Elixir intro
- 1. New software for new hardware Elixir meetup 4.0, Lviv, December 17, 2016 Elixir introduction
- 2. Hello! I am Anton Mishchuk - Ruby developer at Matic Insurance - Elixir fan - Author and maintainer of ESpec BDD test framework Github: antonmi 2
- 3. What I’d tell you ▪ No pictures, sorry ▪ No code samples ▪ No live coding ▪ Lots of philosophy ▪ Few words about Erlang and Elixir 3
- 4. The plan ▪ Software crisis? ▪ Resurrection of FP ▪ Erlang ▪ Erlang concurrency model ▪ Elixir ▪ Elixir is not Ruby ▪ You should try Elixir 4
- 6. Common characteristic of present software ▪ low quality ▪ very inefficient ▪ difficult to maintain 6
- 7. Nathan Myhrvold’s laws ▪ “Software is a gas” ▪ “Software grows until it becomes limited by Moore’s Law” ▪ “Software growth makes Moore’s Law possible” ▪ “Software is only limited by human ambition and expectation” 7 Nathan Myhrvold, formerly CTO at Microsoft
- 8. Niklaus Wirth’s Law: “Software gets slower faster than hardware gets faster” 8 Niklaus Wirth, computer scientist, designer of several PL (incl. Pascal)
- 9. What next? 9
- 10. ▪ Better hardware? or ▪ Better software? 10
- 11. We are experiencing a period of “dramatic changes” ▪ New hardware! ▪ New software! 11
- 12. New hardware: ▪ Multicore systems ▪ ASIC (Application-specific integrated circuit) ▪ FPGA (Field-programmable gate array) 12 “The von Neumann Syndrome”
- 13. New software: Multicore crisis: To take advantage of the additional cores most applications have to be rewritten 13
- 15. Functional programming FP is a programming paradigm that treats computation as the evaluation of mathematical functions. FP: Data structures manipulated through functions OOP: Objects manipulated through methods 15
- 16. FP concepts ▪ High-order functions ▪ Pure functions ▪ Recursions ▪ Lazy evaluations ▪ Currying ▪ Closures 16
- 17. History FP: ▪ Lisp, 1950s ▪ APL, 1960s ▪ Haskell, late 80s ▪ Erlang, 1986 OOP: ▪ Simula, 1960s ▪ Smalltalk, 1970s ▪ Java, Ruby, 1995 17
- 18. CPU: ▪ Single-processor machines ▪ 50%/year clock speed growth ▪ From ~10MHz to ~1GHz RAM: ▪ Relatively expensive and slow ▪ $100/1Mb in 1990. $1/1Mb in 2000 ▪ CPU cache OOP dominating in 90s 18
- 19. ▪ Is OOP the only right way to to create software? ▪ Is OOP the RIGHT way? 19
- 20. Why OO sucks? (c) Joe Armstrong 20
- 21. ▪ Functions do things ▪ Data structures just are Since functions and data structures are completely different types of animal it is fundamentally incorrect to lock them up in the same cage. Objection 1: Data structure and functions should not be bound together 21
- 22. ▪ Should “time” to be an instance of some type? ▪ Should a simple integer number to be an instance of some type? Objection 2: Everything has to be an object 22
- 23. ▪ lots of types with limited amount of specific functions Isn't it better have a smallish number of ubiquitous data types and a large number of small functions that work on them? Objection 3: Data type definitions are spread out all over the place 23
- 24. The state is hidden and visible only through access functions. State is the root of all evil! Objection 4: Objects have private state 24
- 25. Why OO was popular? (c) Joe Armstrong 25
- 26. ▪ It was thought to be easy to learn. ▪ It was thought to make code reuse easier. ▪ It was hyped. ▪ It created a new software industry. Why OO was popular? (c) Joe Armstrong 26
- 27. I see no evidence of 1 and 2. Reasons 3 and 4 seem to be the driving force behind the technology. 27
- 28. If a language technology is so bad that it creates a new industry to solve problems of its own making then it must be a good idea for the guys who want to make money. 28
- 29. 3 Erlang 29
- 30. ▪ Elixir compiler is written in Erlang ▪ Both Elixir and Erlang compiles into BEAM byte code Elixir == Erlang 30
- 31. ▪ Ericsson, 1986 ▪ Joe Armstrong, Robert Virding and Mike Williams ▪ Open-sourced in 1998 ▪ Telecom systems (telephone switches) ▪ The latest version is 19.1 Erlang 31
- 32. ▪ Fault tolerance ▪ Scalability ▪ Distribution ▪ Live update Erlang buzzwords 32
- 33. ▪ Ericsson ▪ WhatsApp ▪ Facebook ▪ Amazon ▪ Bet365 ▪ Demonware ▪ ... Companies using Erlang 33
- 34. ▪ Modules ▪ Functions ▪ Primitive data types: atoms, integers, floats, tuples, lists, maps, strings, binaries, records, references, pids, ports, funs. Erlang components 34
- 36. ▪ Native threads ▪ Green threads ▪ Event loop It is not about: 36
- 37. ▪ lightweight processes ▪ with isolated memory ▪ communication via messages It is about “actors” 37
- 38. In response to a message actor receives, can concurrently: ▪ send a finite number of messages to other actors ▪ create a finite number of new actors ▪ Set the “state” to be used for the next message it receives Actors 38
- 39. ▪ Controlled by BEAM VM ▪ ~ 1µs to create process ▪ ~ 1kB memory per process ▪ ~ 1µs to pass a message Erlang actors 39
- 40. ▪ encapsulate state ▪ communicate with other actors by exchanging messages More OO than objects 40
- 41. ▪ Tools ▪ Libraries Design principles: ▪ Workers and Supervisors ▪ GenServer, GenFSM ▪ Event Handler and Manager ▪ Application Erlang OTP 41
- 42. 5 Elixir 42
- 43. ▪ Jose Valim, 2012 ▪ Current version 1.3 ▪ Hex.pm ▪ >3000 packages, ~20000 down/day ▪ Elixir Conf EU, Elixir Conf US ▪ Elixir Companies ▪ Phoenix web framework (2M websocket conns) Elixir 43
- 44. What is Elixir? 44 Elixir is Erlang with pretty syntax and metaprogramming ▪ Consistent stdlib ▪ Pipe operator ▪ Tools (mix, iex, logger)
- 45. Elixir metaprogramming 45 ▪ Easy code sharing ▪ Perfect Ruby-like DSLs
- 46. 6 Elixir is not Ruby 46
- 47. Elixir is functional 47 ▪ Function is a primary Elixir abstraction. ▪ Data is immutable ▪ Minimal state manipulation
- 48. Ruby is OO language with some FP features 48 ▪ Procs, lambdas ▪ No FP optimizations ▪ Data is mutable
- 49. Concurrency model 49 In Ruby, one moves through a monolithic call stack, manipulating objects. Threads or fibers uses the same memory heap. Elixir processes have their own heaps, and the VM collects process' garbage independently of each other.
- 50. What is the same? 50 “programmer happiness” philosophy: ▪ Pretty syntax ▪ Good documentation ▪ Tools for easy development
- 51. 7 You should try Elixir 51
- 52. Think about future software 52 ▪ Multi-core computers ▪ IoT with up to 30 billion devices to be online by 2020
- 53. Double paradigm switch 53 ▪ Functional programming ▪ Actor concurrency model
- 54. You’ll found it exciting 54 ▪ It’s easy to get started ▪ Lots of tutorials ▪ Nice documentation ▪ Great community