Recently it was made possible to have a directive without selector,
which are referred to as abstract directives. Such directives should not
be registered in an NgModule, but can still contain decorators for
inputs, outputs, queries, etc. The information from these decorators and
the `@Directive()` decorator itself needs to be registered with the
central `MetadataRegistry` so that other areas of the compiler can
request information about a given directive, an example of which is the
template type checker that needs to know about the inputs and outputs of
directives.
Prior to this change, however, abstract directives would only register
themselves with the `MetadataRegistry` as being an abstract directive,
without all of its other metadata like inputs and outputs. This meant
that the template type checker was unable to resolve the inputs and
outputs of these abstract directives, therefore failing to check them
correctly. The typical error would be that some property does not exist
on a DOM element, whereas said property should have been bound to the
abstract directive's input.
This commit fixes the problem by always registering the metadata of a
directive or component with the `MetadataRegistry`. Tests have been
added to ensure abstract directives are handled correctly in the
template type checker, together with tests to verify the form of
abstract directives in declaration files.
Fixes#30080
PR Close#33131
Often the types of an `@Input`'s field don't fully reflect the types of
assignable values. This can happen when an input has a getter/setter pair
where the getter always returns a narrow type, and the setter coerces a
wider value down to the narrow type.
For example, you could imagine an input of the form:
```typescript
@Input() get value(): string {
return this._value;
}
set value(v: {toString(): string}) {
this._value = v.toString();
}
```
Here, the getter always returns a `string`, but the setter accepts any value
that can be `toString()`'d, and coerces it to a string.
Unfortunately TypeScript does not actually support this syntax, and so
Angular users are forced to type their setters as narrowly as the getters,
even though at runtime the coercion works just fine.
To support these kinds of patterns (e.g. as used by Material), this commit
adds a compiler feature called "input coercion". When a binding is made to
the 'value' input of a directive like MatInput, the compiler will look for a
static field with the name ngAcceptInputType_value. If such a field is found
the type-checking expression for the input will use the static field's type
instead of the type for the @Input field,allowing for the expression of a
type conversion between the binding expression and the value being written
to the input's field.
To solve the case above, for example, MatInput might write:
```typescript
class MatInput {
// rest of the directive...
static ngAcceptInputType_value: {toString(): string};
}
```
FW-1475 #resolve
PR Close#33243
Often the types of an `@Input`'s field don't fully reflect the types of
assignable values. This can happen when an input has a getter/setter pair
where the getter always returns a narrow type, and the setter coerces a
wider value down to the narrow type.
For example, you could imagine an input of the form:
```typescript
@Input() get value(): string {
return this._value;
}
set value(v: {toString(): string}) {
this._value = v.toString();
}
```
Here, the getter always returns a `string`, but the setter accepts any value
that can be `toString()`'d, and coerces it to a string.
Unfortunately TypeScript does not actually support this syntax, and so
Angular users are forced to type their setters as narrowly as the getters,
even though at runtime the coercion works just fine.
To support these kinds of patterns (e.g. as used by Material), this commit
adds a compiler feature called "input coercion". When a binding is made to
the 'value' input of a directive like MatInput, the compiler will look for a
static function with the name ngCoerceInput_value. If such a function is
found, the type-checking expression for the input will be wrapped in a call
to the function, allowing for the expression of a type conversion between
the binding expression and the value being written to the input's field.
To solve the case above, for example, MatInput might write:
```typescript
class MatInput {
// rest of the directive...
static ngCoerceInput_value(value: {toString(): string}): string {
return null!;
}
}
```
FW-1475 #resolve
PR Close#33243
Module defs are not considered public API, so the property
that contains them should be prefixed with Angular's marker
for "private" ('ɵ') to discourage apps from relying on def
APIs directly.
This commit adds the prefix and shortens the name from
ngModuleDef to mod. This is because property names
cannot be minified by Uglify without turning on property
mangling (which most apps have turned off) and are thus
size-sensitive.
PR Close#33142
Pipe defs are not considered public API, so the property
that contains them should be prefixed with Angular's marker
for "private" ('ɵ') to discourage apps from relying on def
APIs directly.
This commit adds the prefix and shortens the name from
ngPipeDef to pipe. This is because property names
cannot be minified by Uglify without turning on property
mangling (which most apps have turned off) and are thus
size-sensitive.
PR Close#33142
Directive defs are not considered public API, so the property
that contains them should be prefixed with Angular's marker
for "private" ('ɵ') to discourage apps from relying on def
APIs directly.
This commit adds the prefix and shortens the name from
ngDirectiveDef to dir. This is because property names
cannot be minified by Uglify without turning on property
mangling (which most apps have turned off) and are thus
size-sensitive.
Note that the other "defs" (ngFactoryDef, etc) will be
prefixed and shortened in follow-up PRs, in an attempt to
limit how large and conflict-y this change is.
PR Close#33110
Component defs are not considered public API, so the property
that contains them should be prefixed with Angular's marker
for "private" ('ɵ') to discourage apps from relying on def
APIs directly.
This commit adds the prefix and shortens the name from
`ngComponentDef` to `cmp`. This is because property names
cannot be minified by Uglify without turning on property
mangling (which most apps have turned off) and are thus
size-sensitive.
Note that the other "defs" (ngDirectiveDef, etc) will be
prefixed and shortened in follow-up PRs, in an attempt to
limit how large and conflict-y this change is.
PR Close#33088
Previously, ngtsc attempted to use the .d.ts schema for HTML elements to
check bindings to DOM properties. However, the TypeScript lib.dom.d.ts
schema does not perfectly align with the Angular DomElementSchemaRegistry,
and these inconsistencies would cause issues in apps. There is also the
concern of supporting both CUSTOM_ELEMENTS_SCHEMA and NO_ERRORS_SCHEMA which
would have been very difficult to do in the existing system.
With this commit, the DomElementSchemaRegistry is employed in ngtsc to check
bindings to the DOM. Previous work on producing template diagnostics is used
to support generation of this different kind of error with the same high
quality of error message.
PR Close#32171
A structural directive can specify a template guard for an input, such that
the type of that input's binding can be narrowed based on the guard's return
type. Previously, such template guards could only be methods, of which an
invocation would be inserted into the type-check block (TCB). For `NgIf`,
the template guard narrowed the type of its expression to be `NonNullable`
using the following declaration:
```typescript
export declare class NgIf {
static ngTemplateGuard_ngIf<E>(dir: NgIf, expr: E): expr is NonNullable<E>
}
```
This works fine for usages such as `*ngIf="person"` but starts to introduce
false-positives when e.g. an explicit non-null check like
`*ngIf="person !== null"` is used, as the method invocation in the TCB
would not have the desired effect of narrowing `person` to become
non-nullable:
```typescript
if (NgIf.ngTemplateGuard_ngIf(directive, ctx.person !== null)) {
// Usages of `ctx.person` within this block would
// not have been narrowed to be non-nullable.
}
```
This commit introduces a new strategy for template guards to allow for the
binding expression itself to be used as template guard in the TCB. Now,
the TCB generated for `*ngIf="person !== null"` would look as follows:
```typescript
if (ctx.person !== null) {
// This time `ctx.person` will successfully have
// been narrowed to be non-nullable.
}
```
This strategy can be activated by declaring the template guard as a
property declaration with `'binding'` as literal return type.
See #30235 for an example where this led to a false positive.
PR Close#30248
Previously the template type-checking code only considered the metadata of
directive classes actually referenced in the template. If those directives
had base classes, any inputs/outputs/etc of the base classes were not
tracked when generating the TCB. This resulted in bindings to those inputs
being incorrectly attributed to the host component or element.
This commit uses the new metadata package to follow directive inheritance
chains and use the full metadata for a directive for TCB generation.
Testing strategy: Template type-checking tests included.
PR Close#29698
Previously, metadata registration (the recording of collected metadata
during analysis of directives, pipes, and NgModules) was only used to
produce the `LocalModuleScope`, and thus was handled by the
`LocalModuleScopeRegistry`.
However, the template type-checker also needs information about registered
directives, outside of the NgModule scope determinations. Rather than
reuse the scope registry for an unintended purpose, this commit introduces
new abstractions for metadata registration and lookups in a separate
'metadata' package, which the scope registry implements.
This paves the way for a future commit to make use of this metadata for the
template type-checking system.
Testing strategy: this commit is a refactoring which introduces no new
functionality, so existing tests are sufficient.
PR Close#29698
This refactoring moves code around between a few of the ngtsc subpackages,
with the goal of having a more logical package structure. Additional
interfaces are also introduced where they make sense.
The 'metadata' package formerly contained both the partial evaluator,
the TypeScriptReflectionHost as well as some other reflection functions,
and the Reference interface and various implementations. This package
was split into 3 parts.
The partial evaluator now has its own package 'partial_evaluator', and
exists behind an interface PartialEvaluator instead of a top-level
function. In the future this will be useful for reducing churn as the
partial evaluator becomes more complicated.
The TypeScriptReflectionHost and other miscellaneous functions have moved
into a new 'reflection' package. The former 'host' package which contained
the ReflectionHost interface and associated types was also merged into this
new 'reflection' package.
Finally, the Reference APIs were moved to the 'imports' package, which will
consolidate all import-related logic in ngtsc.
PR Close#27743
To support updating `ModuleWithProviders` calls,
we need to be able to map exported functions between
source and typings files, as well as classes.
PR Close#27326
ngcc would feed ngtsc with the function declaration inside of an IIFE as
that is considered the class symbol's declaration node, according to
TypeScript's `ts.Symbol.valueDeclaration`. ngtsc however only considered
variable decls and actual class decls as potential class declarations,
so given the function declaration node it would fail to generate the
`setClassMetadata` call.
ngtsc no longer makes its own assumptions about what classes look like,
but always asks the reflection host to yield this kind of information.
PR Close#27438
With ngcc's ability to fixup pre-Ivy ModuleWithProviders such that they
include a reference to the NgModule type, the type may become a qualified
name:
```
import {ModuleWithProviders} from '@angular/core';
import * as ngcc0 from './module';
export declare provide(): ModuleWithProviders<ngcc0.Module>;
```
ngtsc now takes this situation into account when reflecting a
ModuleWithProvider's type argument.
PR Close#27562
For ngcc's processing of ES5 bundles, the spread syntax has been
downleveled from `[...ARRAY]` to become `ARRAY.slice()`. This commit
adds basic support for static resolution of such call.
PR Close#27158
When ngtsc compiles @angular/core, it rewrites core imports to the
r3_symbols.ts file that exposes all internal symbols under their
external name. When creating the FESM bundle, the r3_symbols.ts file
causes the external symbol names to be rewritten to their internal name.
Under ngcc compilations of FESM bundles, the indirection of
r3_symbols.ts is no longer in place such that the external names are
retained in the bundle. Previously, the external name `ɵdefineNgModule`
was explicitly declared internally to resolve this issue, but the
recently added `setClassMetadata` was not declared as such, causing
runtime errors.
Instead of relying on the r3_symbols.ts file to perform the rewrite of
the external modules to their internal variants, the translation is
moved into the `ImportManager` during the compilation itself. This
avoids the need for providing the external name manually.
PR Close#27055
The `NgModule` handler generates `R3References` for its declarations, imports,
exports, and bootstrap components, based on the relative import path
between the module and the classes it's referring to. This works fine for
compilation of a .ts Program inside ngtsc, but in ngcc the import needed
in the .d.ts file may be very different to the import needed between .js
files (for example, if the .js files are flattened and the .d.ts is not).
This commit introduces a new API in the `ReflectionHost` for extracting the
.d.ts version of a declaration, and makes use of it in the
`NgModuleDecorationHandler` to write a correct expression for the `NgModule`
definition type.
PR Close#26403
Previously the ReflectionHost API only returned the names of decorators
and not a reference to their TypeScript Identifier. This commit adds
the identifier itself, so that a consumer can write references to the
decorator.
Testing strategy: this commit is trivial, and the functionality will be
exercised by downstream tests.
PR Close#26860
We are close enough to blacklist a few test targets, rather than whitelist targets to run...
Because bazel rules can be composed of other rules that don't inherit tags automatically,
I had to explicitly mark all of our ts_library and ng_module targes with "ivy-local" and
"ivy-jit" tags so that we can create a query that excludes all fixme- tagged targets even
if those targets are composed of other targets that don't inherit this tag.
This is the updated overview of ivy related bazel tags:
- ivy-only: target that builds or runs only under ivy
- fixme-ivy-jit: target that doesn't yet build or run under ivy with --compile=jit
- fixme-ivy-local: target that doesn't yet build or run under ivy with --compile=local
- no-ivy-jit: target that is not intended to build or run under ivy with --compile=jit
- no-ivy-local: target that is not intended to build or run under ivy with --compile=local
PR Close#26471
In some formats variables are declared as `var` or `let` and only
assigned a value later in the code.
The ngtsc resolver still needs to be able to resolve this value,
so the host now provides a `host.getVariableValue(declaration)`
method that can do this resolution based on the format.
The hosts make some assumptions about the layout of the
code, so they may only work in the constrained scenarios that
ngcc expects.
PR Close#26236
Before type checking can be turned on in ngtsc, appropriate metadata for
each component and directive must be determined. This commit adds tracking
of the extra metadata in *DefWithMeta types to the selector scope handling,
allowing for later extraction for type-checking purposes.
PR Close#26203
Previously, if ngtsc encountered a VariableDeclaration without an
initializer, it would assume that the variable was undefined, and
return that result.
However, for symbols exported from external modules that resolve to
.d.ts files, variable declarations are of the form:
export declare let varName: Type;
This form also lacks an initializer, but indicates the presence of an
importable symbol which can be referenced. This commit changes the
static resolver to understand variable declarations with the 'declare'
keyword and to generate references when it encounters them.
PR Close#25775
This commit adds support for enumeration values. An enumeration value
is now a first-class return value of the resolver, which provides both
a Reference to the enum type itself and the name of the value from that
enum. Resolving an enum itself returns a Map<string, EnumValue>.
PR Close#25619
Ivy definitions in .d.ts files often reference the type of a class.
Sometimes, those classes have generic type parameters. When this is
the case, ngtsc needs to emit generic type parameters in the .d.ts
files (usually by passing 'any').
PR Close#25406
In some code formats (e.g. ES5) methods can actually be function
expressions. For example:
```js
function MyClass() {}
// this static method is declared as a function expression
MyClass.staticMethod = function() { ... };
```
PR Close#25406
ngtsc's static resolver can evaluate function calls where parameters
have default values. In TypeScript code these default values live on the
function definition, but in ES5 code the default values are represented
by statements in the function body.
A new ReflectionHost method getDefinitionOfFunction() abstracts over
this difference, and allows the static reflector to more accurately
evaluate ES5 code.
PR Close#25406
This commit creates an API for factory functions which allows them
to be inherited from one another. To do so, it differentiates between
the factory function as a wrapper for a constructor and the factory
function in ngInjectableDefs which is determined by a default
provider.
The new form is:
factory: (t?) => new (t || SomeType)(inject(Dep1), inject(Dep2))
The 't' parameter allows for constructor inheritance. A subclass with
no declared constructor inherits its constructor from the superclass.
With the 't' parameter, a subclass can call the superclass' factory
function and use it to create an instance of the subclass.
For @Injectables with configured providers, the factory function is
of the form:
factory: (t?) => t ? constructorInject(t) : provider();
where constructorInject(t) creates an instance of 't' using the
naturally declared constructor of the type, and where provider()
creates an instance of the base type using the special declared
provider on @Injectable.
PR Close#25392
In some code formats (e.g. ES5) methods can actually be function
expressions. For example:
```js
function MyClass() {}
// this static method is declared as a function expression
MyClass.staticMethod = function() { ... };
```
PR Close#24897
The `ReflectionHost` interface that is being implemented only expects a
return value of `boolean`.
Moreover, if you want to extend this class to support non-TS code formats,
e.g. ES5, the result of this call returning true does not mean that the `node`
is a `ClassDeclaration`. It could be a `VariableDeclaration`.
PR Close#24897
ngtsc used to assume that all .d.ts dependencies (that is, third party
packages) were imported via an absolute module path. It turns out this
assumption isn't valid; some build tools allow relative imports of
other compilation units.
In the absolute case, ngtsc assumes (and still does) that all referenced
types are available through the entrypoint from which an @NgModule was
imported. This commit adds support for relative imports, in which case
ngtsc will use relative path resolution to determine the imports.
PR Close#25080
There is a bug in the existing handling for cross-file references.
Suppose there are two files, module.ts and component.ts.
component.ts declares two components, one of which uses the other.
In the Ivy model, this means the component will get a directives:
reference to the other in its defineComponent call.
That reference is generated by looking at the declared components
of the module (in module.ts). However, the way ngtsc tracks this
reference, it ends up comparing the identifier of the component
in module.ts with the component.ts file, detecting they're not in
the same file, and generating a relative import.
This commit changes ngtsc to track all identifiers of a reference,
including the one by which it is declared. This allows toExpression()
to correctly decide that a local reference is okay in component.ts.
PR Close#25080
When ngtsc encounters a reference to a type (for example, a Component
type listed in an NgModule declarations array), it traces the import
of that type and attempts to determine the best way to refer to it.
In the event the type is defined in the same file where a reference
is being generated, the identifier of the type is used. If the type
was imported, ngtsc has a choice. It can use the identifier from the
original import, or it can write a new import to the module where the
type came from.
ngtsc has a bug currently when it elects to rely on the user's import.
When writing a .d.ts file, the user's import may have been elided as
the type was not referred to from the type side of the program. Thus,
in .d.ts files ngtsc must always assume the import may not exist, and
generate a new one.
In .js output the import is guaranteed to still exist, so it's
preferable for ngtsc to continue using the existing import if one is
available.
This commit changes how @angular/compiler writes type definitions, and
allows it to use a different expression to write a type definition than
is used to write the value. This allows ngtsc to specify that types in
type definitions should always be imported. A corresponding change to
the staticallyResolve() Reference system allows the choice of which
type of import to use when generating an Expression from a Reference.
PR Close#25080
@ContentChild[ren] and @ViewChild[ren] can contain a forwardRef() to a
type. This commit allows ngtsc to unwrap the forward reference and
deal with the node inside.
It includes two modes of support for forward reference resolution -
a foreign function resolver which understands deeply nested forward
references in expressions that are being statically evaluated, and
an unwrapForwardRef() function which deals only with top-level nodes.
Both will be useful in the future, but for now only unwrapForwardRef()
is used.
PR Close#25080
Previously, references had the concept of an identifier, but would not
properly detect whether the identifier should be used or not when
generating an expression. This change fixes that logic.
Additionally, now whenever an identifier resolves to a reference (even
one imported from another module) as part of resolving an expression,
the reference is updated to use that identifier. This ensures that for
a class Foo declared in foo.ts, but referenced in an expression in
bar.ts, the Reference returned includes the identifier from bar.ts,
meaning that writing an expression in bar.ts for the Reference will not
generate an import.
PR Close#24862
Previously, the static resolver did its own interpretation of statements
in the TypeScript AST, which only functioned on TypeScript code. ES5
code in particular would not work with the resolver as it had hard-coded
assumptions about AST structure.
This commit changes the resolver to use a ReflectionHost instead, which
abstracts away understanding of the structural side of the AST. It adds 3
new methods to the ReflectionHost in support of this functionality:
* getDeclarationOfIdentifier
* getExportsOfModule
* isClass
PR Close#24862
