Angular DevTools is working on developing signal debugging support. This commit is a step in the direction of making available debug information to the framework that will allow Angular DevTools to provide users with more accurate information regarding the usage of signals in their applications.
Follow up PRs that will use this arg will:
- Develop a typescript transform that will detect usages of signal functions and attempt to add a debugName without the user needing to specify one directly
- Develop debug APIs for discovering signal graphs within Angular applications (using debugName as a way to label nodes on the graph)
PR Close#57073
Previously, effect() would handle errors differently depending on the effect
type. Root effects had a try/catch that would execute them independently and
report errors to `ErrorHandler`, while component effects would "crash" CD.
This commit switches all effects to use the same error handling (errors
always reach the CD error handler).
An additional unrelated refactoring is thrown in which removes the
`pendingTask` machinery from root effects, since they make `ApplicationRef`
dirty and thus trigger the scheduler.
PR Close#57952
The original effect design for Angular had one "bucket" of effects, which
are scheduled on the microtask queue. This approach got us pretty far, but
as developers have built more complex reactive systems, we've hit the
limitations of this design.
This commit changes the nature of effects significantly. In particular,
effects created in components have a completely new scheduling system, which
executes them as a part of the change detection cycle. This results in
behavior similar to that of nested effects in other reactive frameworks. The
scheduling behavior here uses the "mark for traversal" flag
(`HasChildViewsToRefresh`). This has really nice behavior:
* if the component is dirty already, effects run following preorder hooks
(ngOnInit, etc).
* if the component isn't dirty, it doesn't get change detected only because
of the dirty effect.
This is not a breaking change, since `effect()` is in developer preview (and
it remains so).
As a part of this redesigned `effect()` behavior, the `allowSignalWrites`
flag was removed. Effects no longer prohibit writing to signals at all. This
decision was taken in response to feedback / observations of usage patterns,
which showed the benefit of the restriction did not justify the DX cost.
The new effect timing is not yet enabled - a future PR will flip the flag.
PR Close#56501
We had some tests that were leaking LViews, because we were testing things like `createComponent`, but not destroying them afterwards. These changes clean up most of them, although there are a handful still left that I didn't have time to fully track down.
PR Close#57546
The `afterRender` infrastructure was first implemented around the idea of
independent, singular hooks. It was later updated to support a spec of
multiple hooks that pass values from one to another as they execute, but the
implementation still worked in terms of singular hooks under the hood. This
creates a number of maintenance issues, and a few bugs. For example, when
one hook fails, further hooks in the pipeline should no longer execute, but
this was hard to ensure under the old design.
This refactoring restructures `afterRender` infrastructure significantly to
introduce the concept of a "sequence", a collection of hooks of different
phases that execute together. Overall, the implementation is simplified
while making it more resilient to issues and future use cases, such as the
upcoming `afterRenderEffect`.
As part of this refactoring, the `internalAfterNextRender` concept is
removed, as well as the unused `queueStateUpdate` concept which used it.
PR Close#57453
This commit updates serialization and hydration i18n logic to take into account situations when i18n blocks are located within "skip hydration" blocks.
Resolves#57105.
PR Close#57299
These changes replace most usages of `removeChild` with `remove`. The latter has the advantage of not having to look up the `parentNode` and ensure that the child being removed actually belongs to the specific parent.
The refactor should be fairly safe since all the browsers we cover support `remove`. [Something similar was done in Components](https://github.com/angular/components/pull/23592) some time ago and there haven't been any bug reports as a result.
PR Close#57203
Fixes that the runtime implementation of `ɵɵngDeclareDirective` was interpreting the `hostDirectives` mapping incorrectly. Instead of treating the inputs/outputs as `['binding', 'alias']` arrays, it was parsing them as `['binding: alias']`. This was leading to runtime errors if a user is consuming a partially-compiled library in JIT mode.
Fixes#54096.
PR Close#57002
Adds the implementation of the following new instructions:
* `declareLet` - creation-time instruction that initializes the slot for a let declaration.
* `storeLet` - update-time instruction that stores the current value of a let declaration.
* `readContextLet` - instruction that reads the stored value of a let declaration from a different view.
On top of the instructions, it also introduces a new `LetDeclaration` TNode type.
The new TNode is nececessary for DI to work correctly in pipes inside the let expression,
as well as for proper hydration support.
PR Close#56527
This commit starts exposing `isSignal` for inputs in the
`ComponentMirror`. We initially had this as a draft when rolling out
signal inputs, but there were no good use-cases, so we skipped it.
Now, inside G3, for the testing infrastructure and rolling out
advancements for signal inputs, having this information is necessary and
allows identifying signal inputs without "accessing fields" on the class
that may cause side-effects (like triggering setters).
PR Close#56402
Builds on top of the previous changes to add support for deferred blocks during partial compilation. To do this, the following changes had to be made:
* The metadata passed into `ɵɵngDeclareComponent` has an additional field called `deferBlockDependencies` which has an array of the dependency loading functions for each defer block in the template. During linking, the dependency functions are loaded by matching their template index to the index in the `deferBlockDependencies` array.
* There's a new `ɵɵngDeclareClassMetadataAsync` function that is created for components that have deferred dependencies. It gets transpiled to `setClassMetadataAsync` and works in the same way by capturing a dependency loading function and setting the metadata after the dependencies are resolved. It also has some extra fields for capturing the version which are standard in linker-generated code.
* Deferred import statements are now stripped in partial compilation mode, similar to full compilation.
PR Close#54908
In order to serialize and hydrate i18n blocks, we need to be able to walk an AST for the translated message. This AST is generated during normal parsing of the message.
PR Close#54724
In zone-full applications, this is already true because effects are
scheduled inside a microtask and tracked by ZoneJS. When not using
zones, this should stay consistent on principle and for testability
reasons. A general pattern with zoneless testing will be to update state
and `await` some promise (i.e. `fixture.whenStable`, which will be
linked to `ApplicationRef.isStable`).
PR Close#53835
One downside of implicit dependency tracking in `effect()`s is that it's easy
to for downstream code to end up running inside the effect context by accident.
For example, if an effect raises an event (e.g. by `next()`ing a `Subject`), the
subscribers to that `Observable` will run inside the effect's reactive context,
and any signals read within the subscriber will end up as dependencies of the
effect. This is why the `untracked` function is useful, to run certain
operations without incidental signal reads ending up tracked.
However, knowing when this is necessary is non-trivial. For example, injecting
a dependency might cause it to be instantiated, which would run the constructor
in the effect context unless the injection operation is untracked.
Therefore, Angular will automatically drop the reactive context within a number
of framework APIs. This commit addresses these use cases:
* creating and destroying views
* creating and destroying DI injectors
* injecting dependencies
* emitting outputs
Fixes#54548
There are likely other APIs which would benefit from this approach, but this
is a start.
PR Close#54614
Fixes that `ɵunwrapWritableSignal` inferring getter functions as not matching the interface of `WritableSignal` instead of preserving them.
PR Close#54252
In a previous commit the TCB was changed to cast the assignment to an input in order to widen its type to allow `WritableSignal`. This ended up breaking existing inputs whose setter has a wider type than its getter. These changes switch to unwrapping the value on the binding side.
PR Close#54252
Reworks the TCB for two-way bindings to make them simpler and to avoid regressions for two-way bindings to generic inputs. The new TCB looks as follows:
```
var _t1: Dir;
var _t2 = _t1.input;
(_t1 as typeof _t2 | WritableSignal<typeof _t2>) = expression;
```
PR Close#54252
Similar to signal-based inputs, we support signal-based queries in JIT
by expecting a decorator to be added. This is a consequence of the
design, given that JIT requires query declaration information before
the class is initialized- but ironically there is no way to collect this
information without instantiating the class.
A JIT transform in the Angular CLI will automatically generate these
decorators for testing.
PR Close#54019
This commit changes the `HasTransform` flag to be only concerned with
decorator inputs. This allows us to automatically detect signal input
transforms without reliance on the flag, resulting in less complexity in
the compiler (as outlined in the design doc) and various other places,
while it also allows us to simplify JIT support for signal inputs
because there would be no need to capture the "hasTransform" state in
the decorator so that JIT can generate the according input flags.
`isSignal` will still persist as an input flag to allow for monomorphic
and highly efficient distinguishing at runtime, whether an input is
signal based or not. JIT transform will also need to propagate this
information to the runtime somehow.
PR Close#53808
We generate `advance` instructions before most update instructions and the majority of `advance` calls are advancing by one. We can save some bytes for the most common case by omitting the parameter for `advance(1)` altogether.
PR Close#53845
This PR provides strict type definition for the window.ng object used
for both console debugging and devtools. `GlobalDevModeUtils` now
gathers all type information about all methods exposed on window.ng.
PR Close#53439
This commit removes a hack that deletes `Event` from the global context
when using domino. Instead, it sets the global event to domino's
implementation of `Event`.
PR Close#53659
This commit introduces a new enum for capturing additional metadata
about inputs. Called `InputFlags`. These will be built up at compile
time and then propagated into the runtime logic, in a way that does
not require additional lookup dictionaries data structures, or
additional memory allocations for "common inputs" that do not have any flags.
The flags will incorporate information on whether an input is signal
based. This can then be used to avoid megamorphic accesses when such
input is set- as we'd not need to check the input field value. This also
avoids cases where an input signal may be used as initial value for an
input (as we'd not incorrectly detect the input as a signal input then).
The new metadata emit will be useful for incorporating additional
metadata for inputs, such as whether they are required etc (although
required inputs are a build-time only construct right now- but this is a
good illustration of why input flags can be useful). An alternative
could have been to have an additional boolean entry for signal inputs,
but allocating a number with more flexible input flags seems more future
proof and more reasonable andreadable.
More information on the megamorphic access when updating an input
signal
https://docs.google.com/document/d/1FpnFruviKb6BFTQfMAP2AMEqEB0FI7z-3mT_qm7lzX8/edit.
PR Close#53571
`effect` was expecting an `ErrorHandler` in its constructor which can lead to a circular DI error if an effect is used inside a custom `ErrorHandler`. These changes inject the `ErrorHandler` only when reporting errors.
Fixes#52680.
PR Close#53713
In order to provide a reasonable experience for Angular without Zones,
we need a mechanism to run change detection when we receive a change
notification. There are several existing APIs today that serve as the
change notification: `ChangeDetectorRef.markForCheck`, signal updates,
event listeners (since they mark the view dirty), and attaching a view to
either the `ApplicationRef` or `ChangeDetectorRef`. These operations
are now paired with a notification to the change detection scheduler.
The concrete implementation for this scheduler is still being designed.
However, this gives us a starting point to partner with teams to
experiment with what that might look like.
PR Close#53499
This commit adds the last remaining piece for signal input
type-checking. Bound values to signal inputs are already checked
properly at this point, but inference of generic directive/component
types through their inputs is not implemented.
This commit fixes this. To achieve this, there are a couple of potential
solutions. The generics of a directive are inferred based on input
value expressions using a so-called type constructor. The constructor
looks something like this:
```
const _ctor = <T>(v: Pick<Dir<T>, 'input1', 'input2'>) => Dir<T>;
_ctor({input1: expr1, input2: expr2});
```
This works very well for non-signal inputs where the class member is
directly holding the input values. For signal inputs, this does NOT
work because the class member will actually hold the `InputSignal`
instance. There are a couple of solutions to this:
1. Calling `_ctor` with an `InputSignal<typeof value>`
2. Converting the `_ctor` input signal fields to their write types
(unwrapping the input signals).
We've decided to go with the second option as TypeScript is very
sensitive with assignments and its checks. i.e. co-variance,
contravariance or bivariance. Semantically it makes more sense to unwrap
the input signal "write type" directly and "assign to it". This is safer
and conceptually also easier to follow. A type constructor continues to
only receive the "expresison values". This simplifies code as well.
It's worth noting that the unwrapping as per option 2 also comes at a
cost. We need to be able to generate imports in type constructors. This
was not possible until the previous commit because inline type constructors
did not have an associated type-check block `Environment` and we were
missing access to expression translation and correct import generation.
Overall, solution 2 is now implemented as works as expected. This commit
adds additional unit tests to ensure this.
PR Close#53521
This commit introduces the initial type-checking for signal inputs.
To enable type-checking od signal inputs, there are a couple of tricks
needed. It's not trivial as it would look like at first glance.
Initial attempts could have been to generate additional statements in
type-checking blocks for signal inputs to simply call a method like
`InputSignal#applyNewValue`. This would seem natural, as it would match
what will happen at runtime, but this would break the language-service
auto completion in a highly subtle way. Consider the case where multiple
directives match the same input. Consider the directives have some
overlap in accepted input values, but they also have distinct diverging
values, like:
```ts
class DirA {
value = input<'apple'|'shared'>();
}
class DirB {
value = input<'orange'|'shared'>();
}
```
In such cases, auto completion for the binding expression should suggest
the following values: `apple`, `shared`, `orange` and `undefined`.
The language service achieves this by getting completions in the
type-check block where the user expression would live. This BREAKS if
we'd have multiple places where the expression from the user is used.
Two different places, or more, surface additional problems with
diagnostic collection. Previously diagnostics would surface the union
type of allowed values, but with multiple places, we'd have to work with
potentially 1+ diagnostics. This is non-ideal.
Another important consideration is test coverage. It might sound
problematic to consider the existing test infrastructure as relevant,
but in practice, we have thousands of diagnostic type check block tests
that would greatly benefit if the general emit structure would still
match conceptually. This is another bonus argument on why changing the
way inputs are applied is probably an option we should consider as a
last resort.
Ultimately, there is a good solution where we unwrap directive signal
inputs, based on metadata, and access a brand type field on the
`InputSignal`. This ensures auto-completion continues to work as is, and
also the structure of type check blocks doesn't change conceptually. In
future commits we also need to handle type-inference for generic signal
inputs.
Note: Another alternative considered, in terms of using metadata or not.
We could have type helpers to unwrap signal inputs using type helpers
like: `T extends InputSignal<any, WriteT> ? WriteT : T`. This would
allow us to drop the input signal metadata dependency, but in reality,
this has a few issues:
- users might have `@Input`'s passing around `InputSignal`'s. This is
unlikely, but shows that the solution would not be fully correct.
- we need the metadata regardless, as we plan on accessing it at runtime
as well, to distinguish between signal inputs and normal inputs when
applying new values. This was not clear when this option was
considered initially.
PR Close#53521
This change replaces the implementation of the multi-map used to store
detached views while reconciling lists. The new implementation optimizes
memory allocation for such map and avoid arrays allocation when there are
no duplicated keys.
PR Close#52245
When an effect is created in a component constructor, it might read signals
which are derived from component inputs. These signals may be unreliable or
(in the case of the proposed input signals) may throw if accessed before the
component is first change detected (which is what makes required inputs
available).
Depending on the scenario involved, the effect may or may not run before
this initialization takes place, which isn't a great developer experience.
In particular, effects created during CD (e.g. via control flow) work fine,
as do effects created in bootstrap thanks to the sync CD it performs. When
an effect is created through dynamic component creation outside of CD though
(such as on router navigations), it runs before the component is first CD'd,
causing the issue.
In fact, in the signal components RFC we described how effects would wait
until ngOnInit for their first execution for exactly this reason, but this
behavior was never implemented as it was thought our effect scheduling
design made it unnecessary. This is true of the regular execution of effects
but the above scenario shows that *creation* of the effect is still
vulnerable. Thus, this logic is needed.
This commit makes effects sensitive to their creation context, by injecting
`ChangeDetectorRef` optionally. An effect created with an injector that's
tied to a component will wait until that component is initialized before
initially being scheduled. TestBed effect flushing is also adjusted to
account for the additional interaction with change detection.
PR Close#52473
A runtime error will be thrown if a non-standalone component is being rendered without its NgModule loaded in the browser. This error is thrown only in dev mode and only if the Angular option `forbidOrphanComponents` is set to true. The error contains useful info to find the orphan component in the source code.
PR Close#52061
A new method `isOrphanComponent` is added to the deps tracker API to check if the NgModule declaring this component, if exists, is loaded into the browser.
PR Close#52061
Using `afterRender` schedules long-living lifecycle hooks. Scheduling
such hooks inside reactive contexts could mean that many of the
same hooks would be scheduled, quickly piling up every time a
consumed signal changes. This is likely unintended and could degrade
application performance or result in unexpected behavior.
Additionally, scheduling `afterRender` inside a `computed` is considered
a side effect. Computed expressions are expected to be pure/ i.e. free
of side effects. We can avoid this caveat by detecting the reactive
context in development.
PR Close#52138
Using an `effect` inside a `computed` is a clear violation of
the conceptual idea of computed's being pure/ side-effect free.
Additionally, scheduling new effects from an existing actively
running effect is likely unintended as this could degrage application
performance or result in unintentional behaviors. Multiple long-living
effects would be scheduled every time the effect expressions runs.
For these reasons, we are explicitly preventing this pitfal, by
disallowing using `effect` inside reactive contexts.
PR Close#52138
The error message now contains the code location of the component. It now looks like: "Error: NG01001: Orphan component found! Trying to render the component Main (at $PROJECT_ROOT/src/main.ts:8) without first loading the NgModule ..."
PR Close#51919
The current error stringifier only includes the class name. In this change a new stringifier is added which returns a more helpful string which includes the file path and line number. Note that this is only the case with components, and for other class types (directive, pipes) it will fallback to the current stringifier. Subsequent changes can cover the case of directive and pipes as well.
PR Close#51919
The new list reconcilation algorithm, an alternative to
the DefaultIterableListDiffer. It works by performing updates
in place instead of creating intermediate data describing changes
to apply. For lists expressed as an Array it performs additional
optimizations for the moves and swap scenarios.
The new list diffing approach is meant to be used in the new control
flow and should me much faster as compared to the ngFor with the
DefaultIterableListDiffer.
PR Close#51980
