Prior to this change, any inserted `<style>` nodes into shadow dom trees would be retained
in memory, even after the shadow dom tree has been removed. This commit fixes the memory
leak by tracking the inserted `<style>` nodes per host element, such that removal of the
host element also releases the style nodes.
Fixes#36655
PR Close#42005
When there are multiple attributes that are marked for i18n translation,
which contain expression bindings, we were generating i18n update op-codes
that did not accurately map to the correct value to be bound in the lView.
Each attribute's bindings were relative to the start of the lView first
attributes values rather than to their own.
This fix passes the current binding index to `generateBindingUpdateOpCodes()`
when processing i18n attributes to account for this.
Fixes#41869
PR Close#41912
When including a component in a template, the component's host element
is immediately appended as child of the parent node upon creation.
Hence, `hostElement.parentNode` will be a valid reference. However, if
the parent component is being inserted as an embedded view—through
`ngIf` for example—then the parent's node itself will not have been
inserted yet. This means that we cannot properly determine the position
of the transition namespace, as any `containsElement` check will return
false given that the partial DOM tree has not been inserted yet, even
though it will be contained within an existing transition namespace once
the partial tree is attached.
This commit fixes the issue by not just looking at the existence of a
parent node, but rather a more extensive check using the driver's
`containsElement` method.
PR Close#19854
Fixes an error that will be thrown if `DebugRenderer2.destroyNode` is called with a node that has already been destroyed. The error happened, because we had a non-null assertion, even though the value can be null.
Note that this fix applies only to ViewEngine, because Ivy doesn't provide the `DebugRenderer2`. I decided to resolve it, because it fix is straightforward and this error has been showing up in our logs for a long time now, making actual errors harder to find.
PR Close#41565
* We had a usage of `Observable.subscribe` that uses the deprecated signature with 3 arguments. These changes switch to the non-deprecated version that passes in an `Observer`.
* Avoids always creating a `complete` callback since it isn't required.
* We were repeating all of the internal callbacks twice: once for sync and once for async. These changes move them out into variables so that they're more minifier-friendly. The savings aren't huge (~100 bytes minified), but it doesn't add any maintenance effort on our end so I decided to add it.
PR Close#41450
Introduces an **internal**, **experimental** `profiler` function, which
the runtime invokes around user code, including before and after:
- Running the template function of a component
- Executing a lifecycle hook
- Evaluating an output handler
The `profiler` function invokes a callback set with the global
`ng.ɵsetProfiler`. This API is **private** and **experimental** and
could be removed or changed at any time.
This implementation is cheap and available in production. It's cheap
because the `profiler` function is simple, which allows the JiT compiler
to inline it in the callsites. It also doesn't add up much to the
production bundle.
To listen for profiler events:
```ts
ng.ɵsetProfiler((event, ...args) => {
// monitor user code execution
});
```
PR Close#41421
Currently we normalize all CSS property names in the `StylingBuilder` which breaks custom properties, because they're case-sensitive. These changes add a check so that custom properties aren't normalized.
Fixes#41364.
PR Close#41380
Previously presence of both [class] and [className] bindings on an element was treated as compiler error (implemented in https://github.com/angular/angular/commit/6f203c9575fe9a2da04529fbdf1e5f7b51e85e09). Later, the situation was improved to actually allow both bindings to co-exist (see https://github.com/angular/angular/commit/a153b610983389de6408f7a723fd5714a11c168e), however the compiler check was not removed completely. The only situation where the error is thrown at this moment is when static (but with interpolation) and bound `class` attributes are present on an element, for ex.:
```
<div class="{{ one }}" [class]="'two'"></div>
```
In the current situation the error is acually misleading (as it refers to `[className]`).
This commit removes the mentioned compiler check as obsolete and makes the `class` and `style` attribute processing logically the same (the last occurrence is used to compute the value).
PR Close#41254
This commit makes the `RadioControlRegistry` class tree-shakable by adding the `providedIn` property to its
`@Injectable` decorator. Now if the radio buttons are not used in the app (thus no `RadioControlValueAccessor`
directive is initialized), the `RadioControlRegistry` should not be included into application's prod bundle.
PR Close#41126
Currently the `Validators` class contains a number of static methods that represent different validators as well as some helper methods. Since class methods are not tree-shakable, any reference to the `Validator` class retains all of its methods (even if you've used just one).
This commit refactors the code to extract the logic into standalone functions and use these functions in the code instead of referencing them via `Validators` class. That should make the code more tree-shakable. The `Validators` class still retains its structure and calls these standalone methods internally to keep this change backwards-compatible.
PR Close#41189
PR Close#41220
This commit updates Forms code to avoid direct references to all built-in ControlValueAccessor classes, which
prevents their tree-shaking from production builds. Instead, a new static property is added to all built-in
ControlValueAccessors, which is checked when we need to identify whether a given ControlValueAccessors is a
built-in one.
PR Close#41146
PR Close#41197
Previously, injector definitions contained a `factory` property that
was used to create a new instance of the associated NgModule class.
Now this factory has been moved to its own `ɵfac` static property on the
NgModule class itself. This is inline with how directives, components and
pipes are created.
There is a small size increase to bundle sizes for each NgModule class,
because the `ɵfac` takes up a bit more space:
Before:
```js
let a = (() => {
class n {}
return n.\u0275mod = c.Cb({type: n}),
n.\u0275inj = c.Bb({factory: function(t) { return new (t || n) }, imports: [[e.a.forChild(s)], e.a]}),
n
})(),
```
After:
```js
let a = (() => {
class n {}
return n.\u0275fac = function(t) { return new (t || n) },
n.\u0275mod = c.Cb({type: n}),
n.\u0275inj = c.Bb({imports: [[r.a.forChild(s)], r.a]}),
n
})(),
```
In other words `n.\u0275fac = ` is longer than `factory: ` (by 5 characters)
and only because the tooling insists on encoding `ɵ` as `\u0275`.
This can be mitigated in a future PR by only generating the `ɵfac` property
if it is actually needed.
PR Close#41022
PR Close#41133
One of the main goals of the bundling tests is to verify that unused symbols are tree-shaken away in prod bundles.
Currently both Reactive and Template-driven test apps are merged into one. In order to make these tree-shaking
tests even more useful, this commit splits exiting test app into two, so that we can further optimize sets of
symbols that should be retained in both scenarios.
PR Close#41108
Previously, `ɵɵgetFactoryOf()` was "privately" published from
`@angular/core` since in the past it was assumed that this
might be an instruction generated by the compiler.
This is not currently the case, so this commit removes it from
the private exports and renames it to indicate that it is a local
helper function.
PR Close#41040
Before `unknown` was available, the `never` type was used to discourage
application developers from using "private" properties. The `unknown` type
is much better suited for this.
PR Close#41040
This PR performs a small refactoring to use `RuntimeError` class and corresponding error code (by calling
`throwProviderNotFoundError` which formats the message) to make it more consistent with other places where
similar errors are thrown.
PR Close#40901
This commit implements creating of `ɵɵngDeclarePipe()` calls in partial
compilation, and processing of those calls in the linker and JIT compiler.
See #40677
PR Close#40803
This commit fixes the issue of the ASSERTION ERROR issue when
a projected node(RNode) inside an array is checked against the types
of TNodeType.Element, TNodeType.Container, TNodeType.ElementContainer,
TNodeType.IcuContainer, TNodeType.Projection. As it's inside an array,
it doesn't fall into any of those types, as a result, it throws
the ASSERTION ERROR.
PR Close#37120
PR Close#37167
Prior to this patch, if an element was removed multiple times (due
to the nature of parent/child elements), the leave listeners may
have been fired for an element that was already removed. This patch
adds a guard within the animations code to prevent this.
PR Close#40712
The codebase currently contains several `EMPTY_ARRAY` constants,
and they can end up in the bundle of an application.
A recent commit 6fbe219 tipped us off
as it introduced several `noop` occurrences in the golden symbol files.
After investigating with @petebacondarwin,
we decided to remove the duplicated symbols.
This probably shaves only a few bytes,
but this commit removes the duplicated functions,
by always using the one in `core/src/utils/empty`.
PR Close#40587
Previously, in `_mismatch()`, the `DefaultIterableDiffer` first checks
`_linkedRecords` for `itemTrackBy`, then checks `_unlinkedRecords`.
This cause the `DefaultIterableDiffer` to move "later" items that match the
`itemTrackBy` from the old collection, rather than using the "earlier" one.
Now we check `_unlinkedRecords` first, so that the `DefaultIterableDiffer`
can give a more stable and reasonable result after diffing. For example,
rather than (`a1` and `a2` have same trackById)
```
a1 b c a2 => b a2 c a1
```
we get
```
a1 b c a2 => b a1 c a2
```
where a1 and a2 retain their original order despite both
having the same track by value.
Fixes#23815
PR Close#23941
This commits adds additional expectations to verify that the bloom
filter is able to correctly handle token IDs that exceed the size of
the bloom filter (which is currently 256 bits).
PR Close#40489
Because the query now has `flags` which specify the mode, the static query
instruction can now be remove. It is simply normal query with `static` flag.
PR Close#40091
Previous implementation would fire changes `QueryList.changes.subscribe`
whenever the `QueryList` was recomputed. This resulted in artificially
high number of change notifications, as it is possible that recomputing
`QueryList` results in the same list. When the `QueryList` gets recomputed
is an implementation detail and it should not be the thing which determines
how often change event should fire.
This change introduces a new `emitDistinctChangesOnly` option for
`ContentChildren` and `ViewChildren`.
```
export class QueryCompWithStrictChangeEmitParent {
@ContentChildren('foo', {
// This option will become the default in the future
emitDistinctChangesOnly: true,
})
foos!: QueryList<any>;
}
```
PR Close#40091
This commit updates the logic that calculates `useFactory` function arguments to avoid relying on `instanceof`
checks (thus always retaining symbols) and relies on flags that DI decorators contain (as a monkey-patched property).
Another perf benefit is having less megamorphic reads while calculating args for the `useFactory` call: we used to
check whether a token has `ngMetadataName` property 4 times (in worst case), now we have just 1 megamorphic read in
all cases.
Closes#40143.
PR Close#40145
The `template` and `isInline` fields were previously stored in a nested
object, which was initially done to accommodate for additional template
information to support accurate source maps for external templates. In
the meantime the source mapping has been accomplished in a different
way, and I feel this flattened structure is simpler and smaller so is
preferable over the nested object. This change also makes the `isInline`
property optional with a default value of `false`.
PR Close#40383
When we attach a `ViewRef` to a `ViewContainerRef`, we save a reference to the container
onto the `ViewRef` so that we can remove it when the ref is destroyed. The problem is
that if the container's `hostView` is destroyed first, the `ViewRef` has no way of knowing
that it should stop referencing the container.
These changes remove the leak by not saving a reference at all. Instead, when a `ViewRef`
is destroyed, we clean it up through the `LContainer` directly. We don't need to worry
about the case where the container is destroyed before the view, because containers
automatically clean up all of their views upon destruction.
Fixes#38648.
PR Close#40219
The "monitoring" workflow has been failing since #40127 was merged,
due to a Saucelabs test failure in Internet Explorer 11. The issue is
with the test's expectation which does not account for Ivy instruction
invocations to use "anonymous" instead of the instruction's function
name. This commit changes the test expectation to also accept
"anonymous", which was already the case for similar expectations.
PR Close#40342
Currently when analyzing the metadata of a directive, we bundle together the bindings from `host`
and the `HostBinding` and `HostListener` together. This can become a problem later on in the
compilation pipeline, because we try to evaluate the value of the binding, causing something like
`@HostBinding('class.foo') public true = 1;` to be treated the same as
`host: {'[class.foo]': 'true'}`.
While looking into the issue, I noticed another one that is closely related: we weren't treating
quoted property names correctly. E.g. `@HostBinding('class.foo') public "foo-bar" = 1;` was being
interpreted as `classProp('foo', ctx.foo - ctx.bar)` due to the same issue where property names
were being evaluated.
These changes resolve both of the issues by treating all `HostBinding` instance as if they're
reading the property from `this`. E.g. the `@HostBinding('class.foo') public true = 1;` from above
is now being treated as `host: {'[class.foo]': 'this.true'}` which further down the pipeline becomes
`classProp('foo', ctx.true)`. This doesn't have any payload size implications for existing code,
because we've always been prefixing implicit property reads with `ctx.`. If the property doesn't
have an identifier that can be read using dotted access, we convert it to a quoted one (e.g.
`classProp('foo', ctx['is-foo']))`.
Fixes#40220.
Fixes#40230.
Fixes#18698.
PR Close#40233
In some browsers, notably a mobile version of webkit on iPad, the
result of calling `DOMParser.parseFromString()` returns a document
whose `body` property is null until the next tick of the browser.
Since this is of no use to us for sanitization, we now fall back to the
"inert document" strategy for this case.
Fixes#39834
PR Close#40107
The `ɵɵngDeclareComponent` calls are designed to be translated to fully
AOT compiled code during a build transform, but in cases this is not
done it is still possible to compile the declaration object in the
browser using the JIT compiler. This commit adds a runtime
implementation of `ɵɵngDeclareComponent` which invokes the JIT compiler
using the declaration object, such that a compiled component definition
is made available to the Ivy runtime.
PR Close#40127
The `render3` test targets are currently also executed for ViewEngine
builds, even though the `render3` infrastructure only concerns Ivy
infrastructure. This commit tags the test targets as ivy-only to disable
those tests for View Engine.
PR Close#40127
Currently we check whether a property binding contains an interpolation using a regex so
that we can throw an error. The problem is that the regex doesn't account for quotes
which means that something like `[prop]="'{{ foo }}'"` will be considered an error, even
though it's not actually an interpolation.
These changes build on top of the logic from #39826 to account for interpolation
characters inside quotes.
Fixes#39601.
PR Close#40267
The trustConstantHtml and trustConstantResourceUrl functions are only
meant to be passed constant strings extracted from Angular application
templates, as passing other strings or variables could introduce XSS
vulnerabilities.
To better protect these APIs, turn them into template tags. This makes
it possible to assert that the associated template literals do not
contain any interpolation, and thus must be constant.
Also add tests for the change to prevent regression.
PR Close#40082
There are two parts to this commit:
1. Revert the changes from #38379. This change had an incomplete view of
how things worked and also diverged the implementations of
`applyRedirects` and `recognize` even more.
2. Apply the fixes from the `recognize` algorithm to ensure that named
outlets with empty path parents can be matched. This change also passes
all the tests that were added in #38379 with the added benefit of being
a more complete fix that stays in-line with the `recognize` algorithm.
This was made possible by using the same approach for `split` by
always creating segments for empty path matches (previously, this was
only done in `applyRedirects` if there was a `redirectTo` value). At the
end of the expansions, we need to squash all empty segments so that
serializing the final `UrlTree` returns the same result as before.
Fixes#39952Fixes#10726Closes#30410
PR Close#40029
Prior to this commit, removing `FormControlDirective` and `FormGroupName` directive instances didn't clear
the callbacks previously registered on FromControl/FormGroup class instances. As a result, these callbacks
were executed even after `FormControlDirective` and `FormGroupName` directive instances were destroyed. That was
also causing memory leaks since these callbacks also retained references to DOM elements.
This commit updates the cleanup logic to take care of properly detaching FormControl/FormGroup/FormArray instances
from the view by removing view-specific callback at destroy time.
Closes#20007, #37431, #39590.
PR Close#39235
DI providers can be defined via `useFactory` function, which may have arguments configured via `deps` array.
The `deps` array may contain DI flags represented by DI decorators (such as `@Self`, `@SkipSelf`, etc). Prior to this
commit, having the `@Host` decorator in `deps` array resulted in runtime error in Ivy. The problem was that the `@Host`
decorator was not taken into account while `useFactory` argument list was constructed, the `@Host` decorator was
treated as a token that should be looked up.
This commit updates the logic which prepares `useFactory` arguments to recognize the `@Host` decorator.
PR Close#40122
The `ɵɵngDeclareDirective` calls are designed to be translated to fully
AOT compiled code during a build transform, but in cases this is not
done it is still possible to compile the declaration object in the
browser using the JIT compiler. This commit adds a runtime
implementation of `ɵɵngDeclareDirective` which invokes the JIT compiler
using the declaration object, such that a compiled directive definition
is made available to the Ivy runtime.
PR Close#40101
Internally we store lifecycle hooks in the format `[index, hook, index, hook]` and when
iterating over them, we check one place ahead to figure out whether we've hit found
a hook or an index. The problem is that the loop is set up to iterate up to `hooks.length`
which means that we may go out of bounds on the last iteration, depending on where
we started. This appears to happen under a specific set of circumstances where a
directive calls `detectChanges` from an input setter while it has `ngOnChanges` and
`ngAfterViewInit` hooks.
These changes resolve the issue by only iterating up to `length - 1` which guarantees that
we can always look one place ahead.
This appears to have regressed some time in version 10.
Fixes#38611.
PR Close#40206
PR #39876 introduced an error where the `onDestroy` of `ComponentRef`
would only get called if `ngDevMode` was set to true. This was because
in dev mode we would freeze `TCleanup` to verify that no more
static cleanup would get added to `TCleanup` array. This ensured
that `TCleanup` was always present in dev mode. In production the
`TCleanup` would get created only when needed. The resulting cleanup
code was incorrectly indented and would only run if `TCleanup` was
present causing this issue.
Fix#40105
PR Close#40120
JoostK