Why does $http outside of $apply actually work?

[cburgdorf]

The documentation says that it should not work outside of $apply phase but yet it seems to work fine from outside of an $apply phase as proven by this plunkr:

http://plnkr.co/edit/2EXpWkYQuxCgPZ5x9JOu?p=preview

[0x-r4bbit]

@cburgdorf because it runs an $apply for you internally. Can you point to the place in the docs?

[cburgdorf]

It's not supposed to work like this. Can you point me to the code where it internally calls $apply? Here is how the docs explain that it should not work outside from a $digest.

https://github.com/angular/angular.js/blob/master/src/ng/http.js#L226

And I can confirm that it actually needed to be called from within $apply somewhere before 1.2. So I actually think it has something to do with the new $postDigestQueue (not sure about the correct name).

[caitp]

I believe they're referring to

The $http service will not actually send the request until the next $digest() is executed. Normally this is not an issue, since almost all the time your call to $http will be from within a $apply() block. If you are calling $http from outside Angular, then you should wrap it in a call to $apply to cause a $digest to occur and also to handle errors in the block correctly.

https://github.com/angular/angular.js/blob/master/src/ng/http.js#L227

[revolunet]

The $apply call takes place here, just after the request is finished : https://github.com/angular/angular.js/blob/master/src/ng/http.js#L991

But this is triggered after the $http call. the HTTP call itself wasnt triggered if called from outside Angular before 1.2.0.

When using 1.1.5 in your plunk, the HTTP call is not even triggered so yes, something changed lately and i didnt find the revelant commit or explanation yet. good question :)

[cburgdorf]

@revolunet exactly. I also couldn't find a commit which should have changed it. Also the documentation says it shouldn't be that way. So I assume it has something to do with the $postDigestQueue which was introduced somewhere in between.

[revolunet]

my friend git bisect just told me that it comes from this commit : 6b91aa0

[cburgdorf]

This makes it sound as if it's intended behavior in which we should remove the critical part from the documentation :)

[0x-r4bbit]

Well let's give @IgorMinar the chance to explain what's going on here :)

[btford]

This is intended behavior. 6b91aa0 was introduced to fix an issue with $http response interceptors.

+1 for updating the docs (any volunteers?).

[0x-r4bbit]

Thanks for clarifying @btford

[cburgdorf]

Done #5485