Functional Programming in the Real World

Functional Programming is sold as a silver bullet. Yet most organizations lack real-world experience with the paradigm. It’s hard to adopt functional programming too since most public examples show small and isolated code snippets and it’s not clear how those pieces would fit together to form a system, a functional architecture. A real-world system also has to deal with non-functional side effects and the rapid feature growth of an evolving product. So how well does functional programming work in practice? What happens when you bet your company on a functional programming language? Will the hype transfers into productivity gains and cheaper maintenance?

Transcript

1. @AdamTornhill FUNCTIONAL PROGRAMMING IN THE REAL WORLD http://www.empear.com/
2. @AdamTornhill My Road To Functional Programming
3. Clojure @AdamTornhill My Road To Functional Programming 1980 1990 2000 2010 2015 Assembly C Pascal C++ Java TCL http://www.paulgraham.com/avg.html Erlang J?
4. Less Code, More Features, More Fun @AdamTornhill Erlang Java Lines of Code
5. The 60/60 Rule @AdamTornhill https://www.amazon.com/Facts-Fallacies-Software-Engineering-Robert/dp/0321117425/
6. What’s Functional Programming? @AdamTornhill “If you ask 100 programmers for their definition, you'll likely receive 100 different answers” — Michael Fogus & Chris Houser Immutability Lambda Calculus Recursion Lazy Evaluation Category Theory Monads Pure Functions Referential Transparency Hindley–Milner type inference
7. Immutable State @AdamTornhill i = 41;
8. 5 ? 1 2 8 3 4 6 7 A Quick Reasoning Test @AdamTornhill
9. @AdamTornhill Don’t Turn Your Code Into An IQ Test state Side effects Working memory tax
10. IMMUTABILITY ALLOWS PROGRAMS TO SCALE IN COGNITIVE TERMS @AdamTornhill
11. TDD: A Diminishing Return in FP @AdamTornhill
12. REPL interactions -> soon to become unit tests REPL Driven Development @AdamTornhill
13. The Perils Of Test Setup @AdamTornhill Code from https://github.com/aspnet/Mvc
14. @AdamTornhill “The length of a test’s setup method is inversely related to the readability of the code under test.” —Adam’s Heuristic
15. Unit Tests With Immutable State 0 (ZERO) SETUP METHODS
16. Threading Macros @AdamTornhill (defn costs-by-file
 [costs file-name->ticket]
 (->> costs (map (partial assign-to-files file-name->ticket)) (filter passes-thresholds?) (reduce by-file-name))) Readability and Code Evolution
17. @AdamTornhill (defn costs-by-file
 [costs file-name->ticket]
 (->> costs (map (partial assign-to-files file-name->ticket)) (filter passes-thresholds?) (reduce by-file-name))) Performance And Lazy Evaluation lazy lazy
18. In Introductory Presentations …In The Real World @AdamTornhill Functional Programming
19. A Production System in FP @AdamTornhill https://codescene.io/
20. A Production Systems in FP @AdamTornhill https://codescene.io/ CodeScene on-premises Two
21. Analysing Android @AdamTornhill Analysis of https://github.com/android/platform_frameworks_base
22. @AdamTornhill docs core services packages tools media
23. @AdamTornhill
24. @AdamTornhill
25. @AdamTornhill
26. @AdamTornhill
27. The Need For Concurrency @AdamTornhill . . .<lots of trends>
28. Concurrency Made Easy @AdamTornhill Possible (map complexity-trend files) (pmap complexity-trend files) Concurrency Exception!
29. Functional Architecture @AdamTornhill MACRO LEVEL: same difference SUB-SYSTEM LEVEL: more freedom ISOLATE SIDE EFFECTS Side effects should be clear and kept to a thin layer at the edge of your system.
30. … Social Architectural Pattern: Pipes And Filters Input Sources Git JIRA Teams Transformations Time Users Files Analyses Hotspots Effort Pattern Detectors Risk Neural Nets … Git Time Users Files Hotspots Early Warnings Run-time behaviour Result, purely functional, no side-effects
31. Remember The Pipeline Style? @AdamTornhill (defn costs-by-file
 [costs file-name->ticket]
 (->> costs (map (partial assign-to-files file-name->ticket)) (reduce by-file-name) (filter passes-thresholds?)))
32. … Social Architectural Pattern: Pipes And Filters Input Sources Git JIRA Teams Transformations Time Users Files Analyses Hotspots Effort Pattern Detectors Risk Neural Nets … Git Time Users Files Hotspots Early Warnings Run-time behaviour Result, purely functional, no side-effects
33. Beyond Myths: A Real-World Dirty REST API Domain specific language for describing routes Side-effects: 1. Read from DB 2. Fire-off a pipeline of business logic 3. Write the results to DB, reply
34. Finally A Silver Bullet? @AdamTornhill
35. LAZY EVALUATION IS GREAT @AdamTornhill …EXCEPT FOR WHEN IT’S NOT
36. @AdamTornhill (defn costs-by-file
 [costs file-name->ticket]
 (->> costs (map (partial assign-to-files file-name->ticket)) (reduce by-file-name) (filter passes-thresholds?))) The code we write: clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.AFn.applyTo (AFn.java:144) clojure.core$apply.invokeStatic (core.clj:646) clojure.core$apply.invoke (core.clj:641) codescene.analysis.diagnostics$trace_ns$fn__9583 clojure.lang.RestFn.invoke (RestFn.java:397) clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.RestFn.applyTo (RestFn.java:132) clojure.core$apply.invokeStatic (core.clj:646) clojure.core$apply.invoke (core.clj:641) codescene.analysis.diagnostics$trace_ns$fn__9583 clojure.lang.RestFn.invoke (RestFn.java:397) clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.RestFn.applyTo (RestFn.java:132) clojure.core$apply.invokeStatic (core.clj:646) clojure.core$apply.invoke (core.clj:641) codescene.analysis.diagnostics$trace_ns$fn__9583 clojure.lang.RestFn.invoke (RestFn.java:397) clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.RestFn.applyTo (RestFn.java:132) clojure.core$apply.invokeStatic (core.clj:646) clojure.core$apply.invoke (core.clj:641) codescene.analysis.diagnostics$trace_ns$fn__9583 clojure.lang.RestFn.invoke (RestFn.java:397) clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.RestFn.applyTo (RestFn.java:132) clojure.core$apply.invokeStatic (core.clj:646) clojure.core$apply.invoke (core.clj:641) codescene.analysis.diagnostics$trace_ns$fn__9583 clojure.lang.RestFn.invoke (RestFn.java:397) clojure.lang.AFn.applyToHelper (AFn.java:152) clojure.lang.RestFn.applyTo (RestFn.java:132) clojure.core$apply.invokeStatic (core.clj:646) The exception we get Clojure Specific: The Semantic Gap
37. Java — The Assembly Language of the JVM @AdamTornhill Dealing With Performance Issues
38. Immutability is Key @AdamTornhill Takeaways Think Pipelines - Functions and Architecture No Silver Bullet - Programming is still hard
39. @AdamTornhill Beyond Functional Programming http://www.adamtornhill.com/articles/jlang/beyondfunctional.html The Day I Parsed A Monster http://www.empear.com/blog/parse-a-monster/ Lisp for the Web (free) https://leanpub.com/lispweb Try CodeScene https://codescene.io/ adam.tornhill@empear.com