Prior to this change component styles generated on the server where removed prior to the client side component being rendered and attached it's own styles. In some cases this caused flickering. To mitigate this `initialNavigation: enabledBlocking'` was introduced which allowed the remove of server styles to be defer to a latter stage when the application has finished initialization.
This commit changes the need for this, by not removing the server generated component styles and reuse them for client side rendering.
PR Close#48253
Prior to this change the component IDs where generated based on a counter. This is problematic as there is no guarantee that a component will get the same ID that was assigned on the server when generated on the client side.
This is paramount to be able to re-use the component styles generated on the server.
PR Close#48253
Before this change `ɵɵdefineDirective` called `ɵɵdefineComponent` under the hood. This is problematic for the consistent component id generation as it could result in hash collisions for certain directives. Directives however do not require an id.
This changes moves common definition generation logic into a separate function that is re-used in `ɵɵdefineDirective` and `ɵɵdefineComponent`.
PR Close#49350
The decorator downlevel transform is never used for actual class
decorators because Angular class decorators rely on immediate execution
for JIT. Initially we also supported downleveling of class decorators
for View Engine library output, but libraries are shipped using partial
compilation output and are not using this transform anymore.
The transform is exclusively used for JIT processing, commonly for
test files to help ease temporal dead-zone/forward-ref issues. We can
remove the class decorator downlevel logic to remove technical debt.
PR Close#49351
The private util `isObservable` was actually just testing the same thing as`isSubscribable()`. As the implementation is closer to the function's name, let's only keep ``isSubscribable`.
PR Close#49295
This adds the ngSkipHydration annotation, which allows users to
opt hydration boundaries out of hydration. This enables incremental
adoption of hydration by letting users skip hydration on components
that have implementation issues that conflict with hydration.
co-authored-by: AndrewKushnir <akushnir@google.com>
PR Close#49345
This commit incrementally builds on top of #49285 and adds the logic to hydrate <ng-container>s and their contents. This implementation supports simple <ng-container>s that don't have any Angular features (like *ngIf/*ngFor, etc) and are not content-projected.
The subsequent commits will extend the logic further to support more complex scenarios.
PR Close#49303
This commit incrementally builds on top of https://github.com/angular/angular/pull/49271 and adds the logic to hydrate elements and text nodes that don't have any Angular features (like *ngIf/*ngFor, etc) and are not content-projected.
The subsequent commits will extend the logic further to support more complex scenarios.
Co-authored-by: Jessica Janiuk <jessicajaniuk@google.com>
Co-authored-by: Andrew Kushnir <akushnir@google.com>
PR Close#49285
Close#45698
KeyEventsPlugin now has different behavior with EventsPlugin, it will
always run inside ngZone with ngZone.runGuarded() no matter the
component is initialized inside or outside of NgZone, this PR
make sure KeyEventsPlugin bahave the same with other events.
PR Close#49330
Previously (at the early days of Ivy) a TNode used to keep an array of TViews, but the logic was changed since that time, but the `tViews` field remained on TNode interface (+ corresponding typings).
This commit renames TNode.tViews to TNode.tView and cleans up typings.
PR Close#49313
**Important note**: this is a first commit in a series of commits that will be needed
to support non-destructive hydration. Stay tuned for further updates!
This commit lays the foundation on top of which more hydration logic will be
added in follow up PRs. This PR includes:
* Initial serialization of hydration data
* Data transfer of hydration annotations from server side to client
* Accessing hydration info and populating internal data structures
* Initial APIs (currently private) that enable hydration (in a tree-shakable manner)
* Cleanup of annotations post hydration
* Initial test infrastructure and basic test cases
This commit does **not** expose any public APIs. They'll be exposed later, when
more hydration logic is implemented to a state when it can cover most common
use-cases.
Co-authored-by: Jessica Janiuk <jessicajaniuk@google.com>
Co-authored-by: Andrew Kushnir <akushnir@google.com>
PR Close#49271
We recently migrated the testing example to use the router testing
harness. There was one instance of the previous non-`TestBed` examples
left that still relies on a stub that has already been removed.
This example is not used anywhere and we should rather encourage
a single pattern of testing. i.e. the harness as per recent changes.
This commit removes the broken file.
PR Close#49293
Currently, examples with test commands like `ng build` are *never*
using the local version of `//packages/compiler-cli`. This is because
the CLI is invoked accidentally from within `external/aio_example_deps`.
Since the CLI relies on importing the compiler-cli, it will always
resolve the dependency from that directory- causing it to be always
the version installed via `aio/examples/tools/shared/package.json`.
We should never resolve symlinks and escape the e2e sandbox. That way
the compiler-cli would be resolved properly and could also become the
locally built one, depending on the test mode (i.e. npm or "local").
PR Close#49293
Chromium is launched via Karma from within the Bazel AIO example e2e
tests. This breaks depending on the platform and sandbox mechanism used.
We should never use Chromium's sandbox on top of Bazel's sandbox
invocation. The Angular CLI exposes a browser exactly for this use-case.
PR Close#49293
Whenever we run example tests using the local framework packages, the
e2e tests will have the local framework packages symlinked in the
`node_modules`.
This works well in general, but due to NodeJS by default resolving
symlinks to the target location, NodeJS will end up looking for
transitive dependencies in the `bazel-bin` instead of in the example
`node_modules` folder. This means that we end up incorrectly resolving
older versions of `@angular/core` that end up existing in the main
project dependencies. This causes errors like:
```
Error: ../../home/circleci/.cache/bazel/_bazel_circleci/9ce5c2144ecf75d11717c0aa41e45a8d/execroot/angular/bazel-out/k8-fastbuild/bin/packages/common/npm_package/http/testing/index.d.ts:12:21 - error TS2307: Cannot find module '@angular/common/http' or its corresponding type declarations.
12 import * as i1 from '@angular/common/http';
~~~~~~~~~~~~~~~~~~~~~~
Error: ../../home/circleci/.cache/bazel/_bazel_circleci/9ce5c2144ecf75d11717c0aa41e45a8d/execroot/angular/bazel-out/k8-fastbuild/bin/packages/common/npm_package/index.d.ts:1630:18 - error TS2707: Generic type 'ɵɵDirectiveDeclaration' requires between 6 and 8 type arguments.
1630 static ɵdir: i0.ɵɵDirectiveDeclaration<NgClass, "[ngClass]", never, { "klass": "class"; "ngClass": "ngClass"; }, {}, never, never, true, never>;
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
We can fix this by properly ensuring that NodeJS does not resolve
symlinks, but rather preserves them.
In the error above, the e2e tests end up accidentally resolving
`@angular/core` v14 that comes from `@angular/benchpress`. Angular
Benchpress is installed via `@angular/build-tooling` in the project
root.
PR Close#49293
This causes the e2e tests to fail. The tests were compiled using a
tsconfig with `lib=dom`, but the tests explicitly tried to pull in the
Node types. This causes a conflict for e.g. `AbortController` types.
PR Close#49293
The `run-example-e2e` script does not properly fail if configured
tests of examples are failing. This happens when a CLI example
configures multiple tests in the `example-config`. Due to incorrect
usage of promises in combination with reduce, only the last test
command had an effect on the overall test conclusion.
A similar issue seems to occur with SystemJS Protractor tests.
This commit fixes the problem and also cleans up the code a little
by switching it to `async/await`.
PR Close#49293
