The `generateUniqueIdentifier` helper relies on the internal `identifiers` property of a `ts.SourceFile`.
As this is not a public-facing API, it is not included in the externs provided to Closure Compiler.
Consequently, it is susceptible to being renamed during advanced optimizations, which would lead to runtime failures.
To prevent this, the property is now accessed via a string literal (`sf['identifiers']`). This change ensures that
the property name is preserved through the compilation process. An explanatory comment has been added to clarify the
necessity of this approach for future reference.
PR Close#62517
Currently when a diagnostic happens in a template, we always say "Error occurs in ...", however we have other types of diagnostics as well.
PR Close#62479
The mock file system currently relies on some NodeJS logic. This blocks
the convenient use of the mock file system in the browser. The file
system is useful for creating tests to verify e.g. that our transforms
can work in the browser.
PR Close#62493
The static fields may be part of advanced compilations where the
receiver would change as part of the static field collapsing
optimization.
The optimization requires us to use the explicit class name, over
`this` that would change to e.g. `globalThis`.
PR Close#62493
Fixes that the logic recognizing initializer APIs didn't account for the expression being wrapped in an `as` expresion or in a parenthesized expression. This was already accounted for in the diagnostic so these changes align the behavior between them.
Fixes#62197.
PR Close#62203
Migrates the partial compliance tests to `rules_js`. Also as part of
this, we re-enable RBE to see if that fixed the issues, or in case
they are already resolved from the RBE side.
PR Close#61895
This commit migrates the remaining pieces of `compiler-cli` to
`ts_project`. This involves a few more things during migration:
- the `ng_module` ngc_wrapped rule broke as part of this change, so we
switched it to `ts_project` too. This logic is soon gone anyway.
- we needed an extra pnpm "package.json" for the linker babel test. This test is
loading from the real compiler-cli npm package. Babel needs a real
node module for this, so this solution seems reasonable. It may be
worth exploring in the future to move this test into an integration
test though.
- the older integrationtest in compiler-cli is removed as the coverage
is much better with the compliance test suite and this test.
PR Close#61862
When defer blocks have a reference-based trigger without a parameter, we infer it from the placeholder block. This requires some validations like ensuring that there's only one element in the placeholder. The validations are currently implemented at the parser level which can affect tools like linters that need to pass `preserveWhitespaces: true` in order to get accurate source mappings.
These changes move the validations into the template type checker so that they still get flagged, but much later in the process. Moving them over involves a bit more work, because the template type checker also sets `preserveWhitespaces: true`.
Fixes#61725.
PR Close#61747
Re-exports the `FileSystem` type since it's used by some tsurge migrations and because internally the barrel export from `ngtsc/file_system` is removed.
PR Close#61697
Currently when linking the `@angular/compiler-cli` package, the peer
dependency to `compiler` is not resolved and we are trying to make the
compiler dependency available via `data`. This is unidiomatic and
brittle. This commit fixes this.
Compiler now would have `.js` files. Those aren't picked up as ESM,
unless we install the `package.json` with `type: module`. Sounds great
on paper, but doesn't work in reality because the way the compiler
packages are available to `api-gen/` is via the old `rules_nodejs`
linker, so the `packages/package.json` wouldn't work; nor do the
`package.json`s of the e.g. compiler-cli package work- because those
already contain the `exports` of the built npm package.
We fix this in a much more reasonable way, and the whole module
resolution problem by leveraging the pnpm linking here. This works as
expected.
Replaces the `propertyInterpolateX` instructions with calls to `property` and the `interpolate` helper. This allows us to drop the dedicated interpolation instructions and simplify the runtime for future work.
PR Close#61639
Replaces the `classMapInterpolateX` instructions with `classMap` plus a call to `interpolate` in order to simplify the runtime. The only difference between `classMapInterpolateX` and `classMap` was that the former passes `keyValueArraySet` into `checkStylingMap` while the latter passes `classKeyValueArraySet`. This doesn't appear to matter, because the interpolation instructions always have a string value which means that the function is never called.
PR Close#61639
Replaces the `styleMapInterpolateX` instructions with the existing `styleMap` and a passed-in interpolated value in order to simplify the runtime.
PR Close#61639
The hardcoded config was introduced because suppressing the diagnostic
via `extendedDiagnostics` in the TS config was unreliable in google3.
This has since been fixed and the workaround is no longer needed.
PR Close#61622
Prevents esbuild generated metadata files from being included in build artifacts. This reduces package size and avoids shipping unnecessary internal build data.
PR Close#61636
This commit removes the direct dependency on TypeScript within the linker, addressing a performance overhead that was adding between 500ms to 1s to compilation times for applications.
The primary cause of this overhead was the linker's direct reliance on TypeScript's which was caused by importing from barrel files. While convenient, barrel files are detrimental to code splitting and code motion. They force the bundling of all exported modules, even if only a subset is actually used.
By removing the usage of this barrel file and restructuring the imports to be more granular, we can avoid unnecessary TypeScript imports.
Furthermore, TypeScript has now been changed to an optional peer dependency as using only the linker does not require TypeScript.
(cherry picked from commit 7cf2425fb5)
PR Close#61635
Replaces the attribute interpolation instructions with `attribute` plus the new `interpolateX` instruction. This allows to reduce our overall instruction footprint.
PR Close#61557
Currently we construct the HMR replacement URL inline by calling into the native `URL` constructor. This can cause conflicts with user code that defines a symbol called `URL`.
These changes resolve the issue by moving the URL construction into a separate function. This has a secondary benefit of making the generated code easier to follow and allowing us to update the URL without changing the compiled code.
Fixes#61517.
PR Close#61550
This change ensures that prior results for all files are retained even when
a request is made such that we only need a shim for a single file. Prior
to this change, any prior results that were not part of the request were discarded.
PR Close#61487
Use an unstamped version of the compiler when it runs in `ng_project` as it will get stamped appropriately
whenever the generated code gets stamped after its usage."
PR Close#61479
Fixes the `update_all_goldens` script which was throwing, because the query command was also capturing the call into yarn and Bazel which in turn caused it to throw an error. I've also added some validation for the number of targets.
PR Close#61407
The Angular class metadata emit structure does not support the use of
private fields. If the class metadata emit is enabled and an ECMAScript
private (i.e., `#` prefixed) member contains a decorator, the member will
now be excluded from the emitting `setClassMetadata` call. This prevents
runtime errors due to invalid syntax.
PR Close#61227
The compiler wasn't handling `@let` declarations placed inside i18n blocks. The problem is that `assignI18nSlotDependencies` phase assigns the `target` of i18n ops much earlier than the `@let` optimization. If the `@let` ends up getting optimized because it isn't used in any child views, the pointer in the i18n instruction becomes invalid. This hadn't surfaced so far, because we didn't optimize `declareLet` ops, however once we do, we start hitting assertions that the optimized `declareLet` isn't used anywhere.
These changes resolve the issue by moving the i18n phases after the `@let` optimization.
PR Close#60512
We have some code that avoids `storeLet` calls for declarations only used in the same view, however we didn't previously remove the corresponding `declareLet` calls, because of the following case:
```
@let foo = something$ | async; <!-- First in the template -->
{{foo}}
```
Here we need a `TNode` (created by `declareLet`) in order for DI to work correctly. Since this is only required when using pipes, we can optimize away expressions that do not have pipes.
PR Close#60512
As part of the Bazel toolchain migration we noticed that implicit types
generated by the TypeScript compiler sometimes end up referencing types
from other packages (i.e. cross-package imports).
These imports currently work just because the Bazel `ts_library` and
`ng_module` rules automatically inserted a `<amd-module
name="@angular/x" />` into `.d.ts` of packages. This helped TS figure
out how to import a given file. Notably this is custom logic that is not
occuring in vanilla TS or Angular compilations—so we will drop this
magic as part of the toolchain cleanup!
To improve code quality and keep the existing behavior working, we are
doing the following:
- adding a lint rule that reduces the risk of such imports breaking. The
failure scenario without the rule is that API goldens show unexpected
diffs, and types might be duplicated in a different package!
- keeping the `<amd-module` headers, but we manually insert them into
the package entry-points. This should ensure we don't regress
anywhere; while we also improved general safety around this above.
Long-term, isolated declarations or a lint rule from eslint-typescript
can make this even more robust.
PR Close#61312
In the event of an invalid `schemas` field for an Angular module, an
empty schema array will now be used instead of a fatal error occurring.
A build will still fail in this case with the error reported as a
diagnostic. However, for the language service, this allows the module
to exist in the compiler registry and prevents cascading diagnostics
within an IDE due to "missing" modules/components. The originating
schema related errors will still be reported in the IDE.
PR Close#61220
The `ComponentHandler.resolve` method is ~500 lines and is a bit hard to follow due to some very long `if` statements. These changes split the functionality across several smaller methods to make it easier to manage.
PR Close#61018
Currently to create an `R3TargetBinder`, we have to pass some sort of directive matcher, however we also have a couple of use cases where we use the binder to do analysis that's unrelated to directives (e.g. resolving the `@defer` blocks). In these cases having to create a dummy matcher adds some slight overhead and makes the code harder to reason about since it looks like directive matching may be happening.
These changes update the `R3TargetBinder` to allow for `null` to be passed as the directive matcher.
PR Close#61018
Paul Gschwendtner