Skip to content

Latest commit

 

History

History
341 lines (245 loc) · 12.1 KB

README.rst

File metadata and controls

341 lines (245 loc) · 12.1 KB
Raw

e2e.pom - Page Object Modelling for End-to-End Testing

Use e2e.pom to easily create reusable Page Object Models that offer functionality and error scenarios more naturally than Selenium can.

Extend available locators with plugins like e2e.pom.extjs for ExtJS/Sencha Locators, or write your own!

Eliminate iframe-handling in your tests — let the model take care of where your elements reside. In fact, all elements are guaranteed to be found within their parent model so you can write unique selectors without specifying a long ancestry within the selector itself!

All actions are logged with human-friendly labels, with human-friendly exceptions indicating exactly where things went wrong.

Beta Notice

e2e.pom is still in beta, but the API is expected to be stable. The project is very open to suggestions and feedback at this stage!

This project follows SemVer. It is expected that the first major release (1.x) will not differ in API functionality much from the alpha/beta track (0.x).

Python 3.6+ is currently required. No support for 2.x is ever planned since it is soon to be unsupported. Please feel free to file an issue if you need support for an earlier Python version and simply cannot upgrade.

Modelling Examples

Pages and their components are easy to model on a class-member basis. You do not need to pass around a driver, nor do you need to write properties or methods. There is no need to override __init__ properly either; just author your static model and be done with it!

class LoginForm(pom.Region):
    username_field = pom.Element("Username").by_css('.username')
    password_field = pom.Element("Password").by_css('.password')
    login_button = pom.Element("Login btn").by_css('#login-submit')

class LoginPage(pom.Page):
    login_form = LoginForm("Login form").by_css('.login')

# ...

login_pg = LoginPage("https://my-login-service.local:8443", driver)
login_pg.login_form.wait_for_visible()
login_pg.login_form.username_field.set_text('Admin1')
login_pg.login_form.password_field.set_text('Sup3rS3cret!')
login_pg.login_form.login_button.click()

IFrame Handling

Handling iframes with Selenium is incredibly annoying. e2e.pom provides a special component that you can use as a hidden parent so that test authors do not need to be concerned about the iframe at all.

class LoginForm(pom.Region):
    # Same as Example Model above

class LoginPage(pom.Page):
    # Hide the iframe POM component so the test doesn't need to be aware by
    # overriding the parent of the Form.
    _login_service_iframe = pom.IFrame("LoginService iframe").by_css('iframe#login-service')
    login_form = LoginForm("Login form").by_css('.login', parent=_login_service_iframe)

# Use exactly as before, like the iframe isn't even there!
login_pg.login_form.wait_for_visible()

The POM framework will then handle switching to the correct iframes. Even if a service becomes resident within an iframe that was otherwise inline before, you only need to change the model and not the test code!

In addition, features like is_visible() are mindful of iframes. That is, unlike using Selenium directly, if the iframe is not displayed then e2e.api counts all of its children as invisible.

Hidden Regions

Like the above case for IFrames, you may want to hide a certain part of the model but otherwise use it as a parent to other components. This is often useful for ensuring elements are found within a certain part of the page, but would be too verbose to specify in every test.

class ServicePage(pom.Page):
    _nav_bar = pom.Region("Navigation bar").by_css('.upper-nav-container')
    settings_button = pom.Element("Setting btn").by_css('.gear-icon')

# You can now do this...
service_pg.settings_button.click()

# Instead of having to do this
service_pg.navigation_bar.settings_button.click()

This is especially useful when you have, like in the above example, multiple elements that use .gear-icon but are obviously referring to one particular instance of it. If it is beneficial to the clarity of your model though, you may want to not hide the parent since the increased verbosity may make your test’s actions more clear to others.

Interacting with Elements

While some available methods are similar to Selenium’s API, they have differing requirements and behaviours.

For example, the concept of “visibility” is now more user-friendly. Whether an element is visible to the user depends not only on the display info of the element, but whether the element exists at all. With Selenium, entirely different code is required to handle that difference! With e2e.pom, it is just the visiblility functions. This allows product developers to use techniques like DOM caching without needing to change the test code. This way, e2e.pom is checking user intent rather than implementation details.

# Case 1: With Selenium, element not in DOM. Can also assert NoSuchElement was raised.
assert driver.find_elements_by_xpath(...) == []

# Case 2: With Selenium, element in DOM
assert driver.find_elements_by_xpath(...).is_displayed()

# With happy-path-only POM frameworks (not e2e.pom)
try:
    assert modelled_element.is_displayed()
except NoSuchElementException:
    pass

# With e2e.pom (works for both above cases)
assert modelled_element.is_visible()

Nothing happens until you call a method on an element. This is to give clarity as to when the finding, attribute checking, and interaction takes place within a test.

# Selenium: All of these do something with the driver, even the properties!
element = driver.find_element_by_css_selector(...)
element.text
element.is_displayed()
element.click()
element.clear() and element.send_keys("some text")

# e2e.pom: Consistent API. Methods == driver actions.
element.get_text()
element.is_visible()
element.click()
element.set_text("some text")

All user-like interaction methods require visibility of the element. If the element is not visible, a user cannot interact with it. Exceptions will be raised with the exact details (e.g. if it was because a parent could not be located, the exception will state this).

Sets of Elements

Sometimes you are intentionally referring to a group of similar elements, like rows of a particular type. You can’t use a pom.Element for this because they require uniquely finding an element. In this case, you can use pom.ElementCollection.

Rich Visibility

To-do.

Using Locators Provided by Third-Party Plugins

In certain projects, alternative locator methods may be available that are more deterministic of exact elements.

A first-party example of this, e2e.pom.extjs, provides locators for pages built with ExtJS/Sencha.

This extensibility has a few drawbacks to be aware of:

  • The generic by(...) must be used instead of something like by_css(...)
  • Many setups will fail to type-check or auto-complete this correctly.

Using it is very easy though! Just ensure it is installed, then use the provided Locator class directly in the by() construction.

from e2e import pom
from e2e.pom import extjs

class ServicePage(pom.Page):
    alert_popup = pom.Element("Alert").by(extjs.Locator('panel[title=Alert]'))

Developing Locator Extensions

Please use e2e.pom.extjs as an implementation guide.

The following are recommended:

  • The extension locator class should be called Locator where possible.
  • find_within is the only method needing implementation. It will be passed either a WebDriver or a WebElement. For the WebElement, you must assure that anything returned by your Locator is within the given WebElement.
  • Don’t forget to register the entrypoint!

Pytest Hooks & Fixtures (BETA)

Note: This feature is under development and will change. Feedback is much appreciated!

A information-providing fixture is provided for dumping info useful in test failure analysis.

You must register the drivers you use with this fixture. We recognize users will have their own ways of managing drivers, so this is the best way currently.

def test_my_thing(e2e_pom_dumper: pom.fixtures.DumperFixture):
    my_driver = ...  # However you get this
    e2e_pom_dumper.register_driver(my_driver)
    # Now you can use the driver!

After that, the following features are provided:

  • On test failure, two DOMs are dumped to HTML files:
    • The current frame/iframe at the time of failure (if not the root frame)
    • The root frame
  • A screenshot is taken at the outermost/root frame

Performance Considerations

Certain features like Collections as parents can result in a lot of Selenium requests being issued. If you test against a busy Selenium Grid, this can definitely slow things down!

I do not normally recommend the below workarounds (because they bypass some significant benefits of the framwork), but you may find them necessary in your case.

  1. Avoid a pom.Collection parent if it represents many elements (e.g. rows in a large table). Every found element for the collection is searched for the next child.
  2. Reduce POM depth where possible. Since each child causes a Selenium find request on interaction, reducing the size of that parentage can speed things up.

For the first case, this can be easy to work around with a dynamic model if you are targeting a particular item.

class MyTable(pom.Container):

    rows = pom.Collection('Rows').by_css('tr')

    def row_by_title(self, title: str):
        label = f'Row by title [{title}]'
        xpath = f'.//td[text()="{title}]/ancestor::tr"'
        # Do this IF Selenium performance isn't critical to you (most cases)
        return pom.Element(label).by_xpath(xpath, parent=rows)
        # Do this IF and ONLY IF Selenium performance is very important to you
        return pom.Element(label).by_xpath(xpath)

Note how rows is on the model directly, but also used as a parent override in the method. However, skipping rows as a parent means we can forego some more detailed information on failure while making less Selenium calls.

Making Richer Models

Generic type information can be passed regarding the type of located elements.

This is especially useful for pom.ElementCollections. For example, you may want to have an icon property modelled on each dropdown item returned from a collection. This is normally modelled by a Findable component (e.g. pom.Container or pom.Element), but these cannot be used for found elements (those returned by pom.ElementCollection.get()). Enter, generics.

On construction of any findable POM type, you can specify the concrete type returned by findable.get() or acted on via operations like findable.click(). By default, this is a pom.dom.ElementReference, but you may extend it to provide more functionality.

# TODO: assert Reference parents actually work!

class Dropdown(pom.Container):

    class IconItem(pom.dom.ElementReference):
        icon = pom.Element("Icon").by_css(".option-icon")

    items = pom.ElementCollection("Dropdown items", IconItem).by_css("option")

first_item = ...dropdown.items.get()[0]
assert 'gear-icon' in first_item.icon.get_attribute('class')

The type of items is now a pom.ElementCollection<IconItem>. If it was not specified, it would be a pom.ElementCollection<dom.ElementReference>.

This works because while an ElementReference is not “Findable”, it is “Parentable”. That is, it can be a parent, and thus contain other modelled children.

In the future, a pom.HybridElement may exist that can be used to represent these items both as a Findable and a Found object to avoid code duplication.

Basically, a rich interface is what you write.