DBeaver 'Entity Relationship Diagram':

Tables were designed in a simple way without complex data types. In production, we would probably add JSON columns to each major table called 'attributes' or something similar to allow flexibility. Table definitions can be found here.

• Spring Initializr was used to create a Spring Boot Java application with dependencies managed by the Gradle build tool.

• Spring Data Rest was installed to provide RESTful access if required. See the description for navigating the REST portion of the API).

• GraphQL SPQR was used to develop a GraphQL API using 'code-first' approach in the interest of speed and code simplicity (automatic schema generation). The GraphQL schema can be viewed here or browsed using the following clients:

  • Altair GraphQL client - https://graphql.thomasportwood.com/altair
  • Voyager GraphQL Browser - https://graphql.thomasportwood.com/voyager

• GraphQL operations were included to support:

  • Adding new trucks, beans (Kona, Arabica, etc), preparation types (espresso, french press, etc), and item statuses
  • Creating orders for new or existing customers (distinguished only by phone number in this simple schema)
  • Updating the status of individual items in an order
  • Deleting items from orders
  • Viewing orders for a specific truck

• A GraphQL subscription was included to support the development of the barista app. It works!! It can be tested by:

  • Browsing to Altair
  • Ensuring the subscription URL is correct by selecting 'Subscription URL' on the left toolbar and setting it to wss://graphql.thomasportwood.com/subscriptions
  • Add a subscription to the request like:

    subscription {
      orderSubscription {
        id
        customer {
          phone
        }
        items {
          beanTypeId
          preparationTypeId
        }
      }
    }
    

  • Click on 'Send' (you should see a 'Clear' and a 'Stop' button appear)
  • In another window, submit an order like:

    mutation {
      createOrder(
        input: {
          truckId: 1
          customerPhone: "425-343-3019"
          items: [{ beanTypeId: 1, preparationTypeId: 1 }]
        }
      ) {
        createdAt
        id
      }
    }
    

  • Check the first window for an update! Super powerful!

• This native SQL query reads all items from the previous 30 days for a given truck, groups the items by bean type, and sorts the groups by ascending total count. This will give us a running metric to consider which bean to replace.
• Note that the bean stats for a specific truck are exposed in a GraphQL query

• A Promotion Calculator was created to contain logic that should efficiently produce usable data for distribution of promotional coupons. A group of orders from a given time range is sorted by ascending creation date, and the customer of every order that is eligible for a coupon is included in the response. An enum is used to represent the type of coupon the customer will receive. The calculation is exposed in a GraphQL query:

  query {
    promotionWinners(truckId: 1, days: 7){
      couponType
      phone
    }
  }
  

• I considered using a simpler REST approach by way of Spring Data Rest and following the HATEOAS principle which I find interesting and useful.

• The 'resolver' only deals with a single data source at the moment (MYSQL), but I would still break it up in to separate classes in anticipation of additional data sources.
• The GraphQL tests only served as a debugging tool. I would reorganize them to be more atomic and test failures and edge cases.
• I didn't include any authentication or authorization. I would probably start with Auth0 or a similar solution.