Testing
PA165
Dec 10, 2013
Petr Adámek, Tomáš Pitner
logo_fi.gif logo_mu.gif

Testing of Applications
˂Testing verifies compliance with the specification and implementation of customer expectations.
˂It is an important part of the quality management in software development.
˂Unlike formal verification does not allow to detect all potential errors.
<2>

Základní pravidla
˂Tests should be reproducible.
˂Somebody else should be able to perform the same tests with the same results.
˂Tests should be deterministic, i.e. they should have the same input conditions at the beginning.
˂Tests should be independent, i.e. not to be influenced by each other.
˂Usually by setting the same input conditions for each test
˂Tests should be repeatable cheaply.
˂It usually means running in an automated way.
<3>

Modes of Testing
˂Manual testing:
˂low entry costs;
˂expensive repetitions;
˂difficult to ensure reproducibility, determinism and independence
˂Automated testing:
˂high input costs;
˂cheap repetition;
˂easy to ensure reproducibility, determinism and independence.
<4>

Types of Testing According to Goals
˂Unit testing
˂Does the unit work independently on the context?
˂Integration testing
˂Does the component work integrated in its environment?
˂Functional testing
˂Does it fulfill the functional requirements?
˂Acceptance Testing
˂Is it good for the customers? Will they accept it?
<5>

Types of Testing According to Goals
˂Performance and scalability testing
˂Testing the user-friendliness
˂Security testing
<6>

Unit Testing
˂In unit testing we try to test the individual components of the system being developed at the
lowest level.
˂Individual test components should be isolated from its surroundings to avoid the influence of the
surroundings on the test component.
˂Interaction with the environment is simulated using mock-objects that simulate the behavior of the
neighborhood in a particular test scenario.
˂The better the decomposition is done, the easier the unit testing is.
<7>

Tools for Unit Testing
˂JUnit
˂TestNG
<8>

Example
˂public class CalculatorTest {
˂
˂    private Calculator c;
˂
˂    @Before
˂    public void setUp() {
˂        c = new Calculator();
˂    }
˂
˂    @Test
˂    public void testDivide() {
˂        assertEquals( 9, c.divide( 99, 10));
˂        assertEquals(10, c.divide(100, 10));
˂    }
˂
˂    @Test(expected = IllegalArgumentException.class)
˂    public void testDivideByZero() {
˂        c.divide(100, 0);
˂    }
˂}
<9>

Basic Rules
˂The test outputs are always Yes / No (Boolean)
˂First test, then code (see XP and TDD)
˂When the error is to be corrected: first test, then fix (protection against regression)
˂Trivial get / set methods are not tested
˂Test all non-standard situations and limit values
˂Error messages and comments not always needed
˂Tests runs after every change
<10>

Interactions with Environment
˂Components should be tested in isolation.
˂But it is necessary to simulate the kind of interaction with the environment.
˂That is what Mock objects do.
˂These objects must be type compatible with simulated component:
˂Inheritance
˂Implementing an interface (preferable)
˂Mock objects can be created manually (tedious), or through tooling:
˂Mockito, EasyMock, JMock
<11>

Zjistit a vysvětlit rozdíl mezi Mock Objekty, Stuby, apod…

Example (manually created Mock objects)
˂public class CurrencyConvertorTest {
˂  @Test
˂  public void testConvert() {
˂    ExchangeRateTable exchangeRateTable = new ExchangeRateTable() {
˂
˂      public void setExchangeRate(Currency currency, BigDecimal exchangeRate) {
˂        throw new UnsupportedOperationException("Not supported yet.");
˂      }
˂      public BigDecimal getExchangeRate(Currency currency) {
˂        return BigDecimal.valueOf(28.2);
˂      }
˂    };
˂
˂    CurrencyConvertor convertor = new CurrencyConvertor(exchangeRateTable);
˂    Currency czk = Currency.getInstance("CZK");
˂    BigDecimal actualResult = convertor.convert(czk, BigDecimal.valueOf(10));
˂    BigDecimal expectedResult = BigDecimal.valueOf(282.0);
˂    assertEquals(expectedResult, actualResult)
˂  }
˂}
<12>

Example (Mockito)
˂@RunWith(MockitoJUnitRunner.class)
˂public class CurrencyConvertorTest {
˂
˂  @Mock
˂  ExchangeRateTable exchangeRateTable;
˂
˂  @Test
˂  public void testConvert() {
˂
˂    when(exchangeRateTable.getExchangeRate(czk)).
˂      thenReturn(BigDecimal.valueOf(28.2));
˂
˂    CurrencyConvertor convertor = new CurrencyConvertor(exchangeRateTable);
˂    Currency czk = Currency.getInstance("CZK");
˂    BigDecimal actualResult = convertor.convert(czk, BigDecimal.valueOf(10));
˂    BigDecimal expectedResult = BigDecimal.valueOf(282.0);
˂    assertEquals(expectedResult, actualResult)
˂  }
˂}
<13>

Pro základní kurz to chce více příkladů na mockito

Unit Testing in Java EE
˂For Java EE applications, it is necessary to take into account the existence of the container.
˂Tests outside the container - test only business logic, not behavior depending on the container
(such as transaction management, authorization, etc.)
˂Tests in a container - will test everything, but this kind of testing for unit tests not fit.
˂In testing outside of the container concept is used mock objects which simulate the behavior of
the container.
<14>

Unit Testing - Data
˂How to test data persistence layer:
˂Mock objects (easy with JPA or other libraries and frameworks, complicated by the low-level JDBC).
˂Database is stored in memory (easy for JPA, with low-level JDBC may be a problem with the SQL
dialect).
˂Do not forget to provide the same initial conditions (state database is always the same initial
state).
˂What can help
˂DBUnit
˂Abstract DAO
<15>

What else can help?
˂Tools for measuring test coverage
˂Line Coverage
˂Branch Coverage
˂Tools for generating test data
˂Extended set of assert methods
˂Etc.
<16>

Integration Testing
˂Integration testing is used to verify the correct interaction of individual components that are
assembled and the system behaves as expected in its specification.
˂See also continuous integration
<17>

Functional Testing
˂Functional testing is used to verify the functionality of the end-user perspective.
˂Mostly performed at the user interface level
˂Rational Functional Tester - web GUI+
˂http://www-01.ibm.com/software/awdtools/tester/functional/index.html
˂Selenium IDE – web
˂http://selenium.openqa.org/
˂Marathon – GUI
˂http://marathonman.sourceforge.net/
˂Rational Robot - GUI(for legacy applications), Rational Quality Manager, JWebUnit
<18>

Acceptance Testing
˂Customer acceptance testing verifies that the application meets customer’s requirements and
expectations.
˂Absence of acceptance testing (or its underestimation and lack of design) almost always leads to
future disputes and problems.
˂Customers unfortunately have a tendency to underestimate it. The non-compliance of the
implementation with the customer’s requirements so often comes at the moment of production
deployment :-(.
<19>

Performance and Scalability Testing
˂Performance testing verifies system throughput and response time at high loads.
˂Part of the specification should be the definition of the throughput and response times of the
prescribed load.
˂Rational Performance Tester (+ extensions)
˂http://www-01.ibm.com/software/awdtools/tester/performance/index.htm
˂Rational Service Tester for SOA Quality (functional testing, performance testing +)
˂JMeter - http://jakarta.apache.org/jmeter/
<20>

Usability Testing
˂In the USA a common thing, in Europe still not so obvious and Asia is likely to overtake Europe in
this.
˂The definition of the prototype of the target user.
˂Select a group of test users (test sample).
˂Test user is given a list of tasks that are trying to solve without the help of someone else.
˂His/her behavior is monitored and evaluated.
˂See Štefkovič, M.: Usability of Web applications. https://is.muni.cz/auth/th/166042/fi_b/ (Bc.
Thesis)
<21>

Security Testing
˂Security testing checks resistance against various security attacks.
˂Tools:
˂Rational AppScan - web app security testing
˂http://www-01.ibm.com/software/awdtools/appscan/
<22>

Questions
˂?
<23>