Skip to content

Latest commit

 

History

History
176 lines (127 loc) · 11.3 KB

BuildingPageObjects.md

File metadata and controls

176 lines (127 loc) · 11.3 KB

Introduction

Using the page-model pattern to implement WebDriver automation provides many structural benefits. However, plain-vanilla Selenium doesn't provide much in the way of base-class support for building page objects. Selenium Foundation includes a whole range of building blocks to help you create efficient, reliable WebDriver automation.

Page-model class
package sixth_example;
 
import org.openqa.selenium.By;
import org.openqa.selenium.WebDriver;
import com.nordstrom.automation.selenium.annotations.PageUrl;
import com.nordstrom.automation.selenium.core.ByType.ByEnum;
import com.nordstrom.automation.selenium.interfaces.DetectsLoadCompletion;
import com.nordstrom.automation.selenium.model.Page;
 
@PageUrl("/")
public class GoogleSearchPage extends Page implements DetectsLoadCompletion {
    
    public GoogleSearchPage(WebDriver driver) {
        super(driver);
    }
     
    private enum Using implements ByEnum {
        SEARCH_BOX(By.name("q")),
        SEARCH_BUTTON(By.name("btnG")),
        RESULT_STATS(By.id("resultStats"));
        
        private final By locator;
        
        Using(By locator) {
            this.locator = locator;
        }
        
        @Override
        public By locator() {
            return locator;
        }
    }

    public GoogleSearchPage doGoogleSearch(String searchString) {
        findElement(Using.SEARCH_BOX).sendKeys(searchString);
        findElement(Using.SEARCH_BUTTON).click();
        return new GoogleSearchPage(driver);   
    }
     
    @Override
    public boolean isLoadComplete() {
        return findOptional(Using.RESULT_STATS).hasReference();
    }
}

Declaring Locator Enumerations

To make your code more maintainable and readable, we recommend the use of locator enumerations. The Using enumeration in the previous example demonstrates this pattern. This example also demonstrates implementation of the ByEnum interface. By implementing this interface, we can utilize the web element location methods (e.g. - findElement()) that accept locator constants.

Declaring Web Application Page Path

Typically, each page-model class is associated with a single page of the application under test, located at a fixed path (e.g. - /c/nordstrom). It's also common for the paths of web application pages to conform with a pattern or template (e.g. - /s/<item-description>/<item-identifier>). Selenium Foundation provides a @PageUrl annotation that enables you to associate page-model classes with their corresponding paths.

An example of this annotation is found in the simple page class found above. The specified path ("/") declares that this page-model class is associated with the root of the target web application - in this case, Google. Template-based paths look like this:

Template-based path
@PageUrl(pattern="/s/[a-z]+(-[a-z]+)*/\\d+", value="/s/wolf-blake-watch-roll/3892364")
public class ProductPage extends Page {
    ...
}

In this form of the @PageUrl annotation, the pattern attribute specifies a regular expression that identifies the valid format for the path, and the value attribute provides an exemplar.

Once a page-model class is associated with a fixed path, the class itself is all you need for direct navigation:

Direct navigation to annotated page
...
// perform direct navigation to exemplar product page, re-using the current window
ProductPage productPage = someOtherPage.openAnnotatedPage(ProductPage.class, false);
...

In this example, Selenium Foundation navigates to the path specified by the value attribute of the @PageUrl annotation of the ProductPage class. This is the exemplar page for this class.

Page Transition Synchronization

Whenever a page object method returns a new page object, this informs Selenium Foundation that the method triggered a page transition. In response, Selenium Foundation performs basic synchronization - waiting for a previously-acquired web element reference to go stale. This strategy is effective for basic web applications with simple page load behavior.

For web applications with more complex page load behavior (single-page, dynamic content, etc.), the DetectsLoadCompletion interface enables implementers to provide custom page-load completion detection:

DetectsLoadCompletion
public class OpctPage extends Page implements DetectsLoadCompletion {
  
    ...
 
    @Override
    public boolean isLoadComplete() {
        WebElement body = findElement(By.tagName("body"));
        String cursor = body.getCssValue("cursor");
        return !("wait".equals(cursor));
    }
}

In this example, the page is done loading when the value of the cursor CSS property of the body tag no longer equals "wait". This method, which is declared by the new container object, will be called every 500 mS until it returns 'true' or the page load timeout interval expires. Any type of container class can implement DetectsLoadCompletion - page, component, or frame.

Note that automatic page load synchronization is activated by a method that returns a new page object. No synchronization is performed if a method returns the page object it's standing on (i.e. - 'this'). However, the ContainerMethodInterceptor class includes a static method to invoke the same page-load synchronization logic that Selenium Foundation would:

Explicit synchronization
public class SkuDescriptionSearchDialog extends Dialog {
 
    ...
 
    /**
     * Reset the search criteria.
     */
    public void resetSearch() {
        findElement(Using.RESET_BUTTON.locator).click();
        getParentPage().getWait(WaitType.PAGE_LOAD)
                        .ignoring(PageNotLoadedException.class)
                        .until(ContainerMethodInterceptor.loadIsComplete());
    }
}

In this scenario, the Dialog class doesn't implement DetectsLoadCompletion, but its parent page does. The 'wait' object is acquired in the context of the parent page, so the loadIsComplete() 'wait' proxy will invoke the isLoadComplete() method of the parent page class.

Automatic Driver Targeting

For web applications that use frames or multiple windows, a major source of boilerplate code is management of the driver target. In addition to being extremely repetitive, this code is also surprisingly difficult to implement correctly. Selenium Foundation completely eliminates the need for explicit driver targeting. You get to focus on scenario-specific details instead of low-level plumbing. For more details, see the Automatic Driver Targeting section on front page

Wrapped Element References

One of the most common sources of "noise" failures in WebDriver automation is the StaleElementReferenceException, which occurs when you try to use a web element reference that the driver no longer recognizes. Some action of the user or the application has altered the structure or content of the page, causing the browser to decide that it needs to rebuild its model of the page (the DOM), replacing all of its previous element identifiers with new ones. Even if the structure and content of the page is identical to what it was previously (e.g. - simply refreshing the page), element references that were acquired prior to the browser rebuilding the DOM are unusable. This can be particularly vexing when you're automating a web application with dynamic content. Your automation can fail at any moment, and your code quickly becomes littered with try/catch blocks as you try to prevent these random, pointless failures.

Selenium Foundation saves you from this frustration by producing and using wrapped element references that automatically handle stale element reference failures. The RobustWebElement class is a fault-tolerant wrapper for native WebElement objects. It retains the locator that was used to find the reference and the context in which the search was performed, and every wrapper method includes handling for StaleElementReferenceException failures. Whenever a stale reference is encountered, RobustWebElement uses the original locator and search context to re-acquire the reference that went stale. Prior to locating a new reference for an affected element, the search context hierarchy is refreshed as needed.

  • If a new reference for the affected element is acquired, the request that triggered the StaleElementReferenceException failure will be re-issued. Your automation will continue to run, completely unaware that the entire world shifted underneath.
  • If the attempt to re-acquire the element reference or refresh the search context hierarchy fails, the original exception is re-thrown. This gives you the diagnostic information about what your automation was attempting to accomplish at the point where it failed.

Using Optional Elements

When developing models for web application interfaces, you're likely to encounter elements that only exist in specific scenarios. For example, your web application displays an error message when a required value is omitted, and the element that contains this message is dynamically created when the error is detected. Selenium Foundation provides a convenient, efficient method for handling these sorts of scenarios - the optional element:

Optional elements
public class ItemListRulesDialog extends Dialog {
 
    ...
  
    /**
     * Get the header text for any error messages that are being shown.
     *
     * @return error message header text; 'null' if none is present
     */
    public String getErrorHeader() {
        RobustWebElement header = findOptional(Using.ERROR_HDR.locator);
        return (header.hasReference()) ? header.getText() : null;
    }
}

In this example, the ERROR_HDR element only exists when an error has been detected. This method searches for the optional error header element, returning the error message header text if it's found or 'null' if it's not. Unlike the handling of conventional element search requests, in which the configuration of the "implicit wait" setting can cause the API to poll the DOM repeatedly if a specified element isn't immediately available, optional elements are always only searched for once. This means that you always get control back right away, and you don't need to handle NoSuchElementException failures.

As indicated by the example, the findOptional() method returns a RobustWebElement object. To determine if the specified element was actually found, use the hasReference() method. If the result is 'true', a reference has been acquired for the element (i.e. - the optional element exists).

In almost all respects, optional elements for which references have been acquired behave exactly the same as their conventional counterparts. The one scenario in which they differ is when a stale element reference can't be re-acquired:

  • For conventional RobustWebElement objects, invoking any method will result in a StaleElementReferenceException failure.
  • For optional RobustWebElement objects...
    • ... invoking most methods will result in a NullPointerException failure, caused by NoSuchElementException.
    • ... invoking isDisplayed() or isEnabled() will return 'false'.

Written with StackEdit.