Currently when an instance of the `FormControlName` directive is destroyed, the Forms package invokes
the `cleanUpControl` to clear all directive-specific logic (such as validators, onChange handlers,
etc) from a bound control. The logic of the `cleanUpControl` function should revert all setup
performed by the `setUpControl` function. However the `cleanUpControl` is too aggressive and removes
all callbacks related to the onChange and disabled state handling. This is causing problems when
a form control is bound to multiple FormControlName` directives, causing other instances of that
directive to stop working correctly when the first one is destroyed.
This commit updates the cleanup logic to only remove callbacks added while setting up a control
for a given directive instance.
The fix is needed to allow adding `cleanUpControl` function to other places where cleanup is needed
(missing this function calls in some other places causes memory leak issues).
PR Close#39623
Prior to this commit, the `cleanUpControl` function (responsible for cleaning up control instance)
was not taking validators into account. As a result, these validators remain registered on a detached
form control instance, thus causing memory leaks. This commit updates the `cleanUpControl` function
logic to also run validators cleanup.
As a part of this change, the logic to setup and cleanup validators was refactored and moved to
separate functions (with completely opposite behavior), so that they can be reused in the future.
This commit doesn't add the `cleanUpControl` calls to all possible places, it just fixes the cases
where this function is being called, but doesn't fully perform a cleanup. The `cleanUpControl`
function calls will be added to other parts of code (to avoid more memory leaks) in a followup PR.
PR Close#39234
introduce a boolean to track form groups/arrays own pending async validation to distinguish between pending state due to children and pending state due to own validation
Fixes#10064
PR Close#22575
With these changes, the types are a little stricter now and also not
compatible with Protractor's jasmine-like syntax. So, we have to also
use `@types/jasminewd2` for e2e tests (but not for non-e2e tests).
I also had to "augment" `@types/jasminewd2`, because the latest
typings from [DefinitelyTyped][1] do not reflect the fact that the
`jasminewd2` version (v2.1.0) currently used by Protractor supports
passing a `done` callback to a spec.
[1]: 566e039485/types/jasminewd2/index.d.ts (L9-L15)Fixes#23952Closes#24733
PR Close#19904
All errors for existing fields have been detected and suppressed with a
`!` assertion.
Issue/24571 is tracking proper clean up of those instances.
One-line change required in ivy/compilation.ts, because it appears that
the new syntax causes tsickle emitted node to no longer track their
original sourceFiles.
PR Close#24572
Support for using the `ngModel` input property and `ngModelChange`
event with reactive form directives has been deprecated in
Angular v6 and will be removed in Angular v7.
Now deprecated:
```html
<input [formControl]="control" [(ngModel)]="value">
```
```ts
this.value = 'some value';
```
This has been deprecated for a few reasons. First, developers have
found this pattern confusing. It seems like the actual `ngModel`
directive is being used, but in fact it's an input/output property
named `ngModel` on the reactive form directive that simply approximates
(some of) its behavior. Specifically, it allows getting/setting the
value and intercepting value events. However, some of `ngModel`'s other
features - like delaying updates with`ngModelOptions` or exporting the
directive - simply don't work, which has understandably caused some
confusion.
In addition, this pattern mixes template-driven and reactive forms
strategies, which we generally don't recommend because it doesn't take
advantage of the full benefits of either strategy. Setting the value in
the template violates the template-agnostic principles behind reactive
forms, whereas adding a FormControl/FormGroup layer in the class removes
the convenience of defining forms in the template.
To update your code before v7, you'll want to decide whether to stick
with reactive form directives (and get/set values using reactive forms
patterns) or switch over to template-driven directives.
After (choice 1 - use reactive forms):
```html
<input [formControl]="control">
```
```ts
this.control.setValue('some value');
```
After (choice 2 - use template-driven forms):
```html
<input [(ngModel)]="value">
```
```ts
this.value = 'some value';
```
You can also choose to silence this warning by providing a config for
`ReactiveFormsModule` at import time:
```ts
imports: [
ReactiveFormsModule.withConfig({warnOnNgModelWithFormControl: 'never'});
]
```
Alternatively, you can choose to surface a separate warning for each
instance of this pattern with a config value of `"always"`. This may
help to track down where in the code the pattern is being used as the
code is being updated.
Note: `warnOnNgModelWithFormControl` is set up as deprecated so that it
can be removed in v7 when it is no longer needed. This will not display
properly in API docs yet because dgeni doesn't yet support deprecating
properties in object literals, but we have an open issue to resolve the
discrepancy here: https://github.com/angular/angular/issues/22640.
PR Close#22633
This commit introduces a new option to template-driven forms that
improves performance by delaying form control updates until the
"blur" or "submit" event. To use it, set the `updateOn` property
in `ngModelOptions`.
```html
<input ngModel [ngModelOptions]="{updateOn: blur}">
```
Like in AngularJS, setting `updateOn` to `blur` or `submit` will
delay the update of the value as well as the validation status.
Updating value and validity together keeps the system easy to reason
about, as the two will always be in sync. It's also worth noting
that the value/validation pipeline does still run when the form is
initialized (in order to support initial values).
Upcoming PRs will address:
* Support for setting group-level `updateOn` in template-driven forms
* Option for skipping initial validation run or more global error
display configuration
* Better support of reactive validation strategies
See more context in #18408, #18514, and the [design doc](https://docs.google.com/document/d/1dlJjRXYeuHRygryK0XoFrZNqW86jH4wobftCFyYa1PA/edit#heading=h.r6gn0i8f19wz).
This commit adds support for setting default `updateOn` values
in `FormGroups` and `FormArrays`. If you set `updateOn` to
’blur’` at the group level, all child controls will default to `’blur’`,
unless the child has explicitly specified a different `updateOn` value.
```
const c = new FormGroup({
one: new FormControl()
}, {updateOn: blur});
```
It's worth noting that parent groups will always update their value and
validity immediately upon value/validity updates from children. In other
words, if a group is set to update on blur and its children are individually
set to update on change, the group will still update on change with its
children; its default value will simply not be used.
By default, the value and validation status of a `FormControl` updates
whenever its value changes. If an application has heavy validation
requirements, updating on every text change can sometimes be too expensive.
This commit introduces a new option that improves performance by delaying
form control updates until the "blur" event. To use it, set the `updateOn`
option to `blur` when instantiating the `FormControl`.
```ts
// example without validators
const c = new FormControl(, { updateOn: blur });
// example with validators
const c= new FormControl(, {
validators: Validators.required,
updateOn: blur
});
```
Like in AngularJS, setting `updateOn` to `blur` will delay the update of
the value as well as the validation status. Updating value and validity
together keeps the system easy to reason about, as the two will always be
in sync. It's also worth noting that the value/validation pipeline does
still run when the form is initialized (in order to support initial values).
Closes#7113
This commit fixes a regression where `ngModel` no longer syncs
letter by letter on Android devices, and instead syncs at the
end of every word. This broke when we introduced buffering of
IME events so IMEs like Pinyin keyboards or Katakana keyboards
wouldn't display composition strings. Unfortunately, iOS devices
and Android devices have opposite event behavior. Whereas iOS
devices fire composition events for IME keyboards only, Android
fires composition events for Latin-language keyboards. For
this reason, languages like English don't work as expected on
Android if we always buffer. So to support both platforms,
composition string buffering will only be turned on by default
for non-Android devices.
However, we have also added a `COMPOSITION_BUFFER_MODE` token
to make this configurable by the application. In some cases, apps
might might still want to receive intermediate values. For example,
some inputs begin searching based on Latin letters before a
character selection is made.
As a provider, this is fairly flexible. If you want to turn
composition buffering off, simply provide the token at the top
level:
```ts
providers: [
{provide: COMPOSITION_BUFFER_MODE, useValue: false}
]
```
Or, if you want to change the mode based on locale or platform,
you can use a factory:
```ts
import {shouldUseBuffering} from 'my/lib';
....
providers: [
{provide: COMPOSITION_BUFFER_MODE, useFactory: shouldUseBuffering}
]
```
Closes#15079.
PR Close#15256
Observable subscriptions from previous validation runs should be canceled
before a new subscription is created for the next validation run.
Currently the subscription that sets the errors is canceled properly,
but the source observable created by the validator is not. While this
does not affect validation status or error setting, the source
observables will incorrectly continue through the pipeline until they
complete. This change ensures that the whole stream is canceled.
AsyncValidatorFn previously had an "any" return type, but now it more
explicitly requires a Promise or Observable return type. We don't
anticipate this causing problems given that any other return type
would have caused a runtime error already.