Commit 4213e8d5 introduced shim reference tagging into the compiler, and
changed how the `TypeCheckProgramHost` worked under the hood during the
creation of a template type-checking program. This work enabled a more
incremental flow for template type-checking, but unintentionally introduced
several regressions in performance, caused by poor incrementality during
`ts.Program` creation.
1. The `TypeCheckProgramHost` was made to rely on the `ts.CompilerHost` to
retrieve instances of `ts.SourceFile`s from the original program. If the
host does not return the original instance of such files, but instead
creates new instances, this has two negative effects: it incurs
additional parsing time, and it interferes with TypeScript's ability to
reuse information about such files.
2. During the incremental creation of a `ts.Program`, TypeScript compares
the `referencedFiles` of `ts.SourceFile` instances from the old program
with those in the new program. If these arrays differ, TypeScript cannot
fully reuse the old program. The implementation of reference tagging
introduced in 4213e8d5 restores the original `referencedFiles` array
after a `ts.Program` is created, which means that future incremental
operations involving that program will always fail this comparison,
effectively limiting the incrementality TypeScript can achieve.
Problem 1 exacerbates problem 2: if a new `ts.SourceFile` is created by the
host after shim generation has been disabled, it will have an untagged
`referencedFiles` array even if the original file's `referencedFiles` was
not restored, triggering problem 2 when creating the template type-checking
program.
To fix these issues, `referencedFiles` arrays are now restored on the old
`ts.Program` prior to the creation of a new incremental program. This allows
TypeScript to get the most out of reusing the old program's data.
Additionally, the `TypeCheckProgramHost` now uses the original `ts.Program`
to retrieve original instances of `ts.SourceFile`s where possible,
preventing issues when a host would otherwise return fresh instances.
Together, these fixes ensure that program reuse is as incremental as
possible, and tests have been added to verify this for certain scenarios.
An optimization was further added to prevent the creation of a type-checking
`ts.Program` in the first place if no type-checking is necessary.
PR Close#37641
Previously the `ProgramBasedEntryPointFinder` was parsing all the
entry-points referenced by the program for dependencies even if all the
entry-points had been processed already.
Now this entry-point finder will re-use the `EntryPointManifest` to load
the entry-point dependencies when possible which avoids having to parse
them all again, on every invocation of ngcc.
Previously the `EntryPointManifest` was only used in the
`DirectoryWalkerEntryPointFinder`, which also contained the logic for
computing the contents of the manifest. This logic has been factored out
into an `EntryPointCollector` class. Both the `ProgramBasedEntryPointFinder`
and `DirectoryWalkerEntryPointFinder` now use the `EntryPointManifest` and
the `EntryPointCollector`.
The result of this change is that there is a small cost on the first run of
ngcc to compute and store the manifest - the processing takes 102% of the
processing time before this PR. But on subsequent runs there is a
significant benefit on subsequent runs - the processing takes around 50%
of the processing time before this PR.
PR Close#37665
Source-maps can be linked to from a source-file by a comment at
the end of the file.
Previously the `SourceFileLoader` would read
the first comment that matched `//# sourceMappingURL=` but
this is not valid since some bundlers may include embedded
source-files that contain such a comment.
Now we only look for this comment in the last non-empty line
in the file.
PR Close#32912
Previously localized strings were not mapped to their original
source location, so it was not possible to back-trace them
in tools like the i18n message extractor.
PR Close#32912
Webpack and other build tools sometimes inline the contents of the
source files in their generated source-maps, and at the same time
change the paths to be prefixed with a protocol, such as `webpack://`.
This can confuse tools that need to read these paths, so now it is
possible to provide a mapping to where these files originated.
PR Close#32912
This method will allow us to find the original location given a
generated location, which is useful in fine grained work with
source-mapping. E.g. in `$localize` tooling.
PR Close#32912
The file-writing error in the this commit can also be the result
of the ngcc process dying in the middle of writing files.
This commit improves the error message to offer a resolution
in case this is the reason for the error.
Fixes#36393
PR Close#37672
The `SourceFile` and associated code is general and reusable in
other projects (such as `@angular/localize`). Moving it to `ngtsc`
makes it more easily shared.
PR Close#37114
The `Logger` interface and its related classes are general purpose
and could be used by other tooling. Moving it into ngtsc is a more
suitable place from which to share it - similar to the FileSystem stuff.
PR Close#37114
This dependency host tokenizes files to identify all the imported
paths. This commit calculates the last place in the source code
where there can be an import path; it then exits the tokenization
when we get to this point in the file.
Testing with a reasonably large project showed that the tokenizer
spends about 2/3 as much time scanning files. For example in a
"noop" hot run of ngcc using the program-based entry-point
finder the percentage of time spent in the `scan()` function of
the TS tokenizer goes down from 9.9% to 6.6%.
PR Close#37639
We recently added a transformer to NGC that is responsible for downleveling Angular
decorators and constructor parameter types. The primary goal was to mitigate a
TypeScript limitation/issue that surfaces in Angular projects due to the heavy
reliance on type metadata being captured for DI. Additionally this is a pre-requisite
of making `tsickle` optional in the Angular bazel toolchain.
See: 401ef71ae5 for more context on this.
Another (less important) goal was to make sure that the CLI can re-use
this transformer for its JIT mode compilation. The CLI (as outlined in
the commit mentioned above), already has a transformer for downleveling
constructor parameters. We want to avoid this duplication and exported
the transform through the tooling-private compiler entry-point.
Early experiments in using this transformer over the current one, highlighted
that in JIT, class decorators cannot be downleveled. Angular relies on those
to be invoked immediately for JIT (so that factories etc. are generated upon loading)
The transformer we exposed, always downlevels such class decorators
though, so that would break CLI's JIT mode. We can address the CLI's
needs by adding another flag to skip class decorators. This will allow
us to continue with the goal of de-duplication.
PR Close#37545
Commit 24b2f1da2b introduced an `NgCompiler` which operates on a
`ts.Program` independently of the `NgtscProgram`. The NgCompiler got its
`IncrementalDriver` (for incremental reuse of Angular compilation results)
by looking at a monkey-patched property on the `ts.Program`.
This monkey-patching operation causes problems with the Angular indexer
(specifically, it seems to cause the indexer to retain too much of prior
programs, resulting in OOM issues). To work around this, `IncrementalDriver`
reuse is now handled by a dedicated `IncrementalBuildStrategy`. One
implementation of this interface is used by the `NgtscProgram` to perform
the old-style reuse, relying on the previous instance of `NgtscProgram`
instead of monkey-patching. Only for `NgTscPlugin` is the monkey-patching
strategy used, as the plugin sits behind an interface which only provides
access to the `ts.Program`, not a prior instance of the plugin.
PR Close#37339
Currently the partial evaluator isn't able to resolve a variable declaration that uses destructuring in the form of `const {value} = {value: 0}; const foo = value;`. These changes add some logic to allow for us to resolve the variable's value.
Fixes#36917.
PR Close#37497
Previously, ngcc would only be able to match an ngcc configuration to
packages that were located inside the project's top-level
`node_modules/`. However, if there are multiple versions of a package in
a project (e.g. as a transitive dependency of other packages), multiple
copies of a package (at different versions) may exist in nested
`node_modules/` directories. For example, one at
`<project-root>/node_modules/some-package/` and one at
`<project-root>/node_modules/other-package/node_modules/some-package/`.
In such cases, ngcc was only able to detect the config for the first
copy but not for the second.
This commit fixes this by returning a new instance of
`ProcessedNgccPackageConfig` for each different package path (even if
they refer to the same package name). In these
`ProcessedNgccPackageConfig`, the `entryPoints` paths have been
processed to take the package path into account.
PR Close#37040
This commit adds a `packageName` property to the `EntryPoint` interface.
In a subsequent commit this will be used to retrieve the correct ngcc
configuration for each package, regardless of its path.
PR Close#37040
In order to retrieve the ngcc configuration (if any) for an entry-point,
ngcc has to detect the containing package's version.
Previously, ngcc would try to read the version from the entry-point's
`package.json` file, which was different than the package's top-level
`package.json` for secondary entry-points. For example, it would try to
read it from `node_modules/@angular/common/http/package.json` for
entry-point `@angular/common/http`. However, the `package.json` files
for secondary entry-points are not guaranteed to include a `version`
property.
This commit fixes this by first trying to read the version from the
_package's_ `package.json` (falling back to the entry-point's
`package.json`). For example, it will first try to read it from
`@angular/common/package.json` for entry-point `@angular/common/http`.
PR Close#37040
This commit refactors the way info is retrieved from entry-point
`package.json` files to make it easier to extract more info (such as the
package's name) in the future. It also avoids reading and parsing the
`package.json` file multiple times (as was happening before).
PR Close#37040
Rename the `package` property to `packagePath` on the `EntryPoint`
interface. This makes it more clear that the `packagePath` property
holds the absolute path to the containing package (similar to how `path`
holds the path to the entry-point). This will also align with the
`packageName` property that will be added in a subsequent commit.
This commit also re-orders the `EntryPoint` properties to group related
properties together and to match the order of properties on instances
with that on the interface.
PR Close#37040
Previously, when an entry-point was ignored via an ngcc config, ngcc
would scan sub-directories for sub-entry-points, but would not use the
correct `packagePath`. For example, if `@angular/common` was ignored, it
would look at `@angular/common/http` but incorrectly use
`.../@angular/common/http` as the `packagePath` (instead of
`.../@angular/common`). As a result, it would not retrieve the correct
ngcc config for the actual package.
This commit fixes it by ensuring the correct `packagePath` is used, even
if the primary entry-point corresponding to that path is ignored. In
order to do this, a new return value for `getEntryPointInfo()` is added:
`IGNORED_ENTRY_POINT`. This is used to differentiate between directories
that correspond to no or an incompatible entry-point and those that
correspond to an entry-point that could otherwise be valid but is
explicitly ignored. Consumers of `getEntryPointInfo()` can then use this
info to discard ignored entry-points, but still use the correct
`packagePath` when scanning their sub-directories for secondary
entry-points.
PR Close#37040
Currently Angular internally already handles `InjectionToken` as
predicates for queries. This commit exposes this as public API as
developers already relied on this functionality but currently use
workarounds to satisfy the type constraints (e.g. `as any`).
We intend to make this public as it's low-effort to support, and
it's a significant key part for the use of light-weight tokens as
described in the upcoming guide: https://github.com/angular/angular/pull/36144.
In concrete, applications might use injection tokens over classes
for both optional DI and queries, because otherwise such references
cause classes to be always retained. This was also an issue in View
Engine, but now with Ivy, this pattern became worse, as factories are
directly attached to retained classes (ultimately ending up in the
production bundle, while being unused).
More details in the light-weight token guide and in: https://github.com/angular/angular-cli/issues/16866.
Closes#21152. Related to #36144.
PR Close#37506
Adds @nocollapse to static properties added by ngcc
iff annotateForClosureCompiler is true.
The Closure Compiler will collapse static properties
into the global namespace. Adding this annotation keeps
the properties attached to their respective object, which
allows them to be referenced via a class's constructor.
The annotation is already added by ngtsc and ngc under the
same option, this commit extends the functionality to ngcc.
Closes#36618.
PR Close#36652
It is quite common for the TS compiler to have to add synthetic
types to function signatures, where the developer has not
explicitly provided them. This results in `import(...)` expressions
appearing in typings files. For example in `@ngrx/data` there is a
class with a getter that has an implicit type:
```ts
export declare class EntityCollectionServiceBase<...> {
...
get store() {
return this.dispatcher.store;
}
...
}
```
In the d.ts file for this we get:
```ts
get store(): Store<import("@ngrx/data").EntityCache>;
```
Given that this file is within the `@ngrx/data` package already,
this caused ngcc to believe that there was a circular dependency,
causing it to fail to process the package - and in fact crash!
This commit resolves this problem by ignoring `import()` expressions
when scanning typings programs for dependencies. This ability was
only introduced very recently in a 10.0.0 RC release, and so it has
limited benefit given that up till now ngcc has been able to process
libraries effectively without it. Moreover, in the rare case that a
package does have such a dependency, it should get picked up
by the sync ngcc+CLI integration point.
PR Close#37503
In v7 of Angular we removed `tsickle` from the default `ngc` pipeline.
This had the negative potential of breaking ES2015 output and SSR due
to a limitation in TypeScript.
TypeScript by default preserves type information for decorated constructor
parameters when `emitDecoratorMetadata` is enabled. For example,
consider this snippet below:
```
@Directive()
export class MyDirective {
constructor(button: MyButton) {}
}
export class MyButton {}
```
TypeScript would generate metadata for the `MyDirective` class it has
a decorator applied. This metadata would be needed in JIT mode, or
for libraries that provide `MyDirective` through NPM. The metadata would
look as followed:
```
let MyDirective = class MyDir {}
MyDirective = __decorate([
Directive(),
__metadata("design:paramtypes", [MyButton]),
], MyDirective);
let MyButton = class MyButton {}
```
Notice that TypeScript generated calls to `__decorate` and
`__metadata`. These calls are needed so that the Angular compiler
is able to determine whether `MyDirective` is actually an directive,
and what types are needed for dependency injection.
The limitation surfaces in this concrete example because `MyButton`
is declared after the `__metadata(..)` call, while `__metadata`
actually directly references `MyButton`. This is illegal though because
`MyButton` has not been declared at this point. This is due to the
so-called temporal dead zone in JavaScript. Errors like followed will
be reported at runtime when such file/code evaluates:
```
Uncaught ReferenceError: Cannot access 'MyButton' before initialization
```
As noted, this is a TypeScript limitation because ideally TypeScript
shouldn't evaluate `__metadata`/reference `MyButton` immediately.
Instead, it should defer the reference until `MyButton` is actually
declared. This limitation will not be fixed by the TypeScript team
though because it's a limitation as per current design and they will
only revisit this once the tc39 decorator proposal is finalized
(currently stage-2 at time of writing).
Given this wontfix on the TypeScript side, and our heavy reliance on
this metadata in libraries (and for JIT mode), we intend to fix this
from within the Angular compiler by downleveling decorators to static
properties that don't need to evaluate directly. For example:
```
MyDirective.ctorParameters = () => [MyButton];
```
With this snippet above, `MyButton` is not referenced directly. Only
lazily when the Angular runtime needs it. This mitigates the temporal
dead zone issue caused by a limitation in TypeScript's decorator
metadata output. See: https://github.com/microsoft/TypeScript/issues/27519.
In the past (as noted; before version 7), the Angular compiler by
default used tsickle that already performed this transformation. We
moved the transformation to the CLI for JIT and `ng-packager`, but now
we realize that we can move this all to a single place in the compiler
so that standalone ngc consumers can benefit too, and that we can
disable tsickle in our Bazel `ngc-wrapped` pipeline (that currently
still relies on tsickle to perform this decorator processing).
This transformation also has another positive side-effect of making
Angular application/library code more compatible with server-side
rendering. In principle, TypeScript would also preserve type information
for decorated class members (similar to how it did that for constructor
parameters) at runtime. This becomes an issue when your application
relies on native DOM globals for decorated class member types. e.g.
```
@Input() panelElement: HTMLElement;
```
Your application code would then reference `HTMLElement` directly
whenever the source file is loaded in NodeJS for SSR. `HTMLElement`
does not exist on the server though, so that will become an invalid
reference. One could work around this by providing global mocks for
these DOM symbols, but that doesn't match up with other places where
dependency injection is used for mocking DOM/browser specific symbols.
More context in this issue: #30586. The TL;DR here is that the Angular
compiler does not care about types for these class members, so it won't
ever reference `HTMLElement` at runtime.
Fixes#30106. Fixes#30586. Fixes#30141.
Resolves FW-2196. Resolves FW-2199.
PR Close#37382
The new tooling-cli-shared-api is used to guard changes to packages/compiler-cli/src/tooling.ts
which is a private API sharing channel between Angular FW and CLI.
Changes to this file should be rare and explicitly approved by at least two members
of the CLI team.
PR Close#37467
Now in TS 3.9, classes in ES2015 can be wrapped in an IIFE.
This commit ensures that we still find the static properties that contain
decorator information, even if they are attached to the adjacent node
of the class, rather than the implementation or declaration.
Fixes#37330
PR Close#37436
This finder is designed to only process entry-points that are reachable
by the program defined by a tsconfig.json file.
It is triggered by calling `mainNgcc()` with the `findEntryPointsFromTsConfigProgram`
option set to true. It is ignored if a `targetEntryPointPath` has been
provided as well.
It is triggered from the command line by adding the `--use-program-dependencies`
option, which is also ignored if the `--target` option has been provided.
Using this option can speed up processing in cases where there is a large
number of dependencies installed but only a small proportion of the
entry-points are actually imported into the application.
PR Close#37075
Previously we only checked for static import declaration statements.
This commit also finds import paths from dynamic import expressions.
Also this commit should speed up processing: Previously we were parsing
the source code contents into a `ts.SourceFile` and then walking the parsed
AST to find import paths.
Generating an AST is unnecessary work and it is faster and creates less
memory pressure to just scan the source code contents with the TypeScript
scanner, identifying import paths from the tokens.
PR Close#37075
Previously this host was skipping files if they had imports that spanned
multiple lines, or if the import was a dynamic import expression.
PR Close#37075
This commit will store a cached copy of the parsed tsconfig
that can be reused if the tsconfig path is the same.
This will improve the ngcc "noop" case, where there is no processing
to do, when the entry-points have already been processed.
Previously we were parsing this config every time we checked for
entry-points to process, which can take up to seconds in some
cases.
Resolves#36882
PR Close#37417
`NgCompiler` is the heart of ngtsc and can be used to analyze and compile
Angular programs in a variety of environments. Most of these integrations
rely on `NgProgram` and the creation of an `NgCompilerHost` in order to
create a `ts.Program` with the right shape for `NgCompiler`.
However, certain environments (such as the Angular Language Service) have
their own mechanisms for creating `ts.Program`s that don't make use of a
`ts.CompilerHost`. In such environments, an `NgCompilerHost` does not make
sense.
This commit breaks the dependency of `NgCompiler` on `NgCompilerHost` and
extracts the specific interface of the host on which `NgCompiler` depends
into a new interface, `NgCompilerAdapter`. This interface includes methods
from `ts.CompilerHost`, the `ExtendedTsCompilerHost`, as well as APIs from
`NgCompilerHost`.
A consumer such as the language service can implement this API without
needing to jump through hoops to create an `NgCompilerHost` implementation
that somehow wraps its specific environment.
PR Close#37118
When the compiler encounters a function call within an NgModule imports
section, it attempts to resolve it to an NgModule-annotated class by
looking at the function body and evaluating the statements there. This
evaluation can only understand simple functions which have a single
return statement as their body. If the function the user writes is more
complex than that, the compiler won't be able to understand it and
previously the PartialEvaluator would return a "DynamicValue" for
that import.
With this change, in the event the function body resolution fails the
PartialEvaluator will now attempt to use its foreign function resolvers to
determine the correct result from the function's type signtaure instead. If
the function is annotated with a correct ModuleWithProviders type, the
compiler will be able to understand the import without static analysis of
the function body.
PR Close#37126
Currently, if an ngcc process is killed in a manner that it doesn't clean
up its lock file (or is killed too quickly) the compiler reports that it
is waiting on the PID of a process that doesn't exist, and that it will
wait up to a maximum of N seconds. This PR updates the locking code to
additionally check if the process exists, and if it does not it will
immediately bail out, and print the location of the lock file so a user
may clean it up.
PR Close#37250
Inline source-maps in typings files can impact IDE performance
so ngcc should only add such maps if the original typings file
contains inline source-maps.
Fixes#37324
PR Close#37363
Due to an outage with the proxy we rely on for publishing, we need
to temporarily directly publish to NPM using our own angular
credentials again.
PR Close#37378
In #37221 we disabled tsickle passes from transforming the tsc output that is used to publish all
Angular framework and components packages (@angular/*).
This change however revealed a bug in the ngc that caused __decorate and __metadata calls to still
be emitted in the JS code even though we don't depend on them.
Additionally it was these calls that caused code in @angular/material packages to fail at runtime
due to circular dependency in the emitted decorator code documeted as
https://github.com/microsoft/TypeScript/issues/27519.
This change partially rolls back #37221 by reenabling the decorator to static fields (static
properties) downleveling.
This is just a temporary workaround while we are also fixing root cause in `ngc` - tracked as
FW-2199.
Resolves FW-2198.
Related to FW-2196
PR Close#37317
As of TypeScript 3.9, the tsc emit is not compatible with Closure
Compiler due to
https://github.com/microsoft/TypeScript/pull/32011.
There is some hope that this will be fixed by a solution like the one
proposed in
https://github.com/microsoft/TypeScript/issues/38374 but currently it's
unclear if / when that will
happen.
Since the Closure support has been somewhat already broken, and the
tsickle pass has been a source
of headaches for some time for Angular packages, we are removing it for
now while we rethink our
strategy to make Angular Closure compatible outside of Google.
This change has no effect on our Closure compatibility within Google
which work well because all the
code is compiled from sources and passed through tsickle.
This change only disables the tsickle pass but doesn't remove it.
A follow up PR should either remove all the traces of tscikle or
re-enable the fixed version.
BREAKING CHANGE: Angular npm packages no longer contain jsdoc comments
to support Closure Compiler's advanced optimizations
The support for Closure compiler in Angular packages has been
experimental and broken for quite some
time.
As of TS3.9 Closure is unusable with the JavaScript emit. Please follow
https://github.com/microsoft/TypeScript/issues/38374 for more
information and updates.
If you used Closure compiler with Angular in the past, you will likely
be better off consuming
Angular packages built from sources directly rather than consuming the
version we publish on npm
which is primarily optimized for Webpack/Rollup + Terser build pipeline.
As a temporary workaround you might consider using your current build
pipeline with Closure flag
`--compilation_level=SIMPLE`. This flag will ensure that your build
pipeline produces buildable and
runnable artifacts, at the cost of increased payload size due to
advanced optimizations being disabled.
If you were affected by this change, please help us understand your
needs by leaving a comment on https://github.com/angular/angular/issues/37234.
PR Close#37221
In ES2015 IIFE wrapped classes, the identifier that would reference the class
of the NgModule may be an alias variable. Previously the `Esm2015ReflectionHost`
was not able to match this alias to the original class declaration. This resulted
in failing to identify some `ModuleWithProviders` functions in such case.
These IIFE wrapped classes were introduced in TypeScript 3.9, which is why
this issue is only recently appearing. Since 9.1.x does not support TS 3.9
there is no reason to backport this commit to that branch.
Fixes#37189
PR Close#37206
To better check that the code is working, this commit gives a
distinct name (`DecoratedWrappedClass_1`) to the "adjacent"
class declaration in the tests.
PR Close#37206
Tslib version is bound to the TypeScript version used to compile the library. Thus, we shouldn't list `tslib` as a `peerDependencies`. This is because, a user can install libraries which have been compiled with older versions of TypeScript and thus require multiple `tslib` versions to be installed.
Reference: TOOL-1374 and TOOL-1375
Closes: #37188
PR Close#37198
The work to support case-sensitivity in the `FileSystem` went too far
with the `LogicalFileSystem`, which is used to compute import paths
that will be added to files processed by ngtsc and ngcc.
Previously all logical paths were canonicalised, which meant that on
case-insensitive file-systems, the paths were all set to lower case.
This resulted in incorrect imports being added to files. For example:
```
import { Apollo } from './Apollo';
import { SelectPipe } from './SelectPipe';
import * as ɵngcc0 from '@angular/core';
import * as ɵngcc1 from './selectpipe';
```
The import from `./SelectPipe` is from the original file, while the
import from `./selectpipe` is added by ngcc. This causes the
TypeScript compiler to complain, or worse for paths not to be
matched correctly.
Now, when computing logical paths, the original absolute paths
are matched against rootDirs in a canonical manner, but the actual
logical path that is returned maintains it original casing.
Fixes#36992, #36993, #37000
PR Close#37008
Alex Rickabaugh
Pete Bacon Darwin
Paul Gschwendtner
crisbeto
George Kalpakas
Alan Agius
David Neil
Joey Perrott
Igor Minar
Andrew Scott
Terence D. Honles