Currently there can be cases, exlusively in 3P, where multiple tsconfig
projects have overlap of source files. This is the default setup of new
CLI applications as well.
When this is the case, Tsurge will treat each tsconfig as an isolated
compilation unit (given the concepts and mental model to support
scalable batching). This is wrong though, and the same `.ts` source file
can appear in two migration invocations; resulting in duplicate
replacements or analysis (depending on the migration).
We've worked around this problem in the past by deduplicating
replacements, or migrating to an ID-based approach with natural
deduplication. This worked, but it's just working around the root cause.
This commit attempts to fix the root cause by adjusting Tsurge to ensure
that no source file ever appears in two compilation units. This is
naively achieved by not adding a source file to a migration unit, if it
was part of a previous one. This is expected to be fine given the nature
of Tsurge migrations that are built to operate on isolated pieces
anyway— so it shouldn't be problematic if e.g. `app.component.ts` ends
up being part of the test tsconfig compilation unit (we avoid this order
though by visiting build targets first).
PR Close#61421
PR Close#61612
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
When we switch to relative imports, shared `.d.ts` chunks can be
generated.
We need to also pull these into our mock virtual FS testing
environments. Notably this does not cause a test slow-down because we
are talking about very few extra `.d.ts` chunk files. In our experiments
before, with no dts bundling, we saw test time increase from e.g.
20seconds to 100seconds. The 20s are still the same locally!
In addition, since code for definitions can now reside in shared `.d.ts`
chunks, the language service tests need to be adjusted in cases where
they assert for code definition locations in `@angular/core`. A new
helper prepares for more code to be moved into arbitrary `.d.ts` files;
we should simply assert the definition comes out of
`node_modules/@angular/core`.
PR Close#60487
Instead of relying on Microsoft's API extractor for `d.ts` bundling,
we are switching to Rollup-based `.d.ts` bundling.
This allows us to support code spliting, even for `.d.ts` files,
allowing for relative imports to be used between entry-points, without
ending up duplicating `.d.ts` definitions in two files. This would otherwise cause
problems with assignability of types.
It also nicely integrates into our existing rollup configuration, and
overall simplifies the `ng_package` rule even further!
Notably `tsup` also uses this rollup plugin, and it seems to work well.
Keep in mind that Microsoft's API extractor is pretty hard to integrate,
caused many problems in the past, and isn't capable of code splitting.
This aligns our d.ts bundling with the .mjs bundling (great alignment).
PR Close#60321
PR Close#60332
Currently the language service has some template-specific terminology around type checking, because that's the only place where we had TCB support. These changes make it more generic to accommodate future functionality.
PR Close#60191
Currently only components can have resources, because they're the only symbol kinds being type checked. Since we want to add directives to it, these changes rework the resource handling to accommodate them.
PR Close#60191
Currently the `TemplateSourceManager` is set up to specifically cater to component templates. These changes make it more generic so we can reuse it for directives.
PR Close#60191
Prior to this commit, the tags from the type definition were dropped.
Tags may include, but are not limited to, deprecation information from
the jsdoc.
PR Close#59524
In the past two-way bindings used to be interpreted as `foo = $event` at the parser level. In #54065 it was changed to preserve the actual expression, because it was problematic for supporting two-way binding to signals. This unintentionally ended up causing the TCB to two-way bindings to look something like `someOutput.subscribe($event => expr);` which does nothing. It largely hasn't been a problem, because the input side of two-way bindings was still being checked, except for the case where the input side of the two-way binding has a wider type than the output side.
These changes re-add type checking for the output side by generating the following TCB instead:
```
someOutput.subscribe($event => {
var _t1 = unwrapSignalValue(this.someField);
_t1 = $event;
});
```
PR Close#59002
The error codes that have fixes are present in a Map instance and
can be directly leveraged to determine if an error code can be fixed.
This avoids multiple levels of iteration that would otherwise be needed.
PR Close#58759
Fixes that `getCodeActions` wasn't implemented for the unused imports fixer which meant that it wouldn't show up in the most common cases.
PR Close#58719
Currently the code fixes won't run if the template information can't be resolved on the diagnostic node. This is incorrect for diagnostics that are reported within `.ts` files. These changes make the `templateInfo` nullable and leave the early exit up to the individual fixes.
PR Close#58719
Initially the unused imports check was implemented so that it reports one diagnostic per component with the individual unused imports being highlighted through the `relatedInformation`. This works fine when reporting errors to the command line, but vscode appears to only show `relatedInformation` when the user hovers over a diagnostic which is a sub-par experience.
These changes switch to reporting a diagnostic for each unused import instead.
PR Close#58589
When setting `"useDefineForClassFields": false`, static fields are compiled within a block that relies on the `this` context. This output makes it more difficult for bundlers to treeshake and eliminate unused code.
PR Close#58297
This is necessary given the previous Tsurge refactorings. It should
speed up migrations in the merge phase signficiantly and reduce memory
pressure. In 1P, we already have workers from analyze phase anyway— so
those can be re-used for parallel metadata merging.
PR Close#58280
This is necessary given the previous Tsurge refactorings. It should
speed up migrations in the merge phase signficiantly and reduce memory
pressure. In 1P, we already have workers from analyze phase anyway— so
those can be re-used for parallel metadata merging.
PR Close#58280
This is a follow-up to the VSCode queries code refactoring feature. This
commit adds support for running the refactoring with
`--best-effort-mode`.
PR Close#58168
Instead of skipping queries without any reasoning, we should categorize
fields that couldn't be migrated. This is also important for the VSCode
integration— similar to how it's done with the inputs migration.
We are fully sharing the problematic pattern detection etc. This means
we are also sharing the enum. Not super ideal, but enables the best
sharing of code.
PR Close#58152
This commit moves the incompatibility categorization into a more common
place, and renames it from Input incompatibilities to "field
incompatibilities". This construct can then be used in the queries
migration as well to give insight into why certain fields weren't
migrated.
PR Close#58139
This commit adds support for converting decorator queries to signal queries
via the VSCode extension.
Note that this is not fully finished as we still need to add better
messaging when certain fields could not be migrated.
In addition, it's worth noting that the migration is not as safe as the
input migration because commonly query lists are passed around— this
quickly can break the build— but is an acceptable trade-off for the work
saved. A migration cannot be 100% correct in general; there are always
edge-cases.
PR Close#58106
This allows us to return extra properties along with `#migrate`
replacements. Useful for language service integration or other
integrations of Tsurge migrations in special runners.
PR Close#58106
In order to investigate the performances of SSR, this commit introduces a benchmark suite which will measure several step of the rendering.
PR Close#57647
This commit expands the VScode integration of the signal input migration
to allow migration of full classes and all their inputs. This enables a
faster workflow than just migrating every member individually.
In addition, we now properly support migrating classes that are
unexported and no actual metadata is available in `ngtsc` (but this is
fine for the migration).
PR Close#57975
Paul Gschwendtner