Use selenium to write in and run a code cell in jupyterlab

Question

I would like to test a patched implementation of juptyerlab. I was hoping to use selenium to execute "hello world" in a code cell. So far I can log in and create a new notebook:

from selenium import webdriver

driver = webdriver.Firefox()
# assume jupyterlab is running and serving on localhost at port 8888
driver.get("http://localhost:8888")
elem = driver.find_element_by_id("password_input")
password = ""
elem.send_keys(password)
elem = driver.find_element_by_id("login_submit")
elem.click()
elem = driver.find_element_by_css_selector(".jp-Launcher-cwd+ .jp-Launcher-section .jp-LauncherCard")
elem.click()

This creates a new notebook, but now I'm stuck at the point of entering some code in a cell and running it. If I view the page source I don't see any html elements for the cells. But if I enter print("test") in a cell, then driver.page_source contains this (it's pretty nested in other stuff I've omitted too):

                                
                                    
                                        
                                    
                                        
                                    
                                    
                                        
                                    
                                    
                                    
                                    
                                        
                                            
                                                
                                                    
                                                        
                                                            xxxxxxxxxx
                                                        
                                                        
                                                            print("test")

I can see where the text for print("text") is (i.e. the deepest nested elements in the above html snippet), but I can't figure out which element here I would be able to send text to or send keys to.

I came across robotframework-jupyterlibrary and it has some clues such as this and this . From those links I see

${JLAB CSS ACTIVE INPUT} ${JLAB CSS ACTIVE CELL} .CodeMirror

and

Add and Run JupyterLab Code Cell
    [Arguments]    ${code}=print("hello world")
    [Documentation]    Add a ``code`` cell to the currently active notebook and run it.
    Click Element    css:${JLAB CSS NB TOOLBAR} ${JLAB CSS ICON ADD}
    Sleep    0.1s
    ${cell} =   Get WebElement  css:${JLAB CSS ACTIVE INPUT}
    Click Element    ${cell}
    Set CodeMirror Value    ${JLAB CSS ACTIVE INPUT}  ${code}
    Run Current JupyterLab Code Cell
Click Element ${cell}

which makes me think if I select the .CodeMirror element, then I just need to figure out what Get WebElement does in that weird language and how to do it in selenium.

Any ideas?

I've also tried (based on https://stackoverflow.com/a/48723135/1011724 and https://stackoverflow.com/a/50279295/1011724):

from selenium.webdriver.common.action_chains import ActionChains

actions = action_chains.ActionChains(driver)
textarea = driver.find_elements_by_css_selector('.CodeMirror textarea')[0]  # tried for [0], [1] ,[2] and [3] which is all of them.
actions.move_to_element(textarea).click().send_keys("testing...").perform()

but I keep getting the error

selenium.common.exceptions.WebDriverException: Message: TypeError: rect is undefined

user_9800042 · Answer

Code below tested with Chrome, Firefox and jupyterlab latest versions:

from selenium import webdriver
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support.wait import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.common.by import By

driver = webdriver.Chrome()
wait = WebDriverWait(driver, 10)
driver.get("http://localhost:8888")
token = "0107216930d05db8a7c36ad6a73573dd5349c3dd56fee852"

wait.until(EC.element_to_be_clickable((By.ID, "password_input"))).send_keys(token, Keys.ENTER)

# wait for "Python 3" Notebook menu or CodeMirror element if already launched.
wait.until(EC.presence_of_element_located(
    (By.CSS_SELECTOR, "[title='Python 3'][data-category='Notebook'], .jp-Notebook .CodeMirror")))
# if "Python 3" Notebook menu found click to open new Notebook
if len(driver.find_elements_by_css_selector("[title='Python 3'][data-category='Notebook']")) > 0:
    driver.find_element_by_css_selector("[title='Python 3'][data-category='Notebook']").click()

# wait for CodeMirror and click to focus
code_mirror = wait.until(EC.presence_of_element_located((By.CSS_SELECTOR, ".jp-Notebook .CodeMirror")))
code_mirror.click()
code_mirror.find_element_by_tag_name("textarea").send_keys("print('Hello World!')")
driver.find_element_by_css_selector("[data-icon='run']").click()

output = wait.until(EC.visibility_of_element_located((By.CSS_SELECTOR, ".jp-OutputArea-output")))
print(output.text)
assert output.text.strip() == "Hello World!"

driver.quit()

user_10413749 · Answer

To Open Jupyter-Notebook:

Either Open a command window and navigate to the repository folder or Open anaconda-command-prompt window and simply execute

jupyter notebook --NotebookApp.token='' --NotebookApp.password=''

After loading notebook using driver.get("http://localhost:8888"), Here is the most tricky part, that how to select a Dynamically changing object. Visit Reference: Dynamically Changing IDs.

By using find_element_by_xpath

By using find_element_by_css_selector

These both will give you the same selection-point but it would be better to use Xpath which is more convenient technique. You can move forward like this;

import time
from selenium import webdriver
from selenium.webdriver.support.wait import WebDriverWait

driver = webdriver.Chrome('chromedriver.exe')
wait = WebDriverWait(driver, 20)

driver.maximize_window()
website_url = "http://localhost:8888/"
driver.get(website_url)

# Using Xpath
# I prefer using xpath, because it is simple to understand
# and if you want to dynamically enter data into fields, it would be an awesome approach
driver.find_element_by_xpath("//div[@id='new-buttons']").click()

if len(driver.find_elements_by_xpath("//div[@id='new-buttons']//li[@id='kernel-python3']")) > 0:

   time.sleep(3)
   driver.find_element_by_xpath("//div[@id='new-buttons']//li[@id='kernel-python3']").click()

driver.find_element_by_xpath('//div[@class="cell code_cell rendered selected"]').click()

# Using css_selector
#driver.find_element_by_css_selector("#notebook-container > div").click()

command = 'print("Hello World!")'

#a = driver.find_element_by_css_selector("#notebook-container > div > div.input > div.inner_cell >"
#                                         "div.input_area > div > div:nth-child(1) > textarea")

time.sleep(3)
# To select Note-Book text-area and place command in it.
a = driver.find_element_by_xpath('//div[@class="input_area"]//textarea').click().send_keys(command)

# To run the Code in Selected Cell
time.sleep(3)
driver.find_element_by_xpath("//button[@title='Run']").click()

print("Test is done.")

Use selenium to write in and run a code cell in jupyterlab

2 Answers