Into the Land of lambda, One Programmer's Journey Into Functional Programming

Into the Land of lambda, One Programmer's Journey Into Functional Programming

• 1.
  Into the Landof Lambda (into functional-programming (one (programmers-journey)))
• 2.
  Who am I? Currently, a programmer @livingsocial Previously a software consultant at various Fortune 1000 companies, including banking, pharma, and transportation sectors Lots of work with start-ups since 2005 Speaker at first Rails Conf (‘06), Ruby Conf (‘08) Sarasota resident since ‘96  mike.pence@livingsocial.com @mikepence
• 3.
  The context ofmy experience… Tools I have known and loved (starting out)
• 4.
  The context ofmy experience… Tools I have known and loved (oy! the 90’s!) You had to re-invent yourself every 2 years 
• 5.
  The context ofmy experience… True love since ’05 (well, love/hate with Rails)
• 6.
  ~25 years ofOOP (for me) How come it isn’t getting any easier?
• 7.
  Our relationship withobjects It’s complicated • Graphs (trees) of objects maintain internal state at every level in the call stack • Bugs are the product of programmatic logic + the transient state of the world at the time of the error • Example-based unit or integration tests are of limited usefulness (they prove their example case, but nothing more) • No guarantee that calling the same method with the same arguments will return the same result (aka referential transparency) because variables
• 8.
  Reasoning about objects Feels Like A Shell Game
• 9.
  Attempts to makeOOP better Each with their own trade-offs… • Dependency injection / AOP • Following SOLID design principles • Single Responsibility Principle • Liskov Substitution Principle • Etc. • Hexagonal Architecture • Ports and Adapters • Pipes and Filters Still the fundamental problem remains that state is casually managed, if managed at all. Nevermind concurrency!
• 10.
  This Talk GetsIts Own Slide “Are we there yet?” http://www.infoq.com/presentations/Are-We-There-Yet-Rich-Hickey The conflation of behavior, state, identity and time is a big source of implicit complexity in current object systems. We should primarily be programming with pure functions and immutable values. - Rich Hickey
• 11.
  Is There aDifferent Way? (Spoiler alert: There is!)
• 12.
  Another way Disambiguatingidentity, value, time and mutation
• 13.
  Statefulness, YAGNI Avoidit most of the time. Manage it when you need it. • 99% of the time, you don’t need internal state in functions / methods • As an exercise, try passing in variables and having mutated values returned in <language you use daily> • But Clojure is not about avoiding state, it is about managing it with well-defined semantics • To re-iterate, identity is a sequence of states • America is an identity • The PIC (prez, vice prez) are transient
• 14.
  A couple moreanalogies State is like garbage, or Buddhism (which is not) • Clojure provides "managed time" in much the same way that GC's have come to provide managed memory • * My team at LS offered a lot of this content • They say Buddhists with advanced training don't see a "video" through their eyes, but something like a series of pictures, like a flip book • This "epochal state model" concept is similar, differs in that there are multiple perceivers observing/looking at many flowing identities
• 15.
  Code Examples Let’sSee Some!
• 16.
  Some Code Samples Here they are! • http://nakkaya.com/2009/10/04/fractals-in-clojure-buddhabrot-fractal/ • http://wicketconsole.appspot.com/clojureWithTemplates • http://www.slideshare.net/alexmiller/clojure-the-art-of-abstraction-7161663
• 17.
  A Whole NewWorld Full of win!
• 18.
  Clojure / FPWins Reliability • Code is “provable” because it lacks transient variables – Tests pass in all required state – Example-based (unit) tests are great • Generative tests are better • Stepping debuggers and similar approaches to diagnosing functions with variables are simply not needed (but are available). • My experience so far has been that a whole class of bugs simply don’t occur
• 19.
  Clojure / FPWins Focus • You focus on a small set of hierarchical data structures and the functions that process them – Visualize any JSON nested structure you have seen • Biggest take-away for me – You simply don’t need all of the ceremony of OOP "It is better to have 100 functions operate on one data structure than 10 functions on 10 data structures." — Alan Perlis
• 20.
  Clojure / FPWins Expressivity and elegance • Your functions and the built-in functions, or library functions, are syntactical peers • So no awkwardly implemented DSL’s • No odd, arbitrary-seeming mix of infix, postfix and dot notation • It is functions, all the way down • So you can freely build your own vocabulary of functions to model your problem domain and solution • Which I am really excited to see in practice • Also, macros!
• 21.
  Clojure / FPWins Access to awesome libraries and community https://github.com/trending?l=clojure&since=weekly
• 22.
  Clojure / FPWins More • Concurrent and performant • On the JVM • With all the goodness that brings • Used by LivingSocial for their massive email processing needs • Plays well with Java • But not just for Java! • CLR version • Enables Unity3D and Unreal Engine programming! • JavaScript version • ClojureScript • A whole awesome thing in its own right • Port of Clojure’s immutable data structure libraries included • Om and React.js, OMG! • You could program for tablets, consoles, phones, etc. • LightTable!
• 23.
  Adorable!
• 24.
  Thank you! ForYour Participation
• 25.
  Questions and Discussion This subheading intentionally left blank