Fixed #37154 -- Switched from Selenium to Playwright for integration testing.

[varunkasyap]

Notes:

• selenium find element refer to actual DOM element, but playwright locator is not DOM reference, its lazy selector (lazy evaluation model).

so, Extracting the Tag Name from an Element tag_name selenium possible
example

element = driver.find_element(By.ID, "submit-button")
element_tag = element.tag_name
print(element_tag)  # Output: e.g., "button" or "input"

due to architecture change in Playwright, it doesn't natively expose DOM properties.

playwright translation looks like

element_tag = locator.evaluate("el => el.tagName").lower()

[varunkasyap]

Playwright docs on CI:

https://playwright.dev/python/docs/ci#:~:text=On%20push/pull_request

[varunkasyap]

In Playright:
wait_for_load_state() - Returns when the required load state has been reached.

Arguments

1. state "load" | "domcontentloaded" | "networkidle" (optional)#

example-

• page.wait_for_load_state("load")
• page.wait_for_load_state("domcontentloaded")
• page.wait_for_load_state("networkidle")

Optional load state to wait for, defaults to load. If the state has been already reached while loading current document, the method resolves immediately. Can be one of:

'load' - wait for the load event to be fired.
'domcontentloaded' - wait for the DOMContentLoaded event to be fired.
'networkidle' - DISCOURAGED wait until there are no network connections for at least 500 ms. Don't use this method for testing, rely on web assertions to assess readiness instead.

timeout float (optional)#

[varunkasyap]

In playwright:

expect() is used for assetions

reference - https://playwright.dev/python/docs/test-assertions#list-of-assertions

[varunkasyap]

Playwright uses milliseconds, whereas, selenium used seconds

[varunkasyap]

otherwise, getting this error

File "C:\Users\varun\OneDrive\Desktop\os\django\django\utils\asyncio.py", line 24, in inner
raise SynchronousOnlyOperation(message)
django.core.exceptions.SynchronousOnlyOperation: You cannot call this from an async context - use a thread or sync_to_async.

[varunkasyap]

temporary workaround for now, TODO - should make this compatible for all browsers.

[sarahboyce]

Thank you!
I've added a few suggestions from an initial look at the tests

[sarahboyce]

Suggested change

	if (
	multiprocessing.get_start_method() in {"spawn", "forkserver"}
	and options.parallel != 1
	):
	parser.error(
	"You cannot use --playwright with parallel tests on this system. "
	"Pass --parallel=1 to use --playwright."
	)

I can remove this without errors. I think ideally our tests don't need to have a shared state so would be interested if we can have this be removed and only added if really needed

[sarahboyce]

I don't think we will need a playwright_hub

[sarahboyce]

I think let's reuse the existing headless option

[sarahboyce]

Suggested change

	self.page.click(f"input[value='{login_text}']")
	self.page.wait_for_load_state("load")
	with self.page.expect_navigation():
	self.page.get_by_role("button", name=login_text).click()

Perhaps?

[varunkasyap]

Thanks for the review @sarahboyce

expect_navigation() seems depricated

Playwight is suggesting to use page.wait_for_url() instead.

so,

Suggested change

	self.page.click(f"input[value='{login_text}']")
	self.page.wait_for_load_state("load")
	self.page.get_by_role("button", name=login_text).click()
	self.page.wait_for_url(f"{self.live_server_url}{login_url}")

Perhaps?

[sarahboyce]

I think we can have expect(page).to_have_url(...) and I think the url is the admin index page
I'm not sure what's best practice really, it would be good for us to agree what we should do in most navigation cases

[sarahboyce]

Suggested change

	size = self.page.viewport_size
	self.page.set_viewport_size(
	{"width": size["width"] + 1, "height": size["height"]}
	)
	self.page.wait_for_load_state("domcontentloaded")
	self.page.set_viewport_size(size)
	self.page.wait_for_load_state("domcontentloaded")
	self.page.evaluate(
	"window.dispatchEvent(new Event('resize'))"
	)

And perhaps the selenium version should be:

        self.selenium.execute_script(
            "window.dispatchEvent(new Event('resize'));"
        )

I also think we barely use this so wondering if it is really worth being a helper method 🤔

[sarahboyce]

As agreed, we should decide if each of these helpers are needed

[sarahboyce]

Suggested change

	is_valid = self.page.evaluate(
	"document.getElementById('id_number').checkValidity()"
	)
	is_valid = number_input.evaluate("el => el.checkValidity()")

Perhaps

[sarahboyce]

Is the default_timout logic required? Not sure if we were hacking in things for selenium which playwright solves for us

[sarahboyce]

Suggested change

	self.page.locator("#submit").click()
	self.page.wait_for_load_state("load")
	with self.page.expect_navigation():
	self.page.locator("#submit").click()

[sarahboyce]

Is this needed?

[varunkasyap]

The Selenium test intentionally clicks disabled buttons

Playwright refuses to click disabled elements by default. Hence, added force=True

[varunkasyap]

Safari User Guide says:

Highlight the next field or pop-up menu on a web page: Tab
Highlight the next field, pop-up menu or clickable item (such as a link) on a web page: Option-Tab

reference:
https://support.apple.com/en-in/guide/safari/cpsh003/mac#:~:text=Highlight%20the%20next,of%20Safari%20settings.

[varunkasyap]

The skip link hides via top: -999px CSS, which is off-screen but Playwright's is_visible() still considers it visible. Checking getBoundingClientRect().top >= 0 can detects whether the element is on-screen.

alternatives:
• self.assertFalse(...) → expect(skip_link).not_to_be_in_viewport()
• self.assertTrue(...) → expect(skip_link).to_be_in_viewport()