finraos / jtaf-extwebdriver Goto Github PK

In our projects we use slf4j combined with logback. So we must now mixed two logframeworks (logback, log4j) in the same project. Migration should be an easy task.

if (currentWindowIds == null && currentWindowIds.size() > 0)

Fixed to

if (currentWindowIds != null && currentWindowIds.size() > 0)

These used to be methods in EWD but are no longer there...

Github io pages should be updated to use the latest 1.4 APIs. We should specifically change our sample codes with the By locator.

If #65 is fixed I could integrate an automatic start of jetty in the test. Makes it more comfortable to use in an IDE.

Currently we have PMD turned on with WARNING only. We need to fix all the PMD issues and turn it on with FAILURE.

We need to create unit tests for the following classes:

• org.finra.jtaf.ewd.impl.ClientProperties
• org.finra.jtaf.ewd.properties.GUIProperties
• org.finra.jtaf.ewd.properties.GUIHierarchyConcatenationProperties

Currently, the use of DefaultSessionFactory is hard-coded in SessionManager.getInstance(), such that any custom SessionFactory previously registered with SessionManager.setSessionFactory(...) is discarded.

The relevant code from SessionManager.java can be seen in lines 71-73 here:
https://github.com/FINRAOS/JTAF-ExtWebDriver/blob/master/src/main/java/org/finra/jtaf/ewd/session/SessionManager.java

public static SessionManager getInstance() {
return sessionManager.get().setSessionFactory(new DefaultSessionFactory());
}

The fix is probably to have the SessionManager use its current sessionFactory variable here, which will require sessionFactory (which holds the current factory implementation registered with the SessionManager) to become a ThreadLocal because getInstance() is (and should be) static.

I will submit a pull request shortly.

The conditional at line 295 currently checks for null, then if null checks for emptiness:

if (currentWindowIds == null && currentWindowIds.size() > 0) {...

This will throw a NullPointerException whenever currentWindowIds is null. Switching the operator from an AND to an OR should avoid the NPE.

The second condition doesn't completely make sense in the context of the method, possibly supposed to be testing for equality rather than for greater than.

The constructor of org.finra.jtaf.ewd.widget.element.html.Table will throw a NoSuchElementException if the table does not exist because it calls generateXPathLocator(String locator). It should be possible to at least create the Table object and then later check .isElementPresent(), which is what happens with other widgets.

The system which does our CI builds on commit (Travis CI) takes a custom command for deploy. Travis also has the nice feature of automatic "pull request builds" that kick off when a pull request is made.

According to Travis docs and our experience so far:

• Travis does not make encrypted environment variables available to pull request builds - only to builds triggered by a commit to master.
• As a result, pull request builds do not have authorization to deploy snapshots, which is a good, safe thing.
• However, apart from the environment variable availability, pull request builds use the same Travis configuration for build and deploy steps. Because the deploy fails, the entire build fails.
• Travis makes available an environment variable to help us out: TRAVIS_SECURE_ENV_VARS is set to "true" or "false" depending on the availability of [secure] environment variables.

Given all of that, we should probably wrap our Maven build step with a script that checks this environment variable and avoids deploying altogether if the variables are not available.

Reference:
http://about.travis-ci.org/docs/user/build-configuration/#Set-environment-variables

Latest version of Selenium is 2.39. Should we upgrade from our current version, 2.37.1?

Eclipse creates some warning about unused code. This should be removed except, you use it via reflection.

We need to create tests which only executes on a real browser instead of HtmlUnit only. For instance, we need to create tests for ExtWebDriver interface which deals with browser related activities such as back, forward, dialogs, popups. This suite should also include tests around mouse events such as drag and drop which doesn't run on HtmlUnit.

This suite needs to run against Sauce Labs instead of Travis CI.

It fails with a fatal error when highlight = somevalue is used because of the split statement, and also on highlight.GET, highlight.PUT, highlight.FIND = somecolor.

Also it seems that the ClientProperties readme on the github.io site has incorrect property names for the highlight color properties.

With recent changes made to improve performance we should allow the setting of a default locator so that we people using the pre-existing locator constructors can get the most out of these changes (they can still have the ability to change locator for each widget by just providing the locator type)

In org.finra.jtaf.ewd.impl.DefaultSessionFactory.createInnerDriver(Map<String,String>,DesiredCapabilities), lines 376-405:

else if ((browser.equalsIgnoreCase("firefox") || browser.equalsIgnoreCase("*firefox"))) {
        final String ffProfileFolder = properties.getFirefoxProfileFolder();
        final String ffProfileFile = properties.getFirefoxProfileFile();
        final String path = properties.getBinaryPath();

        if (path != null) {
            // Omitted: FirefoxBinary fireFox and FirefoxProfile ffp are created & configured 
            wd = new FirefoxDriver(fireFox, ffp, desiredCapabilities);
        } else {
            wd = new FirefoxDriver(desiredCapabilities);
        }
}

If the binary path is null, the factory creates a default FirefoxDriver and ignores your profile config.

The old EWD didn't require an explicit binary path to use a profile - is this change intended, or shall I fix it?

We need to have our build integration tests and browser only tests to run against Sauce Labs.

1. We need to create an Open Sauce account
2. Configure Travis CI to run the tests

Whoopsie. It extends Exception

Add to all the pages to support IE compatibility mode.

From the code of Element.java in the ewd.element package, we have the following trace occurring in findElement when highlighting is enabled:

• findElement calls highlight
• highlight calls setBackgrounColor
• setBackgroundColor calls eval
• eval calls getWebElement
• getWebElement calls findElement
• etc.

We need to:

• fix the code to avoid the infinite loop
• add unit tests for highlighting functionality (not just find, but get and put highlighting as well -- see Element.java for uses of highlight(..) )

The following file have CRLF as line encoding. This should be Unix-Style LF
./src/test/java/org/finra/jtaf/ewd/widget/element/html/TableTest.java: ASCII C program text, with CRLF line terminators
./src/test/java/org/finra/jtaf/ewd/widget/element/html/HtmlListTest.java: ASCII C program text, with CRLF line terminators
./src/test/java/org/finra/jtaf/ewd/widget/element/html/InputTest.java: ASCII C program text, with CRLF line terminators
./src/test/java/org/finra/jtaf/ewd/widget/element/html/ImageTest.java: ASCII C program text, with CRLF line terminators
./src/test/java/org/finra/jtaf/ewd/session/SessionManagerTests.java: ASCII C program text, with CRLF line terminators
./src/test/java/org/finra/jtaf/ewd/session/DefaultSessionFactoryTests.java: ASCII C program text, with CRLF line terminators
./src/test/resources/test.properties: ASCII text, with CRLF line terminators
./src/test/resources/log4j.properties: ASCII text, with CRLF line terminators
./src/main/java/org/finra/jtaf/ewd/impl/ClientProperties.java: ASCII C program text, with CRLF line terminators
./src/main/java/org/finra/jtaf/ewd/session/SessionFactory.java: ASCII C program text, with CRLF line terminators

To allow projects currently using string locators to be backwards compatible, ByLegacy should extend StringLocatorAwareBy as mentioned in #70. String locator methods in Element classes should be deprecated.

Currently you have to manually delete these folder when making pull requests. I think the reason for that is there are slashes preceding the dir names in the current gitignore when it it should be after the dir names

Provide widget library for bootstrap components/css. Reference following pages:

http://getbootstrap.com/components/
http://getbootstrap.com/css/

Provide a widget library for AngularJS. Reference the items specified here: http://docs.angularjs.org/api/. For example, different input types and filters.

Element.java findElement is the most used method in the framework and it makes try and error to find the component, first xpath, then byId, then byName .....
This can creates wrong results and is definitely not the fastest way. The main way to find elements should be based on the By-class of selenium. The String location should only be there for backward compatibility..

PhantomJS allows headless test against the QtWebKit.
You can find information how to integrate at

• Blog
• Ghotsdriver

and there is an interesting maven package from arquillian and one for the binaries.

Make the Element base class pluggable so that customizations can be made without having to override it.

Provide a widget library to handle jQuery UI widgets and controls. We should reference the widgets outlined here: http://jqueryui.com/demos/

This shouldn't be done - we should throw the MissingResourceException and IllegalArgumentException instead, or a more specific GUIPropertiesException

When running against Selenium Grid (as well as Sauce Labs), the RemoteWebDriver instance is getting replaced with local driver instance in DefaultSessionFactory. When useGrid is set to true, DefaultSessionFactory should return RemoteWebDriver instance.

It would be good if all integration-test are marked. I see 2 solutions:

1. By name-suffix *IntegrationTest
2. With @Category annotation.

Will start gathering some notes here:
https://github.com/FINRAOS/JTAF-ExtWebDriver/wiki/GWT-Widget-Library----DRAFT

@knappk has looked into this before, has many thoughts on this already.

1. Add PMD to the pom
2. Select basic & strict exceptions rulesets
3. Fix some of the reported errors

So that we can avoid exception handling code in tests and widget class implementations, we should support a getPropertyValueOrDefault(...) method.

This came up in discussion of #24

Upgrade Selenium dependency to 2.42.1+ (latest as of 05/29 is 2.42.1).

There are several CM-related tasks which need to be accomplished for ExtWebDriver:

1. Build on Travis CI
2. Run CI tests on Travis CI (these are headless integration tests)
3. Push snapshots to Sonatype OSS snapshots

About Travis:
http://about.travis-ci.org/docs/user/build-configuration/#.travis.yml-file%3A-what-it-is-and-how-it-is-used

About connecting with Sonatype:
http://blog.xeiam.com/2013/05/configure-travis-ci-to-deploy-snapshots.html

It will have a takeScreenshotOfLocator method which will take the following arguments:
String - locator of element to compare, String - relative folder path to which the bitmap_screenshot.png will be saved, File - control picture to compare against

There will also be an assertSimilar method which will take 2 BufferedImages and a double threshhold. An overloaded version will have a default threshhold of .85

• This method will use a similarity algorithm which returns a double between 0 and 1 that asserts how similar the 2 pictures are

We need to move the 4 failing integration tests to browser only test suite so that our build can pass. We will run the browser only tests against Sauce Labs.

Update folder structure to use multiple maven modules rather than single.

ChromeDriver allows users to set the chrome.switches via DesiredCapabilities. Current, our client properties doesn't support configuring chrome.switches. We should allow users to specify list of chrome.switches via client properties file to apply to ChromeDriver instances.

https://sites.google.com/a/chromium.org/chromedriver/capabilities

If you run, git clone git://github.com/FINRAOS/JTAF-ExtWebDriver.git, as is suggested in the Building section, you get the fatal error, "unable to connect to github.com".

It should be https://github.com...

I can't find any reason, why it is in In InteractiveElement? You can DragAndDrop every WebElement.

Furtermore the code is to complicated

builder.dragAndDrop(this.getWebElement(), 
        new InteractiveElement(element.getByLocator()).getWebElement()).build().perform();

Should be

builder.dragAndDrop(this.getWebElement(), 
            element.getWebElement()).build().perform();

The deploy.sh script may not be sufficient for forked repos to run their builds on travis-ci. The script handles pull requests builds on original repo but forked repos run mvn deploy instead of mvn install. We may need to modify the script to do deploy only when the build originates from FINRAOS project.

My suggestion: Create the private method in Element, isWithinBoundsOfWindow(), and add this check to the private isVisible() method, which is used in waitForElementPresent() and isElementPresent().

The element needs to have scrollTo() called on it before we take a picture of the web page, since RemoteWebDriver screenshots are only a picture of the page as it is scrolled to.

Workaround in the meantime is to call scrollTo() on the element before calling ScreenshotUtils.isSimilarToScreenshot().

This is a workaround for a fixed issue.