def test_execute(self):
# Loads and prepares the graph
res1 = AssignmentResults()
res1.prepare(self.g, self.matrix)
assig1 = allOrNothing(self.matrix, self.g, res1)
assig1.execute()
res2 = AssignmentResults()
res2.prepare(self.g, self.matrix2)
assig2 = allOrNothing(self.matrix2, self.g, res2)
assig2.execute()
load1 = res1.get_load_results()
load2 = res2.get_load_results()
self.assertEqual(list(load1.matrix_tot * 2), list(load2.matrix_tot),
"Something wrong with the AoN")
class DesireLinesProcedure(WorkerThread):
def __init__(self, parentThread, layer: str, id_field: int,
matrix: AequilibraeMatrix, matrix_hash: dict,
dl_type: str) -> None:
WorkerThread.__init__(self, parentThread)
self.layer = layer
self.id_field = id_field
self.matrix = matrix
self.dl_type = dl_type
self.error = None
self.matrix_hash = matrix_hash
self.report = []
self.logger = logging.getLogger('aequilibrae')
self.nodes_to_indices = {
matrix.index[x]: x
for x in range(matrix.zones)
}
self.python_version = (8 * struct.calcsize("P"))
if error:
self.error = 'Scipy and/or Numpy not installed'
self.report.append(self.error)
self.procedure = "ASSIGNMENT"
def doWork(self):
if self.error is None:
# In case we have only one class
unnasigned = 0
classes = self.matrix.matrix_view.shape[2]
layer = get_vector_layer_by_name(self.layer)
idx = layer.dataProvider().fieldNameIndex(self.id_field)
feature_count = layer.featureCount()
self.desire_lines.emit(('job_size_dl', feature_count))
all_centroids = {}
for P, feat in enumerate(layer.getFeatures()):
geom = feat.geometry()
if geom is not None:
point = list(geom.centroid().asPoint())
centroid_id = feat.attributes()[idx]
all_centroids[centroid_id] = point
self.desire_lines.emit(('jobs_done_dl', P))
self.desire_lines.emit(('text_dl', "Loading Layer Features: " +
str(P) + "/" + str(feature_count)))
# Creating resulting layer
EPSG_code = int(layer.crs().authid().split(":")[1])
desireline_layer = QgsVectorLayer(
"LineString?crs=epsg:" + str(EPSG_code), self.dl_type,
"memory")
dlpr = desireline_layer.dataProvider()
base_dl_fields = [
QgsField("link_id", QVariant.Int),
QgsField("A_Node", QVariant.Int),
QgsField("B_Node", QVariant.Int),
QgsField("direct", QVariant.Int),
QgsField("distance", QVariant.Double)
]
if self.dl_type == "DesireLines":
max_zone = self.matrix.index[:].max().astype(np.int64) + 1
items = [(i, j[0], j[1]) for i, j in all_centroids.items()
if i < max_zone]
coords = np.array(items)
coord_index = np.zeros((max_zone, 2))
coord_index[coords[:, 0].astype(np.int64), 0] = coords[:, 1]
coord_index[coords[:, 0].astype(np.int64), 1] = coords[:, 2]
self.desire_lines.emit(
('text_dl', "Manipulating matrix indices"))
zones = self.matrix.index[:].shape[0]
a = np.array(self.matrix.index[:], np.int64)
ij, ji = np.meshgrid(a, a, sparse=False, indexing='ij')
ij = ij.flatten()
ji = ji.flatten()
arrays = [ij, ji]
self.desire_lines.emit(('text_dl', "Collecting all matrices"))
self.desire_lines.emit(
('job_size_dl', len(self.matrix.view_names)))
total_mat = np.zeros((zones, zones), np.float64)
for i, mat in enumerate(self.matrix.view_names):
m = self.matrix.get_matrix(mat)
total_mat += m
arrays.append(m.flatten())
self.desire_lines.emit(('jobs_done_dl', i + 1))
# Eliminates the cells for which we don't have geography
self.desire_lines.emit(
('text_dl', "Filtering zones with no geography available"))
zones_with_no_geography = [
x for x in self.matrix.index[:] if x not in all_centroids
]
if zones_with_no_geography:
self.desire_lines.emit(
('job_size_dl', len(zones_with_no_geography)))
for k, z in enumerate(zones_with_no_geography):
i = self.matrix.matrix_hash[z]
t = np.nansum(total_mat[i, :]) + np.nansum(total_mat[:, i])
unnasigned += t
self.report.append(
'Zone {} does not have a corresponding centroid/zone. Total flow {}'
.format(z, t))
total_mat[i, :] = 0
total_mat[:, i] = 0
self.desire_lines.emit(('jobs_done_dl', k + 1))
self.desire_lines.emit(
('text_dl', "Filtering down to OD pairs with flows"))
field_names = [x for x in self.matrix.view_names]
nonzero = np.nonzero(total_mat.flatten())
arrays = np.vstack(arrays).transpose()
arrays = arrays[nonzero, :]
arrays = arrays.reshape(arrays.shape[1], arrays.shape[2])
base_types = [(x, np.float64) for x in ['from', 'to']]
base_types = base_types + [(x + '_AB', np.float64)
for x in field_names]
dtypes_ab = [(x, np.int64)
for x in ['from', 'to']] + [(x + '_AB', float)
for x in field_names]
dtypes_ba = [(x, np.int64)
for x in ['to', 'from']] + [(x + '_BA', float)
for x in field_names]
ab_mat = np.array(arrays[arrays[:, 0] > arrays[:, 1], :])
ba_mat = np.array(arrays[arrays[:, 0] < arrays[:, 1], :])
flows_ab = ab_mat.view(base_types)
flows_ab = flows_ab.reshape(flows_ab.shape[:-1])
flows_ab = flows_ab.astype(dtypes_ab)
flows_ba = ba_mat.view(base_types)
flows_ba = flows_ba.reshape(flows_ba.shape[:-1])
flows_ba = flows_ba.astype(dtypes_ba)
defaults1 = {x + '_AB': 0.0 for x in field_names}
defaults = {x + '_BA': 0.0 for x in field_names}
defaults = {**defaults, **defaults1}
self.desire_lines.emit(
('text_dl', "Concatenating AB & BA flows"))
flows = rfn.join_by(['from', 'to'],
flows_ab,
flows_ba,
jointype='outer',
defaults=defaults,
usemask=True,
asrecarray=True)
flows = flows.filled()
flows_ab = 0
flows_ba = 0
for f in flows.dtype.names[2:]:
base_dl_fields.extend([QgsField(f, QVariant.Double)])
dlpr.addAttributes(base_dl_fields)
desireline_layer.updateFields()
self.desire_lines.emit(('text_dl', "Creating Desire Lines"))
self.desire_lines.emit(('job_size_dl', flows.shape[0]))
all_features = []
for i, rec in enumerate(flows):
a_node = rec[0]
b_node = rec[1]
a_point = QgsPointXY(*all_centroids[a_node])
b_point = QgsPointXY(*all_centroids[b_node])
dist = QgsGeometry().fromPointXY(a_point).distance(
QgsGeometry().fromPointXY(b_point))
feature = QgsFeature()
feature.setGeometry(
QgsGeometry.fromPolylineXY([a_point, b_point]))
attrs = [i + 1, int(a_node), int(b_node), 0, dist]
attrs.extend([float(x) for x in list(rec)[2:]])
feature.setAttributes(attrs)
all_features.append(feature)
self.desire_lines.emit(('jobs_done_dl', i))
if unnasigned > 0:
self.report.append(
'Total non assigned flows (not counting intrazonals):'
+ str(unnasigned))
if flows.shape[0] > 1:
a = dlpr.addFeatures(all_features)
self.result_layer = desireline_layer
else:
self.report.append('Nothing to show')
elif self.dl_type == "DelaunayLines":
for f in self.matrix.view_names:
base_dl_fields.extend([
QgsField(f + '_ab', QVariant.Double),
QgsField(f + '_ba', QVariant.Double),
QgsField(f + '_tot', QVariant.Double)
])
dlpr.addAttributes(base_dl_fields)
desireline_layer.updateFields()
self.desire_lines.emit(
('text_dl', "Building Delaunay dataset"))
points = []
node_id_in_delaunay_results = {}
i = 0
self.desire_lines.emit(('job_size_dl', len(all_centroids)))
for k, v in all_centroids.items():
self.desire_lines.emit(('jobs_done_dl', i))
points.append(v)
node_id_in_delaunay_results[i] = k
i += 1
self.desire_lines.emit(
('text_dl', "Computing Delaunay Triangles"))
tri = Delaunay(np.array(points))
# We process all the triangles to only get each edge once
self.desire_lines.emit(
('text_dl', "Building Delaunay Network: Collecting Edges"))
edges = []
if self.python_version == 32:
all_edges = tri.vertices
else:
all_edges = tri.simplices
self.desire_lines.emit(('job_size_dl', len(all_edges)))
for j, triangle in enumerate(all_edges):
self.desire_lines.emit(('jobs_done_dl', j))
links = list(itertools.combinations(triangle, 2))
for i in links:
edges.append([min(i[0], i[1]), max(i[0], i[1])])
self.desire_lines.emit(
('text_dl',
"Building Delaunay Network: Getting unique edges"))
edges = OrderedDict((str(x), x) for x in edges).values()
# Writing Delaunay layer
self.desire_lines.emit(
('text_dl', "Building Delaunay Network: Assembling Layer"))
desireline_link_id = 1
data = []
dl_ids_on_links = {}
self.desire_lines.emit(('job_size_dl', len(edges)))
for j, edge in enumerate(edges):
self.desire_lines.emit(('jobs_done_dl', j))
a_node = node_id_in_delaunay_results[edge[0]]
a_point = all_centroids[a_node]
a_point = QgsPointXY(a_point[0], a_point[1])
b_node = node_id_in_delaunay_results[edge[1]]
b_point = all_centroids[b_node]
b_point = QgsPointXY(b_point[0], b_point[1])
dist = QgsGeometry().fromPointXY(a_point).distance(
QgsGeometry().fromPointXY(b_point))
line = []
line.append(desireline_link_id)
line.append(a_node)
line.append(b_node)
line.append(dist)
line.append(dist)
line.append(0)
data.append(line)
dl_ids_on_links[desireline_link_id] = [
a_node, b_node, 0, dist
]
desireline_link_id += 1
self.desire_lines.emit(('text_dl', "Building graph"))
network = np.asarray(data)
del data
# types for the network
self.graph = Graph()
itype = self.graph.default_types('int')
ftype = self.graph.default_types('float')
all_types = [itype, itype, itype, ftype, ftype, np.int8]
all_titles = [
'link_id', 'a_node', 'b_node', 'distance_ab',
'distance_ba', 'direction'
]
dt = [(t, d) for t, d in zip(all_titles, all_types)]
self.graph.network = np.zeros(network.shape[0], dtype=dt)
for k, t in enumerate(dt):
self.graph.network[t[0]] = network[:, k].astype(t[1])
del network
self.graph.type_loaded = 'NETWORK'
self.graph.status = 'OK'
self.graph.network_ok = True
self.graph.prepare_graph(self.matrix.index.astype(np.int64))
self.graph.set_graph(cost_field="distance")
self.graph.set_blocked_centroid_flows(False)
self.results = AssignmentResults()
self.results.prepare(self.graph, self.matrix)
self.desire_lines.emit(('text_dl', "Assigning demand"))
self.desire_lines.emit(
('job_size_dl', self.matrix.index.shape[0]))
assigner = allOrNothing(self.matrix, self.graph, self.results)
assigner.execute()
self.report = assigner.report
print(self.results.link_loads)
self.desire_lines.emit(('text_dl', "Collecting results"))
self.desire_lines.emit(('text_dl', "Building resulting layer"))
features = []
max_edges = len(edges)
self.desire_lines.emit(('job_size_dl', max_edges))
link_loads = self.results.get_load_results()
for i, link_id in enumerate(link_loads.index):
self.desire_lines.emit(('jobs_done_dl', i))
a_node, b_node, direct, dist = dl_ids_on_links[link_id]
attr = [int(link_id), a_node, b_node, direct, dist]
a_point = all_centroids[a_node]
a_point = QgsPointXY(a_point[0], a_point[1])
b_point = all_centroids[b_node]
b_point = QgsPointXY(b_point[0], b_point[1])
feature = QgsFeature()
feature.setGeometry(
QgsGeometry.fromPolylineXY([a_point, b_point]))
for c in self.matrix.view_names:
attr.extend([
float(link_loads.data[c + '_ab'][i]),
float(link_loads.data[c + '_ba'][i]),
float(link_loads.data[c + '_tot'][i])
])
feature.setAttributes(attr)
features.append(feature)
a = dlpr.addFeatures(features)
self.result_layer = desireline_layer
self.desire_lines.emit(('finished_desire_lines_procedure', 0))