There are restrictions on the character set that can be used for xmb and xtb
placeholder names.
However because changing the placeholder names would change the message IDs it
is not possible to add those restrictions to the names used internally. Then we
have to map internal name to public names when generating an xmb file and back
when translating using an xtb file.
Note for implementors of `Serializer`:
- When writing a file, the implementor should take care of converting the
internal names to public names while visiting the message nodes - this is
required because the original nodes are needed to compute the message ID.
- When reading a file, the implementor does not need to take care of the mapping
back to internal names as this is handled in the `I18nToHtmlVisitor` used by the
`TranslationBundle`.
fixes b/34339636
The new view engine allows our codegen to produce less code,
as it can interpret view definitions during runtime.
The view engine is not feature complete yet, but already
allows to implement a tree benchmark based on it.
Part of #14013
- Introduce `InjectionToken<T>` which is a parameterized and type-safe
version of `OpaqueToken`.
DEPRECATION:
- `OpaqueToken` is now deprecated, use `InjectionToken<T>` instead.
- `Injector.get(token: any, notFoundValue?: any): any` is now deprecated
use the same method which is now overloaded as
`Injector.get<T>(token: Type<T>|InjectionToken<T>, notFoundValue?: T): T;`.
Migration
- Replace `OpaqueToken` with `InjectionToken<?>` and parameterize it.
- Migrate your code to only use `Type<?>` or `InjectionToken<?>` as
injection tokens. Using other tokens will not be supported in the
future.
BREAKING CHANGE:
- Because `injector.get()` is now parameterize it is possible that code
which used to work no longer type checks. Example would be if one
injects `Foo` but configures it as `{provide: Foo, useClass: MockFoo}`.
The injection instance will be that of `MockFoo` but the type will be
`Foo` instead of `any` as in the past. This means that it was possible
to call a method on `MockFoo` in the past which now will fail type
check. See this example:
```
class Foo {}
class MockFoo extends Foo {
setupMock();
}
var PROVIDERS = [
{provide: Foo, useClass: MockFoo}
];
...
function myTest(injector: Injector) {
var foo = injector.get(Foo);
// This line used to work since `foo` used to be `any` before this
// change, it will now be `Foo`, and `Foo` does not have `setUpMock()`.
// The fix is to downcast: `injector.get(Foo) as MockFoo`.
foo.setUpMock();
}
```
PR Close#13785
This commit effectively reverts 7e0f02f96e
as it was an invalid fix for #6385, that created a more significant
bug, which was that changes were not always being detected.
Angular 1 digests should be run inside the ngZone to ensure
that async changes are detected.
We don't know how to fix#6385 without breaking change detection
at this stage. That issue is triggered by async operations, such as
`setTimeout`, being triggered inside scope watcher functions.
One could argue that watcher functions should be pure and not do
work such as triggering async operations. It is possible that the
original use case could be supported by moving the debounce
logic into the watch listener function, which is only called if the
watched value actually changes.
Closes#10660, #12318, #12034
PR Close#13812
The change avoids the compiler CLI internal API from mismatching the following case as lazy loading
```
import { NonLazyLoadedModule } from './non-lazy-loaded/non-lazy-loaded.module';
export function getNonLazyLoadedModule() { return NonLazyLoadedModule; }
export const routes = [
{ path: '/some-path', loadChildren: getNonLazyLoadedModule }
];
```
The output of the check is later passed to `RouteDef.fromString()`, so, it makes sense to be only a string.
Fixesangular/angular-cli#3204
- remove outer `<div>` in tests,
- use `<ng-container>` instead of `<template>` where possible,
- use *... instead of template (tag or attr) where possible.
Fixes#13816
With this change the resolver is called when the parameter for the activated and any parent routes change.
ie, switching from `/teams/10/players/5` to `/teams/12/players/5` will now trigger any `PlayerResolver`.
CHANGES:
- Remove unused `onDestroy` method on the `KeyValueDiffer` and
`IterableDiffer`.
DEPRECATION:
- `CollectionChangeRecord` is renamed to `IterableChangeRecord`.
`CollectionChangeRecord` is aliased to `IterableChangeRecord` and is
marked as `@deprecated`. It will be removed in `v5.x.x`.
- Deprecate `DefaultIterableDiffer` as it is private class which
was erroneously exposed.
- Deprecate `KeyValueDiffers#factories` as it is private field which
was erroneously exposed.
- Deprecate `IterableDiffers#factories` as it is private field which
was erroneously exposed.
BREAKING CHANGE:
- `IterableChangeRecord` is now an interface and parameterized on `<V>`.
This should not be an issue unless your code does
`new IterableChangeRecord` which it should not have a reason to do.
- `KeyValueChangeRecord` is now an interface and parameterized on `<V>`.
This should not be an issue unless your code does
`new IterableChangeRecord` which it should not have a reason to do.
Original PR #12570Fixes#13382
This change adds Compiler CLI support for any level of nesting for lazy routes.
For example `{app-root}/lazy-loaded-module-1/lazy-loaded-module-2/lazy-loaded-module-3`
Where `lazy-loaded-module-3` is lazy loaded from `lazy-loaded-module-2`,
and `lazy-loaded-module-2` is lazy loaded from module `lazy-loaded-module-1`,
and `lazy-loaded-module-1` is lazy loaded from `AppModule`
Fixesangular/angular-cli#3663
This was done in order for us to be able to publish tsc-wrapped as @next tag on npm.
The next step is to change the build scripts to version and release @angular/tsc-wrapped
together with all the other packages. I'll create an issue/PR for this.
@angular/language-service now supports using TypeScript 2.1 as the
the TypeScript host. TypeScript 2.1 is now also partially supported
in `ngc` but is not recommended as Tsickle does not yet support 2.1.
Detailed changes:
- remove `UNINITIALIZED`, initialize change detection fields with `undefined`.
* we use `view.numberOfChecks === 0` now everywhere
as indicator whether we are in the first change detection cycle
(previously we used this only in a couple of places).
* we keep the initialization itself as change detection get slower without it.
- remove passing around `throwOnChange` in various generated calls,
and store it on the view as property instead.
- change generated code for bindings to DOM elements as follows:
Before:
```
var currVal_10 = self.context.bgColor;
if (jit_checkBinding15(self.throwOnChange,self._expr_10,currVal_10)) {
self.renderer.setElementStyle(self._el_0,'backgroundColor',((self.viewUtils.sanitizer.sanitize(jit_21,currVal_10) == null)? null: self.viewUtils.sanitizer.sanitize(jit_21,currVal_10).toString()));
self._expr_10 = currVal_10;
}
var currVal_11 = jit_inlineInterpolate16(1,' ',self.context.data.value,' ');
if (jit_checkBinding15(self.throwOnChange,self._expr_11,currVal_11)) {
self.renderer.setText(self._text_1,currVal_11);
self._expr_11 = currVal_11;
}
```,
After:
```
var currVal_10 = self.context.bgColor;
jit_checkRenderStyle14(self,self._el_0,'backgroundColor',null,self._expr_10,self._expr_10=currVal_10,false,jit_21);
var currVal_11 = jit_inlineInterpolate15(1,' ',self.context.data.value,' ');
jit_checkRenderText16(self,self._text_1,self._expr_11,self._expr_11=currVal_11,false);
```
Performance impact:
- None seen (checked against internal latency lab)
Part of #13651
Fixes:
- Inject the i18n specific HtmlParser into the directive normalizer,
- Parse ICU messages while normalizing templates,
- Normalize (visit) the content of ICU messages.
🎄🎁🎅
With the exception of `$onChanges()`, all lifecycle hooks in ng1 are called on
the controller, regardless if it is the binding destination or not (i.e.
regardless of the value of `bindToController`).
This change makes `upgrade` mimic that behavior when calling lifecycle hooks.
Additionally, calling the `$onInit()` hook has been moved before calling the
linking functions, which also mimics the ng1 behavior.
Note: This checks the constructors of `@Injectable` classes more strictly.
E.g this will fail now as the constructor argument has no `@Inject` nor is
the type of the argument a DI token.
```
@Injectable()
class MyService {
constructor(dep: string) {}
}
```
Last part of #12787Closes#12787
Now that rxjs is stable and the rxjs team follows semver, we can update and unpin the dependency safely.
From now on the Angular application/library developers are in charge of controlling the rxjs version as long as it's newer than 5.0.1.
closes#13561closes#13478closes#13572
* doc(compiler): fix the ICU expander API docs
* test(compiler): add lexer and parser specs
* fix(compiler): do not lex `}}` when interpolation is disabled
fix#13525
Fix an issue in `registerForNg1Tests`, where it passes a `null` as
`ng1Injector` to `_bootstrapDone`. This causes a "TypeError: Cannot
read property 'get' of null" to be thrown from `_bootstrapDone`.
This change retracts support for metadata version 2.
The collector used to produce version 2 metadata was incomplete
and can cause the AOT compiler to fail to resolve symbols or
produce other spurious errors.
All libraries compiled and published with 2.3.0 ngc will need
to be recompiled and updated with this change.
- New method `UpgradeAdapter.registerForNg1Tests(modules)` declares the
Angular 1 upgrade module and provides it to the `angular.mock.module()`
helper.
This prevents the need to bootstrap the entire hybrid for every test.
Closes#5462, #12675
- Full support for content projection in downgraded Angular 2
components. In particular, this enables multi-slot projection and
other features on <ng-content>.
- Correctly wire up hierarchical injectors for downgraded Angular 2
components: downgraded components inherit the injector of the first
other downgraded Angular 2 component they find up the DOM tree.
Closes#6629, #7727, #8729, #9643, #9649, #12675
feat(tsc-wrapped): recored when to quote a object literal key
Collecting quoted literals is off by default as it introduces
a breaking change in the .metadata.json file. A follow-up commit
will address this.
Fixes#13249Closes#13356
If you bound an array to select[multiple] via ngModel and subsequently
changed the options to select from, the UI would drop any selections
made since by the user. This was due to
SelectMultipleControlValueAccessor not keeping a reference to the new
model arrays it generated when users interacted with the select control.
Update code to keep the reference.
Closes#12527Closes#12654
When an invalid model value (eg empty string) was preset ngModel on
select[multiple] would throw an error, which is inconsistent with how it
works on other user input elements. Setting the model value to null or
undefined would also have no effect on what was already selected in the
UI. Fix this by clearing selected options when model set to null,
undefined or a type other than Array.
Closes#11926
The existing intl.ts file is not a facade but
rather a set of utils used by i18n-related pipes only.
As such moving it back to common module so those utils
are not used accidently from other places.
Victor Berchet
Pete Bacon Darwin
Dzmitry Shylovich
Matthew Hegarty
Tobias Bosch
Karl Seamon
Jonathan Adamski
Alex Eagle
Miško Hevery
Tim Consolazio
Dzmitry Shylovich
Martin Probst
Vikram Subramanian
Marc Laval
Matias Niemelä
José Nicodemos Maia Neto
Meligy
Pawel Kozlowski
Julien Elbaz
Joao Dias
Chuck Jazdzewski
Ryan Cavanaugh
shlomiassaf
Igor Minar
William KOZA
Jon Walsh
Emanuel Hein
Tsuyoshi Ito
Georgios Kalpakas
Filipe Silva
Chase
Victor Savkin
crisbeto
Brandon
Janne Vanhala
Trotyl Yu
Bowen Ni
Alex Rickabaugh
Eudes Petonnet-Vincent
Bradford C. Smith
gary-b
Hans