feat(ngMock) add atomic flushing strategy for $timeout

[elkeis]

Hi dear Angular community,

During writing unit tests I found one interesting issue. I need to write tests for some code like this:

$timeout(function() {
 action1();
 $timeout(function() {
  action2();
 });
});

And I'm expecting that I need to execute $timeout.flush() 2 times. First time to reach action1 function and the second time to execute action2 function (let's call this kind of flush "atomic" for example). But for now, for the first flush() it executes the both handlers. One after another. Please check this example: http://jsbin.com/fuziri/edit?js,output .

On the other hand we have some cases in tests when this buggy, for the first look, behaviour can simplify our lifes. In such case for example:

 var callback = jasmine.createSpy();
 $timeout(function() {}, 0, false).then(callback);
 $timeout.flush();
 expect(callback).toHaveBeenCalled();

Here $timeout register new defered task to process callback using $browser.defer during flushing. So to execute callback (if you are using that atomic flush strategy) you should call $timeout.flush() twise.

I think that the both strategies of flushing are pretty useful. And may be it will be a good idea to:

• implement that atomic flushing
• to allow to switch between the strategies using some additional flag

What do you think, guys?

P.S. this is a duplicate of #13146, I opened a new one because can't solve the certificate issue for commit with wrong email address.

[atogata]

that were*

[atogata]

+1, using on one of my projects, works great

[elkeis]

@atogata, thanks. Spelling corrected.

[petebacondarwin]

@elkeis - did you know that you can pass a delay to $timeout.flush(delay), which will only flush timeouts that should have completed by the given delay. This is not quite what you want, perhaps, since if you have two timeouts with 0 delay then they would both always be flushed at the same time.
But this can help with testing sequencing issues.

[petebacondarwin]

Oh I just looked at the code. I see that you want to ensure that new calls to timeout that are triggered as part of flushing the current timeouts are not triggered...

[petebacondarwin]

Interesting idea. I think this would need some more API thought as the current atomic parameter is not that intuitive to me.

[petebacondarwin]

This is very vaguely related to #5420. It would be interesting to solve both at once...

[petebacondarwin]

@elkeis once #14686 lands then as long as your timeouts have a non-zero delay, then they will indeed become separable in tests.

Given your original example, if I add 1ms second delays:

$timeout(function() {
 action1();
 $timeout(function() {
  action2();
 }, 1);
}, 1);

Calling $timeout.flush(1) will only trigger action1().
Calling $timeout.flush(1) a second time will trigger action2().

Is this enough for your testing?

[petebacondarwin]

Closing as #14686 has landed and there has been activity on this PR for 10 months.