Hawker finder

This web application finds the nearest hawker center within 20km of the input location (lat, long) within Singapore .

Requirements

1. docker engine or docker desktop on windows/ mac Get docker
2. docker-compose (should come together with docker desktop) Get compose

Starting up

The web application starts by running (from within the project folder)

docker-compose up -d

on your shell with docker installed. Run the following from outside the project folder:

docker-compose up -f <path to folder>/docker-compose.yml -d

The first run might take some time as images are downloaded and built.

Check out the app by going to localhost:8501 in a browser.

Note: You might need to disable firewall/ set firewall rules to allow port 8501.

Components

The application comprises 3 containers

1. ui container developed with Streamlit
2. db container - MongoDB instance for geolocation queries
3. ingest container that downloads dataset from data.gov.sg

Checking out the project

Use the standard docker-compose CLI to explore the project

1. Display running containers: docker-compose ps
2. Display logs: docker-compose logs
3. Run project attached: docker-compose up
4. Run tests
   • UI container docker-compose exec ui pytest
   • Ingest container docker-compose exec ingest pytest

Considerations

Containers for development

This app uses docker containers starting from development. This ensures a consistent, standardised environment for development, testing and running the application.

1. A bridge docker network is used for communication between containers
2. A docker volume is mounted on mongodb container for persistent storage of db data.
3. The <project folder>/compose directory contains the Dockerfiles and other files required for building the images.

Ingestion

Data is brought into the application by the ingest container. The following steps are taken:

1. Data is downloaded in a stream and unzipped.
2. Data is processed line by line - it is loaded as a JSON and a regex pattern is used to extract the PHOTOURL and NAME fields.
3. If the pattern matches and JSON is valid, a type validator is used to validate the data.
4. The data is inserted into MongoDB
5. Geolocation indexes are created on the GeoJSON field
6. Data in DB is persisted in a docker volume

The above steps are written as download_data and extract_data functions and tested independently. The ingest container also exposes these steps as a REST api for the user to reload data.

Querying

The web UI captures users' [lat, long] input values and queries the DB on the geolocation index. The query returns the 5 nearest locations within a (configurable) 20km radius. The UI renders the returned documents and also plots them on a map.