The SAM Pattern: State Machines and Computation
Download as PPTX, PDF3 likes1,129 views
The document discusses the importance of state-action models in programming, emphasizing how they can improve program design and developer thinking. It introduces TLA+ (Temporal Logic of Actions) and its role in describing computations and state machines, showcasing examples like the factorial algorithm. Furthermore, it examines the SAM (State-Action-Model) pattern as a framework for decoupling user interface components from application logic, highlighting its application in modern web development frameworks.
![State machines provide a
framework for much of computer
science
• “Don’t be obsessed by computer languages
[and frameworks]”
• Keep a clear delineation between
o Programming model
o Wiring
o Architecture
-- Dr Lamport](https://image.slidesharecdn.com/sam-state-machines-and-computation-170227110333/85/The-SAM-Pattern-State-Machines-and-Computation-7-320.jpg)
![Introducing TLA+
(Temporal Logic of
Actions)
computing
i = n
f = 1
done
[i==1]
i = i - 1
f = f * i
[i>1]
error
[i==0]](https://image.slidesharecdn.com/sam-state-machines-and-computation-170227110333/85/The-SAM-Pattern-State-Machines-and-Computation-17-320.jpg)
![Primed Variables
computing
i' = n
f' = 1
done
[i==1]
i' = i - 1
f' = f * i'
[i>1]
error
[i==0]
The meaning of primed
variables is the variable's
value in the next state.](https://image.slidesharecdn.com/sam-state-machines-and-computation-170227110333/85/The-SAM-Pattern-State-Machines-and-Computation-18-320.jpg)
![TLA+ is fundamentally
different from classical State
Machines
computing
i = n
f = 1
done
[i==1]
i = i - 1
f = f * i
[i>1]
error
[i==0]
Actions cannot
manipulate State,
and cannot decide
of the Target Stat
e](https://image.slidesharecdn.com/sam-state-machines-and-computation-170227110333/85/The-SAM-Pattern-State-Machines-and-Computation-21-320.jpg)
![View=State(Model.present(Action(event))).then(nap);
“SAM is simple, I [have] never seen a
concept such minimalist and
powerful, [in particular] in its
decoupling potential”
— Daniel
Neveux
next action](https://image.slidesharecdn.com/sam-state-machines-and-computation-170227110333/85/The-SAM-Pattern-State-Machines-and-Computation-42-320.jpg)
The SAM Pattern: State Machines and Computation
- 1. I believe that the best way to get better programs is to teach programmers how to think better -- Dr. Leslie Lamport
- 2. SAMThe State Action Model Pattern Jean-Jacques Dubray © https://creativecommons.org/licenses/by-nc-sa/4.0/
- 4. Action State Thing Relationship Dynamic Static Physical Conceptual The STAR Diagram
- 5. Light Energy Mass Force Dynamic Static Physical Conceptual STAR is as fundamental to Computing as the FLEM Diagram is to Physics
- 7. State machines provide a framework for much of computer science • “Don’t be obsessed by computer languages [and frameworks]” • Keep a clear delineation between o Programming model o Wiring o Architecture -- Dr Lamport
- 8. Classical State Machines A set of states Q = {q0, q1, q2} A start state q0 A set of transitions which are triplets (q,a,q') members of Q x Act x Q
- 9. State Machines can Compute… “the die-hard problem” 5 liters 3 liters 4 liters
- 11. There are several ways to define computation • A computation is a sequence of steps, called a behavior • There are three common choices for what a step is, leading to three different kinds of behavior: • Action Behavior a step is an action, an action behavior is a sequence of actions • State Behavior A step is a pair <Si , Si+1> of states, a state behavior is a sequence s1 → s2 → s3 → · · · of states • State-Action Behavior A step is a triple < Si , α, Si+1 > where S are states and α an action -- Dr. Lamport
- 12. We just created a language that computes volumes of liquid (not sure it will be popular though) 1. Fill Big 2. Fill Small from Big 3. Empty Small 4. Fill small from Big 5. Fill Big 6. Fill small from Big State-Action Behavior Action Behavior 5 liters 3 liters
- 13. Programming Languages are Designed to Make it Easy to Create State Machines 1. Fill Big 2. Fill Small from Big 3. Empty Small 4. Fill small from Big 5. Fill Big 6. Fill small from Big
- 14. How do we describe a “standard” program as a state machine? var fact = function(n) { return (n == 1) ? 1 : n * fact(n-1); } var fact = function(n) { var f = 1, i =1 ; for (i = 1; i <= n; ++i) { f = i * f; } return f ; }
- 15. How do we describe a “standard” program as a state machine? computing i = n f = 1 done i==1 i = i - 1 f = f * i i>1 error i==0
- 16. Recursion and State Machines Computing i = 2 n = 5 f1 = 1 fi = i * fi-1 i = i + 1 i<=n error i==0 f1 = 1 f2 = 2 f3 = 6 f4 = 24 f5 = 120 i = 5 f = f5 !fi && i<=n done i>n
- 17. Introducing TLA+ (Temporal Logic of Actions) computing i = n f = 1 done [i==1] i = i - 1 f = f * i [i>1] error [i==0]
- 18. Primed Variables computing i' = n f' = 1 done [i==1] i' = i - 1 f' = f * i' [i>1] error [i==0] The meaning of primed variables is the variable's value in the next state.
- 19. Next-State Predicate computing i' = n f' = 1 done i' = i - 1 f' = f * i' error error = (i' == 0) done = (i' == 1) computing = !done && !error f = f' i = i'
- 20. What is a Programming Step? computing i' = n f' = 1 done i' = i - 1 f' = f * i' error error = (i' == 0) done = (i' == 1) computing = !done && !error f = f' i = i' 1 Compute next-state property values 2 Update State 3 Decide what happens next
- 21. TLA+ is fundamentally different from classical State Machines computing i = n f = 1 done [i==1] i = i - 1 f = f * i [i>1] error [i==0] Actions cannot manipulate State, and cannot decide of the Target Stat e
- 22. What Changed? Not much… Functions State(s) i' = i - 1 f' = f * i'Action error = (i' == 0) done = (i' == 1) computing = !done && !error f = f' i = i' Next-State Relation (Function) Next-Action Function if (computing) { … } i' = n f' = 1Initial State
- 23. What Changed?
- 24. What is State? What is Type?
- 25. What is State? What is Type? A Type, Thing, Model… is a list of property values { rpm : 2400 } A State is a range of property values (sometimes reduced to one value) started = (rpm > 900)
- 26. TLA+ Specification of Factorial Algorithm Action State Type
- 27. TLA+ Specification of Factorial Algorithm i = 5 5 1 5 mult
- 28. TLA+ Specification of Factorial Algorithm i = 5 5*1 1 5 mult 4
- 29. TLA+ Specification of Factorial Algorithm i = 5 5*1 1 5 mult 4
- 30. TLA+ Specification of Factorial Algorithm i = 5 5 4 mult test
- 31. TLA+ Specification of Factorial Algorithm i = 5 5*4 5 4 mult 3
- 32. How do we use TLA+ in Practice?
- 33. Modern GUIs have become just a Node in a Dynamic Distributed System • your friends, • players • your watch • IoT sensors • …
- 34. Events and Callbacks Composability Encapsulation Subscriptions Separation of Concern Data Consistency from to Maintainability … Ingo Mayer et al (https://infoscience.epfl.ch/record/176887/files/DeprecatingObservers2012.pdf)
- 35. After some intense research … 1. Framing the code we used to write as a “class/component” 2. Adding a “reactive” change detection mechanism to re- render components automatically each time their properties change Leading Web Frameworks like React and Angular came up with a … component Model …
- 36. Functional Reactive Programming: Elm/Redux • Immutable data • Observables • Pure functions • Static typing • Unidirectional data flow Source: http://www.codemag.com/article/1601071 http://guide.elm-lang.org/architecture/user_input/buttons.html
- 37. 33% of the code in Adobe’s desktop applications is devoted to event handling logic 50% of the bugs reported during a product cycle exist in this code Sean Parent , “A Possible Future of Software Development”, Adobe 2007
- 39. ?? The SAM Pattern View Mode l ?? 016 xgen.io
- 40. The SAM Pattern View Mode l V = f(M) Counter = Counter + 1Counter = Counter + 1 var p_counter = counter + 1 counter = p_counter var p_counter = counter + 1 counter.accept(p_counter) var p_counter = counter + 1 counter.accept(p_counter).then(learn) 1 Proposers 2 Acceptor(s) 3 Learners Dr. Leslie Lamport: TLA+ http://research.microsoft.com/en-us/um/people/lamport/pubs/state-machine.pdf { }
- 41. SAM is Unapologetically Focused on Mutation View Model Action s V = f(props) State next-action(Model) 1 Proposers 2 Acceptor(s) next- action(Model) 3Learner(s) 016 xgen.io dispatch(event)
- 42. View=State(Model.present(Action(event))).then(nap); “SAM is simple, I [have] never seen a concept such minimalist and powerful, [in particular] in its decoupling potential” — Daniel Neveux next action
- 43. Unidirectional Data Flow / Single State Tree Create Proposal Modify Data State Representation SAM Pattern next-action React/Redux Angular2 – Two Way Databinding
- 44. How do you write a Web App with SAM? <body> … <div id=“sam”></div> … <script src=“sam.js”></script> <script type=“text/javascript”> var a = { }; initSAM() ; </script> </body> https://github.com/jdubray/sam-samples View = State( Model.present( Action(event) ) ).then(nap);
- 45. view.innerHTML = State( Model.present( Action(event) ) ).then(nap); A simple Blog Example https://fish-trader.gomix.me/ <div id="representation"></div>
- 46. view.innerHTML = State( Model.present( Action(event) ) ).then(nap); A simple Blog Example https://fish-trader.gomix.me/ <div id="representation"></div>
- 47. view.innerHTML = State( Model.present( Action(event) ) ).then(nap); A simple Blog Example https://fish-trader.gomix.me/ <div id="representation"></div>
- 48. view.innerHTML = State( Model.present( Action(event) ) ).then(nap); A simple Blog Example https://fish-trader.gomix.me/ <div id="representation"></div>
- 49. Let’s look at some code © Akveo 2016
- 50. Angular2 Component Coupling the business logic to a Web Framework is an anti- pattern events Propertie s
- 51. Angular2 Unidirectional Data Flow Single State Tree Actions events Propertie s Actions State Model
- 53. Conversational Computing https://www.infoq.com/news/2017/02/mindmeld-guide Amazon Alexa, Apple Siri, Google Assistant, Microsoft Cortana “Since rule-based frameworks are not intended to provide AI capabilities to parse or classify incoming messages, the developer must code all of the necessary message processing and interaction logic by hand. It is not uncommon for even simple applications to require hundreds of rules to handle the different dialogue states in a typical conversational interface.”
- 54. By 2020 there will be 50 billion Internet of Things (IoT) devices Internet of Things
- 55. A new Application Architecture - Services, APIs and Microservices -
- 56. SAM Isolates APIs from the View CUD R There is no need for • an immutable model (Redux) • declarative effects (Elm, ~Redux) They create more problems than they solve Create Proposal Modify Data State Representation SAM Pattern next-action
- 57. SAM + SAM API API API Consistency APIs Services Microservices Systems of Record Services Model ActionsState View Model Dispatche r State View Model Actions View Activity ActionAction
- 58. In Summary • Programming model • Centered on Mutation, not immutability • True Single State Tree, no Sagas/Stateful components • Focused on ”what’s allowed”, not subscriptions • View Components are 100% decoupled from the application business logic • Functional UI/HTML (code generation), not templates • Architecture • Can be implemented with the framework of your choice (React, Angular) • Side-effects friendly • Wiring agnostic • Truly Isomorphic • Action “Hang back” / Generic Cancellations • 3rd party Actions (OAuth)
- 59. References • Computation and State Machines – Dr. Lamport • Thinking for Programmers – Dr. Lamport
Editor's Notes
- #15: var fact = function(n) { return (n == 1) ? 1 : n * fact(n-1); } var fact = function(n) { var f = 1, i =1 ; for (i = 1; i <= n; ++i) { f = i * f; } return f ; }
- #34: Graphical User Interfaces are part of a wide class of software, Reactive Applications, that responds to user input, network messages, and other events Reactive code is asynchronously triggered by event occurrences It is hard to trace and understand the control flow of the entire system