Proposal: reverse geocoder

geocode/geocode.py

@@ -14,6 +14,7 @@
 import hashlib
 import json
 from .flags import flags
+from .reverse_geocoder import ReverseGeocoder
 
 
 logging.basicConfig(level=logging.INFO, format='%(asctime)s [%(levelname)-5.5s] [%(name)-12.12s]: %(message)s')
@@ -312,6 +313,9 @@ def get_arguments_hash(self):
         geocode_args = [str(self.min_population_cutoff), str(self.large_city_population_cutoff)] + self.location_types
         return hashlib.sha256(','.join(geocode_args).encode()).hexdigest()[:15]
 
+    def build_reverse_geocoder(self):
+        return ReverseGeocoder(self.geo_data)
+
     # private
 
     def _get_location_types(self, location_types):

geocode/reverse_geocoder.py

@@ -0,0 +1,18 @@
+import numpy as np
+from sklearn.neighbors import KDTree
+
+class ReverseGeocoder():
+
+    def __init__(self, geo_data):
+        self.geo_data = geo_data
+
+        coord_arr = np.asarray(list((d[3], d[4]) for d in self.geo_data))
+
+        self.tree = KDTree(coord_arr)
+
+    def nearest_geodata(self, lon, lat):
+        value = (lon, lat)
+
+        index = self.tree.query([value])[1][0][0]
+
+        return self.geo_data[index]

requirements.txt

@@ -2,3 +2,4 @@ pandas>=1
 tqdm
 flashtext
 joblib
+scikit-learn