def sharma1998ext(precedents):
"""
Generates a AOA graph (PERT) from successors table
Algorithm sharma1998 extended
returns: pert.Pert() graph data structure
"""
pert_graph = pert.Pert()
successors = graph.reversed_prelation_table(precedents)
# Close the graph (not in sharma1998)
origin = pert_graph.nextNodeNumber()
pert_graph.add_node(origin)
dest = pert_graph.nextNodeNumber()
pert_graph.add_node(dest)
begin_act = graph.begining_activities(successors)
end_act = graph.ending_activities(successors)
begin_end_act = begin_act.intersection(end_act)
# -Creates a common node for starting activities
for act in begin_act - begin_end_act:
pert_graph.addActivity(act, origin)
# -Creates a common node for ending activities
for act in end_act - begin_end_act:
pert_graph.addActivity(act, origin=None, destination=dest)
# -Deals with begin-end activities
if begin_end_act:
act = begin_end_act.pop()
pert_graph.addActivity(act, origin, dest)
for act in begin_end_act:
o, d = pert_graph.addActivity(act, origin)
pert_graph.addActivity("seDummy", d, dest, dummy=True)
# Sharma1998 algorithm
for act in successors:
#print "Processing", act, pert_graph
#window.images.append( graph.pert2image(pert_graph) )
if not pert_graph.activityArc(act):
pert_graph.addActivity(act)
#window.images.append( graph.pert2image(pert_graph) )
a_origin, a_dest = pert_graph.activityArc(act)
#print '(', a_origin, a_dest, ')'
for pre in precedents[act]:
#print pert_graph.successors
#print pre, pre in pert_graph.inActivitiesR(graph.reversed_prelation_table(pert_graph.successors), a_origin)
if pre not in pert_graph.inActivitiesR(a_origin):
if not pert_graph.activityArc(pre):
pert_graph.addActivity(pre)
#window.images.append( graph.pert2image(pert_graph) )
pert_graph.makePrelation(pre, act)
a_origin, a_dest = pert_graph.activityArc(act)
return pert_graph.renumerar()
def cohen_sadeh(prelations):
"""
Build graph PERT using Cohen-Sadeh algorithm
Note: the original algorithm does not consider parallel activities (creates a multigraph)
prelations = {'activity': ['predecesor1', 'predecesor2'...}
return graph pert.Pert()
"""
# Adaptation to avoid multiple end nodes
successors = graph.reversed_prelation_table(prelations)
end_act = graph.ending_activities(successors)
#Step 1. Construct work table with Immediate Predecessors
Columns = namedlist.namedlist('Columns', ['pre', 'blocked', 'dummy', 'suc', 'start_node', 'end_node'])
# [0 Predecesors, 1 Blocked, 2 Dummy, 3 Successors, 4 Start node, 5 End node]
# Blocked = (False or Activity with same precedents)
work_table = {}
for act, predecessors in prelations.items():
work_table[act] = Columns(set(predecessors), False, False, None, None, None)
# print "\n--- Step 1 ---"
# __print_work_table(work_table)
#Step 2. Identify Identical Precedence Constraint of Diferent Activities
visited_pred = {}
for act, columns in work_table.items():
pred = frozenset(columns.pre)
if pred not in visited_pred:
visited_pred[pred] = act
else:
columns.blocked = visited_pred[pred]
# print "\n--- Step 2 ---"
# __print_work_table(work_table)
#Step 3. Identify Necessary Dummy Arcs
dups = set()
visited_act = set()
for columns in work_table.values():
if not columns.blocked:
for act in columns.pre:
if act in visited_act:
dups.add(act)
visited_act.add(act)
# print "\n--- Step 3.1 ---"
# print dups
#Step 3.2, 3.3 and 4. Create rows and information for Dummy Arcs
dummy_counter = collections.Counter()
for _, columns in work_table.items():
# Avoid blocked
if not columns.blocked:
predecessors = columns.pre
if len(predecessors) > 1:
for pre in list(predecessors):
if pre in dups:
predecessors.remove(pre)
dummy_name = pre + '-d' + str(dummy_counter[pre])
dummy_counter[pre] += 1
predecessors.add(dummy_name)
work_table[dummy_name] = Columns(set([pre]), False, True, None, None, None)
# print "\n--- Step 4 ---"
# __print_work_table(work_table)
#Step 5. Creating nodes
node = 0 # instead of 0, can start at 100 to avoid confusion with activities named with numbers when debugging
for act, columns in work_table.items():
if not columns.dummy and not columns.blocked:
columns.start_node = node
node += 1
# print "\n--- Step 5a ---"
# __print_work_table(work_table)
for act, columns in work_table.items():
if not columns.dummy and columns.blocked:
columns.start_node = work_table[columns.blocked].start_node
# print "\n--- Step 5b ---"
# __print_work_table(work_table)
#Step 6. Associate activities with their end nodes
# (a) find one non-dummy successor for each activity
for act, columns in work_table.items():
for suc, suc_columns in work_table.items():
if not suc_columns.dummy and not suc_columns.blocked:
if act in suc_columns.pre:
columns.suc = suc
break
# print "\n--- Step 6a ---"
# __print_work_table(work_table)
# (b) find end nodes
graph_end_node = node # Reserve one node for graph end
node += 1
for act, columns in work_table.items():
suc = columns.suc
if suc:
columns.end_node = work_table[suc].start_node
else:
# Create needed end nodes, avoiding multiple graph end nodes (adaptation)
if act in end_act:
columns.end_node = graph_end_node
else:
columns.end_node = node
node += 1
# print "\n--- Step 6b ---"
# __print_work_table(work_table)
#Step 7. Associate dummy arcs with start nodes
for act, columns in work_table.items():
if columns.dummy:
pred = iter(columns.pre).next()
start_node = work_table[pred].end_node
columns.start_node = start_node
# print "\n--- Step 7 ---"
# __print_work_table(work_table)
#Step 8. Generate the graph
pm_graph = pert.PertMultigraph()
for act, columns in work_table.items():
_, _, dummy, _, start, end = columns
pm_graph.add_arc((start, end), (act, dummy))
p_graph = pm_graph.to_directed_graph()
return p_graph.renumerar()
return data[0]
except IOError:
print 'Error reading file:', filename
sys.exit(1)
if len(sys.argv)==3:
repeticiones= int(sys.argv[2])
filename=sys.argv[1]
data = openProject(filename)
successors = {}
for i in data:
successors[i[1]]=i[2]
prelaciones1 = graph.reversed_prelation_table(successors)
prelaciones = {
'B': [],
'A': [],
'D': ['B'],
'C': [],
'F': ['C'],
'E': ['D'],
'H': ['B'],
'G': ['F'],
'J': ['F'],
'I': ['A'],
'L': ['C', 'E'],
'K': ['I'],
'N': ['B'],
'M': ['H'],
def sysloOptimal(prelations):
"""
Build a PERT graph using Syslo algorithm
return p_graph pert.PertMultigraph()
"""
# Adaptation to avoid multiple end nodes
successors = graph.reversed_prelation_table(prelations)
end_act = graph.ending_activities(successors)
#Kahn1962.check_cycles(successors)
prela = successors.copy()
Columns = namedlist.namedlist('Columns', ['pre', 'blocked', 'dummy', 'suc', 'start_node', 'end_node'])
# [0 Predecesors, 1 Blocked, 2 Dummy, 3 Successors, 4 Start node, 5 End node]
# Blocked = (False or Activity with same precedents)
#Step 0.
grafo = {}
alt = graph.successors2precedents(successors)
grafo = graph.successors2precedents(syslo_table.syslo(prela, grafo, alt))
#Step 1. Save the new prelation table in a work table
work_table = {}
for act, pre in grafo.items():
if not act in prelations:
work_table[act] = Columns(pre, False, True, None, None, None)
else:
work_table[act] = Columns(pre, False, False, None, None, None)
#Step 2. Identify Dummy Activities And Identical Precedence Constraint of Diferent Activities
visited_pred = {}
for act, columns in work_table.items():
pred = frozenset(columns.pre)
if pred not in visited_pred:
visited_pred[pred] = act
else:
columns.blocked = visited_pred[pred]
#Step 3. Creating nodes
# (a) find start nodes
node = 0 # instead of 0, can start at 100 to avoid confusion with activities named with numbers when debugging
for act, columns in work_table.items():
if not columns.blocked:
columns.start_node = node
node += 1
if columns.blocked:
columns.start_node = work_table[columns.blocked].start_node
# Associate activities with their end nodes
for suc, suc_columns in work_table.items():
if not suc_columns.blocked:
if act in suc_columns.pre:
columns.suc = suc
break
# (b) find end nodes
graph_end_node = node # Reserve one node for graph end
node += 1
for act, columns in work_table.items():
suc = columns.suc
if suc:
columns.end_node = work_table[suc].start_node
else:
# Create needed end nodes, avoiding multiple graph end nodes (adaptation)
if act in end_act:
columns.end_node = graph_end_node
else:
columns.end_node = node
node += 1
# Step 4. Remove redundancy of dummy activities
vis = []
for act, columns in work_table.items():
if columns.dummy == False:
for q in work_table[act].pre:
for w in work_table[act].pre:
if q in work_table and w in work_table:
if q != w and work_table[q].pre == work_table[w].pre and work_table[q].dummy==True and work_table[w].dummy==True:
if w not in vis:
del work_table[w]
vis.append(q)
#Step 5. Generate the graph
pm_graph = pert.PertMultigraph()
for act, columns in work_table.items():
_, _, dummy, _, start, end = columns
pm_graph.add_arc((start, end), (act, dummy))
p_graph = pm_graph.to_directed_graph()
return p_graph
return p_graph
def sysloPolynomial(prelations):
# Adaptation to avoid multiple end nodes
successors = graph.reversed_prelation_table(prelations)
end_act = graph.ending_activities(successors)
# Step 0. Construct work table with Immediate Predecessors
Columns = namedlist.namedlist("Columns", ["pre", "blocked", "dummy", "suc", "start_node", "end_node"])
# [0 Predecesors, 1 Blocked, 2 Dummy, 3 Successors, 4 Start node, 5 End node]
# Blocked = (False or Activity with same precedents)
# Step 1. Create the improper covers
work_table_pol = makeCover(prelations, successors)
# Step 2. Syslo Polynomial algorithm
final = successors.copy()
visited = []
for act, pred in prelations.items():
for v in pred:
for u in pred:
if u != v and successors[v] != successors[u] and act not in visited:
# Find activity in the improper cover table
for key, value in work_table_pol.items():
if act in value.w:
w = value.w
# Find each row that belongs to the predecessors of activity
for key, value in work_table_pol.items():
if set(value.u).issubset(prelations[act]) and value.u:
vertex = set(value.u).pop()
# Compare successors of a row with the improper cover of the activity
if successors[vertex] != w:
for q in value.u:
if final.has_key(q):
final[q] = list(
(set(final[q]) - set(w) | set([str(vertex) + separator + str(act)]))
- set([act])
)
else:
final[q] = list(
set(successors[q]) - set(w) | set([str(vertex) + separator + str(act)])
)
final[str(vertex) + separator + str(act)] = [act]
for l in w:
visited.append(l)
final = graph.successors2precedents(final)
work_table = {}
for act, pred in final.items():
work_table[act] = Columns(pred, False, False, None, None, None)
if act not in prelations:
work_table[act].dummy = True
# Step 3. Identify Dummy Activities And Identical Precedence Constraint of Diferent Activities
visited_pred = {}
for act, columns in work_table.items():
pred = frozenset(columns.pre)
if pred not in visited_pred:
visited_pred[pred] = act
else:
columns.blocked = visited_pred[pred]
# Step 4. Creating nodes
# (a) find start nodes
node = 0 # instead of 0, can start at 100 to avoid confusion with activities named with numbers when debugging
for act, columns in work_table.items():
if not columns.blocked:
columns.start_node = node
node += 1
if columns.blocked:
columns.start_node = work_table[columns.blocked].start_node
# Associate activities with their end nodes
for suc, suc_columns in work_table.items():
if not suc_columns.blocked:
if act in suc_columns.pre:
columns.suc = suc
break
# (b) find end nodes
graph_end_node = node # Reserve one node for graph end
node += 1
pm_graph = pert.PertMultigraph()
for act, columns in work_table.items():
suc = columns.suc
if suc:
columns.end_node = work_table[suc].start_node
else:
# Create needed end nodes, avoiding multiple graph end nodes (adaptation)
if act in end_act:
columns.end_node = node
else:
columns.end_node = graph_end_node
node += 1
# Generate the graph
_, _, dummy, _, start, end = columns
pm_graph.add_arc((start, end), (act, dummy))
p_graph = pm_graph.to_directed_graph()
return p_graph
except IOError:
print 'Error reading file:', filename
sys.exit(1)
if len(sys.argv) == 3:
repeticiones = int(sys.argv[2])
filename = sys.argv[1]
data = openProject(filename)
successors = {}
for i in data:
successors[i[1]] = i[2]
prelaciones1 = graph.reversed_prelation_table(successors)
prelaciones = {
'B': [],
'A': [],
'D': ['B'],
'C': [],
'F': ['C'],
'E': ['D'],
'H': ['B'],
'G': ['F'],
'J': ['F'],
'I': ['A'],
'L': ['C', 'E'],
'K': ['I'],
'N': ['B'],
'M': ['H'],
def sysloPolynomial(prelations):
# Adaptation to avoid multiple end nodes
successors = graph.reversed_prelation_table(prelations)
end_act = graph.ending_activities(successors)
#Step 0. Construct work table with Immediate Predecessors
Columns = namedlist.namedlist('Columns', ['pre', 'blocked', 'dummy', 'suc', 'start_node', 'end_node'])
# [0 Predecesors, 1 Blocked, 2 Dummy, 3 Successors, 4 Start node, 5 End node]
# Blocked = (False or Activity with same precedents)
#Step 1. Create the improper covers
work_table_pol = makeCover(prelations, successors)
# Step 2. Syslo Polynomial algorithm
final = successors.copy()
visited = []
for act, pred in prelations.items():
for v in pred:
for u in pred:
if u != v and successors[v] != successors[u] and act not in visited:
# Find activity in the improper cover table
for key, value in work_table_pol.items():
if act in value.w:
w = value.w
# Find each row that belongs to the predecessors of activity
for key, value in work_table_pol.items():
if set(value.u).issubset(prelations[act]) and value.u:
vertex = set(value.u).pop()
# Compare successors of a row with the improper cover of the activity
if successors[vertex] != w:
for q in value.u:
if final.has_key(q):
final[q] = list((set(final[q]) - set(w) | set([str(vertex) + separator + str(act)])) - set([act]))
else:
final[q] = list(set(successors[q]) - set(w) | set([str(vertex) + separator + str(act)]))
final[str(vertex) + separator + str(act)] = [act]
for l in w:
visited.append(l)
final = graph.successors2precedents(final)
work_table = {}
for act, pred in final.items():
work_table[act] = Columns(pred, False, False, None, None, None)
if act not in prelations:
work_table[act].dummy = True
#Step 3. Identify Dummy Activities And Identical Precedence Constraint of Diferent Activities
visited_pred = {}
for act, columns in work_table.items():
pred = frozenset(columns.pre)
if pred not in visited_pred:
visited_pred[pred] = act
else:
columns.blocked = visited_pred[pred]
#Step 4. Creating nodes
# (a) find start nodes
node = 0 # instead of 0, can start at 100 to avoid confusion with activities named with numbers when debugging
for act, columns in work_table.items():
if not columns.blocked:
columns.start_node = node
node += 1
if columns.blocked:
columns.start_node = work_table[columns.blocked].start_node
# Associate activities with their end nodes
for suc, suc_columns in work_table.items():
if not suc_columns.blocked:
if act in suc_columns.pre:
columns.suc = suc
break
# (b) find end nodes
graph_end_node = node # Reserve one node for graph end
node += 1
pm_graph = pert.PertMultigraph()
for act, columns in work_table.items():
suc = columns.suc
if suc:
columns.end_node = work_table[suc].start_node
else:
# Create needed end nodes, avoiding multiple graph end nodes (adaptation)
if act in end_act:
columns.end_node = node
else:
columns.end_node = graph_end_node
node += 1
# Generate the graph
_, _, dummy, _, start, end = columns
pm_graph.add_arc((start, end), (act, dummy))
p_graph = pm_graph.to_directed_graph()
return p_graph
def gento_municio(predecessors):
"""
Creates a PERT graph usign the algorithm defined in:
Gento-Municio, Angel M. “Un Algoritmo Para La Realización de Grafos Con Las Actividades En Los Arcos, Grafos
PERT.” Cuadernos Del CIMBAGE, no. 7 (2004): 103.
"""
nodes = NodeList(predecessors.keys())
successors = graph.reversed_prelation_table(predecessors)
# Generate precedences/successors matrix
matrix = scipy.zeros([nodes.num_real_activities, nodes.num_real_activities], dtype=int)
for activity, successor in successors.items():
for suc in successor:
matrix[nodes.activity_names.index(activity)][nodes.activity_names.index(suc)] = 1
# print "MATRIX filled: \n", matrix
# sum each column
sum_predecessors = scipy.sum(matrix, axis=0)
# print "SUM_PREDECESSORS: ", sum_predecessors
# sum each row
sum_successors = scipy.sum(matrix, axis=1)
# print "SUM_SUCCCESSORS: ", sum_successors
# Step 1. Search initial activities (have no predecessors) [3.1]
beginning, = numpy.nonzero(sum_predecessors == 0)
# print "Beginning: ", beginning
# add begin node to activities that begin at initial node
begin_node = nodes.next_node()
for node_activity in beginning:
nodes[node_activity][0] = begin_node
# Step 2. Search endings activities (have no successors) [3.2]
ending, = numpy.nonzero(sum_successors == 0)
# print "Ending: ", ending
# add end node to activities that end in final node
# note: this step may be replaced by handling them in steps 3 and 4 as stI and stII
end_node = nodes.next_node()
for node_activity in ending:
nodes[node_activity][1] = end_node
# print nodes
# Step 3. Search standard type I (Activity have unique successors) [3.3]
act_one_predeccessor, = numpy.nonzero(sum_predecessors == 1)
stI = collections.defaultdict(list)
for i in act_one_predeccessor:
pred = numpy.nonzero((matrix[:,i]))[0][0]
if (sum_successors[pred] == 1 # this condition is redundant but faster than the following check
or sum_successors[pred] == scipy.sum(matrix[:,numpy.nonzero(matrix[pred])])):
stI[pred].append(i)
# print "stI: ", stI
#Add the same end node of activities to the begin node of its successors activities
for node_activity in stI:
stI_node = nodes.next_node()
nodes[node_activity][1] = stI_node
for successor in stI[node_activity]:
nodes[successor][0] = stI_node
# print nodes
#Step 4. Search standard II(Full) and standard II(Incomplete) [3.4]
# print "--- Step 4 ---"
# dictionary with key: equal successors; value: mother activities
stII = collections.defaultdict(list)
for act in range(nodes.num_real_activities):
stII[frozenset(matrix[act].nonzero()[0])].append(act)
# remove ending activities and those included in type I
del(stII[ frozenset([]) ])
for pred, succs in stI.items():
del(stII[ frozenset(succs) ])
# assigns nodes to type II as indicated in figure 8
mark_complete = []
for succs, preds in stII.items():
u = len(preds)
# if NP[succs] != u (complete)
# print preds, '->', succs,
if not [ i for i in succs if sum_predecessors[i] != u ]:
# print 'complete'
mark_complete.append(succs)
node = nodes.next_node()
for act in preds:
nodes[act][1] = node
for act in succs:
nodes[act][0] = node
else: # (incomplete)
# print 'incomplete'
node = nodes.next_node()
for act in preds:
nodes[act][1] = node
# print nodes
# Step 5. Search for matching successors [3.5]
# print "--- Step 5 ---"
# remove type II complete so that stII becomes MASC
for succs in mark_complete:
del stII[succs]
masc = stII
# print "MASC"
# for succs, preds in stII.items():
# print preds, succs
npc = scipy.zeros([nodes.num_real_activities], dtype=int)
for succs, preds in masc.items():
num_preds = len(preds)
for succ in succs:
npc[succ] += num_preds
# print npc
# Step 6. Identifying start nodes on matching successors
# print "--- Step 6 ---"
act_no_initial = [i for i in range(nodes.num_real_activities) if nodes[i][0] == None]
# print act_no_initial, "<- No initial node"
num_no_initial = len(act_no_initial)
mra = scipy.zeros([num_no_initial, num_no_initial], dtype=int)
for succs, preds in masc.items():
num_preds = len(preds)
for act_i, act_j in itertools.combinations(succs, 2):
mra[act_no_initial.index(act_i), act_no_initial.index(act_j)] += num_preds
mra[act_no_initial.index(act_j), act_no_initial.index(act_i)] += num_preds # Symmetry, any succ order
# print 'MRA'
# print mra
# check matching successors and assign them initial nodes
for i in range(num_no_initial):
for j in range(i+1, num_no_initial):
if mra[i,j] == npc[act_no_initial[i]] and mra[i,j] == npc[act_no_initial[j]]:
# print 'coincidencia', i, j, "(", act_no_initial[i], act_no_initial[j], ")"
if nodes[act_no_initial[i]][0] != None:
node = nodes[act_no_initial[i]][0]
else:
node = nodes.next_node()
nodes[act_no_initial[i]][0] = node
nodes[act_no_initial[j]][0] = node
# assign initial node to the remaining activities (they must be alone, interpreted, not clear on paper)
for node in nodes:
if node[0] == None:
node[0] = nodes.next_node()
# print nodes
# Step 7. String search
# create MNS (to avoid counting matching successors twice)
mns = {}
unconnected = set() # all nodes in MNS
for succs, preds in masc.items():
succ_nodes = set([nodes[succ][0] for succ in succs])
mns[ nodes[preds[0]][1] ] = succ_nodes # as all preds have same successors they will be usign just one node
unconnected.update(succ_nodes)
# print 'MNS'
# for pred, succs in mns.items():
# print pred, '-', succs
# create MRN
unconnected = list(unconnected)
num_unconnected = len(unconnected)
# print unconnected, '<-Unconnected'
appear = scipy.zeros([num_unconnected], dtype=int)
mrn = scipy.zeros([num_unconnected, num_unconnected], dtype=int)
for pred, u_nodes in mns.items():
for node in u_nodes:
appear[ unconnected.index(node) ] += 1
for node_a, node_b in itertools.combinations(u_nodes, 2):
mrn[unconnected.index(node_a), unconnected.index(node_b)] += 1
mrn[unconnected.index(node_b), unconnected.index(node_a)] += 1
# print 'MRN'
# print mrn
# print 'Appear'
# print appear
# create MC
mc = []
for i in range(num_unconnected):
mc.append([j for j in range(num_unconnected) if mrn[i,j] == appear[i] ])
# print 'MC'
# for i in range(num_unconnected):
# print i, '-', mc[i]
# use strings to connect nodes
for i in range(num_unconnected):
following_nodes = sorted(mc[i], key=lambda x : len(mc[x]))
while following_nodes:
# print following_nodes
follower = following_nodes.pop()
# print 'extracted:', follower
# Create dummy i -> follower (unconnected to real)
nodes.append_dummy(unconnected[i], unconnected[follower])
for fol_follower in mc[follower]:
# print 'remove:', fol_follower
try:
following_nodes.remove(fol_follower)
except ValueError:
pass # if it has already been connected, it will not be in list now
# print nodes
# Step 8. Final nodes and dummies
# (note: contrary to what paper says, we have already set final nodes for all
# activities in step 4 as indicated in figure 8. Nevertheless, these nodes are
# unconnected so we replace them here if necessary. Not assigning nodes in step 4
# would break step 7)
for succs, preds in masc.items():
# print "Studying:", preds, '->', succs
if len(succs) == 1: # Case I
# print "Case I"
for pred in preds:
nodes[pred][1] = nodes[next(iter(succs))][0]
else:
# Get follower with lower npc
min_follower = None
min_npc = None
for succ in succs:
if min_npc == None or min_npc > npc[succ]:
min_follower = succ
min_npc = npc[succ]
# print min_follower, 'npc:', min_npc
# Count the number of masc rows containing our successor activities
count = 0
for others in masc:
if succs.issubset(others):
count += len(masc[others])
if count >= min_npc: # Case II (if min_npc==1) and Case III
# print "Case II or III"
for pred in preds:
nodes[pred][1] = nodes[min_follower][0]
else:
for succ in succs:
# note: if there are several predecessors, they have the same end node assigned in step 4
nodes.append_dummy(nodes[next(iter(preds))][1], nodes[succ][0])
# Step 9. Final nodes for type II incomplete
# (note: final nodes have already been assigned in step 8. We think section 3.9 of paper is unnecessary)
return nodes.to_pert_graph().renumerar()
def __init__(self, pert=None):
super(Pert, self).__init__()
self.construct = algoritmoSharma.sharma1998ext
if pert != None:
self.successors, self.arcs = pert
self.predecessors = graph.reversed_prelation_table(self.successors)
def mouhoub(prelations):
"""
Build a PERT graph using Mouhoub algorithm
prelations = {'activity': ['predecesor1', 'predecesor2'...}
return p_graph pert.PertMultigraph()
"""
Columns = namedlist.namedlist('Columns', ['pre', 'su', 'blocked', 'dummy', 'suc', 'start_node', 'end_node', 'aux'])
# [0 Predecesors, 1 Successors, 2 Blocked, 3 Dummy, 4 Blocked successor, 5 Start node, 6 End node, 7 Auxiliar ]
# Blocked = (False or Activity with same precedents)
# Adaptation to avoid multiple end nodes
successors = graph.reversed_prelation_table(prelations)
successors_copy = graph.reversed_prelation_table(prelations.copy())
end_act = graph.ending_activities(successors)
# Step 0. Remove Z Configuration. Update the prelation table in complete_bipartite dictionary
complete_bipartite = successors
complete_bipartite.update(zConfiguration.zconf(successors))
# STEPS TO BUILD THE PERT GRAPH
#Step 1. Save the prelations in the work table
complete_bipartite = graph.successors2precedents(complete_bipartite)
work_table = {}
for act, sucesores in complete_bipartite.items():
work_table[act] = Columns(set(sucesores), successors[act], None, False, None, None, None, None)
if act not in prelations:
work_table[act].dummy = True
#Step 2. Identify Identical Precedence Constraint of Diferent Activities
visited_pred = {}
for act, columns in work_table.items():
pred = frozenset(columns.pre)
if pred not in visited_pred:
visited_pred[pred] = act
else:
columns.blocked = visited_pred[pred]
#Step 3. Creating nodes
# (a) Find start nodes
node = 0 # instead of 0, can start at 100 to avoid confusion with activities named with numbers when debugging
for act, columns in work_table.items():
if not columns.blocked:
columns.start_node = node
node += 1
if columns.blocked:
columns.start_node = work_table[columns.blocked].start_node
# Associate activities with their end nodes
for suc, suc_columns in work_table.items():
if not suc_columns.blocked:
if act in suc_columns.pre:
columns.suc = suc
break
# (b) Find end nodes
graph_end_node = node # Reserve one node for graph end
node += 1
for act, columns in work_table.items():
suc = columns.suc
if suc:
columns.end_node = work_table[suc].start_node
else:
# Create needed end nodes, avoiding multiple graph end nodes (adaptation)
if act in end_act:
columns.end_node = graph_end_node
else:
columns.end_node = node
node += 1
#Step 4. MOUHOUB algorithm rules to remove extra dummy activities
mouhoubRules.rule_1(successors_copy, work_table)
G2 = mouhoubRules.rule_2(prelations, work_table)
G3 = mouhoubRules.rule_3(G2, work_table)
G4 = mouhoubRules.rule_4(G3, work_table)
G5_6 = mouhoubRules.rule_5_6(successors_copy, work_table, G4)
G3a = mouhoubRules.rule_3(G5_6, work_table)
G4a = mouhoubRules.rule_4(G3a, work_table)
G7 = mouhoubRules.rule_7(successors_copy, successors, G4a, node)
work_table_final = {}
for act, sucesores in G7.items():
work_table_final[act] = Columns([], [], [], sucesores.dummy, sucesores.suc, sucesores.start_node, sucesores.end_node, [])
#Step 5. Delete Dummy Cycles
for act, sucesores in work_table_final.items():
for act2, sucesores2 in work_table_final.items():
if act != act2:
if sucesores.end_node == sucesores2.end_node and sucesores.start_node == sucesores2.start_node:
if act not in successors:
del work_table_final[act]
#Step 6. Generate the graph
pm_graph = pert.PertMultigraph()
for act, columns in work_table_final.items():
_, _, _, dummy, _, start, end, _ = columns
pm_graph.add_arc((start, end), (act, dummy))
p_graph = pm_graph.to_directed_graph()
return p_graph
def __init__(self, pert=None):
super(Pert, self).__init__()
self.construct = algoritmoSharma.sharma1998ext
if pert != None:
self.successors, self.arcs = pert
self.predecessors = graph.reversed_prelation_table(self.successors)
def main():
"""
Test AOA (PERT) network generation algorithms with some given project files
"""
# Parse arguments and options
parser = argparse.ArgumentParser(
description='Test AOA graph generation algorithms with given files')
parser.add_argument('infiles', nargs='*', help='Project files to test')
parser.add_argument(
'--table-file',
'-t',
default='resultados.csv',
help=
'Name of file to append test results in CSV format (default: resultados.csv)'
)
parser.add_argument('-r',
'--repeat',
default=1,
type=int,
help='Number of repetitions (default: 1)')
parser.add_argument('--SVG',
action='store_true',
help='Draw the graph in a SVG file')
parser.add_argument('--no-stop',
action='store_true',
help='Do not stop when an algorithm fails')
parser.add_argument('-c',
'--CohenSadeh',
action='store_true',
help='Test Cohen Sadeh algorithm')
parser.add_argument('-s',
'--Sharma',
action='store_true',
help='Test Sharma algorithm')
parser.add_argument('-l',
'--Salas',
action='store_true',
help='Test Lorenzo Salas algorithm')
parser.add_argument('-g',
'--GentoMunicio',
action='store_true',
help='Test Gento Municio algorithm')
parser.add_argument('-o',
'--Optimal',
action='store_true',
help='Test set based optimal algorithm')
parser.add_argument('-m',
'--Mouhoub',
action='store_true',
help='Test Mouhoub algorithm')
parser.add_argument('-p',
'--Syslo_Polynomial',
action='store_true',
help='Test Syslo Polynomial algorithm')
parser.add_argument('-y',
'--Syslo_Optimal',
action='store_true',
help='Test Syslo Optimal algorithm')
args = parser.parse_args()
if args.repeat < 1:
print 'Number of repetitions must be > 0'
return 1
try:
f_csv = open(args.table_file, "a")
except IOError:
print 'Can not open table file (%s) to append results in CSV format' % (
args.table_file, )
return 1
# List of name and function of each algorithm to test
algorithms = []
if args.CohenSadeh:
algorithms.append(('CohenSadeh', algoritmoCohenSadeh.cohen_sadeh))
if args.Sharma:
algorithms.append(('Sharma', algoritmoSharma.sharma1998ext))
if args.Optimal:
algorithms.append(('Conjuntos', algoritmoConjuntos.algoritmoN))
if args.GentoMunicio:
algorithms.append(
('GentoMunicio', algoritmoGentoMunicio.gento_municio))
if args.Salas:
algorithms.append(('Salas', algoritmoSalas.salas))
if args.Mouhoub:
algorithms.append(('Mouhoub', algoritmoMouhoub.mouhoub))
if args.Syslo_Polynomial:
algorithms.append(
('Syslo Polinomico', algoritmoSysloPolynomial.sysloPolynomial))
if args.Syslo_Optimal:
algorithms.append(('Syslo Optimo', algoritmoSysloOptimal.sysloOptimal))
# Perform tests on each file
for filename in args.infiles:
print "\nFilename: ", filename
data = openProject(filename)
if not data:
print 'Can not read or understand file'
else:
# XXX Aqui habria que cortar si falla el checkeo del fichero
check_activities(data)
# Test each algorithm
for name, alg in algorithms:
print name
# Get successors from activities table
successors = {}
for i in data:
successors[i[1]] = i[2]
# Count prelations
list_of_predecessors = successors.values()
num_of_predecessors = 0
for predecessors in list_of_predecessors:
num_of_predecessors += len(predecessors)
# Get predecessors from successors
prelaciones = graph.reversed_prelation_table(successors)
# Run algorithm
pert_graph = None
itime = os.times()
for i in range(args.repeat):
try:
pert_graph = alg(prelaciones)
except Exception:
print traceback.format_exc()
print " --- Algorithm failed! --- "
if not args.no_stop:
return 1
break
if pert_graph:
ftime = os.times()
utime = ftime[0] - itime[0]
# Print test results
print "utime %.4f" % (utime)
print "utime: ", utime
print "numero de nodos: ", pert_graph.number_of_nodes()
print "numero de arcos: ", pert_graph.number_of_arcs()
print "numero de arcos reales: ", pert_graph.numArcsReales(
)
print "numero de arcos ficticios: ", pert_graph.numArcsFicticios(
)
print "numero de predecesors/sucesores: ", num_of_predecessors
print "Validation: "
if not validation.check_validation(
successors, pert_graph) and not args.no_stop:
return 1
print ""
# XXX ??Falta incluir aqui el numero de actividades??
result_line = '"' + filename + '",' + '"' + name + '",' + str(len(data)) + ',' + str(num_of_predecessors) + ',' + \
str(pert_graph.number_of_nodes()) + ',' + str(pert_graph.number_of_arcs()) + ',' + \
str(pert_graph.numArcsReales()) + ',' + str(pert_graph.numArcsFicticios()) + ',' + "%.4f"%(utime)
f_csv.write(result_line + "\n")
if pert_graph == 1:
print "No hay resultados que mostrar"
# Draw graph and save in a file (*.svg)
if args.SVG:
image_text = graph.pert2image(pert_graph)
fsalida = open(
os.path.split(filename)[1] + '_' + name + '.svg',
'w')
fsalida.write(image_text)
fsalida.close()
f_csv.close()
return 0
def gento_municio(predecessors):
"""
Creates a PERT graph usign the algorithm defined in:
Gento-Municio, Angel M. “Un Algoritmo Para La Realización de Grafos Con Las Actividades En Los Arcos, Grafos
PERT.” Cuadernos Del CIMBAGE, no. 7 (2004): 103.
"""
nodes = NodeList(predecessors.keys())
successors = graph.reversed_prelation_table(predecessors)
# Generate precedences/successors matrix
matrix = scipy.zeros(
[nodes.num_real_activities, nodes.num_real_activities], dtype=int)
for activity, successor in successors.items():
for suc in successor:
matrix[nodes.activity_names.index(activity)][
nodes.activity_names.index(suc)] = 1
# print "MATRIX filled: \n", matrix
# sum each column
sum_predecessors = scipy.sum(matrix, axis=0)
# print "SUM_PREDECESSORS: ", sum_predecessors
# sum each row
sum_successors = scipy.sum(matrix, axis=1)
# print "SUM_SUCCCESSORS: ", sum_successors
# Step 1. Search initial activities (have no predecessors) [3.1]
beginning, = numpy.nonzero(sum_predecessors == 0)
# print "Beginning: ", beginning
# add begin node to activities that begin at initial node
begin_node = nodes.next_node()
for node_activity in beginning:
nodes[node_activity][0] = begin_node
# Step 2. Search endings activities (have no successors) [3.2]
ending, = numpy.nonzero(sum_successors == 0)
# print "Ending: ", ending
# add end node to activities that end in final node
# note: this step may be replaced by handling them in steps 3 and 4 as stI and stII
end_node = nodes.next_node()
for node_activity in ending:
nodes[node_activity][1] = end_node
# print nodes
# Step 3. Search standard type I (Activity have unique successors) [3.3]
act_one_predeccessor, = numpy.nonzero(sum_predecessors == 1)
stI = collections.defaultdict(list)
for i in act_one_predeccessor:
pred = numpy.nonzero((matrix[:, i]))[0][0]
if (sum_successors[pred] ==
1 # this condition is redundant but faster than the following check
or sum_successors[pred] == scipy.sum(
matrix[:, numpy.nonzero(matrix[pred])])):
stI[pred].append(i)
# print "stI: ", stI
#Add the same end node of activities to the begin node of its successors activities
for node_activity in stI:
stI_node = nodes.next_node()
nodes[node_activity][1] = stI_node
for successor in stI[node_activity]:
nodes[successor][0] = stI_node
# print nodes
#Step 4. Search standard II(Full) and standard II(Incomplete) [3.4]
# print "--- Step 4 ---"
# dictionary with key: equal successors; value: mother activities
stII = collections.defaultdict(list)
for act in range(nodes.num_real_activities):
stII[frozenset(matrix[act].nonzero()[0])].append(act)
# remove ending activities and those included in type I
del (stII[frozenset([])])
for pred, succs in stI.items():
del (stII[frozenset(succs)])
# assigns nodes to type II as indicated in figure 8
mark_complete = []
for succs, preds in stII.items():
u = len(preds)
# if NP[succs] != u (complete)
# print preds, '->', succs,
if not [i for i in succs if sum_predecessors[i] != u]:
# print 'complete'
mark_complete.append(succs)
node = nodes.next_node()
for act in preds:
nodes[act][1] = node
for act in succs:
nodes[act][0] = node
else: # (incomplete)
# print 'incomplete'
node = nodes.next_node()
for act in preds:
nodes[act][1] = node
# print nodes
# Step 5. Search for matching successors [3.5]
# print "--- Step 5 ---"
# remove type II complete so that stII becomes MASC
for succs in mark_complete:
del stII[succs]
masc = stII
# print "MASC"
# for succs, preds in stII.items():
# print preds, succs
npc = scipy.zeros([nodes.num_real_activities], dtype=int)
for succs, preds in masc.items():
num_preds = len(preds)
for succ in succs:
npc[succ] += num_preds
# print npc
# Step 6. Identifying start nodes on matching successors
# print "--- Step 6 ---"
act_no_initial = [
i for i in range(nodes.num_real_activities) if nodes[i][0] == None
]
# print act_no_initial, "<- No initial node"
num_no_initial = len(act_no_initial)
mra = scipy.zeros([num_no_initial, num_no_initial], dtype=int)
for succs, preds in masc.items():
num_preds = len(preds)
for act_i, act_j in itertools.combinations(succs, 2):
mra[act_no_initial.index(act_i),
act_no_initial.index(act_j)] += num_preds
mra[act_no_initial.index(act_j),
act_no_initial.
index(act_i)] += num_preds # Symmetry, any succ order
# print 'MRA'
# print mra
# check matching successors and assign them initial nodes
for i in range(num_no_initial):
for j in range(i + 1, num_no_initial):
if mra[i, j] == npc[act_no_initial[i]] and mra[i, j] == npc[
act_no_initial[j]]:
# print 'coincidencia', i, j, "(", act_no_initial[i], act_no_initial[j], ")"
if nodes[act_no_initial[i]][0] != None:
node = nodes[act_no_initial[i]][0]
else:
node = nodes.next_node()
nodes[act_no_initial[i]][0] = node
nodes[act_no_initial[j]][0] = node
# assign initial node to the remaining activities (they must be alone, interpreted, not clear on paper)
for node in nodes:
if node[0] == None:
node[0] = nodes.next_node()
# print nodes
# Step 7. String search
# create MNS (to avoid counting matching successors twice)
mns = {}
unconnected = set() # all nodes in MNS
for succs, preds in masc.items():
succ_nodes = set([nodes[succ][0] for succ in succs])
mns[nodes[preds[0]][
1]] = succ_nodes # as all preds have same successors they will be usign just one node
unconnected.update(succ_nodes)
# print 'MNS'
# for pred, succs in mns.items():
# print pred, '-', succs
# create MRN
unconnected = list(unconnected)
num_unconnected = len(unconnected)
# print unconnected, '<-Unconnected'
appear = scipy.zeros([num_unconnected], dtype=int)
mrn = scipy.zeros([num_unconnected, num_unconnected], dtype=int)
for pred, u_nodes in mns.items():
for node in u_nodes:
appear[unconnected.index(node)] += 1
for node_a, node_b in itertools.combinations(u_nodes, 2):
mrn[unconnected.index(node_a), unconnected.index(node_b)] += 1
mrn[unconnected.index(node_b), unconnected.index(node_a)] += 1
# print 'MRN'
# print mrn
# print 'Appear'
# print appear
# create MC
mc = []
for i in range(num_unconnected):
mc.append(
[j for j in range(num_unconnected) if mrn[i, j] == appear[i]])
# print 'MC'
# for i in range(num_unconnected):
# print i, '-', mc[i]
# use strings to connect nodes
for i in range(num_unconnected):
following_nodes = sorted(mc[i], key=lambda x: len(mc[x]))
while following_nodes:
# print following_nodes
follower = following_nodes.pop()
# print 'extracted:', follower
# Create dummy i -> follower (unconnected to real)
nodes.append_dummy(unconnected[i], unconnected[follower])
for fol_follower in mc[follower]:
# print 'remove:', fol_follower
try:
following_nodes.remove(fol_follower)
except ValueError:
pass # if it has already been connected, it will not be in list now
# print nodes
# Step 8. Final nodes and dummies
# (note: contrary to what paper says, we have already set final nodes for all
# activities in step 4 as indicated in figure 8. Nevertheless, these nodes are
# unconnected so we replace them here if necessary. Not assigning nodes in step 4
# would break step 7)
for succs, preds in masc.items():
# print "Studying:", preds, '->', succs
if len(succs) == 1: # Case I
# print "Case I"
for pred in preds:
nodes[pred][1] = nodes[next(iter(succs))][0]
else:
# Get follower with lower npc
min_follower = None
min_npc = None
for succ in succs:
if min_npc == None or min_npc > npc[succ]:
min_follower = succ
min_npc = npc[succ]
# print min_follower, 'npc:', min_npc
# Count the number of masc rows containing our successor activities
count = 0
for others in masc:
if succs.issubset(others):
count += len(masc[others])
if count >= min_npc: # Case II (if min_npc==1) and Case III
# print "Case II or III"
for pred in preds:
nodes[pred][1] = nodes[min_follower][0]
else:
for succ in succs:
# note: if there are several predecessors, they have the same end node assigned in step 4
nodes.append_dummy(nodes[next(iter(preds))][1],
nodes[succ][0])
# Step 9. Final nodes for type II incomplete
# (note: final nodes have already been assigned in step 8. We think section 3.9 of paper is unnecessary)
return nodes.to_pert_graph().renumerar()
