Code Metrics For Improving Design

Using code metrics, we can find problems in our software that make it hard for our design to change. This presentation shows an example of using complexity metrics to find a trouble spot, and a few first steps in the refactoring process.

Transcript

1. Metrics!
2. Metrics? …analytical gobbledygook
3. Metrics!
6. …metrics help you analyze Object Oriented Design principles.
7. Practical Object-Oriented Design in Ruby: An Agile Primer Sandi Metz!
8. SOLID Single responsibility, dependency inversion...
9. SOLID Single responsibility, dependency inversion... DRY Don’t repeat yourself
10. SOLID Single responsibility, dependency inversion... DRY Don’t repeat yourself TRUE Transparent, reasonable, usable, exemplary
11. SOLID Single responsibility, dependency inversion... DRY Don’t repeat yourself Law of Demeter Loose coupling, avoid train wrecks TRUE Transparent, reasonable, usable, exemplary
12. SOLID Single responsibility, dependency inversion... DRY Don’t repeat yourself Law of Demeter Loose coupling, avoid train wrecks TRUE Transparent, reasonable, usable, exemplary
13. SOLID Single responsibility, dependency inversion... DRY Don’t repeat yourself Law of Demeter Loose coupling, avoid train wrecks TRUE Transparent, reasonable, usable, exemplary “Alexander, I think the Law of Demeter is shit and never follow it.” --DHH
14. How easily can our design change?
15. Tight coupling = makes our design hard to change
16. Tight coupling = makes our design hard to change Duplication = makes our design hard to change
18. Tight coupling = makes our design hard to change Duplication = makes our design hard to change Complex methods = make our design hard to change
19. Tight coupling = makes our design hard to change Duplication = makes our design hard to change Complex methods = make our design hard to change …etc = makes our design hard to change
20. What is complexity?
21. |ABC| = Math.sqrt((A*A)+(B*B)+(C*C)) Assignment • Branches • Conditionals — http://c2.com/cgi/wiki?AbcMetric
22. |ABC| = Math.sqrt((A*A)+(B*B)+(C*C)) Assignment • Branches • Conditionals
24. Complex Methods Complex class/module de nitions High total complexity of a class or module Duplication
25. Complex Methods Complex class/module de nitions High total complexity of a class or module Duplication
26. og: github.com/seattlerb/flog
30. Score of Means 0-10 Awesome 11-20 Good enough 21-40 Might need refactoring 41-60 Possible to justify 61-100 Danger 100-200 Whoop, whoop, whoop 200 + Someone please think of the children — Jake Skruggs, author of metric_fu http://jakescruggs.blogspot.com/2008/08/whats-good- og-score.html
31. Score of Means 0-10 Awesome 11-20 Good enough 21-40 Might need refactoring 41-60 Possible to justify 61-100 Danger 100-200 Whoop, whoop, whoop 200 + Someone please think of the children — Jake Skruggs, author of metric_fu http://jakescruggs.blogspot.com/2008/08/whats-good- og-score.html poor metrics = makes our design hard to change
32. reek: github.com/troessner/reek
33. Attribute Class Variable Boolean Parameter Control Parameter Data Clump Duplicate Method Call Feature Envy Utility Function Irresponsible Module Long Parameter List Long Yield List Nested Iterators Simulated Polymorphism Nil Check Repeated Conditional Too Many Instance Variables Too Many Methods Too Many Statements Uncommunicative Name Unused Parameters
36. Duplicate Method Call
37. describe #completed?
38. Extract method
39. Extract method...
40. Extract method...
41. Before: A er:
44. Simulated Polymorphism
45. Simulated Polymorphism Train wrecks
46. Simulated Polymorphism Train wrecks …more work to do (later!)
47. Metrics can help you... Identify problems Reveal Interactions Measure pain Discover code smells Expose poor design Expose di culty: In changing objects In testing, In understanding In reusing
48. Flog ABC metric with a ruby bias against metaprogramming… Reek Detect code smells. Flay Analyzes code for structural similarities (duplication). Rails Best Practices Checks the quality of your Rails code. Brakeman A static analysis security vulnerability scanner. Cane Fails your build if code quality thresholds are not met. Simplecov Test coverage analysis.
49. anks!