FeedMe Tech Test

Running The Stack In Docker

All elements of this tech test can be spun up inside docker containers.

I have provided a Dockerfile for each service, a shell script to simplify building the images (build_docker_images.sh), and have updated the docker-compose.yml to include these services.

To build and run the stack in docker:

./build_docker_images.sh
docker compose up

Note:

• The reader and consumer containers are configured to wait for rabbitmq to pass healthchecks before starting.
• Both reader and consumer have configurable log levels (info, warn and error). This is set to warn in docker-compose.yml but can be changed to info for more verbose logging on a per-message basis.

Components

reader

This application connects to the TCP socket and parses the packets into JavaScript objects according to the provided schema.

These JavaScript objects represent new or updated Fixtures, Markets and Outcomes. These updates are then published to a RabbitMQ via RabbitPublisherClient.publishMessage which takes a parsed Message, JSON serialises it and then pushing it to Rabbit.

RabbitPublisherClient uses a private constructor to allow the creation of a static async constructor - new. This was done to simplify initialisation of the client.

The URL of the rabbit vhost and the name of the queue are configurable using environment variables, along with the hostname and port of the TCP Provider.

consumer

This application connects to the RabbitMQ queue and consumes JSON encoded Messages via RabbitConsumerClient. On receiving a Message it is JSON parsed and then persisted into MongoDB via MongoService, and then the message is acknowledged.

The consumer is set up to only receive one message at a time. This, along with the manual acknowledgement once MongoService.handleMessage has completed ensure messages are not missed and are share equally between consumers.

Multiple consumers can be spun up as required.

feedme-viewer

Visible at http://localhost when the stack is running in docker.

This application is a NextJS web application that reads from the MongoDB database populated by reader and consumer and presents this information in a user-friendly way.

Any data with displayed: false is ignored and cannot be seen by the user. If an event, market, or outcome is marked as suspended: true, then the child outcomes are shown, but cannot be selected by the user.

React has been used for the UI, and for styling I have used CSS Modules. The styling is based off the Sky Bet website.

The next features I would implement is some form of client-side store for the users Bet Slip. This would then allow users to populate the bet slip using the buttons on the event's page. After this I would look into user login and account functionality, along with placing of bets.

Note: When first starting the stack, the MongoDB is empty, so the UI won't be able to show much until it is populated with events/markets/outcomes that have displayed set to true.

Technology Choices

I have used ts-node and TypeScript for development on reader and consumer, feedme-viewer uses Typescript and NextJS.

Where unit testing has been done, Jest has been used.

RabbitMQ has been used for message queueing. This was selected as I have more experience with RabbitMQ than Kafka.

MongoDB was used as the NoSQL database, as suggested by the docker-compose.yml file.

Dockerfiles

The dockerfiles for reader and consumer both build from a base dockerfile that contains ts-node and the common library package.