Scala Days Chicago 2017: Building a Company on Scala
- 1. Building a Company on Scala Dag Liodden
- 2. 2
- 4. Scaling from 1 dev to 50?
- 5. You need a FRAMEWORK
- 6. Having too many degrees of freedom hurts productivity Constraints will: • Make developers focus on more on design than implementation details • Make your code base more approachable • Make your codebase easier to maintain and refactor when you change the constraintsConstraining patterns are good
- 7. Evolving patterns Step 1: What can this thing do? Step 2: How do I do X? Step 3: What other ways can I do X? Step 4: What’s the best way to do X? Step 5: Formalize 4 into pattern Step 6a: Get work done Step 6b: Revisit 4 at appropriate freq Scala makes it particularly easy to get stuck in 3 and 4!
- 8. Scala is incredibly flexible That comes at a price
- 9. A framework is a formalization of patterns and best practices It’s necessary to stay productive
- 10. There are few frameworks in Scala-land Toolkits rule the Scala domain
- 11. Constructing a framework Because you have to
- 12. “FP” Mastery is not required to Master Scala Scala is not Java, but it’s also not Haskell
- 13. Don’t let the possibilities intimidate or seduce you Just because it’s possible, doesn’t make it right
- 14. Decide early how you want to apply Scala Figuring out where on the OO-FP scale you want to be along the way is risky
- 15. Tools and toolkits are means to an end Eager awareness, careful adoption Everything has trade-offs
- 16. The Actor model “Akka is a toolkit and runtime for building highly concurrent, distributed, and resilient message-driven applications on the JVM.” • Know what it is and isn’t • Leverage it only if appropriate!
- 17. Reactive and Async vs Sync Make a firm decision early, middle-ground does not work Reactive programming comes with: • Cognitive overhead • Backpressure, anyone? • Performance gotchas • Compatibility issues
- 18. Don’t obsess about terseness Focus on clarity and approachability What do I need to import to make this work? 1 second // vs FiniteDuration.seconds(1) db ? (sql, "name" like "%Joe%") // vs db.selectOneWhere( sql, SqlField("name").like("%Joe%") )
- 19. Premature and one-off abstractions Building abstractions is SOO tempting in Scala, but: • A pattern is not a pattern until it has been observed several times • Assuming that you capture the pattern right away is a bold one • A function is often sufficient! • Be very careful when promoting patterns into the framework
- 20. Don’t reinvent everything in Scala Borrow from the rest of the JVM ecosystem It’s fun to implement RFCs in Scala, but is it productive? • Dispatch vs Dispatch Reboot • Static singletons -> Cake pattern -> Guice
- 21. When to begin?
- 22. Allow evolution to happen Step 1: What can this thing do? Step 2: How do I do X? Step 3: What other ways can I do X? Step 4: What’s the best way to do X? Step 5: Promote pattern to framework` Step 6a: Get work done Step 6b: Revisit 4 at appropriate freq In step 5, invest time in going back and removing the anti-patterns (this is hard)
- 23. Don’t leave it to a single developer
- 24. Let the tools fade into the background Spend more time on higher level design
- 25. Train, explain, scrutinize, evolve As a team and a community
- 26. • You don’t have to master FP to master Scala • Expect your mindset to evolve, but be rigorous when you adopt new patterns • Treat your tools as means to an end and allow them to fade into the background Key takeaways
- 27. Questions?