Skip to content

Instantly share code, notes, and snippets.

@JoshData JoshData/
Last active Aug 29, 2015

What would you like to do?
DCHenge (like Manhattanhenge but for every day of the year in DC)
#Get the Census's TIGER/Line roads database for the District of Columbia and convert it into GeoJSON:
# wget
# unzip
# ogr2ogr -f geojson dcroads2.geojson tl_2013_11001_roads.shp
import sys, json, datetime, math
from astral import Astral
astral = Astral()
def main():
city = astral["Washington DC"]
# Load in the JSON data containing roads.
road_data = json.load(sys.stdin)
# For the next 365 days, show the best henge-road.
for day_shift in range(366):
date = ( + datetime.timedelta(days=day_shift)).date()
print(date, end="\t")
for sunrise_sunset in ("sunrise", "sunset"):
time, target_vector = get_solar_vector(date, city, sunrise_sunset)
roads = []
for feature in road_data["features"]:
roadname = feature["properties"]["FULLNAME"]
if roadname is None: continue
if feature["geometry"]["type"] != "LineString": continue # There is one MultiLineString, for Moreland St NW, which we'll ignore.
for length, location in get_road_segments(feature, target_vector):
roads.append((length, roadname, location))
# Get the road segment of the longest length.
road = roads[0]
print (road[1], ",%f,14/%s/%s" % (road[2][1], road[2][0],
date.isoformat().replace("-", "."), str(time.strftime("%H:%M"))),
def get_solar_vector(date, city, sunrise_sunset):
# Gets a vector in lng/lat coordinates pointing toward either sunrise or sunset
# on the given date.
# Get the azimuth.
if sunrise_sunset == "sunrise":
time = city.sunrise(date=date, local=True)+datetime.timedelta(minutes=20)
azimuth = astral.solar_azimuth(time, city.latitude, city.longitude)
elif sunrise_sunset == "sunset":
time = city.sunset(date=date, local=True)-datetime.timedelta(minutes=30)
azimuth = astral.solar_azimuth(time, city.latitude, city.longitude)
raise ValueError()
# Convert the azimuths (in clockwise degrees from north) into a vector on the lng/lat plane.
from math import sin, cos, radians
return (time, (sin(radians(azimuth)), cos(radians(azimuth))))
def haversine(p1, p2):
# We'll need to compute distances. The haversine function computes the great
# circle distance between two points (given in decimal degree lat/long). The
# return value is in kilometers.
from math import radians, cos, sin, asin, sqrt
lon1, lat1, lon2, lat2 = map(radians, [p1[0], p1[1], p2[0], p2[1]]) # deg => rad
dlon = lon2 - lon1
dlat = lat2 - lat1
a = sin(dlat/2)**2 + cos(lat1) * cos(lat2) * sin(dlon/2)**2
c = 2 * asin(sqrt(a))
km = 6367 * c # 6367 km is the radius of the Earth
return km
# And we'll need to get the cosine of the angle between two vectors.
def vector_dot(v1, v2):
return (v1[0]*v2[0] + v1[1]*v2[1])
def vector_cosine(v1, v2):
from math import sqrt
return vector_dot(v1, v2) / sqrt( vector_dot(v1,v1) * vector_dot(v2,v2) )
def middle(x):
return x[int(len(x)/2)]
def get_road_segments(feature, target_vector):
# Each road is made up of a number of line segments. It may not be straight.
# Return tuples of (length, location) for segments of this road that are
# pointed toward the target vector.
segments = [[0.0,[]]]
pt0 = feature["geometry"]["coordinates"][0]
for pt1 in feature["geometry"]["coordinates"][1:]:
# Get the length of this segment.
seg_length = haversine(pt0, pt1)
# Instead of getting the angle of the segment and subtracting it from
# the solar azimuth, we'll go directly to computing the cosine of the
# angle between the azimuth and the road. Roads pointing exactly toward
# or away from the azimuth will have cosines of +1 and -1. Whether it's
# toward or away is meaningless, so take the absolute value.
v = (pt1[0]-pt0[0], pt1[1]-pt0[1])
seg_angle = abs(vector_cosine(v, target_vector))
# We want segments very close to +1.
if seg_angle > .9999:
segments[-1][0] += seg_length
elif segments[-1] != 0.0:
# reset
pt0 = pt1
# Return non-zero-length segments, and the middle point in the segment.
segments = [(x[0], middle(x[1])) for x in segments if x[0] > 0]
return segments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.