This reverts commit d0bc83ca27.
Protractor-based prerendering is flakey on Travis and takes several minutes to
complete, slowing down the build. Prerendering has a lower impact now that we
use a ServiceWorker. We will revisit in the future (probably using a
`PlatformServer`-based approach).
PR Close#15346
`ngc` would look for flat module resources relative to the flat module index.
`ngc` now looks for flat module resources relative to the `.d.ts` file that declarates
the component.
Fixes#15221
PR Close#15367
Added an "origins" section to the flat module `.metadata.json` files
that records where the original symbols was declared. This allows
correctly calculating relative path references recorded in metadata.
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
This shouldn't change anything. But it's interesting that we used to have this import
that seemed bogus, but there were no compilation or rutime errors.
Content pages like `tutorial/index.md` were being mapped to `tutorial.index.json`,
which meant that they could only be rendered if you browsed to `/tutorial/index`.
This didn't sit well so now these pages are mapped to `tutorial.json`, which
means that you browser to them via `/tutorial/` or just `/tutorial`.
Fixed#15335
The navigation.json is now passed through the dgeni pipeline.
The source file has been moved to `aio/content/navigation.json`
but the generated file will now appear where the original source file
was found, `aio/src/content/navigation.json`.
Everything inside `aio/src/content` is now generated and ignored by git.
The `processNavigationMap` processor in this commit adds the current version
information to the navigation.json file and verifies the relative urls in
the file map to real documents.
The navigationService exposes the versionInfo as an observable, which the
AppComponent renders at the top of the sidenav.
The navigation.json is now passed through the dgeni pipeline.
The source file has been moved to `aio/content/navigation.json`
but the generated file will now appear where the original source file
was found, `aio/src/content/navigation.json`.
Everything inside `aio/src/content` is now generated and ignored by git.
The `processNavigationMap` processor in this commit adds the current version
information to the navigation.json file and verifies the relative urls in
the file map to real documents.
The navigationService exposes the versionInfo as an observable, which the
AppComponent renders at the top of the sidenav.
The original Rho is too strict when it comes to markdown headings.
It requires that there be a blank line separating the heading and the
next paragraph. The forked version here fixes that; but the Rho project
will not merge it as it goes against there basic rules.
Unlike in the browser, on the server there is no concept of a document origin.
Thus, it is illegal to make requests for relative URLs against Http on platform-server.
Currently this fails with a vague error:
Error: Uncaught (in promise): Error at resolvePromise
This change adds explicit validation and a friendlier error message:
Error: URLs requested via Http on the server must be absolute. URL: /testing
Another option considered was to track the concept of an origin for the platform
and automatically prepend it to relative URLs. This would cause automatic "local
RPCs" to be made, though, which would be an unexpected and undesirable default
behavior.
Fixes#15349
PR Close#15357
This change reduces the amount of generated code by only adding `log`
calls for elements and text nodes.
We need the `log` calls to allow users to jump to the right place
in the template via source maps. However, we only need it for element
and text nodes, but not for directives, queries, … as for them we
first locate the corresponding element or text node.
Related to #15239
PR Close#15350
This is needed to support the corner cases:
- usage of a `ComponentFactory` that was created on the fly via `Compiler`
- overwriting of the `NgModuleRef` that is associated to a
`ComponentFactory` by the `ComponentFactoryResolver` from
which it was read.
Fixes#15241
We extracted ids from i18n attributes but forgot to use them when merging the translations, resulting in an error about missing translations even when they were correctly defined.
Fixes#15234
PR Close#15302
If a directive has not bindings nor has a `ngDoCheck` / `ngOnInit`
lifecycle hook, don’t generate a `check` call.
This does not have an impact on the behavior, but produces
less code.
PR Close#15322
This reverts commit 8b5c6b2732.
This feature is not compatible with the `Injector.get` which now only
takes `Type` or `InjectableToken`. `Symbol` is not a valid type.
Closes#15183
PR Close#15319
The Router use the type `Params` for all of:
- position parameters,
- matrix parameters,
- query parameters.
`Params` is defined as follow `type Params = {[key: string]: any}`
Because parameters can either have single or multiple values, the type should
actually be `type Params = {[key: string]: string | string[]}`.
The client code often assumes that parameters have single values, as in the
following exemple:
```
class MyComponent {
sessionId: Observable<string>;
constructor(private route: ActivatedRoute) {}
ngOnInit() {
this.sessionId = this.route
.queryParams
.map(params => params['session_id'] || 'None');
}
}
```
The problem here is that `params['session_id']` could be `string` or `string[]`
but the error is not caught at build time because of the `any` type.
Fixing the type as describe above would break the build because `sessionId`
would becomes an `Observable<string | string[]>`.
However the client code knows if it expects a single or multiple values. By
using the new `ParamMap` interface the user code can decide when it needs a
single value (calling `ParamMap.get(): string`) or multiple values (calling
`ParamMap.getAll(): string[]`).
The above exemple should be rewritten as:
```
class MyComponent {
sessionId: Observable<string>;
constructor(private route: ActivatedRoute) {}
ngOnInit() {
this.sessionId = this.route
.queryParamMap
.map(paramMap => paramMap.get('session_id') || 'None');
}
}
```
Added APIs:
- `interface ParamMap`,
- `ActivatedRoute.paramMap: ParamMap`,
- `ActivatedRoute.queryParamMap: ParamMap`,
- `ActivatedRouteSnapshot.paramMap: ParamMap`,
- `ActivatedRouteSnapshot.queryParamMap: ParamMap`,
- `UrlSegment.parameterMap: ParamMap`