[1.3-beta16] ngModel $parsers regression vs. v1.2

[mgcrea]

I'm having issues with the AngularStrap datepicker in the latest v1.3 beta.

Basically, I've found out that the ngModel $parsers won't be called if $setViewValue is called multiple times with the same value/reference. It was the case in the v1.2 branch.

This is be problematic when we are using objects that change values but keep the same reference (such as Dates).

Check the following plunkers (console.warn):

• v1.2: http://plnkr.co/edit/PuleeGHkLHj3RvYSobs5?p=preview
• v1.3: http://plnkr.co/edit/yRL4JR0aVY3kDGgwLXX2?p=preview

Might be related to this commit

[Narretz]

Yes, it looks like facd904#diff-c244afd8def7f268b16ee91a0341c4b2R1759 introduced the change. It works with 1.3.0-beta.9, and there were no other changes related to forms.

[mgcrea]

1f6a5a1 might provide a workaround with revalidate. Not sure this is good enough though as this is a breaking change.

[shahata]

yes, of course. since we are testing if (ctrl.$$lastCommittedViewValue === value) { return; } we will never commit a new value in case $$lastCommittedViewValue is an object that keeps being reused by later updates.

I can look into this, but I'm not sure that this is the only point in ngModelController that assumes that the viewValue is not a reference to some object that can update independently after $setViewValue is called. I might be wrong, but it means for example, that when this object gets updated independently, all the scope variables that are bound to it get updated as well. Which means that updates cannot be delayed by ngModelOptions. WDYT? Is this really a legitimate way to use this API?

[shahata]

The example provided by @mgcrea for example can be fixed easily by doing controller.$setViewValue(angular.copy(date)) instead of controller.$setViewValue(date)
http://plnkr.co/edit/u44HOYpWrrQQyEFDszV5?p=preview

[mgcrea]

@shahata, yup, did patch AngularStrap with a proper workaround. Thanks!

I do agree that the API usage was a bit far-fetched and I'm OK with the new behavior.

[shahata]

@Narretz WDYT? I believe this is an acceptable breaking change and can be closed...