Reverse Geocoder

This repository holds a small web service that performs reverse geocoding to determine whether a user specified location is within the geographic bounds of a country. If it is then the response will contain attributes associated with the matched country such as name, iso label, etc.

I created this simple demonstrator app to learn about FastAPI and PostGIS. It is built using Python and a PostGIS database loaded with country outline polygons obtained from shape files.

API Endpoints

All endpoints are located at /v1/reverse-geocoder/ and are accessible by HTTP.

The OpenAPI specification can be read from /v1/reverse-geocoder/openapi.json.

A SwaggerUI that renders the OpenAPI specification can be found at /v1/reverse-geocoder/docs. The root / will also redirect to the SwaggerUI docs page.

Reverse Geocoder Endpoint

To access the reverse geocoder send a POST to /v1/reverse-geocoder/ with a payload containing the location of interest. See further down for examples using curl.

When the user supplied point is within a country then the response contains attributes of the country that matched. When the user supplied point is not within a country then the response indicates that no country matched.

The response will always contain a copy of the query parameters so that it provides some context for the data. When the location is within a country bounds then the response payload will look like this:

{
  "location":{
    "latitude":-42.239392,
    "longitude":146.558384,
    "altitude":null
  },
  "country":{
    "name":"Australia",
    "iso2":"AU",
    "iso3":"AUS"
  }
}

If the point is not within a country boundary then the response will look like this:

{
  "location":{
    "latitude":-35.031741,
    "longitude":138.119541,
    "altitude":null
  },
  "country":null
}

Demo

To simplify running the reverse geocoding service a Docker compose configuration example is included. However, a few set up steps need to be run first to prepare data that will go into the database - as it is not stored in this repository.

Clone Repo

Start by cloning this repo.

$ git clone https://github.com/claws/reverse-geocoder.git
$ cd reverse-geocoder

Preparation

This demonstration uses world country boundary outlines in shape file format (.shp, .dbf, .shx files) which contain encoded polygons along with other attributes. However, the files are not stored in this repo so they need to be downloaded and then converted into SQL statements.

The following steps show how to download the content from here, decompress it and then convert it into SQL statements that can later be run to insert the contents into the database.

It is important that the name of the generated SQL file is 90-shapes.sql to ensure it gets run after the builtin PostGIS initialization script (which is 10_postgis.sh). This file will be used as a volume mount in the Docker compose configuration.

shp2pgsql is a command line tool that comes with PostGIS. It converts shape files into a SQL format that can be imported into a PostGIS database. The '-G' specifies the use of the geography data type. The '-I' option creates a spatial index after the table is created. This is strongly recommended for improved performance. To run the last command you may need to install PostGIS on your host machine to get the shp2pgsql tool - there may even be a Docker container that has it too.

$ cd database
$ wget http://thematicmapping.org/downloads/TM_WORLD_BORDERS-0.3.zip
$ unzip TM_WORLD_BORDERS-0.3.zip
$ shp2pgsql -G -I TM_WORLD_BORDERS-0.3.shp countries > 90-shapes.sql
$ cd ..

The shape file produces columns containing the following structures:

Start services

Start the database and web server using Docker compose. As part of the startup steps the database runs initialisation script - which will run SQL file created in the setup steps above.

$ docker-compose up --build

The FastAPI framework is built upon OpenAPI and it supports a builtin viewer for the interface specification. These allow developers to perform manual tests.

Once the services start up you should be able to access the SwaggerUI web user interface here which should redirect you to the docs site.

Click the POST /v1/reverse-geocoder/ row to expose details about the endpoint. This endpoint is implemented to accept a payload, rather than URL parameters, as it can then rely on the specification and Pydantic to simplify the interface between the API endpoint and the database layer.

Click the Try it out button. This changes the form to allow user input. Leave the default payload structure then click Execute. The Responses section should get populated with the data returned from the web service - which in this case indicates that the point was in Australia.

Try changing the location to some of those in the table below. The last two should return 'null' for the country as they are located in a sea.

The Swagger UI also shows the equivalent curl command to use too.

Check using curl

The Reverse Geocoder Service REST API can be used from curl too.

When the location is within a country bounds then the response payload will contain country attributes.

$ curl -X POST "http://localhost:8000/v1/reverse-geocoder/" \
  -H "accept: application/json" \
  -H "Content-Type: application/json" \
  -d "{\"location\":{\"longitude\":146.558384,\"latitude\":-42.239392}}"

{
  "location":{
    "latitude":-42.239392,
    "longitude":146.558384,
    "altitude":null
  },
  "country":{
    "name":"Australia",
    "iso2":"AU",
    "iso3":"AUS"
  }
}

When the user supplied point is not within a country then the response indicates that no country matched.

$ curl -X POST "http://localhost:8000/v1/reverse-geocoder/" \
  -H "accept: application/json" \
  -H "Content-Type: application/json" \
  -d "{\"location\":{\"longitude\":138.119541,\"latitude\":-35.031741}}"

{
  "location":{
    "latitude":-35.031741,
    "longitude":138.119541,
    "altitude":null
  },
  "country":null
}

Developer Notes

Data Server

It can be useful during development to run the Python web server locally. The instructions below show how to do that.

Create a Python virtual environment.

$ cd web-service
$ python3.8 -m venv venv --prompt fast
$ source venv/bin/activate
(fast) $ pip install pip -U
(fast) $ pip install -r requirements.dev.txt
(fast) $ pip install -r requirements.txt

Apply code style

$ black app

Run the data server. Use --reload to enable automatic reloads on code changes.

(fast) $ uvicorn app.main:app --reload --log-level info
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO:     Started reloader process [2617]
INFO:     Started server process [2619]
INFO:     Waiting for application startup.
INFO:     Connected to database postgresql://postgres:********@localhost:54321/postgres
INFO:     Application startup complete.

Database Inspection

It can be useful to attach to the database directly to test out queries. Use docker-compose to connect to the running database by attaching to the container. Once attached the psql tool can be used to execute a query identical to that done by the web service to check if a point lies within a polygon.

The PostGIS ST_Covers function which returns TRUE if A covers B (i.e. no points of B are outside A).

$ docker-compose run database bash
postgres@4f5d94706a9b:/$
postgres@4f5d94706a9b:/$ psql --host database -U postgres postgres
Password for user postgres:
postgres=#
postgres=# SELECT name,fips,iso2,iso3 FROM countries WHERE ST_Covers(countries.geog, ST_GeographyFromText('POINT(146.558384 -42.239392)'));
   name    | fips | iso2 | iso3
-----------+------+------+------
 Australia | AS   | AU   | AUS
(1 row)
postgres=#
postgres=# SELECT name,fips,iso2,iso3 FROM countries WHERE ST_Covers(countries.geog, ST_GeographyFromText('POINT(138.119541 -35.031741)'));
 name | fips | iso2 | iso3
------+------+------+------
(0 rows)
postgres=# \q
postgres@4f5d94706a9b:/$ exit