Scala Bay Meetup - The State of Scala Code Style

Questions:
• What are the current Scala code styles?
• Are we respecting them? An analysis of Scala Open Source projects using Codacy
• What might become best practices/standard?

Transcript

1. State of Scala code style
2. Hello! • Jaime Jorge from Codacy • Master thesis & company with Scala • Scala lover for 4 years
3. • Why? • Questions • Code styles • OSS Analysis • Brief overview of observations
4. Why care? • Expressive language, haters like to call ‘complex’ • We’re reviewing code everyday • Reduce number of moving parts • Onboarding becomes easier
5. Questions • What are the current code styles? • Are we respecting them? • What might become best practices/standard?
6. What code styles are currently out there?
7. Gathered code styles • Official Scala Style • Twitters Effective Scala • Apache Spark • Bizo Scala Style • Kiji project • Vertx mod lang • Apache Kafka
8. Method • Categories • Extract rules from all • Understand what they have in common/different
9. Rules extracted • 127 rules total • All taken from guidelines
13. Major differences • Max 100 character line • Do not use relative imports from other packages • Always use braces for ifs except for ternary operator behavior • Do not use infix notation if method is not an operator (map, filter, flatMap)
15. Major differences • Max 80 character line • Declarations: • Use vals when possible. • Use private when possible for member variables. • Use named arguments when passing in literal values • Options should be used instead of null whenever possible • Prefer case classes over tuples
17. Major differences • Adds more than 40 rules to original style guide • Introduces Collections, Object Oriented and Functional style advice • Shows preferences over APIs
18. Conclusion • Built upon Scala Style Guide • 127 rules you can choose for your project • Some differences are important enough to include in Official Scala Style
19. Are we respecting them?
20. Answer by • Analyzing open source projects for violations of Official Scala Style • Understand % of compliance • Understand number of code style violations • Per category • Per code pattern
21. OSS Analysis • Analyzed 50 most popular Scala projects
22. Tools • ScalaStyle • HairyFotr/linter • WartRemover • Abide • scapegoat • Codacy
24. Results • 51% code style compliance on average • Inverse correlation between Age of project and Code style compliance of project • Correlation between number of committers and number of violations (expected)
25. Breakdown Curly braces Line max Method naming catching Fatal exceptions Mutable fields object naming convention class naming convention others
26. Other interesting facts • Significant violations of Option.get • Significant violations of nulls
27. Limitations of analysis • Most popular tend to be libraries • May not represent reality in industry perfectly
28. Conclusions • 51% code style compliance on average • Naming and formatting biggest culprits • Older projects have less compliance • Are we respecting them? • Answer: we could do a better job.
29. What might become standard?
30. Areas • Collection of rules we see enforced humanly in Pull requests and commits
31. Areas • Collections • Testing • Object oriented programming • Functional programming
32. Collapsing of containers Options if (startField.isEmpty && endField.isEmpty) Seq("foo", "bar") else if (startField.isEmpty && !endField.isEmpty) Seq("foo", endField.get) else if (!startField.isEmpty && endField.isEmpty) Seq(startField.get, "bar") else Seq(startField.get, endField.get) Seq(startField.getOrElse("foo"),endField.getOrElse("bar")) Vs
33. Collapsing of containers Await.result(client.hGet(foo, bar)).get client.hGet(foo, bar).map(f => ..) Vs Futures
34. Collections: optimizations • exists(x => x == b) replaceable with contains(b) • .filter(x => ).head can be replaced with find(x => ) match { .. } • .filter(x =>).headOption can be replaced with find(x => ) • .filter(x => Bool).isEmpty can be replaced with !exists(x => Bool) • .filter(_.isDefined).map(_.get) can be replaced with flatten • .filter(x => Bool).size can be replaced more concisely with with count(x => Bool) • sort.filter can be replaced with filter.sort for performance • !Traversable.isEmpty can be replaced with Traversable.nonEmpty • !Traversable.nonEmpty can be replaced with Traversable.isEmpty From Scapegoat
35. Akka • Immutable messages • If state becomes complex, context.become • Not expose any state to outside Guidelines
36. Unit testing • Consistency in tests • Encapsulate test state in a Context object. Testing with state
38. Object oriented programming • Use dependency injection for program modularization • The use of traits are highly encouraged • Do not use Exceptions for commonplace errors • Encode commonplace errors explicitly: using Option or (scala.util./com.twitter.util.)Try
39. Functional programming • Options should be used instead of null whenever possible • Do not overuse Option: if there is a sensible default — a Null Object — use that instead. • Don’t use pattern matching for conditional execution • Only use call-by-name for creating new control constructs such as DSLs • Prefer case classes over tuples (specially no ._1)
40. Note: Microservices • Style of thinking and code in ‘Your Server as a Function’ • Server operations through future combinators • Declarative programming
41. Microservices: declarative programming recordHandletime andThen traceRequest andThen collectJvmStats andThen parseRequest andThen logRequest andThen recordClientStats andThen sanitize andThen respondToHealthCheck andThen applyTrafficControl andThen virtualHostServer
43. Thanks! QA
44. Exclusive Offer Use Codacy for free for 3 months on private repositories Promo Code: SCALABAY-LLHLKS