I have come across few articles which talk of page object returning another page object being bad practice. The reason cited are following -
- Tight coupling between page objects and less flexibility
- Duplication of code as two variant of a method need to be created returning two different page object depending on if operation was valid or not, i.e. -
‘loginAs‘ and ‘loginAsExpectingError‘
Let’s see following example of page objects wherein a page object class does not return another page object. For the sake of simplicity, I have consider three page object classes -
- LoginPage class which lets user login
- LoggedInUserHomePage class which provides services when user is logged in
- InvoiceDetailPage class using which user invoice details
A simplified version of classes would be -
public class LoginPage {
public LoginPage() {
if(!WebUIDriver.getWebDriver().getTitle().equalsIgnoreCase("login page")) {
throw new IllegalStateException("Not Login Page");
}
}
private TextFieldElement loginTextBox = new TextFieldElement("login text box", Locator.locateById("login"));
private TextFieldElement passwordTextBox = new TextFieldElement("password text box", Locato.locateById("password"));
private ButtonElement submitButton = new ButtonElement("password text box", Locator.locateById("password"));
public void login(User user) {
loginTextBox.sendKeys(user.getUserName());
passwordTextBox.sendKeys(user.getPassword());
submitButton.click();
}
}
public class LoggedInUserHomePage {
public LoggedInUserHomePage() {
if(!WebUIDriver.getWebDriver().getTitle().equalsIgnoreCase("logged in user page")) {
throw new IllegalStateException("not logged in user page");
}
}
public void getInvoiceDetailsOfOrderNumber(Object orderNumber) {
}
}
And
public class InvoiceDetailPage {
public InvoiceDetailPage() {
if(!WebUIDriver.getWebDriver().getTitle().equalsIgnoreCase("invoice detail page")) {
throw new IllegalStateException("Not invoice detail Page");
}
}
public BigDecimal getOrderAmount(Object orderNumber) {
return orderAmount;
}
}
A test method may look as -
@Test
public void shouldDisplayOrderAmountOfGivenOrderNumber() {
int orderNumber = 12345;
BigDecimal expectedOrderAmount = new BigDecimal(50.0);
LoginPage loginPage = new LoginPage();
loginPage.login(new User());
LoggedInUserHomePage loggedInUserHomePage = new LoggedInUserHomePage();
loggedInUserHomePage.getInvoiceDetailsOfOrderNumber(orderNumber);
InvoiceDetailPage invoiceDetailPage = new InvoiceDetailPage();
BigDecimal orderAmount = invoiceDetailPage.getOrderAmount(orderNumber);
assertThat("Wrong order amount", orderAmount, is(expectedOrderAmount));
}
Notice that how each page object class had to be instantiated each time before being able to use the corresponding page object. That’s because none of our page object method return a page object.
Now let’s consider the old page object approach wherein each page object method return object of another page if it brings control on it. In this case test method would looks as -
@Test
public void shouldDisplayOrderAmountOfGivenOrderNumber() {
int orderNumber = 12345;
BigDecimal expectedOrderAmount = new BigDecimal(50.0);
BigDecimal orderAmount = new LoginPage()
.login(new User())
.getInvoiceDetailsOfOrderNumber(orderNumber)
.getOrderAmount(orderNumber);
assertThat("Wrong order amount", orderAmount, is(expectedOrderAmount));
}
You can see how succinct and comprehensible test is when page object method return other objects. This is an example of just three page objects but real world applications have more deep linked pages than this. And the advantages I cited earlier begin to diminish when pages are deep linked. Do you page object methods also return other page objects? What is your opinion about this?