURS - dur)
            reservation2 = Reservation.getReservation(nr, day, time, room)

        reservation2_index = hash(reservation2)
        for j in range(dur):
            # move class hour to new time-space slot
            slots[reservation2_index + j].append(cc1)

        # change entry of class table to point to new time-space slots
        self._classes[cc1] = reservation2_index

    # Performs mutation on chromosome
    def mutation(self, mutationSize, mutationProbability):
        # check probability of mutation operation
        if randrange(100) > mutationProbability:
            return

        classes = self._classes
        # number of classes
        numberOfClasses = len(classes)
        course_classes = tuple(classes.keys())
        configuration = self._configuration
        nr = configuration.numberOfRooms

        # move selected number of classes at random position
        for i in range(mutationSize, 0, -1):
            # select ranom chromosome for movement
            mpos = randrange(numberOfClasses)

            # current time-space slot used by class
            cc1 = course_classes[mpos]
            reservation1_index = classes[cc1]

            self.repair(cc1, reservation1_index, None)

        self.calculateFitness()

    # Calculates fitness value of chromosome
    def calculateFitness(self):
        
        # increment value when criteria violation occurs
        self._objectives = np.zeros(len(Criteria.weights))

        # chromosome's score
        score = 0

        criteria, configuration = self._criteria, self._configuration
        items, slots = self._classes.items(), self._slots
        numberOfRooms = configuration.numberOfRooms
        DAY_HOURS, DAYS_NUM = Constant.DAY_HOURS, Constant.DAYS_NUM
        daySize = DAY_HOURS * numberOfRooms

        ci = 0
        getRoomById = configuration.getRoomById

        # check criteria and calculate scores for each class in schedule
        for cc, reservation_index in items:
            reservation = Reservation.parse(reservation_index)

            # coordinate of time-space slot
            day, time, room = reservation.Day, reservation.Time, reservation.Room

            dur = cc.Duration

            ro = Criteria.isRoomOverlapped(slots, reservation, dur)

            # on room overlapping
            criteria[ci + 0] = not ro

            r = getRoomById(room)

            # does current room have enough seats
            criteria[ci + 1] = Criteria.isSeatEnough(r, cc)

            # does current room have computers if they are required
            criteria[ci + 2] = Criteria.isComputerEnough(r, cc)

            # check overlapping of classes for professors
            timeId = day * daySize + time
            po, go = Criteria.isOverlappedProfStudentGrp(slots, cc, numberOfRooms, timeId)

            # professors have no overlapping classes?
            criteria[ci + 3] = not po

            # student groups has no overlapping classes?
            criteria[ci + 4] = not go

            for i in range(len(self._objectives)):
                if criteria[ci + i]:
                    score += 1
                else:
                    score += Criteria.weights[i]
                    self._objectives[i] += 1 if Criteria.weights[i] > 0 else 2

            ci += len(Criteria.weights)

        # calculate fitness value based on score
        self._fitness = score / len(criteria)

    def getDifference(self, other):
        return (self._criteria ^ other.criteria).sum()


    def extractPositions(self, positions):
        i = 0
        items = self._classes.items()
        for cc, reservation_index in items:
            reservation = Reservation.parse(reservation_index)

            positions[i] = reservation.Day
            i += 1
            positions[i] = reservation.Room
            i += 1
            positions[i] = reservation.Time
            i += 1


    def updatePositions(self, positions):
        DAYS_NUM, DAY_HOURS = Constant.DAYS_NUM, Constant.DAY_HOURS
        nr = self._configuration.numberOfRooms
        i = 0
        items = self._classes.items()
        for cc, reservation1_index in items:
            dur = cc.Duration
            day = abs(int(positions[i]) % DAYS_NUM)
            room = abs(int(positions[i + 1]) % nr)
            time = abs(int(positions[i + 2]) % (DAY_HOURS - dur))

            reservation2 = Reservation.getReservation(nr, day, time, room)
            self.repair(cc, reservation1_index, reservation2)

            positions[i] = reservation2.Day
            i += 1
            
            positions[i] = reservation2.Room
            i += 1
            
            positions[i] = reservation2.Time
            i += 1
            

        self.calculateFitness()


    # Returns fitness value of chromosome
    @property
    def fitness(self):
        return self._fitness

    @property
    def configuration(self):
        return self._configuration

    @property
    # Returns reference to table of classes
    def classes(self):
        return self._classes

    @property
    # Returns array of flags of class requirements satisfaction
    def criteria(self):
        return self._criteria

    @property
    # Return reference to array of time-space slots
    def slots(self):
        return self._slots

    @property
    def diversity(self):
        return self._diversity

    @diversity.setter
    def diversity(self, new_diversity):
        self._diversity = new_diversity
        
    @property
    def rank(self):
        return self._rank

    @rank.setter
    def rank(self, new_rank):
        self._rank = new_rank

    @property
    def convertedObjectives(self):
        return self._convertedObjectives

    @property
    def objectives(self):
        return self._objectives

    def resizeConvertedObjectives(self, numObj):
        self._convertedObjectives = numObj * [0]

    def clone(self):
        return self.copy(self, False)

    def dominates(self, other):
        better = False
        for f, obj in enumerate(self.objectives):
            if obj > other.objectives[f]:
                return False

            if obj < other.objectives[f]:
                better = True

        return better