feat(forms): implement setErrors

[vsavkin]

Example:

var login = new Control("someLogin");
c.setErrors({"notUnique": true});
expect(c.valid).toEqual(false);
expect(c.errors).toEqual({"notUnique": true});

c.updateValue("newLogin");
expect(c.valid).toEqual(true);

BREAKING CHANGE:

Before:

ControlGroup.errors and ControlArray.errors returned the reduced values of their children controls' errors.

After:

ControlGroup.errors and ControlArray.errors return the errors of the group and array.
And ControlGroup.controlsErrors and ControlArray.controlsErrors return the reduce values of their children controls' errors.

Notes:

• I had to separate errors and controlsErrors. This is to provide the following guarantee: A control group with invalid children is invalid.
• Calling setErrors on a control will update the validity of its ancestors.

[mhevery]

Can we add docs? It is easier to do now, then during a big docs push.

[mary-poppins]

Merging PR #4917 on behalf of @vsavkin to branch presubmit-vsavkin-pr-4917.

[TedSander]

I wonder if it makes sense to move group, and array from Validators? I'm worried that a user might try to pass them in the constructor, or pass them in as a combined validator. Should these functions be used as a regular validator at this point?

[TedSander]

LGTM I like it turned out much cleaner than the original design.

[Krustie101]

I notice the following things in the source code:

• The group/array validator only take into account the errors, not the controlErrors of the abstract controls in the group/array. Since the group/array validator are used to compute the controlErrors of the group/array, I expect the validy status of the group to be incorrect in the following case:
  control A in controlGroup B in controlGroup C
  Assume control A is invalid (either by internal validation or manual validation) and that we do not set the errors object of either B or C.
  The controlGroup B is invalid because the controlErrors object of B contains the error of A, but controlGroup C is valid because the error of A is not "passed through" to C.
• It is a good idea that setting the errors property of a control overrides the errors that arose from the validator? I presume the underlying reasoning is that the manual validation should not have been performed because the control contained a validator error so it does not make a difference.
  But what if you have a numeric input with a maxValue validator and there are some other controls in the form and depending on their values, a min value constrained kicks in. I might try to write a validator for this or I might conclude that a manual validation (before submit?) is more appropriate in this case. Then any max value error (message) is lost if the min value constraint is not satisfied.
  Not overriding the errors arising from the validator in the control case (combined with clearing of manual errors if the value is updated) would also be more consistent with the group/array behaviour.
  Even simpler, a developer can clear all the errors (validator & manual) of a control by calling setError(null). This seems strange.

[angular-automatic-lock-bot]

This issue has been automatically locked due to inactivity.
Please file a new issue if you are encountering a similar or related problem.

Read more about our automatic conversation locking policy.

_{This action has been performed automatically by a bot.}