Test-Driven Development

A presentation on TDD as I was presenting it in 2008.

Transcript

1. Matteo Vaccari m.vaccari@sourcesense.com http://matteo.vaccari.name/blog BIt Systems, 4 June 2008 Test-Driven Development
2. Three kinds of test • Unit test • Integration test • Customer test
3. Unit tests • Focus on a class, method or function • Run entirely in memory • Extremely fast!
4. Unit tests run fast. A test is not a unit test if: 1. It talks to a database 2. It communicates across the network 3. It touches the file system 4. You have to do things to your environment to run it (eg, change config files) Tests that do this are integration tests Michael Feathers
5. Integration tests • Unit tests prove our logic is correct • Integration test prove our program communicates correctly with the outside world
6. System under test DB Billing GUI End-to-end tests
7. System under test DB Billing GUI End-to-end tests • ...are the least convenient kind of integration tests
8. System under test DB Billing GUI Focused integration tests • Each test deals with one external interaction
9. Customer tests • Also known as “Acceptance tests” • Motto: We implement tricky domain concepts correctly Example from Mugridge & Cunningham Fit For Developing Software
10. Customer tests • Customer test are customer-provided examples • Often captured in table form • Expressed in the language of the business • Automated Example from Mugridge & Cunningham Fit For Developing Software
12. Test-Driven Development
13. Clean code that works.
14. Clean code that works • is out of reach of even the best programmers, some of the time, • and out of reach of most programmers (like me) most of the time -- Kent Beck
15. Simple design The code is simple enough when it: 0. Runs all the tests 1. Expresses every idea that we need to express 2. Contains no duplication 3. Has the minimum number of classes and functions (In this order) Adapted from Extreme Programming Installed by Ron Jeffries et al.
16. Clean code that works • First we'll solve the “that works” part • Then we'll solve the “clean code” part
17. m.vaccari@sourcesense.com Write a test public class AdderTest { @Test public void testTwoPlusThree() { Adder a = new Adder(); assertEquals(5, a.add(2, 3)); } }
18. m.vaccari@sourcesense.com Now it compiles public class AdderTest { @Test public void testTwoPlusThree() { Adder a = new Adder(); assertEquals(5, a.add(2, 3)); } } public class Adder { public int add(int a, int b) { return 0; } }
19. m.vaccari@sourcesense.com Red bar! public class AdderTest { @Test public void testTwoPlusThree() { Adder a = new Adder(); assertEquals(5, a.add(2, 3)); } } public class Adder { public int add(int a, int b) { return 0; } } Expected 5, was 0
20. m.vaccari@sourcesense.com Just pretend public class AdderTest { @Test public void testTwoPlusThree() { Adder a = new Adder(); assertEquals(5, a.add(2, 3)); } } public class Adder { public int add(int a, int b) { return 5; } }
21. m.vaccari@sourcesense.com Remove the duplicated “5” public class AdderTest { @Test public void testTwoPlusThree() { Adder a = new Adder(); assertEquals(5, a.add(2, 3)); } } public class Adder { public int add(int a, int b) { return a+b; } }
22. m.vaccari@sourcesense.com The procedure 1. Write a test 2. Make it compile 3. Make it pass 4. Remove duplication Expected 5, was 0
23. m.vaccari@sourcesense.com Red Green Refactor Repeat every 2-10 min.
24. m.vaccari@sourcesense.com Clean code, why? • Design is the great accelerator: • If you drop quality for speed, you will get neither • If you aim for quality... • ... and you know how to get it... • ... you will also be fast!
25. m.vaccari@sourcesense.com Test first, why? • You think code from the point of view of the caller • This perspective makes for better design • Test coverage is a useful byproduct
26. m.vaccari@sourcesense.com Refactor, why? • Because I can: the tests support refactoring • Refactoring is when I do design • I don’t claim I can guess the right design at first • Design emerges, with thought, care and small steps
27. m.vaccari@sourcesense.com The Bowling Score By Robert Martin “Uncle Bob” http://butunclebob.com/ArticleS.UncleBob.TheBowlingGameKata
28. m.vaccari@sourcesense.com Il punteggio del bowling Ci sono 10 frame. In ogni frame il giocatore ha due possibilità di abbattere 10 birilli (pins). Il punteggio per il frame è il numero di birilli abbattuti, più i bonus per stike o spare. Uno spare è quando il giocatore abbatte 10 birilli in due tiri. Il bonus per quel frame è il numero di birilli abbattuti al tiro successivo. Nel frame 3 dell'esempio, il punteggio è 10 (i birilli abbattuti) più il bonus di 5 (abbattuti nel tiro successivo.) Uno strike è quando il giocatore abbatte tutti i birilli in un solo tiro. Il bonus per quel frame è il numero di birilli abbattuti nei due tiri successivi Nel decimo frame, se il giocatore fa uno strike o spare può fare i tiri necessari per completare il frame. In ogni caso, al decimo frame non vengono fatti più di tre tiri.
29. m.vaccari@sourcesense.com The requirements • Write class “Game” with two methods: • void roll(int pins); call when the player rolls a ball. The argument is the number of pins knocked down. • int score(); called when the game is ended. Returns the final score.
30. m.vaccari@sourcesense.com Let’s think about design? • A quick object-oriented analysis leads us to think we need • class Game • class Frame • class Roll • class TenthFrame extending Frame • ...
31. m.vaccari@sourcesense.com Let’s think about design? • A quick object-oriented analysis leads us to think we need • class Game • class Frame • class Roll • class TenthFrame extending Frame • ... forget about all that!
32. m.vaccari@sourcesense.com Demo time • ...eclipse!
33. m.vaccari@sourcesense.com Ancora, perché test first? • per concentrarsi su quello che serve veramente (no gold plating) • good enough! quando il test passa so che posso fermarmi • perché penso al codice come un cliente di questo codice • perché ottengo codice testabile, e il codice testabile • ha uno scopo preciso • è disaccoppiato dal resto del sistema • è più generale • il design emerge mano a mano che capisco meglio il problema
34. m.vaccari@sourcesense.com Ancora, perché refactoring? • simplicity is key • il design nel tempo si imbastardisce • fare il design prima significa farlo nel momento peggiore: quando ne so di meno • molto meglio fare design mentre sviluppo
35. m.vaccari@sourcesense.com Do the simplest thing that can possibly work
36. m.vaccari@sourcesense.com Do the simplest thing 1. Build the quickest code that will pass the tests 2. Refactor the code to have the simplest design possible
37. m.vaccari@sourcesense.com What is simple design? • The code passes all tests • There is no duplication • The code expresses the programmer’s intention • Using the smallest number of classes and methods In this order
38. m.vaccari@sourcesense.com TDD is a key practice • Defects kill predictability no predictability, no planning! • Test-driven is predictable • Hardly ever use the debugger
39. m.vaccari@sourcesense.com No silver bullet • Needs lots of practice • Requires discipline • Must think and be alert at all times!
40. Debugging Sucks Testing Rocks
42. Supermarket checkout • Compute the total price • Scan items one at a time • In any order Source: Dave Thomas, http://codekata.pragprog.com/2007/01/kata_nine_back_.html
43. Back to the supermarket