def query_processing_algo4(u, m, cell_u_position_info, query_obj, learned_dict_start, learned_dict_arrival):
"""
Algorithm 4 :: Query processing using two-level friend list
:param u: userID
:param m: the number of seats of a vehicle
:param cell_u_position_info: client information
:param query_obj: query system(DB system)
:return:
"""
#IO_count
IO_count=0
#저번 epoch 흔적 지우기
cell_u_position_info.VH = set()
cell_u_position_info.CS = set()
p = query_obj.SELECT_position_WHERE_userid(u)
t = query_obj.SELECT_time_WHERE_userid(u)
# Line 3
best_cost=100000
# new exit condition
group_generating_flag=False
top_m_minus_1 = list()
top_m_minus_1_mc = list()
threshold_cost = 10000
threshold = 0
first_vertex = True
candidate_cost = 0
# defining necessary minheap
cell_u_position_info.VH=tree_algorithm.MinHeap_moving_cost(p, query_obj)
candidate_list=tree_algorithm.MinHeap_moving_cost(p, query_obj)
# computing group obj
# group 연산에 대한 메소드를 가진 오브젝트
compute_group_obj = compute_group_algo4.ComputeGroup_algo4(m, p, query_obj)
optimal_group=set()
#셀들 전부 다 넣고 시작
DH, AH = cell_u_position_info.find_cover_cell(u, query_obj, learned_dict_start, learned_dict_arrival, 0.01) # return (SH, EH)
print("[simulator] complete to make cell heap")
start_time = time.time()
# Line 5
if True: #originally this statement is for the number of friend inspection
# Line 6 : 친구들이 커버하고 있는 셀 찾기. 여기를 고쳐야한다. 1. 모든 셀 다 넣어버리던가
# Line 7
while not DH.is_empty() or not AH.is_empty():
# Line 8
if not DH.is_empty():
cd=DH.delete()
# Line 9
if not AH.is_empty():
ca=AH.delete()
# Line 10
if group_generating_flag:
if best_cost - candidate_cost < DH.mindist_user_and_cell(cd) + AH.mindist_user_and_cell(ca): #>>overhead
# Line 11
break
IO_flag = cell_u_position_info.search_cell_data_algo4(u, query_obj, cd, t, True) # it is from start cell
if IO_flag:
IO_count=IO_count+1
# Line 13
IO_flag = cell_u_position_info.search_cell_data_algo4(u, query_obj, ca, t, False) # it is from end cell
if IO_flag:
IO_count=IO_count+1
# Line 14
while not cell_u_position_info.is_VH_empty():
# Line 15
selected_vertex=cell_u_position_info.VH.delete()
# Line 16
selected_vertex_position = query_obj.SELECT_position_WHERE_userid(selected_vertex)
selected_vertex_mc = cell_u_position_info.VH.moving_cost(selected_vertex_position)
if m!=2:
if len(top_m_minus_1)<m-2:
top_m_minus_1.append(selected_vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
if first_vertex:
first_vertex = False
threshold = selected_vertex
threshold_cost = selected_vertex_mc
if threshold_cost < selected_vertex_mc:
threshold = selected_vertex
threshold_cost = selected_vertex_mc
else: #m-1이 되면!
if threshold_cost > selected_vertex_mc:
top_m_minus_1.remove(threshold)
top_m_minus_1_mc.remove(threshold_cost)
top_m_minus_1.append(selected_vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
candidate_cost = compute_group_obj.group_moving_cost(set(top_m_minus_1))
#threshold를 정하는 반복문
max = top_m_minus_1_mc[0]
max_index = 0
for index, mc in enumerate(top_m_minus_1_mc):
if max < mc:
max = mc
max_index = index
threshold_cost = top_m_minus_1_mc[max_index]
threshold = top_m_minus_1[max_index]
if group_generating_flag:
if best_cost - candidate_cost < selected_vertex_mc:
# Line 17
break
# Line 18
if selected_vertex not in candidate_list.heap:
computed_group=compute_group_obj.find_min_mc_group(selected_vertex, candidate_list, m)
# Line 19
if len(computed_group)+1 >= m: #user도 포함해서 계산해야한다.
group_moving_cost = compute_group_obj.group_moving_cost(computed_group)
if best_cost > group_moving_cost:
# Line 24
optimal_group=computed_group.copy()
# Line 25
best_cost=group_moving_cost
# Line 26
group_generating_flag = True
result_time=time.time()-start_time
print(candidate_cost, best_cost)
return (optimal_group.union({u}), result_time, IO_count)
def query_processing_algo4(u, l, m, cell_u_position_info, query_obj, friend_obj):
"""
Algorithm 4 :: Query processing using two-level friend list
:param u: userID
:param l: an acceptable social boundary
:param m: the number of seats of a vehicle
:return:
"""
#IO_count
IO_count=0
#저번 epoch 흔적 지우기
cell_u_position_info.VH = set()
cell_u_position_info.CS = set()
p = query_obj.SELECT_position_WHERE_userid(u)
t = query_obj.SELECT_time_WHERE_userid(u)
# Line 3
best_cost=10000
# new exit condition
group_generating_flag=False
top_m_minus_1 = list()
top_m_minus_1_mc = list()
threshold_cost = 10000
threshold = 0
first_vertex = True
candidate_cost = 0
# defining necessary minheap
cell_u_position_info.VH=tree_algorithm.MinHeap_moving_cost(p, query_obj)
candidate_list=tree_algorithm.MinHeap_moving_cost(p, query_obj)
# computing group obj
# group 연산에 대한 메소드를 가진 오브젝트
compute_group_obj = compute_group_algo4.ComputeGroup_algo4(m, friend_obj, p, query_obj)
optimal_group=set()
start_time = time.time()
# Line 4
user_friend_list=friend_obj.friend_list_with_SCF(u, l)[0] #이거 return 값 업데이트 되었다.
# Line 5
if m<=len(user_friend_list.union({u})):
# print("length of user list: "+str(len(user_friend_list)))
# Line 6 : 친구들이 커버하고 있는 셀 찾기. >> overhead
start_time1=time.time()
DH, AH = cell_u_position_info.find_cover_cell(u, user_friend_list, query_obj) # return (SH, EH)
# print("finding cover cell: "+str(time.time()-start_time1))
# Line 7
while not DH.is_empty() or not AH.is_empty():
# Line 8
if not DH.is_empty():
cd=DH.delete()
# Line 9
if not AH.is_empty():
ca=AH.delete()
# Line 10
if group_generating_flag:
if best_cost - candidate_cost < DH.mindist_user_and_cell(cd) + AH.mindist_user_and_cell(ca): #>>overhead
# Line 11
break
#print(DH.mindist_user_and_cell(cd) , AH.mindist_user_and_cell(ca))
IO_flag = cell_u_position_info.search_cell_data_algo4(query_obj, cd, user_friend_list, t, True) # it is from start cell
if IO_flag:
IO_count=IO_count+1
# Line 13
IO_flag = cell_u_position_info.search_cell_data_algo4(query_obj, ca, user_friend_list, t, False) # it is from end cell
if IO_flag:
IO_count=IO_count+1
# Line 14
while not cell_u_position_info.is_VH_empty():
# Line 15
selected_vertex=cell_u_position_info.VH.delete()
# Line 16
selected_vertex_position = query_obj.SELECT_position_WHERE_userid(selected_vertex)
selected_vertex_mc = cell_u_position_info.VH.moving_cost(selected_vertex_position)
#candidate 갱신이 여기서 이루어져야 한다.
if m!=2:
if len(top_m_minus_1)<m-2:
top_m_minus_1.append(selected_vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
if first_vertex:
first_vertex = False
threshold = selected_vertex
threshold_cost = selected_vertex_mc
if threshold_cost < selected_vertex_mc:
threshold = selected_vertex
threshold_cost = selected_vertex_mc
else: #m-1이 되면!
if threshold_cost > selected_vertex_mc:
top_m_minus_1.remove(threshold)
top_m_minus_1_mc.remove(threshold_cost)
top_m_minus_1.append(selected_vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
candidate_cost = compute_group_obj.group_moving_cost(set(top_m_minus_1))
#threshold를 정하는 반복문
max = top_m_minus_1_mc[0]
max_index = 0
for index, mc in enumerate(top_m_minus_1_mc):
if max < mc:
max = mc
max_index = index
threshold_cost = top_m_minus_1_mc[max_index]
threshold = top_m_minus_1[max_index]
if group_generating_flag:
if best_cost - candidate_cost < selected_vertex_mc:
# Line 17
break
# Line 18
if selected_vertex not in candidate_list.heap:
computed_group=compute_group_obj.find_min_mc_group_with_MFL(selected_vertex, candidate_list, l, m)
# Line 19
if len(computed_group)+1 >= m: #user도 포함해서 계산해야한다.
group_moving_cost = compute_group_obj.group_moving_cost(computed_group)
if best_cost > group_moving_cost:
# Line 24
optimal_group=computed_group.copy()
# Line 25
best_cost=group_moving_cost
# Line 26
group_generating_flag = True
# candiCost 갱신
# copied_heap = copy.deepcopy(candidate_list)
# top_m_minus_1.clear()
# for count in range(m-2):
# best_vertex = copied_heap.delete()
# top_m_minus_1.append(best_vertex)
# candidate_cost = compute_group_obj.group_moving_cost(set(top_m_minus_1))
result_time=time.time()-start_time
print(candidate_cost)
print(best_cost)
return (optimal_group.union({u}), result_time, IO_count)
def query_processing_algo4(u, m, cell_u_position_info, query_obj, grid_length):
"""
Algorithm 4 :: Query processing using two-level friend list
:param u: userID
:param m: the number of seats of a vehicle
:param cell_u_position_info: client information
:param query_obj: query system(DB system)
:param grid_length: grid length(only for v.naive)
:return:
"""
#count IO
IO_count = 0
best_cost = 100000
#저번 epoch 흔적 지우기
cell_u_position_info.VH = set()
cell_u_position_info.CS = set()
#candidate cost 추출을 위한 변수
group_generating_flag = False
top_m_minus_1 = list()
top_m_minus_1_mc = list()
threshold_cost = 10000
threshold = 0
first_vertex = True
candidate_cost = 0
p = query_obj.SELECT_position_WHERE_userid(u)
t = query_obj.SELECT_time_WHERE_userid(u)
cell_u_position_info.VH = tree_algorithm.MinHeap_moving_cost(p, query_obj)
# CPM 객체 생성
cpm = pseudo_pcm.CPM(u, query_obj, grid_length, cell_u_position_info)
best_group = set()
# DECLARE SET
departure_set = set()
arrival_set = set()
intermediate_set = tree_algorithm.MinHeap_moving_cost(p, query_obj)
# 그룹을 생성하는 객체 생성
compute_group_obj = compute_group_algo4.ComputeGroup_algo4(m, p, query_obj)
#start
start_time = time.time()
if True: #originally this statement is for the number of friend inspection
while True: #not DH.is_empty() and not AH.is_empty():
# CPM으로 vertex하나씩 꺼내는 코드
start_vertex = cpm.pop(True)
while start_vertex == -1 or start_vertex == u: #-1이 return되면 계속해서 탐색해야하고, user와 같은 vertex는 무시한다.
start_vertex = cpm.pop(True)
end_vertex = cpm.pop(False)
while end_vertex == -1 or end_vertex == u: #위와 같은 이유이다.
end_vertex = cpm.pop(False)
if start_vertex == -2 and end_vertex == -2: # 한 쪽에서라도 더 이상 탐색할 cell이 없다면, break한다
break
#line 8-9 : step of filtering veretx
if group_generating_flag:
if best_cost - candidate_cost < cpm.start_min_dist_cell + cpm.end_min_dist_cell:
break
# getting start-info phase
start_vertex_position = query_obj.SELECT_position_WHERE_userid(
start_vertex)
mc_start_vertex = cpm.start_vertex_minheap.moving_cost(
start_vertex_position)
start_vertex_time = query_obj.SELECT_time_WHERE_userid(
start_vertex)
dist_vertex_position = cpm.start_vertex_minheap.dist(
start_vertex_position[0], start_vertex_position[1])
if (mc_start_vertex < dist_vertex_position
) and check_time_condition(t, start_vertex_time):
departure_set.add(start_vertex)
# getting end-info phase
end_vertex_position = query_obj.SELECT_position_WHERE_userid(
end_vertex)
mc_end_vertex = cpm.start_vertex_minheap.moving_cost(
end_vertex_position)
end_vertex_time = query_obj.SELECT_time_WHERE_userid(end_vertex)
dist_vertex_position = cpm.start_vertex_minheap.dist(
end_vertex_position[0], end_vertex_position[1])
if (mc_end_vertex<dist_vertex_position)\
and check_time_condition(t, end_vertex_time):
arrival_set.add(end_vertex)
# line 12- : 그룹 검사하기
added_vertex_set = set()
added_vertex_set.add(start_vertex)
added_vertex_set.add(end_vertex)
for vertex in added_vertex_set:
selected_vertex_position = query_obj.SELECT_position_WHERE_userid(
vertex)
selected_vertex_mc = cell_u_position_info.VH.moving_cost(
selected_vertex_position)
if m != 2:
if len(top_m_minus_1) < m - 2:
top_m_minus_1.append(vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
if first_vertex:
first_vertex = False
threshold = vertex
threshold_cost = selected_vertex_mc
if threshold_cost < selected_vertex_mc:
threshold = vertex
threshold_cost = selected_vertex_mc
else: # m-1이 되면!
if threshold_cost > selected_vertex_mc:
top_m_minus_1.remove(threshold)
top_m_minus_1_mc.remove(threshold_cost)
top_m_minus_1.append(vertex)
top_m_minus_1_mc.append(selected_vertex_mc)
candidate_cost = compute_group_obj.group_moving_cost(
set(top_m_minus_1))
# threshold를 정하는 반복문
max = top_m_minus_1_mc[0]
max_index = 0
for index, mc in enumerate(top_m_minus_1_mc):
if max < mc:
max = mc
max_index = index
threshold_cost = top_m_minus_1_mc[max_index]
threshold = top_m_minus_1[max_index]
if (vertex in departure_set) or (vertex in arrival_set):
if vertex not in intermediate_set.heap: #intermediate_set에 새로운 vetex가 추가 된다면 그룹을 만들어야한다.
computed_group = compute_group_obj.find_min_mc_group(
vertex, intermediate_set,
m) #메소드 안 쪽에서 vertex를 intermediate set에 추가해준다.
if len(computed_group) + 1 >= m:
current_cost = compute_group_obj.group_moving_cost(
computed_group)
if current_cost < best_cost:
best_group = computed_group
best_cost = compute_group_obj.group_moving_cost(
best_group)
group_generating_flag = True
print(best_cost, candidate_cost)
result_time = time.time() - start_time
IO_count = cpm.IO_count
return (best_group.union({u}), result_time, IO_count)