When a template element matches an imported foreign component, the compiler
omits standard element instructions (`ɵɵelementStart`/`ɵɵelement`) and instead
generates a single `ɵɵforeignComponent` call. The call passes the exact foreign
import wrapper expression defined in `@Component.foreignImports` along with an
aggregated object literal containing all static attributes and property
bindings.
The instruction itself is currently a no-op.
We extract the identifier name from the `foreignImports` expression in
`ComponentDecoratorHandler` and use a `SelectorlessMatcher` to match element
tags against these names. If an element matches both a regular Angular
directive and a foreign component, a conflict error is thrown.
In addition, we implement strict template semantic validation for these matched
foreign components within `TemplateSemanticsChecker`. Elements matched as
foreign components only support static attributes and property bindings. Any
event bindings, template references, or non-property input bindings (e.g.
class, style, or attribute bindings) trigger a semantic error diagnostic.
Finally, we skip standard DOM schema checks for foreign components to prevent spurious
validation errors since foreign components are not defined in standard HTML schemas.
PR Close#68674
Ensures that namespaced <script> elements (such as :svg:script) are correctly classified as PreparsedElementType.SCRIPT by the template preparser and stripped during compilation to prevent potential XSS vulnerabilities. Consequently, obsolete security schema mappings and runtime sanitization checks for <script> attributes have been removed since these elements are never present in compiled template outputs.
Currently if a `@for` loop doesn't have a `track` expression we don't produce an AST for it at all which means no type checking and language service support for it.
These changes make it so we produce the AST anyways since it gives the user more tools to resolve the issue (e.g. autocompletion when writing the `track` expression).
This allows us to show the API docs when the jsdoc block is at the top of a overloaded function (and not on the implementation signature).
eg: `injectAsync`
Langauge service previously never compiled non-exported classes. This avoided issues
where test modules would cause diagnostic noise due to components appearing in
declarations of two modules, for example. This change updates the logic to ensure
non-exported, but standalone classes _are_ still compiled.
fixes#65515
PR Close#68454
* Test that `minDate`/`maxDate` binds to `min`/`max` on date and time inputs
* Test that `min`/`max` attribute can be set directly on date and time inputs
* Relax type checker to allow `min`/`max` bindings on date and time inputs
PR Close#68001
There was not a test demonstrating local compilation with the
'bootstrap' param on NgModule. This test adds one, among other NgModule
fields in one. These other fields are broadly covered already, but this
rolls them into one test exercising all fields.
This commit updates `@defer` logic related to incremental hydration to be tree-shakable.
If hydrate triggers are used in a `@defer` block, the compiler emits a single top-level call to `ɵɵenableIncrementalHydrationRuntime`, placed once per create block before the first `ɵɵdefer` that requires it.
As a result, the incremental hydration runtime is only included in the bundle when hydrate is explicitly used.
Removes the `checkTwoWayBoundEvents` flag since the code it generates is quite breaking and we never got the chance to enable it. Also it caused our tests to misrepresent how the compiler behaves for actual users.
These changes introduce the new `@Service` decorator which is a more ergonomic alternative to `@Injectable`. The reason we're adding a new decorator is that `@Injectable` has been around since the beginning of Angular and it has a lot of baggage that adds unnecessary overhead for users that generally want to define a singleton service, available in their entire app. The key differences between `@Service` and `@Injectable` are:
1. `@Service` is `providedIn: 'root'` by default. You can opt into providing the service yourself by setting `autoProvided: false` on it.
2. `@Service` doesn't allow constructor-based injection, only the `inject` function.
3. `@Service` doesn't support the complex type signature of `@Injectable` (`useClass`, `useValue` etc.). Instead it supports a single `factory` function.
Example:
```ts
import {Service} from '@angular/core';
import {HttpClient} from '@angular/common/http';
import {AuthService} from './auth';
@Service()
export class PostService {
private readonly httpClient = inject(HttpClient);
private readonly authService = inject(AuthService);
getUserPosts() {
return this.httpClient.get('/api/posts/' + this.authService.userId);
}
}
```
The commit updates the TCB for safe navigation expressions to allow for correct narrowing of nullables.
This will trigger the `nullishCoalescingNotNullable` and `optionalChainNotNullable` diagnostics on exisiting projects.
You might want to disable those 2 diagnotiscs in your `tsconfig` temporarily if you want to update your project without having to fix all the issues at once.
Narrowing can be disabled altogether with `strictSafeNavigationTyes: false`.
fixes#37619
BREAKING CHANGE: This change will trigger the `nullishCoalescingNotNullable` and `optionalChainNotNullable` diagnostics on exisiting projects.
You might want to disable those 2 diagnotiscs in your `tsconfig` temporarily.
Fixes an error that was heppning when a generic param has type parameters of its own. There were a few different issues going on:
1. In #67707 I had changed a bit how we pass the `genericContextBehavior` which ended up ignoring the `useContextGenericType` option from the environment.
2. All directives depend on themselves, but we were overridding the `genericContextBehavior` for the directive being processed.
3. The type translator wasn't handling type parameter declarations. Technically we shouldn't be able to hit a code path that has a type parameter, however it's also easy enough to handle so we might as well.
Relates to #67704.
Fixes that we were parsing bindings in the `host` object with a regex that didn't account for nested brackets which may come up with something like Tailwind.
Fixes#68039.
Implements the logic at the compiler level that will de-duplicate host directives and merge them together. It will also report if a conflict is detected during merging.
The default change detection strategy is now OnPush.
BREAKING CHANGE: Component with undefined `changeDetection` property are now `OnPush` by default. Specify `changeDetection: ChangeDetectionStrategy.Eager` to keep the previous behavior.
Add NG8023 extended diagnostic to report duplicate component selectors
during compilation.
This replaces the former NG0300 runtime error, ensuring the failure
occurs at build time instead of runtime.
Closes angular#48377
BREAKING CHANGE: Elements with multiple matching selectors will now throw at compile time.
`<input type="number">` often does not provide the desired user experience when editing numbers in
a form. MDN even [describes](https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Elements/input/number#using_number_inputs)
how text inputs should be used in many cases instead, via `<input type="text" inputmode="numeric">`
or similar configurations. Previously, this did not work with Signal Forms without a custom input
component/directive.
This PR builds support for binding `number|null` models directly to `<input type="text">` native
controls via `[formField]`. When a model has a number or `null` value, signal forms will preserve
that status when the user makes edits/changes. Empty string values are converted to `null`, other
values are parsed as numbers, and a parse error is raised when a non-numeric value is entered.
Note that it's up to the UI developer to configure additional UI affordances such as setting an
appropriate `inputmode`, rejecting non-numeric keypresses, etc.
Fixes#66903Fixes#66157
Fixes a regression caused by the recent TCB changes where we moved the type parameter processing earlier in the pipeline and stopped properly accounting for the `TcbGenericContextBehavior`.
Fixes#67704.
Implement support for `FormUiComponent`s in both Reactive and Template-driven
forms. This allows components that use the new signal-based form control
architecture to be used seamlessly within existing Angular form paradigms.
Key changes:
- Integrated `ɵngControlCreate` and `ɵngControlUpdate` lifecycle hooks into
`NgModel`, `FormControlDirective`, and `FormControlName`.
- Implemented branching logic to choose between the traditional `ControlValueAccessor` (CVA) path and the new FVC path based on the host element's capabilities.
- Added comprehensive unit tests for FVC integration in both Reactive (`reactive_fvc.spec.ts`) and Template-driven (`template_fvc.spec.ts`) forms, covering:
- Value synchronization (model -> view and view -> model).
- Status synchronization (touched, dirty, valid, invalid, pending, required).
- Error propagation and `parseErrors` support.
- Fallback behavior to native DOM properties (disabled, required) when FVC inputs are missing.
- Graceful fallback to CVA when no FVC pattern is detected.
- Refined `NgModel` to correctly handle `required` validation via its existing `RequiredValidator` directive while supporting FVC for other properties.
In some cases the `debugName` transform generates a spread into the signal function parameters. This can cause compiler errors, because the functions don't have rest parameters.
These changes work around it by adding a `@ts-ignore` above it.
The Angular compiler unconditionally adds a debug name transform for signals
which generates a conditional on `ngDevMode` (e.g., `ngDevMode ? { debugName: "xyz" } : []`).
During testing, `ngDevMode` is true, so the true branch executes but the
false branch is never executed. Consequently, coverage tools report the
false branch as an untested line/branch, preventing 100% test coverage.
This commit adds a synthetic `/* istanbul ignore next */` comment to the
generated false branch so that Istanbul ignores it. We only include the
istanbul comment (instead of additionally including c8) to focus on the
established standard for Angular CLI/Karma coverage while maintaining
compatibility with modern Vitest setups, since @vitest/coverage-v8 now
natively respects the fallback istanbul comment.
Fixes#64583
Allows specifying a timeout parameter for idle-based deferred triggers, enabling more granular control over when deferred actions are executed.
Closes angular#67187
inputs & outputs cannot be binded to 2 different directives/components properties
Eg
```
data = model();
dataChange = output(); // throws because model already emits on the `dataChange` output
userSomething = input({alias 'user'});
user = input(); // throws because userSomething already binds to the `user` input
````
fixes#65844
BREAKING CHANGE: The compiler will throw when there a when inputs, outputs or model are binding to the same input/outputs.
Wrap `@defer` trigger expressions (`when`, `prefetch when`, `hydrate when`)
in a conditional context within the TCB to enable TypeScript's TS2774
diagnostic for detecting functions used without invocation.
Previously, signals and functions passed to `when` triggers without
parentheses would silently evaluate to truthy, causing unexpected behavior.
Now the compiler reports an error when a function is used as a condition
without being called.
`FieldTree` was an unnecessarily specific type for the `[formField]`
input. It forced the directive to care about what _kind_ of `FieldTree`
was bound–specifically whether it was Reactive Forms compatible or not.
This made it difficult to author forms system-agnostic components with
passthrough `[formField]` inputs.
