def calc_speed(self):
segment_speeds = []
# ignore the first point if it's not within the moving threshold
points = self[1:] if len(self) > 1 and not Path.linked(self[0], self[1]) else self[:]
# dont consider paths that we really didnt capture (aka teleporting)
if len(points) < 2:
self.speed = None
return
# calc segment speeds
for i, point in enumerate(points[:-1]):
next = points[i+1]
d = science.geo_distance(point.location, next.location, True)
t = abs(point.t - next.t)
if t > 0.0:
segment_speeds.append(d / t)
# we are using the second-to-fastest segment, if possible, as the speed for this path. cars also go slow; by not using the fastest we're filtering subway travel.
segment_speeds.sort()
self.speed = None
index = -2
while self.speed is None and index <= 0:
try:
self.speed = segment_speeds[index] * 60 * 60 # mph
except IndexError as e:
index += 1
def bake(self):
self.start_point = self[0]
self.end_point = self[-1]
self.distance = science.geo_distance(self.start_point.location, self.end_point.location)
self.start_time = self[1].date if len(self) > 1 and not Path.linked(self[0], self[1]) else self[0].date # dont use the start_point date as the path start_time if we've been at the start_point for awhile
self.end_time = self[-1].date
self.calc_speed()
def get_closest(point):
# in theory, a point in a cluster could be closer to the centroid of a different cluster (maybe?). however, I dont think doing things this way will be detrimental
min_d = float('inf')
for cluster in clusters:
d = science.geo_distance((point.lon, point.lat), cluster.vector)
if d < min_d:
closest_cluster = cluster
min_d = d
return closest_cluster
def build(cls, data, start, end, center=None):
log.info("Loading OpenPaths data...")
if center is not None:
points = [Point(i, float(d['lon']), float(d['lat']), d['t']) for i, d in enumerate(data) if d['t'] >= start and d['t'] <= end and science.geo_distance((float(d['lon']), float(d['lat'])), center) <= RADIUS]
else:
points = [Point(i, float(d['lon']), float(d['lat']), d['t']) for i, d in enumerate(data) if d['t'] >= start and d['t'] <= end]
points.sort(key=lambda p: p.t)
log.info("Generating graph...")
start_clock = time.clock()
paths = Path.find_paths(points)
places = Place.find_places(paths, CLUSTER_RADIUS)
cities = City.find_cities(places, CITY_RADIUS)
almanac = Almanac(points, paths, places, cities)
# almanac.label()
print("(%ss)" % (time.clock() - start_clock))
return almanac
