The website I’m trying to test is full of custom html-elements and therefore makes extensive use of the shadow dom. This makes it difficult to use locators. So I came up with the following workaround (which I’m still not quite happy with).

In short, I setup a utils.feature where I define a JS-Function that returns the string of the actual function that I want.

@ignore
Feature:

  Scenario:
    # we compose a function using another function (the one above)
    * def setupFindAnywhere =
      """
      function setupFindAnywhere() {
          let findAnywhere = function(selector, domParent) {
            //code that traverses both DOM and ShadowDom and returns found element
          }
        }
        return findAnywhere + ''; //returns the function as a string
      }
      """
    * def setupText = call setupFindAnywhere # this is now a string of just the function
    * script('let findAnywhere = ' + setupText) # injects the function into the browser

And then in my actual test.feature:

Feature: does stuff
  Background:
    * ... stuff
    * call read('classpath:utils.feature') # sets up 'findAnywhere' in the browser

  Scenario: stuff
    * ... stuff
    # because the browser knows `findAnywhere` I can just call it.
    * script('findAnywhere("tbody > tr", document).click();')

This seems like a quite round-about way of doing this so I was wondering if there’s cleaner methods to interact with ShadowDom. Because otherwise I see myself quickly re-writing a lot of the pre-existing functions and that kind of defeats the purpose of a testing framework.