feat(ngMock.$componentController): add helper to instantiate controllers for components

[petebacondarwin]

Closes #13683

[gkalpak]

I'd add a test with controller: 'TestController as test'

[petebacondarwin]

If you like but the implementation is just proxying to $controller anyway so it doesn't test any new paths.

[Narretz]

Wouldn't it be nice to have that for plain directives, too?

[gkalpak]

There will be some slight inconsistency between using controller: 'SomeCtrl as identifier' and controller: 'SomeCtrl' or controller: SomeCtrl. In the former case, the controller will be added as an alias on the scope, whereas in other two cases, it won't (when usign $componentController, I mean). Likewise, failing to provide a $scope in locals will throw an error in the former case (because it will try to assign the controller instance on the scope), but not in the other two cases.

[petebacondarwin]

@Narretz - it is much more complicated to do it for plain directives because we don't actually call the directive factory function until the directive is matched in some HTML. This means that the controller is not easily available. Components are simpler because the controller is not wrapped in a factory function.

[petebacondarwin]

There will be some slight inconsistency between using controller: 'SomeCtrl as identifier' and controller: 'SomeCtrl' or controller: SomeCtrl. In the former case, the controller will be added as an alias on the scope, whereas in other two cases, it won't (when usign $componentController, I mean). Likewise, failing to provide a $scope in locals will throw an error in the former case (because it will try to assign the controller instance on the scope), but not in the other two cases.

@gkalpak this is also true of ngMock.$controller right now.

[gkalpak]

@petebacondarwin, true, but things are more explicit there. You are are the one passing the argument to $controller() so the outcome should be less surprizing. In the case of $componentController() this happens indirectly; I would expect the same outcome when instantiating a component controller.

I wouldn't go to great lengths trying to "fix" this. If there's no easy fix (e.g. extracting the controller name and ignoring the alias or something), I would at least document that when using controller: 'SomeCtrl as identifier', you need to supply a $scope in locals.

No biggie anyway - it only affects test code. Feel free to leave as is, if you think it's OK.
LGTM otherwise 👍

[petebacondarwin]

@gkalpak - OK, I see what you mean. But since this is for people testing controllers that (hopefully) they have actually written, then it shouldn't be too surprising!

[petebacondarwin]

Actually in light of the latest discussions about controller as we should probably provide the ident param, if it is not explicitly provided, to be in line with the values of the component definition. I.E. if none is given the controller is attached to the scope via $scope.$ctrl.

[gkalpak]

Now it LRGTM ('R' for "really") 😃