Downloaded 40 times
Uploaded byAndy Zaidman
First steps in testing analytics: Does test code quality matter?
The document discusses the importance of test code quality in software development, emphasizing that high-quality tests can lead to increased productivity despite not necessarily easing bug fixing. It presents findings from analytics in testing, highlighting metrics such as coverage and assertion density as indicators of test quality. The author encourages collaboration on research related to testing practices and their impact on development efficiency.
More Related Content
![[Paul Holland] Trends in Software Testing](https://cdn.slidesharecdn.com/ss_thumbnails/paulhollandtrendsinswtesting-140717230533-phpapp01-thumbnail.jpg?width=640&height=640&fit=bounds)
Similar to First steps in testing analytics: Does test code quality matter?
First steps in testing analytics: Does test code quality matter?
- 1. First Steps inTesting Analytics: Does Test Code Quality Matter? Andy Zaidman Delft University of Technology Nederlandse Testdag 17 november 2014, Amersfoort
- 2. Joint work withMoritz Beller & Georgios Gousios (TU Delft) Dimitrios Athanasiou, Ariadi Nugroho & Joost Visser (Software Improvement Group)
- 3. Who am I? Software engineering teacher by day Testing evangelist Software engineering researcher by night
- 4. What would youlike to know? What analytics? • ???
- 6. WHAT EVER HAPP ENS…
- 7.
- 8. TODAY 2 testanalytics stories
- 9. STORY 1 TestCode Quality Matters
- 10. Andy teaching Firstyear programming course @ TU Delft In week 3 everyone is unit testing “I don’t have time for testing” “I don’t like testing, my code is good anyway” “How much should I test?” “How good should my test code be?”
- 11. @Test public voidtestConstructor() { assertTrue(true) } Solution 1
- 13. Solution 2 @Test public void testConstructor() { House h = new House(“Street”, “9”); assertEquals(“Street 9”, h.toString()); }
- 14. Solution 3 @Test public void testConstructor() { House h = new House(“Street”, “9”); assertEquals(“Street”, h.getStreet()); assertEquals(9, h.getNumber()); }
- 15. Solution 4 @Test public void testConstructorStreet() { House h = new House(“Street”, “9”); assertEquals(“Street”, h.getStreet()); } @Test public void testConstructorNumber() { House h = new House(“Street”, “9”); assertEquals(9, h.getNumber()); }
- 16.
- 17.
- 18.
- 19.
- 20.
- 25. Thou shalt not have less than 70% coverage
- 27. What makes a good unit test? What makes a good integration test?
- 28. What makes a good developer test?
- 29. What makes a good developer test? Complete Effective Maintainable
- 30. Complete • Howmuch of the code is covered by the tests? • How many of the decision points in the code are tested? Code coverage #assertions/cyclomatic complexity
- 31. Effective • Howable is the test code to detect defects? • How able is the test code to locate the cause? Assertion density #assertions / LOCtest Directness unit versus integration testing
- 32. Maintainable • Howmaintainable is the system’s test code? Software Improvement Group quality model
- 33. Test Code QualityModel* Complete + Effective + Maintainable * Calibrated using 86 systems
- 34.
- 36. Better tests Easier development ?
- 37. Observation 1* Highquality tests do not lead to quicker/easier bug fixing Why? The low-hanging fruit is already solved if you have good tests… the difficult bugs remain * While studying 75 years of combined development of 18 software systems
- 38. High quality testsdo lead to higher productivity • More issues fixed per month per developer • More issues fixed per month per KLOC Why? Implementing new features does seem to benefit from having a safety net for existing functionality. Observation 2
- 39. High quality testsdo make your team more productive!
- 40.
- 41. #assertions McCabe cyclomaticcomplexity ≥ 1 The single most important thing for test code quality
- 42.
- 43. STORY 2 Testingtakes 50% of your time
- 44. Says this guy • Study from 1975 • Questionable research approach
- 45. Differences? • Waterfall • Developer testing • Programming language • Tools (e.g. IDE)
- 46.
- 48.
- 49. Recording 1. Timespent on developing production code 2. Time spent on developing test code 3. How frequently are Junit tests executed? 4. How frequently do these test runs fail?
- 50.
- 52.
- 53.
- 54.
- 55. My ultimate dream? Continuous background testing
- 57. Take home •Test code quality matters for the productivity of your team • #assertions/cyclomatic complexity ≥ 1 • Help us out with our research! a.e.zaidman@tudelft.nl http://azaidman.wordpress.com httpp://www.st.ewi.tudelft.nl/~zaidman @azaidman