Gett Delivery - Breaking the State Machine

Learn How Gett Delivery leveraged Firebase and RxSwift to break its state machine, creating an architecture suited for the delivery world.

Transcript

1. Breaking the State Machine Oren Kosto
2. About Myself • Oren Kosto ! • Embarrassing Husband/Dad • Communication Systems Engineer • Writing code for a living for the last 5 years $ • iOS & Web Developer @ Gett for 1 year • Multi-disciplinary, Generalist • First time speaking in a tech meetup... orenk86 oren.kosto oren-kosto
3. Gett Growth • Expanding Gett's capabilities and services beyond the world of taxis • Developing additional products on top of Gett’s existing infrastructure
4. What are we going to talk about? • Gett Delivery Reboot • (Breaking) State machines
5. The “Traditional” Taxi Ride • On demand ride • Each passenger needs a dedicated driver • Deterministic process: • Order • Find a driver • Pickup • Drop Off • Charge • Built with state machines
6. Gett Delivery • On Demand deliveries • Every Package needs a dedicated courier • Same deterministic process as a taxi • Built using Gett’s existing taxi infrastructure • Taxi ride state machine
7. Gett Delivery - DaaS • Delivery As A Service • Same hour, same day, next day • Parcels are delivered from a depot to multiple drop-off points during the day • Higher paying multi-point future jobs for couriers • Same supply, much higher capacity
8. Gett Delivery - DaaS • Delivery As A Service • Same hour, same day, next day • Parcels are delivered from a depot to multiple drop-off points during the day • Higher paying multi-point future jobs for couriers • Same supply, much higher capacity
9. Journeys Optimization Overview Bundle Deliveries
10. Challenges Optimistic UI Offline Functionality Changes On The Fly Process Not Deterministic Conflict Resolution Event Management Connectivity & Sync Recovery
11. Building A New Product
12. Delivery One-To-Many Driver Actions Shared Model, Realtime Flexible & Expandable
13. Structure • Firebase Realtime Database & Cloud Functions • RxSwift: Firebase events wrapped in Observables • MVVM: View Model for every UI component - per action • Dependency Injection with Swinject • Protocol Oriented (action types, services) • (Almost) Stateless
15. Driver Action Protocols Exchangeable Action Contactable Action Mappable Action
16. Return Driver Actions Flow Delivery Service Cloud Functions Realtime Database Drive Contact Deliver
17. Implementation Map View Drive To Pickup VM Complete Action View MappableAction Presentable CompletableAction Presentable ContactableAction Presentable Call Recipient Component MappableAction ViewModeling CompletableAction ViewModeling ContactableAction ViewModeling Map View VM Complete Action VM Call Recipient VM Drive VC
18. Implementation Firebase Service Driver Actions Service Analytics Manager Delivery Service GetTogether MultiPickup VM Delivery Collect VM Deliver VM Return VM Drive To Pickup VM Drive To Drop-Off VM
19. Implementation Firebase Service Driver Actions Service Analytics Manager Delivery Service GetTogether MultiPickup VM Delivery Collect VM Deliver VM Return VM Drive To Pickup VM Drive To Drop-Off VM
20. Our Experience • The Good • Firebase Realtime DB - Just works • Isolated the “problem” to one place, single source of truth • Extremely modular, highly reduced the cost of adding features • Minimized regression bugs • The Ugly • Firebase - Easy to abuse, expensive. Use it wisely • Multiple copies of the data on the server - must avoid conflicts • Hard learning curve, might seem like over-engineering at first • Cloud functions - hard to debug, hard to change, no version control
21. Thank You! • Icons made by Freepik and SmashIcons from www.flaticon.com