Question: use of Syn instead of native selenium calls

[robsquires]

I'm using this driver to test an large angularJS application. A fair amount of our UI behaviour is delivered using CSS as well as javascript...:hover being an example.

I've been experiencing issues when elements aren't available for clicking because they are not displayed. I presume this is because javascript events are being used (via syn) instead of the native mouse movements available in selenium.

I imagine you've a very good reason for doing this (i've seen https://github.com/Behat/MinkSelenium2Driver/issues/51) relating to compatibility across browsers? Does the selenium server not provide this abstraction very effectively?

Anyhow I've forked and replaced a few commands with native webdriver calls as I don't want our frontend devs to get too wrapped up in the internals of the webdriver api - they're pretty happy with the mink api.

I'll continue to use my fork but I'm wondering if you've plans to adopt more of the native api calls into the mink API?

[aik099]

Hi,

can you be more specific about stuff, that doesn't work for you? If we're looking at situation in general, then it's up to individual driver (e.g. Selenium, Sahi, etc.) to to use one or another approach in implementing specific user interaction with a browser.

If there is a proof, that Selenium native implementation of some functionality does call all the JS events properly (in all browsers of course), then we surely can skip using Syn there.

But I recommend to be careful with massive changes. Instead let's just focus on stuff that doesn't work for you and analyze alternative (you have PR ready already?), that would do what you need plus all tests will be still passing.

[robsquires]

I've created a project to test this specific case: https://github.com/robsquires/Selenium-Tests

If you run the feature suite you'll see a failing test + the screenshot generated in /screenshot confirms the CSS :hover state is not being triggered.

I'm using phantomjs in webdriver mode. I've not tested cross browser as I'm not sure what support you are going for - could you confirm?

The quick + dirty change I mentioned above can be seen here

[aik099]

I see, current mouseOver implementation does emulate the mouseover event, but doesn't really move mouse anywhere 😄

What if you, in current mouseOver method implementation, change code to do $this->wdSession->moveTo ...? If that would:

1. move the mouse to desired element
2. fire mouseover events along the way
3. cause :hover css selector to work

then you can safely send a PR for this case.

[robsquires]

Yup thanks, that's all cool. The code I posted uses moveTo + I'll happily submit a PR for that change.

In relation to this question - what was behind the implementation decision to use syn as opposed to the native selenium calls?

[aik099]

I wasn't there at decision (whatever to use Syn or not) making time, but looking at Selenium server development it looks like, that it had problems with firing native JavaScript events when being used.

Then things have changed (for Selenium server), but Mink driver wasn't updated to reflect that. The related #51 issue, that you mentioned, proves that. I'll watch closely for Syn to be replaced with native calls for cases, where Syn won't be doing it's job properly in emulating real user action.

[robsquires]

Ok thanks

[aik099]

@robsquires , why you closed this task? Shouldn't you be submitting PR instead and associating it with this task?

[robsquires]

yeah you're right...sorry :)

[robsquires]

Excuse this slightly nooby question..What's the best way for me to get my proposed changes into the codebase.

I added a web-fixture for this issue to the Mink package. Now I realise I'm not going to be able to update Mink through composer as the dependancy is ~1.5.0 :(

It doesn't look like any test cases have been added to Selenium2DriverTest for a while - what's the best approach to get these changes in?

thanks

[aik099]

It doesn't look like any test cases have been added to Selenium2DriverTest for a while - what's the best approach to get these changes in?

This is because all drivers should identically simulate user actions and all tests are placed in following folder: https://github.com/Behat/Mink/tree/develop/tests/Behat/Mink/Driver

I guess you need to add new test to JavascriptDriverTest.php. This test would cover that :hover effect testing. However you must also test that both Selenium1 and Sahi driver would pass that new test too.

[aik099]

Here is how I usually do this:

1. add all needed code to Mink, but don't commit yet
2. have working repo clone for each of drivers under test (selenium1/2, sahi)
3. copy updated mink code to vendor folder within drivers replacing 1.5.0 version of Mink, which there already
4. run tests for each driver

[robsquires]

thanks. I've taken a similar approach in my working repo. I was worried about travis really. Composer will install mink 1.5.0 during the build so my tests won't be there...any ideas?

In terms of adding a test to JavascriptDriverTest.php I don't think that's advisable. This issue actually relates to CSS and not javascript - i.e the driver's ability for behaviour implemented in CSS to be accurately tested. If this test was added, the suite would always fail for the zombie driver.

I can add a specific test case for the issue (378) to the selenium 1 and sahi drivers to ensure they work in the same way.