State Machines and How they Shape Mobile Architectures

At Sendwave, we wanted to rebuild our data syncing mechanism on Android. That's when we came up with the idea of using a State Machine which helped the team build a much more solid and scalable data syncing system. State Machines are a very good solution for making sense of things that can be quite complex. So many companies are using State Machines as part of their architecture (like Tinder, GoDaddy, Square and more). Some popular architecture patterns like MVI or Redux were inspired by the concept of State Machines. And it has even made some language features possible (Kotlin Coroutines). We will start by understanding State Machines from the ground up and then we will go through all the examples above and will finish by showing how we used a State Machine at Sendwave

Transcript

1. STATE MACHINES
 AND HOW THEY SHAPE
 MOBILE ARCHITECTURES DROIDCON KENYA By Rygel Louv
2. Rygel Louv Senior Android Engineer Sendwave
3. Agenda 1 2 3 4 Data Sync Problem State Machines What is it? Expect some maths Real World Applications of State Machines Solving the Data sync problem 3
4. State Machines A fi nite-state machine (FSM) or fi nite-state automaton (FSA, plural: automata), fi nite automaton, or simply a state machine, is a mathematical model of computation. It is an abstract machine that can be in exactly one of a fi nite number of states at any given time. The FSM can change from one state to another in response to some inputs; the change from one state to another is called a transition - Wikipedia
5. Why Am I talking about this?
6. Data sync problem
7. Data sync problem Sync some user information Sync countries con fi guration data Sync user recipients Sync remote con fi g Etc
8. Data sync problem: constraints Is it a new session? Is everything synced already? What if there is a failure? Should we allow the user to interact? What if the user logs out while syncing? What if we have more constraints to add to this?
9. Architecture pattern and feature complexity Sometimes MV* is not enough Some features are more complex than others Some feature have no UI Third party integration, global app component, internal reusable library, a feature with multiple states/statuses, 
 or shared data across screens Is MV* always how we should approach problems? Leverage a speci fi c technique or design pattern to solve a problem
10. A State Machine as solution to some problems State Machines could be a solution to reduce code complexity Make sense of things that could be complicated to understand Self-document or visualize a speci fi c fl ow
11. Does this sound interesting? Then lets do some maths
12. State Machines: Theory Math de fi nition of an FSM
13. State Machines: Theory FSM: our Finite State Machine Q: fi nite set of States Σ: fi nite set of Inputs δ: the transition function q0: initial state F: set of fi nal states
14. State Machines: Theory q0 is element of Q F is a subset or equal to Q δ: the transition function is a function that takes
 a state within Q and an Input within Σ to produce 
 a new state which should be within F The rules: q0 ∈ Q F ⊆ Q
15. State Machines: Theory Example: the light switch
16. State Machines: Theory Example: the light switch
17. State Machines: Theory Example: the light switch Where
18. State Machines: Theory Example: the light switch
19. State Machines: From Theory to Kotlin Constructor de fi nition
20. State Machines: Naive Implementation
21. State Machines: Naive Implementation
22. State Machines: Naive Implementation
23. State Machines: Naive Implementation
24. But how can I know when to build a State Machine?
25. Recognizing a State Machine pattern Adding states or status to your program Growing complexity Having many functions in the form of isSomething Many booleans in the program could also be an indicator All of these toghether would de fi nitely be the perfect indicator
26. But who is using this in the real world?
27. Shopify Login complexity
28. Shopify Used a State Machine system to reduce complexity and make the login scalable
29. Shopify If you want to know more:
30. Tinder Scaled the Payment Flow using State Machines
31. Tinder Happy Flow
32. Tinder Failure Flows
33. Tinder Side Effect to handle business logic
34. Tinder Open source declarative State Machine https://github.com/Tinder/StateMachine
35. Square Work fl ow https://github.com/square/work fl ow A library that supports writing business logic and complex UI navigation logic as state machines, thereby enabling con fi dent reasoning about state and validation of correctness.
36. Redux: glori fi ed state machine
37. MVI: glori fi ed state machine
38. Kotlin Coroutines: Suspending State Machines Kotlin Coroutines let you create asynchronous code that looks like synchronous This is possible thanks to a technique called “Continuation Passing Style” Suspend functions generate a State Machine to enable the Continuation object to resume the execution
39. Kotlin Coroutines: Suspending State Machines Direct Style
40. Kotlin Coroutines: Suspending State Machines Continuation Passing Style
41. Kotlin Coroutines: Suspending State Machines Suspend State Machine
42. Kotlin Coroutines: Suspending State Machines Example:
43. Kotlin Coroutines: Suspending State Machines Example:
44. Kotlin Coroutines: Suspending State Machines Example:
45. Kotlin Coroutines: Suspending State Machines
46. Kotlin Coroutines: Suspending State Machines
47. So… how about Sendwave?
48. Reminder: Data sync problem Sync some user information Sync countries con fi guration data Sync user recipients Sync remote con fi g Etc Is it a new session? Is everything synced already? What if there is a failure? Should we allow the user to interact? What if the user logs out while syncing? What if we have more constraints to add to this?
49. Data sync problem: Solution
50. Sendwave Data sync problem
51. Data sync State Machine Implementation
52. Data sync State Machine Implementation
53. Data sync State Machine Implementation
54. Data sync State Machine Implementation
55. Data sync State Machine Implementation
56. Data sync State Machine Implementation Reducer
57. Data sync State Machine Implementation
58. Data sync State Machine Implementation
59. Data sync State Machine Implementation
60. Data sync problem: Solution
61. What have we learned about State Machines? State Machines could be a solution to reduce code complexity Make sense of things that could be complicated to understand Self-document or visualize a speci fi c fl ow Not that dif fi cult to build and integrate in your app Improve testability Improve scalability Improve debuggability/maintainability Go beyond the MV* architectural pattern and give it a different shape to better solve problems.
62. What have we NOT learned about State Machines? Not a solution for all your problems Not a solution to solve all types of complexity Not necessarily a replacement to your state management approach Not a dark tool/technique only for “experts”
63. Asante sana Matondi mingi Jërëjëf
64. @Rygellouv @rygelouv @rygel Questions? Teranga Engineering Podcast
65. @Rygellouv @rygelouv @rygel https://sp4ghetticode.medium.com/ fi nite-state-machines-in-kotlin-part-1-57e68d54d93b https://medium.com/tinder/scaling-out-tinder-android-payment- fl ow-using-state-machine-e14ef0591b6 https://medium.com/swlh/mvi-architecture-with-android-fcde123e3c4a https://www.droidcon.com/2018/06/27/the-rise-of-the-state-machines/ https://www.droidcon.com/2019/08/07/square-work fl ow-android-and-ios-apps-composed-of-state-machines-de fi ning-ui-and-navigation/ Resources https://medium.com/google-developer-experts/coroutines-suspending-state-machines-36b189f8aa60 https://labs.pedrofelix.org/guides/kotlin/coroutines/coroutines-and-state-machines