Gazetteer

A simple, fully offline, accurate boundary based reverse geocoding library in python

Docs - https://gazetteer.readthedocs.io/en/latest/

Reverse Geocoding

It is a process of finding valid address out of coordinates (lat, long), that includes county,state,country etc..

Advantage

• This is a boundary based geocoding library which provides significantly more accurate result than the point based computing counterparts

• Provide data upto 3 administrative divisions

• ADM0 - Country

• ADM1 - State

• ADM2/ADM3 - Towns, Cities, Regions (boundaries which does not have ADM3, ADM2 (districts) will be provided)

• Boundary validation is computed on the fly, which makes it extremely fast and efficient

• No max usage, rate limits since everything runs on yours device fully offline

• No boundary ambiguity or invalid address near border regions, since it is validated by boundary itself

• Covers over 145,000+ boundaries of 210+ countries

How to use gazetteer

Installation

pip install python-gazetteer

Usage

from gazetteer import Gazetteer #import the library
gz = Gazetteer() #gazetteer instance
coordinates = [(-74.0060,40.7128),(76.2673,9.9312)] # list of tuples, in order longitude,latitude
for data in gz.search(coordinates): # gz.search() returns a generator
print(data) # GeocoderResultBaseModel

Result

lat=40.7128 lon=-74.006 result=LocationBaseModel(lat=40.77488420403987, lon=-73.97077035729478, name='New York', admin1='United States', admin2='New York')
lat=9.9312 lon=76.2673 result=LocationBaseModel(lat=9.96268442863286, lon=76.32995967426373, name='Kanayannur', admin1='India', admin2='Kerala')

Return Base Model

from pydantic import BaseModel, Field
class LocationBaseModel(BaseModel):
lat: float = Field(..., description="Centroid latitude of the nearest neighbor")
lon: float = Field(..., description="Centroid longitude of the nearest neighbor")
name: str = Field(..., description="Name of the nearest ADM3(eg towns,cities)/ADM2(eg districts) neighbour")
admin1: str = Field(..., description="Name of the primary administrative division ADM0 (country)")
admin2: str = Field(
...,
description="Name of the secondary administrative division ADM1 (eg: state)",
)


class GeocoderResultBaseModel(BaseModel):
lat: float = Field(..., description="Given latitude")
lon: float = Field(..., description="Given longitude")
result: LocationBaseModel

For large dataset (100000+) coordinate pairs use

from gazetteer import Gazetteer
gz = Gazetteer(mode=2)

This switches to multiprocessing mode, utilizing all cores in your system

If the given coordinates are not found or found inside a water boundary like sea/ocean, it should return None instead of LocationBaseModel

How it works

Data Management

• Boundary Data is sourced from Geoboundaries
• Data is stored in two places, sqlite db and a csv file
• A simplified boundary is stored inside sqlite db in the format of WKB(Well Known Library)
• The metadata for locations is stored as csv

Computation

• Uses KDTree nearest neighbour algorithm from scipy to find the closest boundary point
• Validates the nearest neighbour using the boundary provided in the sqlite

Note: Regions that doesn’t have an available ADM3 boundary will return ADM2 instead as the nearest neighbour

Acknowledgements

• Geoboundaries - A big shout out for making their amazing dataset free and open-source
• Ajay Thampi - A part of this implementation was inspired by his reverse_geocoder

License

This project is licensed under the GNU Lesser General Public License (LGPL v2.1). See the LICENSE file for details.