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#61316
We don't need this tooling anymore because we are already validating
that there are no circular dependencies via the `ng-dev` tooling that
checks `.ts` files directly.
Also these tests never actually failed to my knowledge.
PR Close#61209
The `TemplateLiteralElementExpr` has some logic where it tries to estimate the `rawText` if one isn't provided by looking at the node's source span. The problem with this approach is that we have some long-standing issues with our expression AST parser (see https://github.com/angular/angular/pull/60267#discussion_r1986402524) where it might not produce accurate spans if escape sequences are involved. This in turn can lead to unrecoverable errors, because TypeScript will throw an error if the raw string doesn't match the cooked one when constructing a TypeScript AST node.
These changes remove the logic that depends on the source span and relies purely on the secondary fallback that inserts escaped characters manually.
It's also worth noting that the `rawText` doesn't seem to matter much at this point, because the main usage of it is when downlevelling template literals to ES5 which we no longer support.
Fixes#60528.
PR Close#60529
PR Close#60753
Note that this is not a public-facing change because the compiler is not
supposed to be used directly. See `public-api-surface.md`
We are changing this because in the current state, for advanced setups
like our Bazel infrastructure in Material, this peer dependency can
result in a cycle. That is because we may decide to link the
`@angular/core` package using the compiler-cli/compiler; while
transitively the compiler has a dependency on core. That's a cycle
(surfacing in Bazel w/ `rules_js`).
Dropping this optional peer dependency is acceptable because `core`,
since recently, also has an optional peer dependency on the compiler.
This one makes more sense and is also more user-facing (people generally
never install the compiler directly).
Notably the compiler does not have any runtime dependency on `core` and
this was purely added for some safety checking. See:
8c171da29f
PR Close#60437
Fixes that we were producing zero-length spans for template literals and template literal elements.
Fixes#60320.
Fixes#60319.
PR Close#60323
PR Close#60331
Currently `R3TargetBinder.bind` gets a set of data back from `DirectiveBinder.apply` and `TemplateBinder.applyWithScope`. This will be annoying if we have multiple sources of data, because we'd have to do merge them at the end.
These changes switch to constructing the various data structures ahead of time and passing them into the binders to populate them instead.
I also extracted some of the less trivial types into type aliases so we don't have to repeat them.
PR Close#60191
Historically we've had to be VERY cautious about the way we import
things between entry-points. That is because the `ng_package` rule
bundling is subject to silently introducing code duplication, breaking
singletons etc. We've had this surface a couple of times already, and
dev-infra tried to help detect such cases by adding safety analysis into
`ng_package`.
Long-term we want to get to an approach where it's easy to simply share
code between chunks. Precisely, with the upcoming `rules_js` migration,
this will be necessary as we will have different import "guidelines"
that would currently, before this commit, result in code duplication, or
trigger our "safety check/lint".
This commit prepares `ng_package` to support relative imports between
entry-points, so that we only need the safety check for cross-package
imports/exports. The result is that `ng_package`/APF is now smartly able
to generate shared chunks for things that are needed between multiple
entry-points. Yay!
Note that those shared chunks still remain private, and are guarded by
our `package.json` "exports"; so no new public API surface is
exposed.
PR Close#60241
Currently when we generate the tracking expression for a `@for` block, we process its expression in the context of the creation block. This is incorrect, because the expression may require ops of its own for cases like nullish coalescing or safe reads. The result is that while we do generate the correct variable, they're added to the creation block rather than the tracking function which causes an error at runtime.
These changes address the issue by keeping track of a separate set of ops for the `track` expression that are prepended to the generated function, similarly to how we handle event listeners.
Fixes#56256.
PR Close#58520
Adjusts the code we generate for HMR so that it passes in the HMR ID and `import.meta` to the `replaceMetadata` call. This is necessary so we can do better logging of errors.
PR Close#59854
When we generate an HMR replacement function, we determine which locals from the file are used and we pass them by reference. This works fine in most cases, but breaks down for const enums which don't have a runtime representation.
These changes work around the issue by passing in all the values as an object literal.
Fixes#59800.
PR Close#59815
Makes the following cleanups in the output AST:
* The `TemplateLiteral` and `TemplateLiteralElement` nodes have been renamed to `TemplateLiteralExpr` and `TemplateLiteralElementExpr` respectively for consistency and to avoid overlaps with the expression AST nodes.
* The `TemplateLiteralExpr` and `TemplateLiteralElementExpr` have been refactored to be `Expression`s for correctness. This involves updating some existing code.
* The `TaggedTemplateExpr` has been renamed to `TaggedTemplateLiteralExpr` for consistency.
PR Close#59230
Reworks the lexer's scanner to produce more than one token at a time. This can be useful for the cases where one token means the end of another one.
Also cleans up the scanner by making all non-essential methods private and using strict equality everywhere.
PR Close#59230
Both `:host` and `:host-context` work by looking for a specific character sequence that is terminated by `,` or `{` and replacing selectors inside of it with scoped versions. This is implemented as a regex which isn't aware of things like nested selectors. Normally this is fine for `:host`, because each `:host` produces one scoped selector which doesn't affect any child selectors, however it breaks down with `:host-context` which replaces each instance with two selectors. For example, if we have a selector in the form of `:host-context(.foo) a:not(.a, .b)`, the compiler ends up determining that `.a,` is the end selector and produces `.foo[a-host] a[contenta]:not(.a, .foo [a-host] a[contenta]:not(.a, .b) {}`.
These changes resolve the issue by splitting the CSS alogn top-level commas, processing the `:host-context` in them individually, and stiching the CSS back together.
PR Close#59276
The URL that is dynamically imported to fetch a potential component update
for HMR is now based on the value of `import.meta.url`. This ensures that
the request is sent to the same location that was used to retrieve the
application code. For some development server setups the HTML base HREF
may not be the location of the Angular development server. By using the
application code location which was generated by the development server,
HMR requests can continue to work as expected in these scenarios. In
most common cases, this change will not have any effect as the HTML base
HREF aligns with the location of the application code files.
PR Close#59459
Some time ago we narrowed down the expressions we support in two-way bindings, because in most cases any apart from property reads doesn't make sense. This ended up preventing users from using `$any` in the binding since it's considered a function call.
These changes update the validation logic to allow `$any`.
Fixes#51165.
PR Close#59362
'parameter' was spelled as 'paramater'.
Fix spelling error in Update r3_control_flow.ts
'parameter' was spelled as 'paramater'.
Fix spelling error in r3_template_transform_spec.ts
'parameter' was spelled as 'paramater'.
PR Close#59289
Adds the implementation of the `ɵɵattachSourceLocations` instruction that will add the `data-ng-source-location` attribute to nodes to indicate where they were defined.
PR Close#58982
Deletes the `allowInvalidAssignmentEvents` which was added to facilitate a migration away from invalid two-way bindings. Since the migration doesn't exist anymore, we don't need the flag either.
PR Close#58988
The current HMR compiler assumes that there will only be one namespace import in the generated code (`@angular/core`). This is incorrect, because the compiler may need to generate additional imports in some cases (e.g. importing directives through a module). These changes adjust the compiler to capture all the namespaces in an array and pass them along.
Fixes#58915.
PR Close#58924
This cleans up the triggering code base and consolidates it down to one
function that outlines the logic. This also resolves the `hydrate when`
behavior issue.
fixes: #58709
PR Close#58833
Currently host bindings are in a bit of a weird state, because their source spans all point to the root object literal, rather than the individual expression. This is tricky to handle at the moment, because the object is being passed around as a `Record<string, string>` since the compiler needs to support both JIT and non-JIT environments, and because the AOT compiler evaluates the entire literal rather than doing it expression-by-expression. As a result, when we report errors in one of the host bindings, we end up highlighting the entire expression which can be very noisy in an IDE.
These changes aim to report a more accurate error for the most common case where the `host` object is initialized to a `string -> string` object literal by matching the failing expression to one of the property initializers. Note that this isn't 100% reliable, because we can't map cases like `host: SOME_CONST`, but it's still better than the current setup.
PR Close#58870
fix transformation logic for `:where` and `:is` pseudo-selectors
when these selectors were used in a chain. results were often broken,
the last letter of the selector was incorrectly trimmed.
see tests for examples
Fixes#58226
PR Close#58681
fix several use-cases where `:host` was used in or around pseudo-selectors
- `:host` followed by a comma inside pseudo-selectors
- `:host` outside of pseudo-selectors when another `:host` is present within
see tests for examples
PR Close#58681
Fixes an edge case where a control flow node that has non-projectable nodes followed by an element node at the end would cause the entire control flow node to be project. For example if we have a projection target of `Main: <ng-content/> Slot: <ng-content select="[foo]"/>`, inserting a node of `@if (true) {Hello <span foo>world</span>}` would project the entire `Hello world` into the `[foo]` slot.
In the process of working on the issue, I also found that `@let` declarations at the root of the control flow node would prevent content projection as well.
PR Close#58607
Adjusts the HMR initialization to avoid the edge case where a developer makes change to a non-rendered component that exists in a lazy loaded chunk that has not been loaded yet. The changes include:
* Moving the `import` statement out into a separate function.
* Adding a null check for `d.default` before calling `replaceMEtadata`.
* Triggering the `import` callback eagerly on initialization.
Example of the new generated code:
```js
(() => {
function Cmp_HmrLoad(t) {
import(
/* @vite-ignore */ "/@ng/component?c=test.ts%40Cmp&t=" + encodeURIComponent(t)
).then((m) => m.default && i0.ɵɵreplaceMetadata(Cmp, m.default, [/* Dependencies go here */]));
}
(typeof ngDevMode === "undefined" || ngDevMode) && Cmp_HmrLoad(Date.now());
(typeof ngDevMode === "undefined" || ngDevMode) &&
import.meta.hot &&
import.meta.hot.on("angular:component-update", (d) => {
if (d.id === "test.ts%40Cmp") {
Cmp_HmrLoad(d.timestamp);
}
});
})();
```
PR Close#58465
