app/soc/logic/allocations.py
author Sverre Rabbelier <srabbelier@gmail.com>
Sat, 24 Oct 2009 16:23:31 -0700
changeset 3043 187c1709756b
parent 2366 1a954e80d444
permissions -rw-r--r--
GSoC Views

#!/usr/bin/python2.5
#
# Copyright 2009 the Melange authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Slot allocation logic.
"""

__authors__ = [
  '"Sverre Rabbelier" <sverre@rabbelier.nl>',
  ]


import math


class Error(Exception):
  """Error class for the Allocation module.
  """

  pass


class Allocator(object):
  """A simple student slots allocator.

  The buildSets method is used to validate the allocation data as well as
  construct the sets that the algorithm then uses to distribute the slots.
  By separating these steps it is possible to write a different allocation
  algorithm but re-use the sets and validation logic.
  """

  # I tried to write explicit code that does not require any
  # additional comments (with the exception of the set notation for
  # the convenience of any mathematicians that happen to read this
  # piece of code ;).

  def __init__(self, orgs, popularity, max, slots,
               max_slots_per_org, min_slots_per_org, algorithm):
    """Initializes the allocator.

    Args:
      orgs: a list of all the orgs that need to be allocated
      popularity: the amount of applications per org
      max: the amount of assigned mentors per org
      slots: the total amount of available slots
      max_slots_per_org: how many slots an org should get at most
      min_slots_per_org: how many slots an org should at least get
      algorithm: the algorithm to use
    """

    self.locked_slots = {}
    self.adjusted_slots = {}
    self.adjusted_orgs = []
    self.locked_orgs = []
    self.unlocked_orgs = []
    self.unlocked_applications = []
    self.slots = slots
    self.mentors = {}
    self.max_slots_per_org = max_slots_per_org
    self.min_slots_per_org = min_slots_per_org
    self.orgs = set(orgs)
    self.popularity = None
    self.total_popularity = None
    self.initial_popularity = popularity
    self.max = max
    self.algorithm = algorithm

  def allocate(self, locked_slots):
    """Allocates the slots and returns the result.

    Args:
      locked_slots: a dict with orgs and the number of slots they get
    """

    self.locked_slots = locked_slots

    self.buildSets()

    if not sum(self.popularity.values()) or not sum(self.max.values()):
      return dict([(i, 0) for i in self.orgs])

    if self.algorithm == 1:
      return self.preprocessingAllocation()

    if self.algorithm == 2:
      return self.reliableAlgorithm()

    return self.iterativeAllocation()

  def buildSets(self):
    """Allocates slots with the specified constraints.
    """

    popularity = self.initial_popularity.copy()

    # set s
    locked_slots = self.locked_slots

    # set a and b
    locked_orgs = set(locked_slots.keys())

    # set a' and b'
    unlocked_orgs = self.orgs.difference(locked_orgs)

    # a+o and b+o should be o
    locked_orgs_or_orgs = self.orgs.union(locked_orgs)

    total_popularity = sum(popularity.values())

    # a+o should be o, testing length is enough though
    if len(locked_orgs_or_orgs) != len(self.orgs):
      raise Error("Unknown org as locked slot")

    self.unlocked_orgs = unlocked_orgs
    self.locked_orgs = locked_orgs
    self.popularity = popularity
    self.total_popularity = total_popularity

  def rangeSlots(self, slots, org):
    """Returns the amount of slots for the org within the required bounds.
    """

    slots = int(math.floor(float(slots)))
    slots = min(slots, self.max_slots_per_org)
    slots = max(slots, self.min_slots_per_org)
    slots = min(slots, self.max[org])

    return slots

  def iterativeAllocation(self):
    """A simple iterative algorithm.
    """

    adjusted_orgs = self.adjusted_orgs
    adjusted_slots = self.adjusted_slots
    locked_orgs = self.locked_orgs
    locked_slots = self.locked_slots

    unallocated_popularity = self.total_popularity - len(locked_slots)

    available_slots = self.slots
    allocations = {}

    for org in self.orgs:
      popularity = self.popularity[org]
      mentors = self.mentors[org]

      if org in locked_orgs:
        slots = locked_slots[org]
      elif unallocated_popularity:
        weight = float(popularity) / float(unallocated_popularity)
        slots = int(math.floor(weight*available_slots))

      if org in adjusted_orgs:
        slots += adjusted_slots[org]

      slots = min(slots, self.max_slots_per_org)
      slots = min(slots, mentors)
      slots = min(slots, available_slots)

      allocations[org] = slots
      available_slots -= slots
      unallocated_popularity -= popularity

    return allocations

  def preprocessingAllocation(self):
    """An algorithm that pre-processes the input before running as normal.
    """

    adjusted_orgs = self.adjusted_orgs
    adjusted_slots = self.adjusted_slots
    locked_orgs = self.locked_orgs
    locked_slots = self.locked_slots
    unlocked_orgs = self.unlocked_orgs
    total_popularity = self.total_popularity

    available_slots = self.slots
    allocations = {}
    slack = {}

    for org in locked_orgs:
      popularity = self.popularity[org]
      slots = locked_slots[org]
      slots = self.rangeSlots(slots, org)

      total_popularity -= popularity
      available_slots -= slots
      allocations[org] = slots
      del self.popularity[org]

    # adjust the orgs in need of adjusting
    for org in adjusted_orgs:
      slots = float(adjusted_slots[org])

      adjustment = (float(total_popularity)/float(available_slots))*slots
      adjustment = int(math.ceil(adjustment))
      self.popularity[org] += adjustment
      total_popularity += adjustment

    # adjust the popularity so that the invariants are always met
    for org in unlocked_orgs:
      popularity = self.popularity[org]
      # mentors = self.mentors[org]

      slots = (float(popularity)/float(total_popularity))*available_slots
      slots = self.rangeSlots(slots, org)

      popularity = (float(total_popularity)/float(available_slots))*slots

      self.popularity[org] = popularity

    total_popularity = sum(self.popularity.values())

    # do the actual calculation
    for org in unlocked_orgs:
      popularity = self.popularity[org]
      raw_slots = (float(popularity)/float(total_popularity))*available_slots
      slots = int(math.floor(raw_slots))

      slack[org] = raw_slots - slots
      allocations[org] = slots

    slots_left = available_slots - sum(allocations.values())

    # add leftover slots, sorted by slack, decending
    for org, slack in sorted(slack.iteritems(), 
        key=lambda (k, v): v, reverse=True):
      if slots_left < 1:
        break

      current = allocations[org]
      slots = self.rangeSlots(current + 1, org)

      slots_left += slots - current
      allocations[org] = slots

    return allocations

  def reliableAlgorithm(self):
    """An algorithm that reliable calculates the slots assignments.
    """

    # adjusted_orgs = self.adjusted_orgs
    # adjusted_slots = self.adjusted_slots
    locked_orgs = self.locked_orgs
    locked_slots = self.locked_slots
    unlocked_orgs = self.unlocked_orgs
    total_popularity = self.total_popularity

    available_slots = self.slots
    allocations = {}
    # slack = {}

    # take out the easy ones
    for org in locked_orgs:
      popularity = self.popularity[org]
      slots = locked_slots[org]
      slots = float(slots)
      slots = self.rangeSlots(slots, org)

      total_popularity -= popularity
      available_slots -= slots
      allocations[org] = slots
      del self.popularity[org]

    total_popularity = sum(self.popularity.values())

    # all orgs have been locked, nothing to do
    if total_popularity <= 0:
      return allocations

    pop_per_slot = float(available_slots)/float(total_popularity)

    # slack = 0
    wanted = {}

    # filter out all those that deserve more than their maximum
    for org in unlocked_orgs:
      popularity = self.popularity[org]
      raw_slots = float(popularity)*pop_per_slot
      slots = int(math.floor(raw_slots))
      slots = self.rangeSlots(slots, org)
      max = self.max[org]

      if max > slots:
        wanted[org] = max - slots

      allocations[org] = slots

    available_slots = self.slots - sum(allocations.values())

    # distribute the slack
    while available_slots > 0 and (sum(wanted.values()) > 0):
      for org, _ in wanted.iteritems():
        available_slots = self.slots - sum(allocations.values())
        if available_slots <= 0:
          break

        if wanted[org] <= 0:
          continue

        current = allocations[org]
        slots = self.rangeSlots(current + 1, org)
        extra = current - slots

        wanted[org] += extra
        allocations[org] = slots

    return allocations