Adding new commands to testdriver.js

Assumptions

We assume the following in this writeup:

  • You know what web-platform-tests is and you have a working checkout and can run tests
  • You know what WebDriver or Selenium is
  • Familiarity with JavaScript and Python

Introduction!

Let’s implement window resizing. We can do this via the Set Window Rect command in WebDriver.

First, we need to think of what the API will look like a little. We will be using Selenium and Marionette for this, so we can look and see that they take in x, y coordinates, width and height integers.

The first part of this will be browser agnostic, but later we will need to implement a specific layer for each browser (here we will do Firefox and Chrome).

Code!

resources/testdriver.js

This is the main entry point the tests get. Here we need to add a function to the test_driver object that will call the test_driver_internal object.

window.test_driver = {

    // other commands...

    /**
    * Triggers browser window to be resized and relocated
    *
    * This matches the behaviour of the {@link
    * https://w3c.github.io/webdriver/webdriver-spec.html#dfn-set-window-rect|WebDriver
    * Set Window Rect command}.
    *
    * @param {Integer} x - The x coordinate of the top left of the window
    * @param {Integer} y - The x coordinate of the top left of the window
    * @param {Integer} width - The width of the window
    * @param {Integer} height - The width of the window
    * @returns {Promise} fulfilled after window rect is set occurs, or rejected in
    *                    the cases the WebDriver command errors
    */
    set_window_rect: function(x, y, width, height) {
        return window.test_driver_internal.set_element_rect(x, y, width, height);
    }

In the same file, lets add to the internal object. ( do we need to do this?) (make sure to do this if the internal call has different arguments than the external call, especially if it calls multiple internal calls)

window.test_driver_internal = {

    // other commands...

    /**
     * Triggers browser window to be resized and relocated
     *
     * This matches the behaviour of the {@link
     * https://w3c.github.io/webdriver/webdriver-spec.html#dfn-set-window-rect|WebDriver
     * Set Window Rect command}.
     *
     * @param {Integer} x - The x coordinate of the top left of the window
     * @param {Integer} y - The x coordinate of the top left of the window
     * @param {Integer} width - The width of the window
     * @param {Integer} height - The width of the window
     * @returns {Promise} fulfilled after window rect is set occurs, or rejected in
     *                    the cases the WebDriver command errors
     */
    set_window_rect: function(x, y, width, height) {
        return Promise.reject(new Error("unimplemented"))
    }

We will leave this unimplemented and override it in another file. Lets do that now!

wptrunner/wptrunner/testdriver-extra.js

This will be the default function called when invoking the test driver commands (sometimes it is overridden by testdriver-vendor.js, but this is outside the scope of this writeup).

window.test_driver_internal.set_element_rect = function(x, y, width, height) {
    const pending_promise = new Promise(function(resolve, reject) {
        pending_resolve = resolve;
        pending_reject = reject;
    });
    window.opener.postMessage(
        {"type": "action", "action": "set_window_rect", "x": x, "y": y, "width": width, "height": height}, "*");
    return pending_promise;
};

The main thing here is the postMessage argument. The first argument is an object with properties

  • type: this always has to be the string "action"
  • action: the name of the testdriver command this defines (in this case, set_window_rect)
  • any other things you want to pass to the next point of execution (in this case, the x, y coordinates and the width and height)

The pending promise is out of scope of this function and is resolved when the window recieves a completion message from the executor. This happens here in the same file:

    let pending_resolve = null;
    let pending_reject = null;
    window.addEventListener("message", function(event) {
        const data = event.data;

        if (typeof data !== "object" && data !== null) {
            return;
        }

        if (data.type !== "testdriver-complete") {
            return;
        }

        if (data.status === "success") {
            pending_resolve();
        } else {
            pending_reject();
        }
    });

One limitation this introduces is that only one testdriver call can be made at one time since the pending_resolve and pending_reject variables are in an outer scope.

Next, this is passed to the executor and protocol in wptrunner. Time to switch to Python!

tools/wptrunner/wptrunner/executors/protocol.py

class SetWindowRectProtocolPart(ProtocolPart):
    """Protocol part for resizing and changing location of window"""
    __metaclass__ = ABCMeta

    name = "set_window_rect"

    @abstractmethod
    def set_window_rect(self, x, y, width, height):
        """Change the window rect

        :param x: The x coordinate of the top left of the window.
        :param y: The y coordinate of the top left of the window.
        :param width: The width of the window.
        :param height: The height of the window."""
        pass

Next we change the base executor.

tools/wptrunner/wptrunner/executors/base.py

class CallbackHandler(object):
    """Handle callbacks from testdriver-using tests.

    The default implementation here makes sense for things that are roughly like
    WebDriver. Things that are more different to WebDriver may need to create a
    fully custom implementation."""

    def __init__(self, logger, protocol, test_window):
        self.protocol = protocol
        self.test_window = test_window
        self.logger = logger
        self.callbacks = {
            "action": self.process_action,
            "complete": self.process_complete
        }

        self.actions = {
            "click": ClickAction(self.logger, self.protocol),
            "send_keys": SendKeysAction(self.logger, self.protocol),
            {other actions},
            "set_window_rect": SetWindowRectAction(self.logger, self.protocol) # add this!
        }
class SetWindowRectAction(object):
    def __init__(self, logger, protocol):
        self.logger = logger
        self.protocol = protocol

    def __call__(self, payload):
        x, y, width, height = payload["x"], payload["y"], payload["width"], payload["height"]
        self.logger.debug("Setting window rect to be: x=%s, y=%s, width=%s, height=%s"
                          .format(x, y, width, height))
        self.protocol.set_window_rect.set_window_rect(x, y, width, height)

Don’t forget to write docs in testdriver.md. Now we write the browser specific implementations.

Chrome

We will use executorselenium and use the Selenium API (in the future there are plans to use the WebDriver API directly).

There isn’t too much work to do here, we just need to define a subclass of the protocol part we defined earlier.

class SeleniumSetWindowRectProtocolPart(SetWindowRectProtocolPart):
    def setup(self):
        self.webdriver = self.parent.webdriver

    def set_window_rect(self, x, y, width, height):
        return self.webdriver.set_window_rect(x, y, width, height)

Make sure to import the protocol part too!

from .protocol import (BaseProtocolPart,
                       TestharnessProtocolPart,
                       Protocol,
                       SelectorProtocolPart,
                       ClickProtocolPart,
                       SendKeysProtocolPart,
                       {... other protocol parts}
                       SetWindowRectProtocolPart, # add this!
                       TestDriverProtocolPart)

Here we have the setup method which just redefines the webdriver object at this level. The important part is the set_window_rect function (and it’s important it is named that since we called it that earlier). This will be call the Selenium API for set window rect (self.webdriver is a Selenium WebDriver instance here).

Finally, we just need to tell the SeleniumProtocol to implement this part.

class SeleniumProtocol(Protocol):
    implements = [SeleniumBaseProtocolPart,
                  SeleniumTestharnessProtocolPart,
                  SeleniumSelectorProtocolPart,
                  SeleniumClickProtocolPart,
                  SeleniumSendKeysProtocolPart,
                  {... other protocol parts}
                  SeleniumSetWindowRectProtocolPart,
                  SeleniumTestDriverProtocolPart]

Firefox

We use the set window rect Marionette command.

We will use executormarionette and use the Marionette Python API.

We have little actual work to do here! We just need to define a subclass of the protocol part we defined earlier.

class MarionetteSetWindowRectProtocolPart(SetWindowRectProtocolPart):
    def setup(self):
        self.marionette = self.parent.marionette

    def set_window_rect(self, x, y, width, height):
        return self.marionette.set_window_rect(x, y, width, height)

Make sure to import the protocol part too!

from .protocol import (BaseProtocolPart,
                       TestharnessProtocolPart,
                       Protocol,
                       SelectorProtocolPart,
                       ClickProtocolPart,
                       SendKeysProtocolPart,
                       {... other protocol parts}
                       SetWindowRectProtocolPart, # add this!
                       TestDriverProtocolPart)

Here we have the setup method which just redefines the webdriver object at this level. The important part is the set_window_rect function (and it’s important it is named that since we called it that earlier). This will be call the Marionette API for set window rect (self.marionette is a marionette instance here).

Finally, we just need to tell the SeleniumProtocol to implement this part.

class MarionetteProtocol(Protocol):
    implements = [MarionetteBaseProtocolPart,
                  MarionetteTestharnessProtocolPart,
                  MarionettePrefsProtocolPart,
                  MarionetteStorageProtocolPart,
                  MarionetteSelectorProtocolPart,
                  MarionetteClickProtocolPart,
                  MarionetteSendKeysProtocolPart,
                  {... other protocol parts}
                  MarionetteSetWindowRectProtocolPart # add this
                  MarionetteTestDriverProtocolPart]

Other Browsers

Other browsers may also use executorselenium (such as safari), or a completely new executor (such as servo). For these, you must change the executor in the same way as we did with chrome and firefox.

Write an infra test

Make sure to add a test to infrastructure/testdriver :)

Here is some template code!

<!DOCTYPE html>
<meta charset="utf-8">
<title>TestDriver set window rect method</title>
<script src="/resources/testharness.js"></script>
<script src="/resources/testharnessreport.js"></script>
<script src="/resources/testdriver.js"></script>
<script src="/resources/testdriver-vendor.js"></script>

<script>
promise_test(async t => {
  await test_driver.set_window_rect(100, 100, 100, 100);
  // do something
}
</script>

What about testdriver-vendor.js?

The file testdriver-vendor.js is the equivalent to testdriver-extra.js above, except is run instead of testdriver-extra.js in browser specific test environments. For example, in Chromium LayoutTests.

What if I need to return a value from my testdriver API?

We currently don’t have this capability, but it is coming soon and will be documented. The bug is here