"Playwright", which was an E2E automatic test tool for JavaScript, ** It was preview-released as "Playwright for Python" for Python **, so I immediately touched it.
Announcing Playwright for Python: Reliable end-to-end testing for the web
Playwright is a library that automates Chromium, Firefox, and Webkit with a single API. This allows you to perform fast and reliable cross-browser tests.
The Official Document states the following four points. Below is my translation.
--Supports Chromium, Firefox, Webkit (Safari). --Cross-platform testing of Webkit is possible. (Windows, Mac, Linux) --You can test your browser for mobile. --It can be executed in either headless mode or headful mode.
--The API will wait automatically. ――Automation without timeout is possible. --Parallel execution is possible using Browser Context. --You can use an element selector that is resistant to changes.
-Supports multiple domains, pages, and frames. --Powerful network control is possible. --You can use modern Web functions. (Shadow DOM, geolocation, permissions, etc ...) ――It has a function to cover all scenarios. (From file DL / UL to dark mode)
--Can be installed in one line. --Supports TypeScript. --Can be integrated with debugging tools. --Python and C # bindings are available. --You can deploy the test to CI. (Official Docker image, provided by GitHub Actions)
From here, I will actually write a test case. I used the HOTEL PLANISPHERE room reservation (plan with special benefits) page for the test target. .. The test case was a normal test, "Access the page to be tested, set each parameter, then move to the reservation confirmation page, and if the amount displayed there is 15,000 yen, pass".
Install according to the description on the official GitHub. Also, this time we will use pytest as a unit test framework, so Install the pytest plugin.
pip install playwright
python -m playwright install
pip install pytest-playwright
It will not be executed this time, but if you want to execute asynchronous processing, please perform pip install asyncio separately.
Before touching Playwright, I wrote a test case that I want to realize this time with Selenium Webdriver as a reference. It's quite difficult to manage the number of import statements, wait processing for dynamic content, and Webdriver management for each browser.
selenium.py
import time
from selenium import webdriver
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.ui import Select, WebDriverWait
from webdriver_manager.chrome import ChromeDriverManager
class TestPlanisphere:
def test_reserve_otoku(self):
driver = webdriver.Chrome(ChromeDriverManager().install())
wait = WebDriverWait(driver, 10)
driver.get("https://hotel.testplanisphere.dev/ja/reserve.html?plan-id=0")
# NOTE: wait.until(EC.presence_of_all_elements_located) <-Doesn't work
time.sleep(1) # HACK
date = driver.find_element_by_id("date") #Accommodation date
date.clear()
date.send_keys("2020/12/24")
date.send_keys(Keys.TAB)
term = driver.find_element_by_id("term") #number of stays
term.clear()
term.send_keys("2")
head_count = driver.find_element_by_id("head-count") #Number of people
head_count.clear()
head_count.send_keys("1")
sightseeing = driver.find_element_by_id("sightseeing") #Select "Great Sightseeing Plan"
if not sightseeing.is_selected():
sightseeing.click()
driver.find_element_by_id("username").send_keys("Beri-bo") #name
Select(driver.find_element_by_id("contact")).select_by_value("tel") #"I would like to be contacted by phone"
driver.find_element_by_id("tel").send_keys("80120828828") #phone number
driver.find_element_by_id("submit-button").click() #Transition to reservation confirmation page
wait.until(EC.presence_of_all_elements_located)
driver.save_screenshot("selenium.png ") #Take screenshot
assert "15,000" in driver.find_element_by_id("total-bill").text
driver.quit()
Then I wrote the same test in Playwright for Python. The methods used this time are summarized on the following pages of the official document. Although it is described only in JavaScript, it can be used almost entirely in Python.
play.py
from playwright.sync_api import Page
class TestPlanisphere:
def test_reserve_otoku(self, page: Page):
page.goto("https://hotel.testplanisphere.dev/ja/reserve.html?plan-id=0")
page.waitForLoadState("networkidle")
page.fill("#date", "2020/12/24") #Accommodation date
page.press("#date", "Tab")
page.fill("#term", "2") #number of stays
page.fill("#head-count", "1") #Number of people
page.check("#sightseeing") #Plan selection
page.fill("#username", "Beri-bo") #name
page.selectOption("#contact", "tel") #"I would like to be contacted by phone"
page.fill("#tel", "80120828828") #phone number
page.click("#submit-button") #Transition to reservation confirmation page
page.waitForLoadState("networkidle")
page.screenshot(path="playwright.png ") #Take screenshot
assert "15,000" in page.innerText("#total-bill")
page.close()
Compared to Selenium
--Simple module configuration, almost no import statement required --If you import the Page type and combine it with pylance, you can complete the method search on VS Code, which is very convenient. --Cross-browser support (the browser to start can be specified when running pytest) --Each browser is built into the package, so there is no need for webdriver version control. --Excellent standby processing for dynamic content
You can see that Selenium has solved the difficult part by just looking at the script.
All you have to do now is run the above play.py in pytest. You can specify headless / headful and execution browser as a run-time option.
You can also save the boot options in the file pytest.ini. ](Https://github.com/microsoft/playwright-pytest#usage)
#Test run(Headless Chromium is the default)
pytest play.py
#Test run in headful mode
pytest --headful play.py
#Test execution by specifying a browser(chromium, firefox, webkit)
pytest --browser firefox play.py
#Test run in multiple browsers
pytest --browser chromium --browser webkit play.py
I think there are two directions for the evolution of automation tools, I think that one is to evolve as a multifunctional test platform that integrates reporting functions and CI functions, and the other is to evolve as a lightweight, high-speed, programmable tool. I felt that Playwright would be a powerful option for the latter.
The backing of a giant called Microsoft (and the origin is Google's Puppeteer ...) is very powerful as OSS, and future enhancements can be expected, so it is better to start from now on the preview release. think.
To be honest, I personally felt that it was much easier and less expensive to learn than Selenium, so I will give priority to Selenium when it can be used!
Recommended Posts