Agile Developers Go Testing Part II Darcy Casselman @flying_squirrel Alexei Zheglov @az1 November 2010 What are we talking about today? • A lot of things… • They may seem disconnected, but they all: – improve software quality – involve testing – done by developers • None of them can do the job alone • We cannot: – Talk about all of them – Give them equal time Unit Testing • We showed you a very simple unit test example earlier • We’ll show you a more difficult example later • Let’s step back first Focus on Quality quotes "...in 2000, during the dot-com bubble, ... software quality for North American teams... ranged from 6 defects per function point to less than 3 per 100 function point, a range of 200 to 1. The midpoint is approximately 1 defect per 0.6 to 1.0 function points. This implies that it is common for teams to spend more than 90 percent of their effort fixing defects.” http://www.amazon.com/Kanban-David-J-Anderson/dp/0984521402/ http://www.amazon.com/Software-Assessments-Benchmarks-Best-Practices/dp/0201485427/ Logical Connection? “The midpoint is approximately 1 defect per 0.6 to 1.0 function points. This implies that it is common for teams to spend more than 90 percent of their effort fixing defects.” How does he know what they are spending their time on? Capacity Allocation To Fixing Defects Capacity Allocation To Fixing Defects = Defect Injection Rate Bugs, Features http://www.flickr.com/photos/dratz/1045336659/ Another View • Root-cause analysis of established practices • E.g. unit testing • Five-whys applied to the solution, not the problem • The link between the third and fourth whys? Source: Karl Scotland http://availagility.co.uk/2010/10/04/a-root-cause-analysis-of-agile-practices/ Compare Two Teams Team A • Just finished 40 “features” (units of work) • Introduced 10 “bugs” (units of rework) Team B • 10 features • 8 bugs Team A Progression 1 40 10 2.5 ... 53 3 • • • • • Terms beyond the first three are negligible Most defects eliminated in the next 2 releases The 2nd maintenance release scope is very small Illusion that any team can do it Time spent fixing defects: 25% Team B Progression 10 8 6.4 5.12 ... 50 • The sum of the first 3 terms: only 24.4 (less than half of the total!) • The sum of the first 10: approx. 45 • Time spent fixing defects: 80% We want to be more like Team A (or better) and certainly not like Team B Testing • Is what we are talking about today • Leaving out other technical practices • Remembering that testing itself does not provide quality Test Matrix Why? Unit Did we build the code right? Does it work as specified? Test Matrix Why? Unit Acceptance Did we build the code right? Does it work as specified? Did we build the right code? Does it do what the user wants? Test Matrix Why? Unit Did we build the code right? Does it work as specified? Acceptance Did we build the right code? Does it do what the user wants? Integration Do system components work together? What about performance? Test Matrix Why? Unit Did we build the code right? Does it work as specified? Acceptance Did we build the right code? Does it do what the user wants? Integration Do system components work together? What about performance? System Is the system installed and configured correctly? Test Matrix Why? Unit Did we build the code right? Does it work as specified? Acceptance Did we build the right code? Does it do what the user wants? Integration Do system components work together? What about performance? System Exploratory Is the system installed and configured correctly? What can we find out about the system that is not specified by these tests? Test Matrix Coverage % Unit Acceptance Integration System Exploratory Auto ? Tools Developers’ Role Test Matrix Coverage % Auto ? Tools Unit Yes NUnit/JUnit and similar tools isolation (mocking) frameworks Acceptance Yes acceptance testing frameworks (Cucumber, FitNesse, Robot etc.) domain-specific languages (DSLs) Integration Yes UI testing frameworks (e.g. Selenium) System Yes same Exploratory No - Developers’ Role Sources: Robert C. Martin, Rob Diana http://regulargeek.com/2009/06/12/how-much-testing-is-enough/ Test Matrix Coverage % Auto ? Unit 90% Yes NUnit/JUnit and similar tools isolation (mocking) frameworks Acceptance 50% Yes acceptance testing frameworks (Cucumber, FitNesse, Robot etc.) domain-specific languages (DSLs) Integration 20% Yes UI testing frameworks (e.g. Selenium) System 5% Yes same - No - Exploratory Tools Developers’ Role Sources: Robert C. Martin, Rob Diana http://regulargeek.com/2009/06/12/how-much-testing-is-enough/ Test Matrix Tools Developers’ Role Yes NUnit/JUnit and similar tools isolation (mocking) frameworks take personal responsibility 50% Yes acceptance testing frameworks (Cucumber, FitNesse, Robot etc.) domain-specific languages (DSLs) Integration 20% Yes UI testing frameworks (e.g. Selenium) System 5% Yes same - No - Coverage % Auto ? Unit 90% Acceptance Exploratory collaborate - Sources: Robert C. Martin, Rob Diana http://regulargeek.com/2009/06/12/how-much-testing-is-enough/ A Very Short Introduction to Acceptance Testing Consider This Test Script • A new customer registers • The customer places three books in the shopping cart • The customer goes to check-out • The customer fills in their delivery address (province: Ontario) • The system offers free delivery to the customer Source: Gojko Adzic. Bridging the Communication Gap, pp. 78-79 http://www.amazon.com/gp/product/0955683610 Compare the Test Script to This The system offers free delivery to Ontario customers when they place their first order and their order contains at least three books Convert to the Tabular Form number of books order number ship to free delivery? 3 1 Ontario Yes number of books order number ship to free delivery? 3 1 Ontario Yes 2 1 Ontario No number of books order number ship to free delivery? 3 1 Ontario Yes 2 1 Ontario No 3 2 Ontario No number of books order number ship to free delivery? 3 1 Ontario Yes 2 1 Ontario No 3 2 Ontario No 3 1 Alberta No How To Do It • Collaborate: – developers – business analysts – testers – product owner • Which tool to use is not as important • Don’t look up terminology in Wikipedia Books http://www.amazon.com/gp/product/0955683610 http://manning.com/adzic/ Unit-Testing When System Under Test (SUT) Has Dependencies Break the dependencies What If We Don’t? • • • • • • • Configuration Database data Remote services Permissions Time to run Infrequent errors Dependencies in development Can You Run Unit Tests Like This? http://www.flickr.com/photos/theredproject/2878006012/ Code Example Caller StudentRepository GradingEngine NotificationService Code Example Unit Test Stub System Under Test Mock (Open IDE) Popular Frameworks (.NET) 0.92% 1.21% 0.57% 3.97% 1.77% 1.99% 3.12% 4.05% 1.35% 37.62% 9.01% 34.42% Moq RhinoMocks Hand-rolled Typemock Isolator Moles JustMock NSubstitute NMock2 NMock FakeItEasy None Other Source http://www.osherove.com/blog/2010/9/10/2010-poll-which-isolation-framework-do-you-use-in-net.html September 2010 Two Camps Design for Testability: Moq, RhinoMocks, etc. Mock Anything, e.g. Typemock Isolator, JustMock • Dynamically fake objects implementing interfaces/base classes in their own way • Require software under test to be designed to be testable • Programmers to follow SOLID principles (esp. Dependency Inversion Principle) • Rely on Profiler API • Can intercept anything and run something else instead • Can fake statics, sealed classes, etc. • Don’t force design for testability Popular Frameworks (Java) • “Conventional”: – Mockito http://code.google.com/p/mockito/ – EasyMock http://easymock.org/ – JMock http://www.jmock.org/ • Alternative: – JMockit http://code.google.com/p/jmockit/ Roy Osherove • The Chief Architect at TypeMock • Wrote: “A customer switches to Moq – and I’m happy” • And: “This is pure awesome. This is why I come to work every day.” • How is it possible? • Read it: http://www.osherove.com/blog/2010/9/29/acustomer-switches-to-moq-and-im-happy.html Back to Our Code Example • Grading engine operates on Student and Grade • But Student has a Department • And Grade has a Course and Course has a Semester • We start mocking • Soon, we have a lot of mocks “Every time a mock returns another mock a fairy dies” • Who said it? • Not the guy whose tweet comes up first in Google search results • I am not 100% sure • What the heck does it mean anyway? “Every time a mock returns another mock a fairy dies” • Who said it? • Not the guy whose tweet comes up first in Google search results • I am not 100% sure • What the heck does it mean anyway? • We don’t have time for it in this presentation http://www.amazon.com/Growing-Object-Oriented-Software-Guided-Tests/dp/0321503627/ Creating Objects • • • • Constructor injection (as in our code example) Factory What if dependencies have dependencies? What if dependencies of dependencies have dependencies? Pretty Soon, Things Begin To Look Like This (and this is a simple example) (different box color = different life span) Source: http://ayende.com/Blog/archive/2010/01/22/rejecting-dependency-injection-inversion.aspx These Tools Can Help • • • • • StructureMap Castle Windsor Unity NInject autofac Conclusions Quality: More Important Than This http://www.flickr.com/photos/markhillary/2681349855/ More Conclusions • • • • • • • • You have to test No single testing strategy is nearly enough Employ multiple strategies Each of them is difficult to master We barely scratched the surface today Agile brings testing into focus Combine hard and soft skills Collaborate Q&A http://creativecommons.org/licenses/by-nd/3.0/