engine: vectorise distance matrix calc

engine.py

@@ -3,7 +3,6 @@
 import overpy
 import numpy as np
 import pandas as pd
-import geopy.distance as distance
 
 from tqdm import tqdm
 
@@ -12,10 +11,12 @@ BRANDS: dict[str, str] = {
     "greggs": "[\"brand:wikidata\"=\"Q3403981\"]",
 }
 
+DATA_FOLDER = ""
+
 EncodedLocation = list[tuple[float, list[float]]]
 
 
-def fetch_data(brand: str) -> list[tuple[float | None, float | None]]:
+def fetch_data(brand: str) -> list[tuple[float, float]]:
     """Fetch a list of locations from OSM."""
     api = overpy.Overpass()
 
@@ -25,10 +26,10 @@ def fetch_data(brand: str) -> list[tuple[float | None, float | None]]:
     result = []
 
     for way in query.ways:
-        result.append((way.center_lat, way.center_lon))
+        result.append((float(way.center_lat), float(way.center_lon)))
 
     for node in query.nodes:
-        result.append((node.lat, node.lon))
+        result.append((float(node.lat), float(node.lon)))
 
     for (lat, lon) in result:
         if (lat is None) or (lon is None):
@@ -37,31 +38,37 @@ def fetch_data(brand: str) -> list[tuple[float | None, float | None]]:
     return result
 
 
+def spherical_dist(pos1, pos2, r=6378137):
+    """Calculate sperical distances between two arrays of coordinates."""
+    pos1 = pos1 * np.pi / 180
+    pos2 = pos2 * np.pi / 180
+
+    cos_lat1 = np.cos(pos1[..., 0])
+    cos_lat2 = np.cos(pos2[..., 0])
+    cos_lat_d = np.cos(pos1[..., 0] - pos2[..., 0])
+    cos_lon_d = np.cos(pos1[..., 1] - pos2[..., 1])
+    return r * np.arccos(cos_lat_d - cos_lat1 * cos_lat2 * (1 - cos_lon_d))
+
+
 def encode(location: tuple[float, float]) -> EncodedLocation:
     """Encode a location."""
 
-    print("fetching")
+    greggs = np.array(fetch_data("greggs"))
-    greggs = fetch_data("greggs")
-    print("fetched")
-    #greggs = [(55.85,-4.02),(52.443,-1.833),(51.28,-1.0)]
 
-    #relative distance matrix
+    repeat_rows = np.tile(greggs, (len(greggs), 1, 1))
-    dist_matrix = np.zeros((len(greggs),len(greggs)))
+    repeat_cols = np.transpose(repeat_rows, (1, 0, 2))
-    for i in tqdm(range(len(greggs))):
+    dist_matrix = spherical_dist(repeat_rows, repeat_cols)
-        first=greggs[i]
-        for j in range(len(greggs)):
-            second=greggs[j]
-            #calculate the distance between i and j coordinates
-            #dist_matrix[i,j] = np.sqrt((first[0]-second[0])**2 + (first[1]-second[1])**2 )
-            dist_matrix[i,j] = distance.distance(distance.lonlat(*first), distance.lonlat(*second)).km*1000
-    print(dist_matrix)
 
     #find closest greggs
-    distances = pd.Series(np.zeros(len(greggs)))
+    # distances = pd.Series(np.zeros(len(greggs)))
-    for i in range(len(greggs)):
+    # for i in range(len(greggs)):
-        current = greggs[i]
+    #     current = greggs[i]
-        #distances[i] = np.sqrt((current[0]-location[0])**2 + (current[1]-location[1])**2 
+    #     #distances[i] = np.sqrt((current[0]-location[0])**2 + (current[1]-location[1])**2
-        distances[i] = distance.distance(distance.lonlat(*current), distance.lonlat(*location)).km*1000
+    #     distances[i] = distance.distance(distance.lonlat(*current), distance.lonlat(*location)).km*1000
+
+    repeated = np.tile(location, (len(greggs), 1))
+    distances = spherical_dist(repeated, greggs)
+    distances = pd.Series(distances)
 
     print(distances)
     distances = distances.sort_values()
@@ -104,7 +111,7 @@ def main():
     #greggs = fetch_data("greggs")
     #print(f"Query done - got {len(greggs)} Greggs!")
 
-    print(format_location(encode((0.091659, 52.210796))))
+    print(format_location(encode((52.210796, 0.091659))))
 
 
 if __name__ == "__main__":

requirements.txt

@@ -2,4 +2,3 @@ overpy
 django
 numpy
 pandas
-geopy