Some versions of rxjs cause the algorithm used in the Router to not recognize Route configs correctly.
This commit updates the algorithm to be compatible in the same way as other code locations internally.
Context:
https://github.com/angular/angular/commit/1160b8194f78141b40d1d4885ec182d033ff2659fixes#47089
Note: This does not have a test because I was unable to identify the
version of rxjs that would cause a failure here.
PR Close#47098
This brings the typing of ngStyle into parity with ngClass since commit
e2ab99b95e. Should help with some strict template checking edge cases.
PR Close#47069
Users using the "disabled" property binding on reactive form controls would want to know how to dynamically update the disabled state of a form control when they get a console warning.
PR Close#47041
We used to track all modules in a top-level constant called `autoRegisterModuleById` which was used by `getRegisteredNgModuleType`. As of #45024 the constant isn't being used anymore so we can remove it.
PR Close#47066
The current Router APIs require guards/resolvers to be present in the DI tree. This is because we want to treat all guards/resolvers equally and some may require dependencies. This requirement results in quite a lot of boilerplate for guards. Here are two examples:
```
const MY_GUARD = new InjectionToken<any>('my_guard');
…
providers: {provide: MY_GUARD, useValue: () => window.someGlobalState}
…
const route = {path: 'somePath', canActivate: [MY_GUARD]}
```
```
@Injectable({providedIn: 'root'})
export class MyGuardWithDependency {
constructor(private myDep: MyDependency) {}
canActivate() {
return myDep.canActivate();
}
}
…
const route = {path: 'somePath', canActivate: [MyGuardWithDependency]}
```
Notice that even when we want to write a simple guard that has no dependencies as in the first example, we still have to write either an InjectionToken or an Injectable class.
With this commit router guards and resolvers can be plain old functions.
For example:
```
const route = {path: 'somePath', component: EditCmp, canDeactivate: [(component: EditCmp) => !component.hasUnsavedChanges]}
```
Additionally, these functions can still use Angular DI with `inject` from `@angular/core`.
```
const route = {path: 'somePath', canActivate: [() => inject(MyDependency).canActivate()]}
```
PR Close#46684
The `RootContext` implementation contained a number of fields that were needed to support an experimental `renderComponent` function. The `renderComponent` function was removed, which allows us to cleanup the `RootContext` further.
The only field that remains on the `RootContext` is the list of bootstrapped components. This list is presumably mostly unused right now (it just contains the current component) and further refactoring can happen in a followup PR.
PR Close#46806
This commit updates the `renderModule` and `renderApplication` functions to also accept a document reference (in addition to the serialized document contents as a string). This should provide the necessary flexibility to NgUniversal (and other API consumers) to structure the logic to avoid serializing and parsing document multiple times during the SSR request.
PR Close#47032
Updates the tsickle version in the repository and accounts for its changes in
the `compiler-cli` package. Tsickle made a breaking change in the minor version
segment bump that would break the use with `@angular/compiler-cli`
Additionally the tsickle version for `@angular/bazel` is updated since
we updated `@bazel/typescript` to also account for the breaking changes.
See: https://github.com/bazelbuild/rules_nodejs/commit/78a05281075c50ee31658bc48dd852e35b973496
PR Close#47018
In Angular 14, we introduced the `loadComponent` API for a `Route` to
allow lazy loading of a routed component in addition to the existing
`loadChildren` which allows lazy loading of child routes. As a result,
the `preload` method of the `PreloadingStrategy` needs to sometimes be
called even when there is a `canLoad` guard on the `Route`. `CanLoad`
guards block loading of child routes but _do not_ block loading of the
component.
This change updates the conditional checks in the internal preloader to
skip calling the `PreloadingStrategy.preload` when there is only a
`loadChildren` callback with a `canLoad` guard an no `loadComponent`.
In this case, the callback passed to the `preload` method is already
effectively a no-op so it's not necessary to call it at all.
resolves#47003
PR Close#47007
The dev-infra build tooling is now decoupled from `ng-dev`. This will
make it easier to update `ng-dev` without necessarily needing to upgrade
the whole build system, Bazel etc. This is useful when e.g. new release
tool features have been added and should also be ported to active LTS
branches.
PR Close#46976
When using the Angular Router, one of `APP_BASE_HREF` or a `<base>` in
the header must be provided. When _not_ using the `RouterModule`,
injecting the `LocationStrategy` will result in the
`PathLocationStrategy` being provided with a default value used in place
of `APP_BASE_HREF` that is `document?.location?.origin ?? ''`.
It can be quite surprising and annoying that once you add `RouterModule`
to the application, suddenly the `APP_BASE_HREF` must be specifically
provide something new when it could use a sensible default instead.
The current behavior (before this commit) is as follows:
* When `RouterModule` is not provided (or the dev doesn't specifically provide
`PathLocationStrategy`): use `DOCUMENT.location?.origin ?? ''`.
Note that the base href in the dom and `APP_BASE_HREF` are not used.
* When `RouterModule` _is_ provided:
1. APP_BASE_HREF if defined
2. Get base href from DOM
3. throw if neither of the two above are defined
This commit updates this behavior to be aligned regardless of `RouterModule`
usage. The order (by default) is now:
1. Developer provided `APP_BASE_HREF`
2. base href from the DOM
3. `location.origin`
4. If none of the above exist, use `''`
This is slightly different than the behavior before. However, I believe
it is more appropriate. For the case without `RouterModule`, it would
likely be surprising that `APP_BASE_HREF` and the base href from the DOM
are ignored by default. For the case with `RouterModule`, we now have a
more sensible fallback/default when neither `APP_BASE_HREF` nor `<base>`
are defined (instead of just throwing an error).
PR Close#46929
With strict template type checking, a null/undefined value will raise an
error. However the implementation is completely fine with it, and it
would be pointless to "fix" it at the callsite and convert to e.g. an
empty string. Allow all of the "supported types" to be passed in
directly to ngClass.
Fixes#39280
PR Close#46906
This commit adds the `defaultUrlMatcher` from the Router to the public
API. `UrlMatcher` and `UrlMatchResult` are already in the public api so
the signature of the function as well as the return value are already
exposed. Any change to those or the implementation of `defaultUrlMatcher`
would already be breaking so there's no additional risk in exposing the
default matcher.
This function can be useful for developers who want to create a custom
matcher which builds on the default matcher of the Router. Currently,
the only way to do this would be to copy-paste the implementation.
fixes#35928
PR Close#46913
Pull request #46672 added some debugging code to trace down the
root cause of its bug, but parts of the debugging code has never
been cleaned up and ended up landing as part of the PR.
This commit removes the code as it might cause unexpected issues.
Likely when e.g. `Error` is patched and would perform XHRs in testing,
unveiling e.g. CORS issues. See #46989.
PR Close#46989
Since the `init` parameter has a default value of `{}`, it can never be
`!== undefined`. Thus, it is not necessary to account for that case.
PR Close#46912
Implement a new `notificationclick` action, `sendRequest`, which sends a
GET request to the specified URL, without opening a new window. This can
be useful for hitting an API endpoint.
PR Close#46912
This commit adds the `isEmpty` method to the `TransferState` class to make it possible to check whether the state is empty or not. This is helpful in situations when the `TransferState` should be serialized and the content is transferred to the client (if the state is empty - certain operations can be omitted).
PR Close#46915
This updates the internal use of INITIAL_NAVIGATION to do two things:
1. Explicitly provide `Enabled` as the default for the token factory
2. Use an enum instead of a string to reduce bundle size
PR Close#46925
This commit updates the code to include the TransferState providers (used for serialization) into the `ServerModule` instead of having the need to import the `ServerTransferStateModule` separately.
The list of providers in the `ServerTransferStateModule` is now empty and importing it is a noop. This is not a breaking change, since the `ServerModule` must be included anyways to make server rendering work correctly.
PR Close#46899
Similarly to what has been done in #46846 for the extended diagnostics about missing control flow directive that was only mentioning that the `CommonModule` should be imported, this commit improves the validation done by the JiT compiler.
Now that the control flow directives are available as standalone, the message mentions that directive itself can be imported.
The message now also mentions which import should be used for the directive (as it can be tricky to figure out that `NgForOf` is the directive corresponding to `*ngFor`).
PR Close#46903
This commit updates the `TransferState` to make it `providedIn: 'root'`. This makes the entire `BrowserTransferStateModule` module unnecessary, so it got deprecated as well.
The `ServerTransferStateModule` is still retained, but the `renderApplication` function now also includes the necessary tokens to serialize the `TransferState` automatically, so when using the `renderApplication` function, there is no need to include `ServerTransferStateModule` as well.
This change is a part of the ongoing efforts to update the shape of the FW APIs to make them standalone-friendly (so there is no need to import any NgModules).
PR Close#46879
improve the error message for non-standalone components which are not
exported from their module, and that are also imported directly as if
they were standalone
this change simply adds the suggestion to the developer to import the
ngModule instead
resolves#46004
PR Close#46114
This commit refactors the `NgModuleRef` implementation to drop functions required by the `EnvironmentInjector` interface. Previously the idea was that the `NgModuleRef` can act as an Injector to facilitate easier transition to standalone. However, from the mental model perspective, the `NgModuleRef` has the `injector` field, which is the correct injector reference and can be used is needed as an `EnvironmentInjector`.
PR Close#46896
The source map tests rely on asynchronous logic from the `source-map`
package. The tests itself are written using `fakeAsync` but this
unnecessarily complicates the interaction with the asynchronous
source-map helpers/package.
To fix this, we just make the tests use async/await as we don't intend
to test fakeAsync in this `describe` block..
PR Close#46888
The source-map package now requires the
`SourceMapConsumer`/`SourceMapGenerator` classes to be instantiated
asynchronously. This commit updates our tests to account for that.
PR Close#46888
The source-map package is no longer explicitly used in the
`compiler-cli` package and therefore can be removed from the ESBuild
bundling, denoting it as external. This should be a noop.
PR Close#46888
The implementation of the `DefaultTitleStrategy` was modeled after the
existing strategy patterns in the Router. These patterns were developed
before the `providedIn` syntax for injectables. We can simplify the
model a lot by providing the default in the factory of the abstract
class.
Note that the other strategy patterns aren't touched in this PR due to
how long they've existed. Because they have been there for such a long
time, it's possible there will need to be some adjustments to code
if/when they are refactored to do the same.
PR Close#46876
style values get added to the `_globalTimelineStyles` map in order to keep
them so that they can be used across different timelines
`_globalTimelineStyles` was previously a plain object but has been
refactored to a map in #44482, as part of the update a check has been
changed from a ternary operation to an or (||), causing falsy values (as 0)
not to be added to the map anymore, apply the nullish coalescing operator (??)
instead to make sure only `undefined` and `null` are filtered out
also since this aspect was clearly not covered by tests, add a new test
to ensure that such regression doesn't happen in the future
resolves#46833
PR Close#46863
Andrew Scott