inv-track

• a simple inventory tracking web application with
  • GraphQL queries, mutations and subscriptions (with Redis PubSub)
  • batching dataloaders (N+1 problem)
  • Docker Compose for development and test, with minimised Docker image for production
• see Extensibility for future improvements
• feature: create locations and assign inventory

Running

• note docker compose v3 is required, but this is the only dependency to serve this project (because docker <3)

development: docker compose up
integration tests: docker compose -f docker-compose.yml -f integration-test.yml up
production: docker compose -f docker-compose.yml -f production.yml up

• navigate to http://localhost:8080 for the UI or use the endpoints at http://localhost:8000/graphql and http://localhost:8000/subscriptions

Using

• using the playground link allows creating GraphQL requests and receiving responses

• using the (very basic) front-end allows for all the basic operations
  • the front-end includes an inefficient implementation of auto-updating from subscriptions

• clicking on the links on the right-hand side of the tables on the list pages give a detailed view and options to edit or delete

Technologies Used

• Rust + actix-web + async-graphql + sqlx
• Postgresql + Redis PubSub
• Docker Compose

GraphQL

• chose GraphQL because of the amount of querying done (also because it's interesting)
• API endpoint is /graphql, playground IDE is /playground
• source code in /server/src/graphql

Batching Dataloaders (N+1 Problem)

• batching dataloaders to mitigate the N+1 Problem
• source code in /server/src/batcher

Subscriptions

• subscriptions for real-time data updates
• endpoint is /subscriptions
• source code in /server/src/graphql/subscription.rs

Transactions

• a transaction is a change in quantity for an item (optionally at a location)
• note if we were to add shipments, they would comprise multiple transactions

Testing

• tests are located in /server/src/main.rs and /server/src/batcher/id_loader.rs in mod test

Extensibility

• add pagination with cursors (https://async-graphql.github.io/async-graphql/en/cursor_connections.html)
• add query complexity and depth limits (https://async-graphql.github.io/async-graphql/en/depth_and_complexity.html)
• use SeaORM instead of SQLx (not an ORM) from experience I'd like to stick with raw queries
  • used SQLx since it was interesting to write queries out instead of using an ORM language
• we explicitly choose to allow for negative quantities
  • in the future, warnings can be added to the response
• use interfaces to return errors and make errors more up-to-spec
• ideally we would be using something like Nginx, but since we're not as focused on deployment, the current setup works
  • in fact, Kubernetes would be a sensible choice, along with making sure everything here is stateless to make this highly available - which for subscriptions, is implemented using Redis

Third-party Crates and Packages

• multiple open-source crates and packages were used to build this. Their licences are included below

Crates

Crate	Licence
actix	MIT/Apache 2.0
async-graphql	MIT/Apache 2.0
async-trait	MIT/Apache 2.0
chrono	MIT/Apache 2.0
dataloader	MIT/Apache 2.0
derive_more	MIT
futures	MIT/Apache 2.0
redis-rs	BSD-3
serde	MIT/Apache 2.0
sqlx	MIT
tokio	MIT

Packages

Packages	Licence
bulma	MIT
core-js	MIT
graphql.js	MIT
graphql-ws	MIT
vue	MIT

Note several dev dependencies were also used.