The decorator downleveling transform patches `ts.EmitResolver.isReferencedAliasDeclaration`
to prevent elision of value imports that occur only in a type-position, which would
inadvertently install the patch repeatedly for each source file in the program.
This could potentially result in a stack overflow when a very large number of files is
present in the program.
This commit fixes the issue by ensuring that the patch is only applied once.
This is also a slight performance improvement, as `isReferencedAliasDeclaration`
is no longer repeatedly calling into all prior installed patch functions.
Fixes#40276
PR Close#40374
Now that `ReadonlyFileSystem` and `PathManipulation` interfaces are
available, this commit updates the compiler-cli to use these more
focussed interfaces.
PR Close#40281
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
This commit changes the `PartialComponentLinker` to use the original source
of an external template when compiling, if available, to ensure that the
source-mapping of the final linked code is accurate.
If the linker is given a file-system and logger, then it will attempt
to compute the original source of external templates so that the final
linked code references the correct template source.
PR Close#40237
Now, if a source-mapping compliance test fails, the message displays both
the path to the generated file, and more helpfully the path to the expected
file.
PR Close#40237
Previously the names of the source and expectation files were often reused,
which caused potential confusion.
There is now a single source file for
each test-case, which is important when they are being compiled with different
compiler options, since the GOLDEN_PARTIAL file will only contain one copy
per file name.
The names of the expectation files have now been changed so that is clearer
which test-case they are related to.
PR Close#40237
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
The linker is implemented using a Babel transform such that Babel needs
to parse and walk a source file to find the declarations that need to be
compiled. If it can be determined that a source file is known not to
contain any declarations the parsing and walking can be skipped as a
performance improvement. This commit adds an exposed function for tools
that integrate the linker to use to allow short-circuiting of the linker
transform.
PR Close#40137
The types of directives and pipes that are used in a component's
template may be emitted into the partial declaration wrapped inside a
closure, which is needed when the type is declared later in the module.
This poses a problem for JIT compilation of partial declarations, as
this closure is indistinguishable from a class reference itself. To mark
the forward reference function as such, this commit changes the partial
declaration codegen to emit a `forwardRef` invocation wrapped around
the closure, which ensures that the closure is properly tagged as a
forward reference. This allows the forward reference to be treated as
such during JIT compilation.
PR Close#40117
Currently when `ɵɵtemplate` and `ɵɵelement` instructions are generated by compiler, all static attributes are
duplicated for both instructions. As a part of this duplication, i18n translation blocks for static i18n attributes
are generated twice as well, causing duplicate entries in extracted translation files (when Ivy extraction mechanisms
are used). This commit fixes this issue by introducing a cache for i18n translation blocks (for static attributes
only).
Also this commit further aligns `ɵɵtemplate` and `ɵɵelement` instruction attributes, which should help implement
more effective attributes deduplication logic.
Closes#39942.
PR Close#40077
Durring analysis we find template parse errors. This commit changes
where the type checking context stores the parse errors. Previously, we
stored them on the AnalysisOutput this commit changes the errors to be
stored on the TemplateData (which is a property on the shim). That way,
the template parse errors can be grouped by template.
Previously, if a template had a parse error, we poisoned the module and
would not procede to find typecheck errors. This change does not poison
modules whose template have typecheck errors, so that ngtsc can emit
typecheck errors for templates with parse errors.
Additionally, all template diagnostics are produced in the same place.
This allows requesting just the template template diagnostics or just
other types of errors.
PR Close#40026
Refactors the i18n error tests to be unit tests in ngtsc_spec.ts. There
is two reasons for doing this.
First is that the tests in compliace_old expected an expection to be be
thrown but did not fail the test if no exception was thrown. That means
that this test could miss catching a bug. It is also a big hacky to call
compile directly and expect an exception to be thrown for diagnostics.
Also, this can easily be unit tested and an end-to-end test is not
necessary since we are not making use of the goldfiles for these tests.
It is easier to maintain and less hacky to validate that we get helpful
error messages when nesting i18n sections by calling getDiagnostics
directly.
PR Close#40026
This commit temporarily excludes classes declared in .d.ts files from checks
regarding whether providers are actually injectable.
Such classes used to be ignored (on accident) because the
`TypeScriptReflectionHost.getConstructorParameters()` method did not return
constructor parameters from d.ts files, mostly as an oversight. This was
recently fixed, but caused more providers to be exposed to this check, which
created a breakage in g3.
This commit temporarily fixes the breakage by continuing to exclude such
providers from the check, until g3 can be patched.
PR Close#40118
This commit replaces `bazel` with `yarn bazel` in the error message (that instructs to regenerate golden file)
thrown while executing compliance tests. We use `yarn bazel` in other places (so we use the local version of bazel,
not the global one).
PR Close#40078
Prior to this change, the `setClassMetadata` call would be invoked
inside of an IIFE that was marked as pure. This allows the call to be
tree-shaken away in production builds, as the `setClassMetadata` call
is only present to make the original class metadata available to the
testing infrastructure. The pure marker is problematic, though, as the
`setClassMetadata` call does in fact have the side-effect of assigning
the metadata into class properties. This has worked under the assumption
that only build optimization tools perform tree-shaking, however modern
bundlers are also able to elide calls that have been marked pure so this
assumption does no longer hold. Instead, an `ngDevMode` guard is used
which still allows the call to be elided but only by tooling that is
configured to consider `ngDevMode` as constant `false` value.
PR Close#39987
These tests started failing because they had type-check
errors in their templates, and a recent commit turned on
full template type-checking by default.\
This commit fixes those templates and updates the expected
files as necessary.
PR Close#40040
These tests do not pass the typecheck phase of the compiler and fail.
The option to disable typechecking was removed recently so these tests
need to be fixed to be valid applications.
PR Close#40033
A couple reasons to justify removing the flag:
* It adds code to the compiler that is only meant to support test cases
and not any production. We should avoid code in that's only
meant to support tests.
* The flag enables writing tests that do not mimic real-world behavior
because they allow invalid applications
PR Close#40013
The golden files for the partial compliance tests need to be updated
with individual Bazel run invocations, which is not very ergonomic when
a large number of golden files need to updated. This commit adds a
script to query the Bazel targets that update the goldens and then runs
those targets sequentially.
PR Close#39989
This test migrates source-mapping tests to the new compliance test framework.
The original tests are found in the file at:
`packages/compiler-cli/test/ngtsc/template_mapping_spec.ts`.
These new tests also check the mappings resulting from partial compilation
followed by linking, after flattening the pair of source-maps that each
process generates.
Note that there are some differences between the mappings for full compile
and linked compile modes, due to how TypeScript and Babel use source-span
information on AST nodes. To accommodate this, there are two expectation
files for most of these source files.
PR Close#39939
This commit allows compliance test-cases to be written that specify
source-map mappings between the source and generated code.
To check a mapping, add a `// SOURCE:` comment to the end of a line:
```
<generated code> // SOURCE: "<source-url>" <source code>
```
The generated code will still be checked, stripped of the `// SOURCE` comment,
as normal by the `expectEmit()` helper.
In addition, the source-map segments are checked to ensure that there is a
mapping from `<generated code>` to `<source code>` found in the file at
`<source-url>`.
Note:
* The source-url should be absolute, with the directory containing the
TEST_CASES.json file assumed to be `/`.
* Whitespace is important and will be included when comparing the segments.
* There is a single space character between each part of the line.
* Newlines within a mapping must be escaped since the mapping and comment
must all appear on a single line of this file.
PR Close#39939
Previously one could set a flag in a `TEST_CASES.json` file to exclude
the test-cases from being run if the input files were being compiled
partially and then linked.
There are also scenarios where one might want to exclude test-cases
from "full compile" mode test runs.
This commit changes the compliance test tooling to support a new
property `compilationModeFilter`, which is an array containing one or
more of `"full compile"` and `"linked compile"`. Only the tests
whose `compilationModeFilter` array contains the current compilation
mode will be run.
PR Close#39939
Previously files were serialized with an extra newline seperator that
was not removed when parsing. This caused the parsed file to start with
an extra newline that invalidated its source-map.
Also, the splitting was producing an empty entry at the start of the extracted
golden files which is now ignored.
PR Close#39939
The schema accidentally included the `expectedErrors` and `extraCheck`
properties below the `files` property instead of below the `expectations`
property.
PR Close#39939
The partial compiler will add a version number to the objects that are
generated so that the linker can select the appropriate partial linker
class to process the metadata.
Previously this version matching was a simple number check. Now
the partial compilation writes the current Angular compiler version
into the generated metadata, and semantic version ranges are used
to select the appropriate partial linker.
PR Close#39847
The newly built compliance test runner was not using the shared source
file cache that was added in b627f7f02e,
which offers a significant performance boost to the compliance test
targets.
PR Close#39956
Previously, if a component had an external template with a hard error, the
compiler would "forget" the link between that component and its NgModule.
Additionally, the NgModule would be marked as being in error, because the
template issue would prevent the compiler from registering the component
class as a component, so from the NgModule it would look like a declaration
of a non-directive/pipe class. As a combined result, the next incremental
step could fix the template error, but would not refresh diagnostics for the
NgModule, leading to an incrementality issue.
The various facets of this problem were fixed in prior commits. This commit
adds a test verifying the above case works now as expected.
PR Close#39923
Previously, if a trait's analysis step resulted in diagnostics, the trait
would be considered "errored" and no further operations, including register,
would be performed. Effectively, this meant that the compiler would pretend
the class in question was actually undecorated.
However, this behavior is problematic for several reasons:
1. It leads to inaccurate diagnostics being reported downstream.
For example, if a component is put into the error state, for example due to
a template error, the NgModule which declares the component would produce a
diagnostic claiming that the declaration is neither a directive nor a pipe.
This happened because the compiler wouldn't register() the component trait,
so the component would not be recorded as actually being a directive.
2. It can cause incorrect behavior on incremental builds.
This bug is more complex, but the general issue is that if the compiler
fails to associate a component and its module, then incremental builds will
not correctly re-analyze the module when the component's template changes.
Failing to register the component as such is one link in the larger chain of
issues that result in these kinds of issues.
3. It lumps together diagnostics produced during analysis and resolve steps.
This is not causing issues currently as the dependency graph ensures the
right classes are re-analyzed when needed, instead of showing stale
diagnostics. However, the dependency graph was not intended to serve this
role, and could potentially be optimized in ways that would break this
functionality.
This commit removes the concept of an "errored" trait entirely from the
trait system. Instead, analyzed and resolved traits have corresponding (and
separate) diagnostics, in addition to potentially `null` analysis results.
Analysis (but not resolution) diagnostics are carried forward during
incremental build operations. Compilation (emit) is only performed when
a trait reaches the resolved state with no diagnostics.
This change is functionally different than before as the `register` step is
now performed even in the presence of analysis errors, as long as analysis
results are also produced. This fixes problem 1 above, and is part of the
larger solution to problem 2.
PR Close#39923
If the testcase has not specified that errors were expected, then any
errors that have occurred should be reported. These errors may have
prevented an output file from being generated, which resulted in hard
to debug test failures due to missing files.
PR Close#39862
Previously this would have just printed that `false` was not equal to
`true`, which, although true, is not very helpful. This commit adds
details about which special check failed together with the generated
code, for easier debugging.
PR Close#39863
This commit provides the machinery for the new file-based compliance test
approach for i18n tests, and migrates the i18n tests to this new format.
PR Close#39661
This commit implements partial compilation of components, together with
linking the partial declaration into its full AOT output.
This commit does not yet enable accurate source maps into external
templates. This requires additional work to account for escape sequences
which is non-trivial. Inline templates that were represented using a
string or template literal are transplated into the partial declaration
output, so their source maps should be accurate. Note, however, that
the accuracy of source maps is not currently verified in tests; this is
also left as future work.
The golden files of partial compilation output have been updated to
reflect the generated code for components. Please note that the current
output should not yet be considered stable.
PR Close#39707
In production mode this flag defaults to `true`, but the compliance
tests override this to `false` unless it is provided. As such, the
linker should also adhere to this default as otherwise the compilation
output would not align with the output of the full tests.
There are still tests that exercise the value of this flag, together
with it being `undefined` to verify the behavior of the actual default
value.
PR Close#39707
The linker does not currently support outputting ES5 syntax, so any
compliance tests that request ES5 output cannot be run in partial
compilation mode. This commit marks these tests as pending.
PR Close#39707
The JSON schema reference was off-by-one, preventing IDEs from finding
the file and offering suggestions and documentation. Additionally the
name of the golden file was slightly off.
PR Close#39707
If a template declares a reference to a missing target then referring to
that reference from elsewhere in the template would crash the template
type checker, due to a regression introduced in #38618. This commit
fixes the crash by ensuring that the invalid reference will resolve to
a variable of type any.
Fixes#39744
PR Close#39805
When the `preserveWhitespaces` is not true, the template parser will
process the parsed AST nodes to remove excess whitespace. Since the
generated `goog.getMsg()` statements rely upon the AST nodes after
this whitespace is removed, the i18n extraction must make a second pass.
Previously this resulted in innacurrate source-spans for the i18n text and
placeholder nodes that were extracted in the second pass.
This commit fixes this by reusing the source-spans from the first pass
when extracting the nodes in the second pass.
Fixes#39671
PR Close#39717
ngtsc has a robust suite of testing utilities, designed for in-memory
testing of a TypeScript compiler. Previously these utilities lived in the
`test` directory for the compiler-cli package.
This commit moves those utilities to an `ngtsc/testing` package, enabling
them to be depended on separately and opening the door for using them from
the upcoming language server testing infrastructure.
As part of this refactoring, the `fake_core` package (a lightweight API
replacement for @angular/core) is expanded to include functionality needed
for Language Service test use cases.
PR Close#39594
Currently when we encounter an implicit method call (e.g. `{{ foo(1) }}`) and we manage to resolve
its receiver to something within the template, we assume that the method is on the receiver itself
so we generate a type checking code to reflect it. This assumption is true in most cases, but it
breaks down if the call is on an implicit receiver and the receiver itself is being invoked. E.g.
```
<div *ngFor="let fn of functions">{{ fn(1) }}</div>
```
These changes resolve the issue by generating a regular function call if the method call's receiver
is pointing to `$implicit`.
Fixes#39634.
PR Close#39686
In order to more accurately map from a node in the TCB to a template position,
we need to provide more span information in the TCB. These changes are necessary
for the Language Service to map from a TCB node back to a specific
locations in the template for actions like "find references" and
"refactor/rename". After the TS "find references" returns results,
including those in the TCB, we need to map specifically to the matching
key/value spans in the template rather than the entire source span.
This also has the benefit of producing diagnostics which align more
closely with what TypeScript produces.
The following example shows TS code and the diagnostic produced by an invalid assignment to a property:
```
let a: {age: number} = {} as any;
a.age = 'laksjdf';
^^^^^ <-- Type 'string' is not assignable to type 'number'.
```
A corollary to this in a template file would be [age]="'someString'". The diagnostic we currently produce for this is:
```
Type 'number' is not assignable to type 'string'.
1 <app-hello [greeting]="1"></app-hello>
~~~~~~~~~~~~~~
```
Notice that the underlined text includes the entire span.
If we included the keySpan for the assignment to the property,
this diagnostic underline would be more similar to the one produced by TypeScript;
that is, it would only underline “greeting”.
[design/discussion doc]
(https://docs.google.com/document/d/1FtaHdVL805wKe4E6FxVTnVHl38lICoHIjS2nThtRJ6I/edit?usp=sharing)
PR Close#39665
ngtsc will avoid emitting generated imports that would create an import
cycle in the user's program. The main way such imports can arise is when
a component would ordinarily reference its dependencies in its component
definition `directiveDefs` and `pipeDefs`. This requires adding imports,
which run the risk of creating a cycle.
When ngtsc detects that adding such an import would cause this to occur, it
instead falls back on a strategy called "remote scoping", where a side-
effectful call to `setComponentScope` in the component's NgModule file is
used to patch `directiveDefs` and `pipeDefs` onto the component. Since the
NgModule file already imports all of the component's dependencies (to
declare them in the NgModule), this approach does not risk adding a cycle.
It has several large downsides, however:
1. it breaks under `sideEffects: false` logic in bundlers including the CLI
2. it breaks tree-shaking for the given component and its dependencies
See this doc for further details: https://hackmd.io/Odw80D0pR6yfsOjg_7XCJg?view
In particular, the impact on tree-shaking was exacerbated by the naive logic
ngtsc used to employ here. When this feature was implemented, at the time of
generating the side-effectful `setComponentScope` call, the compiler did not
know which of the component's declared dependencies were actually used in
its template. This meant that unlike the generation of `directiveDefs` in
the component definition itself, `setComponentScope` calls had to list the
_entire_ compilation scope of the component's NgModule, including directives
and pipes which were not actually used in the template. This made the tree-
shaking impact much worse, since if the component's NgModule made use of any
shared NgModules (e.g. `CommonModule`), every declaration therein would
become un-treeshakable.
Today, ngtsc does have the information on which directives/pipes are
actually used in the template, but this was not being used during the remote
scoping operation. This commit modifies remote scoping to take advantage of
the extra context and only list used dependencies in `setComponentScope`
calls, which should ameliorate the tree-shaking impact somewhat.
PR Close#39662
This commit adds bazel rules to test whether linking the golden partial
files for test cases produces the same output as a full compile of the
test case would.
PR Close#39617
This commit contains the basic runner logic and a couple of sample test cases
for the "full compile" compliance tests, where source files are compiled
to full definitions and checked against expectations.
PR Close#39617
This commit renames the original `compliance` test directory to `compliance_old`.
Eventually this directory will be deleted once all the tests have been
migrated to the new test case based compliance tests.
PR Close#39617
There is a compiler transform that downlevels Angular class decorators
to static properties so that metadata is available for JIT compilation.
The transform was supposed to ignore non-Angular decorators but it was
actually completely dropping decorators that did not conform to a very
specific syntactic shape (i.e. the decorator was a simple identifier, or
a namespaced identifier).
This commit ensures that all non-Angular decorators are kepts as-is
even if they are built using a syntax that the Angular compiler does not
understand.
Fixes#39574
PR Close#39577
In PR #38938 an additional Bazel target was introduced for the compliance
tests, as preparation to run the compliance tests in partial compilation
mode and then apply the linker transform. The linker plugin itself was
not available at the time but has since been implemented, so this commit
updates the prelink target of the compliance tests to apply the linker
transform using the Babel plugin.
Actually emitting partial compilations to be transformed will be done in
follow-up work.
PR Close#39518
When a class with a custom decorator is transpiled to ES5, it looks something like this:
```
var SomeClass = (function() {
function SomeClass() {...};
var SomeClass_1 = __decorate([Decorator()], SomeClass);
SomeClass = SomeClass_1;
return SomeClass;
})();
```
The problem is that if the class also has an Angular decorator that refers to the class itself
(e.g. `{provide: someToken, useClass: SomeClass}`), the generated `setClassMetadata` code will
be emitted after the IIFE, but will still refer to the intermediate `SomeClass_1` variable from
inside the IIFE. This happens, because we generate the `setClassMetadata` call directly from
the source AST which contains identifiers that TS will rename when it emits the ES5 code.
These changes resolve the issue by looking through the metadata AST and cloning any `Identifier`
that is referring to the class. Since TS doesn't have references to the clone, it won't rename
it when transpiling to ES5.
Fixes#39509.
PR Close#39527
Currently expressions `$event.foo()` and `this.$event.foo()`, as well as `$any(foo)` and
`this.$any(foo)`, are treated as the same expression by the compiler, because `this` is considered
the same implicit receiver as when the receiver is omitted. This introduces the following issues:
1. Any time something called `$any` is used, it'll be stripped away, leaving only the first parameter.
2. If something called `$event` is used anywhere in a template, it'll be preserved as `$event`,
rather than being rewritten to `ctx.$event`, causing the value to undefined at runtime. This
applies to listener, property and text bindings.
These changes resolve the first issue and part of the second one by preserving anything that
is accessed through `this`, even if it's one of the "special" ones like `$any` or `$event`.
Furthermore, these changes only expose the `$event` global variable inside event listeners,
whereas previously it was available everywhere.
Fixes#30278.
PR Close#39323
Currently `i18n` attributes are treated the same no matter if they have data bindings or not. This
both generates more code since they have to go through the `ɵɵi18nAttributes` instruction and
prevents the translated attributes from being injected using the `@Attribute` decorator.
These changes makes it so that static translated attributes are treated in the same way as regular
static attributes and all other `i18n` attributes go through the old code path.
Fixes#38231.
PR Close#39408
Previously directive "queries" that relied upon a namespaced type
```ts
queries: {
'mcontent': new core.ContentChild('test2'),
}
```
caused an error to be thrown. This is now supported.
PR Close#38959
Previously the `ConcreteDeclaration` and `InlineDeclaration` had
different properties for the underlying node type. And the `InlineDeclaration`
did not store a value that represented its declaration.
It turns out that a natural declaration node for an inline type is the
expression. For example in UMD/CommonJS this would be the `exports.<name>`
property access node.
So this expression is now used for the `node` of `InlineDeclaration` types
and the `expression` property is dropped.
To support this the codebase has been refactored to use a new `DeclarationNode`
type which is a union of `ts.Declaration|ts.Expression` instead of `ts.Declaration`
throughout.
PR Close#38959
Removes `ViewEncapsulation.Native` which has been deprecated for several major versions.
BREAKING CHANGES:
* `ViewEncapsulation.Native` has been removed. Use `ViewEncapsulation.ShadowDom` instead. Existing
usages will be updated automatically by `ng update`.
PR Close#38882
To verify the correctness of the linker output, we leverage the existing
compliance tests. The plan is to test the linker by running all compliance
tests using a full round trip of pre-linking and subsequently post-linking,
where the generated code should be identical to a full AOT compile.
This commit adds an additional Bazel target that runs the compliance
tests in partial mode. Follow-up work is required to implement the logic
for running the linker round trip.
PR Close#38938
With the introduction of incremental type checking in #36211, an
intermediate `ts.Program` for type checking is only created if there are
any templates to check. This rendered some tests ineffective at avoiding
regressions, as the intermediate `ts.Program` was required for the tests
to fail if the scenario under test would not be accounted for. This
commit adds a single component to these tests, to ensure the
intermediate `ts.Program` is in fact created.
PR Close#39011
Prior to this fix, incremental rebuilds could fail to type check due to
missing ambient types from auto-discovered declaration files in @types
directories, or type roots in general. This was caused by the
intermediary `ts.Program` that is created for template type checking,
for which a `ts.CompilerHost` was used which did not implement the
optional `directoryExists` methods. As a result, auto-discovery of types
would not be working correctly, and this would retain into the
`ts.Program` that would be created for an incremental rebuild.
This commit fixes the issue by forcing the custom `ts.CompilerHost` used
for type checking to properly delegate into the original
`ts.CompilerHost`, even for optional methods. This is accomplished using
a base class `DelegatingCompilerHost` which is typed in such a way that
newly introduced `ts.CompilerHost` methods must be accounted for.
Fixes#38979
PR Close#39011
We weren't resolving a path correctly which resulted in an error on Windows.
For reference, here's the error. Note the extra slash before `C:`:
```
Error: ENOENT: no such file or directory, scandir '/C:/bazel_output_root/yxvwd24o/external/npm/node_modules/typescript'
at Object.readdirSync (fs.js:854:3)
```
PR Close#39005
Prior to this change, each invocation of `loadStandardTestFiles` would
load the necessary files from disk. This function is typically called
at the top-level of a test module in order to share the result across
tests. The `//packages/compiler-cli/test/ngtsc` target has 8 modules
where this call occurs, each loading their own copy of
`node_modules/typescript` which is ~60MB in size, so the memory overhead
used to be significant. This commit loads the individual packages into
a standalone `Folder` and mounts this folder into the filesystem of
standard test files, such that all file contents are no longer
duplicated in memory.
PR Close#38909
Some compiler tests take a long time to run, even using multiple
executors. A profiling session revealed that most time is spent in
parsing source files, especially the default libraries are expensive to
parse.
The default library files are constant across all tests, so this commit
introduces a shared cache of parsed source files of the default
libraries. This achieves a significant improvement for several targets
on my machine:
//packages/compiler-cli/test/compliance: from 23s to 5s.
//packages/compiler-cli/test/ngtsc: from 115s to 11s.
Note that the number of shards for the compliance tests has been halved,
as the extra shards no longer provide any speedup.
PR Close#38909
In Ivy, template type-checking has 3 modes: basic, full, and strict. The
primary difference between basic and full modes is that basic mode only
checks the top-level template, whereas full mode descends into nested
templates (embedded views like ngIfs and ngFors). Ivy applies this approach
to all of its template type-checking, including the DOM schema checks which
validate whether an element is a valid component/directive or not.
View Engine has both the basic and the full mode, with the same distinction.
However in View Engine, DOM schema checks happen for the full template even
in the basic mode.
Ivy's behavior here is technically a "fix" as it does not make sense for
some checks to apply to the full template and others only to the top-level
view. However, since g3 relies exclusively on the basic mode of checking and
developers there are used to DOM checks applying throughout their template,
this commit re-enables the nested schema checks even in basic mode only in
g3. This is done by enabling the checks only when Closure Compiler
annotations are requested.
Outside of g3, it's recommended that applications use at least the full mode
of checking (controlled by the `fullTemplateTypeCheck` flag), and ideally
the strict mode (`strictTemplates`).
PR Close#38943
When the target of the compiler is ES2015 or newer then we should
be generating `let` and `const` variable declarations rather than `var`.
PR Close#38775
Common AST formats such as TS and Babel do not use a separate
node for comments, but instead attach comments to other AST nodes.
Previously this was worked around in TS by creating a `NotEmittedStatement`
AST node to attach the comment to. But Babel does not have this facility,
so it will not be a viable approach for the linker.
This commit refactors the output AST, to remove the `CommentStmt` and
`JSDocCommentStmt` nodes. Instead statements have a collection of
`leadingComments` that are rendered/attached to the final AST nodes
when being translated or printed.
PR Close#38811
The type-to-value conversion could previously crash if a symbol was
resolved that does not have any declarations, e.g. because it's imported
from a missing module. This would typically result in a semantic
TypeScript diagnostic and halt further compilation, therefore not
reaching the type-to-value conversion logic. In Bazel however, it turns
out that Angular semantic diagnostics are requested even if there are
semantic TypeScript errors in the program, so it would then reach the
type-to-value conversation and crash.
This commit fixes the unsafe access and adds a test that ignores the
TypeScript semantic error, effectively replicating the situation as
experienced under Bazel.
Fixes#38670
PR Close#38684
Previously, localized strings had very limited or incorrect source-mapping
information available.
Now the i18n AST nodes and related output AST nodes include source-span
information about message-parts and placeholders - including closing tag
placeholders.
This information is then used when generating the final localized string
ASTs to ensure that the correct source-mapping is rendered.
See #38588 (comment)
PR Close#38645
Previously, the compiler was not able to display template parsing errors as
true `ts.Diagnostic`s that point inside the template. Instead, it would
throw an actual `Error`, and "crash" with a stack trace containing the
template errors.
Not only is this a poor user experience, but it causes the Language Service
to also crash as the user is editing a template (in actuality the LS has to
work around this bug).
With this commit, such parsing errors are converted to true template
diagnostics with appropriate span information to be displayed contextually
along with all other diagnostics. This majorly improves the user experience
and unblocks the Language Service from having to deal with the compiler
"crashing" to report errors.
PR Close#38576
Previously, the `sourceSpan` and `startSourceSpan` were the same
object, which meant that you had the following situation:
```
element = <div>some content</div>
sourceSpan = <div>
startSourceSpan = <div>
endSourceSpan = </div>
```
This made `sourceSpan` redundant and meant that if you
wanted a span for the whole element including its content
and closing tag, it had to be computed.
Now `sourceSpan` is separated from `startSourceSpan`
resulting in:
```
element = <div>some content</div>
sourceSpan = <div>some content</div>
startSourceSpan = <div>
endSourceSpan = </div>
```
PR Close#38581
The HTML parser gets an element's namespace either from the tag name
(e.g. `<svg:rect>`) or from its parent element `<svg><rect></svg>`) which
breaks down when an element is inside of an SVG `foreignElement`,
because foreign elements allow nodes from a different namespace to be
inserted into an SVG.
These changes add another flag to the tag definitions which tells child
nodes whether to try to inherit their namespaces from their parents.
It also adds a definition for `foreignObject` with the new flag,
allowing elements placed inside it to infer their namespaces instead.
Fixes#37218.
PR Close#38477
We had a couple of places where we were assuming that if a particular
symbol has a value, then it will exist at runtime. This is true in most cases,
but it breaks down for `const` enums.
Fixes#38513.
PR Close#38542
This commit updates the code to move generated i18n statements into the `consts` field of
ComponentDef to avoid invoking `$localize` function before component initialization (to better
support runtime translations) and also avoid problems with lazy-loading when i18n defs may not
be present in a chunk where it's referenced.
Prior to this change the i18n statements were generated at the top leve:
```
var I18N_0;
if (typeof ngI18nClosureMode !== "undefined" && ngI18nClosureMode) {
var MSG_X = goog.getMsg(“…”);
I18N_0 = MSG_X;
} else {
I18N_0 = $localize('...');
}
defineComponent({
// ...
template: function App_Template(rf, ctx) {
i0.ɵɵi18n(2, I18N_0);
}
});
```
This commit updates the logic to generate the following code instead:
```
defineComponent({
// ...
consts: function() {
var I18N_0;
if (typeof ngI18nClosureMode !== "undefined" && ngI18nClosureMode) {
var MSG_X = goog.getMsg(“…”);
I18N_0 = MSG_X;
} else {
I18N_0 = $localize('...');
}
return [
I18N_0
];
},
template: function App_Template(rf, ctx) {
i0.ɵɵi18n(2, 0);
}
});
```
Also note that i18n template instructions now refer to the `consts` array using an index
(similar to other template instructions).
PR Close#38404
The compiler does not currently report errors when there's an `@Input()`
for a `private`, `protected`, or `readonly` directive/component class member.
This change adds an option to enable reporting errors when a template
attempts to bind to one of these restricted input fields.
PR Close#38249
Prior to this change, the template type checker would always use a
type-constructor to instantiate a directive. This type-constructor call
serves two purposes:
1. Infer any generic types for the directive instance from the inputs
that are passed in.
2. Type check the inputs that are passed into the directive's inputs.
The first purpose is only relevant when the directive actually has any
generic types and using a type-constructor for these cases inhibits
a type-check performance penalty, as a type-constructor's signature is
quite complex and needs to be generated for each directive.
This commit refactors the generated type-check blocks to only generate
a type-constructor call for directives that have generic types. Type
checking of inputs is achieved by generating individual statements for
all inputs, using assignments into the directive's fields.
Even if a type-constructor is used for type-inference of generic types
will the input checking also be achieved using the individual assignment
statements. This is done to support the rework of the language service,
which will start to extract symbol information from the type-check
blocks.
As a future optimization, it may be possible to reduce the number of
inputs passed into a type-constructor to only those inputs that
contribute the the type-inference of the generics. As this is not a
necessity at the moment this is left as follow-up work.
Closes#38185
PR Close#38249
In TypeScript 3.8 support was added for type-only imports, which only brings in
the symbol as a type, not their value. The Angular compiler did not yet take
the type-only keyword into account when representing symbols in type positions
as value expressions. The class metadata that the compiler emits would include
the value expression for its parameter types, generating actual imports as
necessary. For type-only imports this should not be done, as it introduces an
actual import of the module that was originally just a type-only import.
This commit lets the compiler deal with type-only imports specially, preventing
a value expression from being created.
Fixes#37900
PR Close#37912
When using the safe navigation operator in a binding expression, a temporary
variable may be used for storing the result of a side-effectful call.
For example, the following template uses a pipe and a safe property access:
```html
<app-person-view [enabled]="enabled" [firstName]="(person$ | async)?.name"></app-person-view>
```
The result of the pipe evaluation is stored in a temporary to be able to check
whether it is present. The temporary variable needs to be declared in a separate
statement and this would also cause the full expression itself to be pulled out
into a separate statement. This would compile into the following
pseudo-code instructions:
```js
var temp = null;
var firstName = (temp = pipe('async', ctx.person$)) == null ? null : temp.name;
property('enabled', ctx.enabled)('firstName', firstName);
```
Notice that the pipe evaluation happens before evaluating the `enabled` binding,
such that the runtime's internal binding index would correspond with `enabled`,
not `firstName`. This introduces a problem when the pipe uses `WrappedValue` to
force a change to be detected, as the runtime would then mark the binding slot
corresponding with `enabled` as dirty, instead of `firstName`. This results
in the `enabled` binding to be updated, triggering setters and affecting how
`OnChanges` is called.
In the pseudo-code above, the intermediate `firstName` variable is not strictly
necessary---it only improved readability a bit---and emitting it inline with
the binding itself avoids the out-of-order execution of the pipe:
```js
var temp = null;
property('enabled', ctx.enabled)
('firstName', (temp = pipe('async', ctx.person$)) == null ? null : temp.name);
```
This commit introduces a new `BindingForm` that results in the above code to be
generated and adds compiler and acceptance tests to verify the proper behavior.
Fixes#37194
PR Close#37911
For attribute bindings that target a directive's input, the template
type checker is able to verify that the type of the input expression is
compatible with the directive's declaration for said input. This
checking adheres to the `strictNullChecks` flag as configured in the
TypeScript compilation, such that errors are reported for expressions
that include `undefined` or `null` in their type if the input's
declaration does not include those types.
There was a bug with this level of type-checking for directives that
also declare coercion members, where binding an expression that includes
the `undefined` type to a directive's input that does not include the
`undefined` type would not be reported as error.
This commit fixes the bug by changing the type-constructor in type-check
code to use an intersection type of regular inputs and coerced inputs,
instead of a union type. The union type would inadvertently allow
`undefined` types to be assigned into the regular inputs, as that would
still satisfy the characteristics of a union type.
As a result of this change, you may start to see build failures if
`strictTemplates` is enabled and `strictInputTypes` is not disabled.
These errors are legitimate and some action is required to achieve a
successful build:
1. Update the templates for which an error is reported and introduce the
non-null assertion operator at the end of the expression. This
removes the `undefined` type from the expression's type, making it
appear as a valid assignment.
2. Disable `strictNullInputTypes` in the compiler options. This will
implicitly add the non-null assertion operators similar to option 1,
but all templates in the compilation are affected.
3. Update the directive's input declaration to include the `undefined`
type, if the directive is not implemented in an external library.
PR Close#38273
Roll forward of #38147.
This allows Closure compiler to tree shake unused constructor calls to `NgModuleFactory`, which is otherwise considered
side-effectful. The Angular compiler generates factory objects which are exported but typically not used, as they are
only needed for compatibility with View Engine. This results in top-level constructor calls, such as:
```typescript
export const FooNgFactory = new NgModuleFactory(Foo);
```
`NgModuleFactory` has a side-effecting constructor, so this statement cannot be tree shaken, even if `FooNgFactory` is
never imported. The `NgModuleFactory` continues to reference its associated `NgModule` and prevents the module and all
its unused dependencies from being tree shaken, making Closure builds significantly larger than necessary.
The fix here is to wrap `NgModuleFactory` constructor with `noSideEffects(() => /* ... */)`, which tricks the Closure
compiler into assuming that the invoked function has no side effects. This allows it to tree-shake unused
`NgModuleFactory()` constructors when they aren't imported. Since the factory can be removed, the module can also be
removed (if nothing else references it), thus tree shaking unused dependencies as expected.
The one notable edge case is for lazy loaded modules. Internally, lazy loading is done as a side effect when the lazy
script is evaluated. For Angular, this side effect is registering the `NgModule`. In Ivy this is done by the
`NgModuleFactory` constructor, so lazy loaded modules **cannot** have their top-level `NgModuleFactory` constructor
call tree shaken. We handle this case by looking for the `id` field on `@NgModule` annotations. All lazy loaded modules
include an `id`. When this `id` is found, the `NgModuleFactory` is generated **without** with `noSideEffects()` call,
so Closure will not tree shake it and the module will lazy-load correctly.
PR Close#38320
Currently the `getInheritedFactory` function is implemented to allow
closure to remove the call if the base factory is unused. However, this
method does not work with terser. By adding the PURE annotation,
terser will also be able to remove the call when unused.
PR Close#38291
This reverts commit 7f8c2225f2.
This commit caused test failures internally, which were traced back to the
optimizer removing NgModuleFactory constructor calls when those calls caused
side-effectful registration of NgModules by their ids.
PR Close#38303
This allows Closure compiler to tree shake unused constructor calls to `NgModuleFactory`, which is otherwise considered
side-effectful. The Angular compiler generates factory objects which are exported but typically not used, as they are
only needed for compatibility with View Engine. This results in top-level constructor calls, such as:
```typescript
export const FooNgFactory = new NgModuleFactory(Foo);
```
`NgModuleFactory` has a side-effecting constructor, so this statement cannot be tree shaken, even if `FooNgFactory` is
never imported. The `NgModuleFactory` continues to reference its associated `NgModule` and prevents the module and all
its unused dependencies from being tree shaken. This effectively prevents all components from being tree shaken, making
Closure builds significantly larger than they should be.
The fix here is to wrap `NgModuleFactory` constructor with `noSideEffects(() => /* ... */)`, which tricks the Closure
compiler into assuming that the invoked function has no side effects. This allows it to tree-shake unused
`NgModuleFactory()` constructors when they aren't imported. Since the factory can be removed, the module can also be
removed (if nothing else references it), thus tree shaking unused components as expected.
PR Close#38147
JoostK
Pete Bacon Darwin
Kristiyan Kostadinov
Bjarki
Andrew Kushnir
Zach Arend
Alex Rickabaugh
Andrew Scott
Charles Lyding
Alan Agius
Andrew Scott
Doug Parker