-
-
Save adrienpoly/862846f5882796fdeb4fc85b260b3c5a to your computer and use it in GitHub Desktop.
<!DOCTYPE html> | |
<html> | |
<head> | |
... | |
</head> | |
<body data-controller="js"> | |
<%= yield %> | |
</body> | |
</html> |
import { Controller } from 'stimulus' | |
export default class extends Controller { | |
static values = { loaded: Boolean } | |
connect() { | |
this.loadedValue = true | |
} | |
} |
# ... | |
scenario 'test some js behavior', js: true do | |
visit some_path | |
ensure_js_is_ready | |
# JS is fully loaded you can start interacting with the page | |
end | |
# ... |
def ensure_js_is_ready | |
expect(page).to have_css('[data-js-loaded-value="true"]') | |
end |
This works but it doesn't really scale up well when your page starts having more and more stimulus controllers attaching to it. It starts to get really messy when all controllers need this added to them. I really wish there was a way to handle this for all controllers even if it's only in test mode.
There is only one controller needed in my example. Once a controller is connected it means all controllers on your page are connected.
There is only one controller needed in my example. Once a controller is connected it means all controllers on your page are connected.
Part of the issue I have is I’m using turbo streams to dynamically load additional parts of the page which have stimulus controllers making it all more complicated. In my case one controller being connected doesn’t necessarily mean all of them are.
Lazy loaded frames have a complete
attribute when. they are loaded
Turbo Stream channel have a connected
attribute once they are ready to receive messages
given that the Stimulus application is already started on the page when new content is added it gets connected almost immediately. It never created a flaky on my side.
Can you give a more detailed example that provides unreliable test?
Sure, I have a flaky test at the moment that uses a toggle
controller to show or hide a dropdown element. The controller is extremely basic:
toggle_controller.js
import {Controller} from "@hotwired/stimulus";
export default class extends Controller {
static targets = ["content"]
static classes = ["visibility"]
toggle(event) {
event.preventDefault();
this.contentTargets.forEach(t => t.classList.toggle(this.visibilityClass));
}
}
On the page I have the following:
div data-controller="toggle" data-toggle-visibility-class="hidden"
button#dropdown-menu data-action="toggle#toggle"
.w-6 = "Dropdown"
.relative
.hidden data-toggle-target="content"
= link_to some_path, "Some Action"
...
I then have a test the does the following:
visit "/"
ensure_js_is_ready
click_on "dropdown-menu"
click on "Some Action"
I get the following failure sometimes:
Failure/Error: click_on "Some Action"
Capybara::ElementNotFound:
Unable to find visible link or button "Some Action"
This test intermittently fails for me since even though the JS controller is loaded, the toggle controller has not attached to the button yet. I can see from the test screenshot that the click has fired since the button is selected but the dropdown is missing. In order to fix this I have been having to add this.loadedValue = true
individually to all of my stimulus controllers, which isn't ideal but it is the only way I can get somewhat reliable tests.
Interesting I never had any issue with similar test
here is what you could try
connect() {
requestAnimationFrame(() => {
this.loadedValue = true
})
}
loaded value will be set in the next tick after connect that could help hopefully
@adrienpoly Thanks for the suggestion! I'll have to try that out and report back. For now I've just been adding this.loadedValue = true
on connect for all my controllers where this problem presents itself. It works but it really litters up all my stimulus controllers. It would really be nice if stimulus were just able to handle this instead of requiring workarounds like this.
I've been playing around with this as well. I noticed when i load all my controllers using import "controllers"
, they appear to load alphabetically when i put console.log
in the various connect()
function. So i renamed the js_controller
to zz_js_controller
and i'm going to try that out.
Given it's an intermittent issue only in my system tests, i can't confirm if it will fix anything. Thanks for this post though!
@andrewyoo Did it work?
@alan-pie Yes, it works very well. I have no intermittent issues after implementing it in my CI tests and identifying the problem areas (right after a visit).
Thanks for this idea!
We're lazy loading Stimulus controllers and this didn't fully solve our flaky tests. Since lazy loading controllers get loaded in the order in which they occur in the DOM, this solution seems to work best when referencing the controller at the very end of the document. Something like this:
<!DOCTYPE html>
<html>
<head>
...
</head>
<body>
<%= yield %>
</body>
<js-testing data-controller="js" />
</html>
I know this kind of self-closing is invalid HTML, but for some reason the tag needs to be self closing for the controller to be executed. Haven't had the time to dig into that.
This works but it doesn't really scale up well when your page starts having more and more stimulus controllers attaching to it. It starts to get really messy when all controllers need this added to them. I really wish there was a way to handle this for all controllers even if it's only in test mode.