Feature Suggestion: Support multiple browser instances from single test

[bobthekingofegypt]

Currently I am working on some socket based multi-player games using angular. In order to perform an e2e test I need to have two browsers talk to each other as the games wont start till 2 people join.

I was wondering if this could be something that is added on top of protractor.
I kinda hacked up a proof of concept to allow me to test users joining and chatting on the game.
bobthekingofegypt@8ac849a

As a concept it appears to work ok, it allows you to write tests like

    var browserOne = protractors[0].browser;
    var browserTwo = protractors[1].browser;

    var elementTwo = protractors[1].element;

    var byOne = protractors[0].by;
    var byTwo = protractors[1].by;

    browserOne.get('http://127.0.0.1:4999/');
    browserTwo.get('http://127.0.0.1:4999/');

    var joinButton = browserOne.element(byOne.css('[value="Join"]'));
    joinButton.click();

    element(by.model('input.name')).sendKeys('Bob');
    browserOne.element(byOne.css('[value="OK"]')).click();

    var playerList = element.all(by.repeater('player in players'));
    expect(playerList.count()).toBe(1);

    var playerListOther = elementTwo.all(byTwo.repeater('player in players'));
    expect(playerListOther.count()).toBe(1);

This change breaks certain things like saucelabs support and it's not very complete but it is meant as a conversation starter.

It is quite a niche use case but is this something you would consider supporting in protractor?

[juliemr]

I'm so happy you brought this up, because a while ago we'd had a discussion and decided that no one would ever actually use that functionality!

So, you could do this now by basically rewriting runner.js (which is what you've started). I don't think there's any way to generalize your use case enough that it would be worth it to make it reusable, so I'd encourage you to just go ahead and write your own runner.js.

If anyone else would use this though, I'd love for them to chime in here!

[tennisgent]

We would definitely use this. We are writing a web app that allows insurance policy holders to communicate with construction contractors who are supposed to do repairs on the policy holders' homes.

Currently, we have to write e2e tests for each side of the app separately. We have to write one test that signs in as a policy holder, does some setup and then sends off a communication to the contractor. And then we have to log out as a policy holder, log in as a contractor, and then make sure that the communication arrived in the contractor's inbox as expected.

However, if the functionality described above were possible, then we could run two parallel versions of the browser at the same time where one is signed in as a policy holder and the other is signed in as a contractor and we could instantly verify that the communications can go back and forth as expected.

That would be a huge benefit to us. I can imagine there are many other such websites out there that could use this type of functionality.

[christothes]

I'd like to use this as well. I'd like to test a game that needs multiple instances to exercise all scenarios

[hanssgo]

Would definitely use this as well! Been logging out and logging in as different users as well, with this support we would no longer have to do that =)

[christothes]

@juliemr Curious why this couldn't be generalized?

I was thinking it might make more sense if the concept of a "primary driver" remained, with a config option to specify n number of "helper drivers" that could be kept in an array similar to the example from @bobthekingofegypt. Would this approach make it any less intrusive to the mainline use case while allowing the multi-driver flexibility?

Granted I don't understand how saucelab compatibility comes into play here. However, if two two concepts simply aren't compatible, that could be accounted for in the config logic.

I wouldn't mind attempting a PR.

[elgalu]

I need this to test instant messaging and notifications between two logged in users.

My suggestion is similar to @bobthekingofegypt but using blocks instead of elementOne, elementTwo

And having browser1, browser2, ... kind of globals:

By default browser would point to browser1 and all expectations would run by default on browser1 unless browser2 is specified with some DSL like browser2.run(..{...}...); making the feature backward compatible.

Sample usage of browser2

it('works with 2 logged in users', function() {
  browser1.get('page1');
  browser2.get('page2');
  expect($('.div').getText()).toContain('hi'); // always defaults to browser1
  browser2.run(function() {
    $('button#send').click();
    expect($('.div').getText()).toContain('hello you');
  });
});

[adamduffy]

+1 need for this.

[hanssgo]

+1. Would be able to use this as well. Saves having to log out and log in
as different users.
On Mar 12, 2014 10:22 AM, "adamduffy" notifications@github.com wrote:

+1 need for thtis.

Reply to this email directly or view it on GitHubhttps://github.com//issues/381#issuecomment-37437150
.

[caitp]

/sub

[elgalu]

Will leave a reminder here that this feature won't work on IE:
https://code.google.com/p/selenium/wiki/FrequentlyAskedQuestions#Q:_Can_I_run_multiple_instances_of_the_WebDriver_sub-classes?

Not sure about Safari, probably won't support this either.

[bobthekingofegypt]

It was a while ago that I did this, and I haven't been working with Angular/protractor since so it is a bit hazy.

But looking at my diff I don't think I did anything beyond what was required to make a proof of concept work. All the promise work is because you need to block till all sessions are ready, and then make sure all sessions terminate at the end of the test. I also left in some code so single browser tests would work as normal. The fork does work, I used it for two projects; it does have flaws though.

I haven't tracked the changes in protractor but if what you want to do is simple you can try checking out my fork and npm link it, I changed the name from protractor to protractor-multi so I could run them both at the same time.

[toedter]

+1 for this. I would like to test asynchronous pub/sub with 2 browsers.

[Mikrovolnovka]

You can generalize this with implementing the ability of parallelization tests with scaling that is not limited to 2.

Sometimes we implement tests that are not affecting each other and can be run in parallel.
We would like to have ability of specifying different sets of specs.
I guess this should be possible both with Selenium Standalone server and direct usage of drivers.
I tried to run 2 different instances of protractor on single local machine (but in different bash windows, hehe) at a time using one instance of Selenium Standalone server and they worked in parallel well.

So, the POC is to implement running parallel tests within single protractor instance (with single output and ability to collect results to single report). Maybe there can be way to use something like Async.JS to run different specs in threads.

Maybe there is a way to do that with some Grunt package, but it is not convenient to have many Protractor config files, we would like to have all configs in one file.

Thanks in advance for your reply.

[orcaman]

+1 for this (writing a websockets app that could benefit from this)