With the work done in #41291, the compiler always tracks the last known
program, so there's no need to track the program in the compiler factory
anymore.
PR Close#41517
`NgCompiler` previously had a notion of the "next" `ts.Program`, which
served two purposes:
* it allowed a client using the `ts.createProgram` API to query for the
latest program produced by the previous `NgCompiler`, as a starting
point for building the _next_ program that incorporated any new user
changes.
* it allowed the old `NgCompiler` to be queried for the `ts.Program` on
which all prior state is based, which is needed to compute the delta
from the new program to ultimately determine how much of the prior
state can be reused.
This system contained a flaw: it relied on the `NgCompiler` knowing when
the `ts.Program` would be changed. This works fine for changes that
originate in `NgCompiler` APIs, but a client of the `TemplateTypeChecker`
may use that API in ways that create new `ts.Program`s without the
`NgCompiler`'s knowledge. This caused the `NgCompiler`'s concept of the
"next" program to get out of sync, causing incorrectness in future
incremental analysis.
This refactoring cleans up the compiler's `ts.Program` management in
several ways:
* `TypeCheckingProgramStrategy`, the API which controls `ts.Program`
updating, is renamed to the `ProgramDriver` and extracted to a separate
ngtsc package.
* It loses its responsibility of determining component shim filenames. That
functionality now lives exclusively in the template type-checking package.
* The "next" `ts.Program` concept is renamed to the "current" program, as
the "next" name was misleading in several ways.
* `NgCompiler` now wraps the `ProgramDriver` used in the
`TemplateTypeChecker` to know when a new `ts.Program` is created,
regardless of which API drove the creation, which actually fixes the bug.
PR Close#41291
Currently, we throw a FatalDiagnosticError when we fail to load a resource
(`templateUrl` or `styleUrl`) at various stages in the compiler. This prevents
analysis of the component from completing. This will result in in users not being
able to get any information in the component template when there is a missing
`styleUrl`, for example.
This commit simply tracks the diagnostic, marks the component as poisoned, and
continues merrily along. Environments configured to use poisoned data
(like the language service) will then be able to use other information from the analysis.
Fixes https://github.com/angular/vscode-ng-language-service/issues/1241
PR Close#41403
When possible, the @angular/language-service should only provide
information related to Angular. When there is an embedded language, like
inline templates, editor extensions should have the ability to create
virtual documents and forward the requests to the relevant providers for
that language type (see https://github.com/angular/vscode-ng-language-service/pull/1212).
This commit removes all dom schema completions in both inline and
external templates and provides only the Angular syntax for property completions
on elements.
PR Close#41278
Adds perf tracing for the public methods in LanguageService. If the log level is verbose or higher,
trace performance results to the tsServer logger. This logger is implemented on the extension side
in angular/vscode-ng-language-service.
PR Close#41319
ngtsc has an internal performance tracing package, which previously has not
really seen much use. It used to track performance statistics on a very
granular basis (microseconds per actual class analysis, for example). This
had two problems:
* it produced voluminous amounts of data, complicating the analysis of such
results and providing dubious value.
* it added nontrivial overhead to compilation when used (which also affected
the very performance of the operations being measured).
This commit replaces the old system with a streamlined performance tracing
setup which is lightweight and designed to be always-on. The new system
tracks 3 metrics:
* time taken by various phases and operations within the compiler
* events (counters) which measure the shape and size of the compilation
* memory usage measured at various points of the compilation process
If the compiler option `tracePerformance` is set, the compiler will
serialize these metrics to a JSON file at that location after compilation is
complete.
PR Close#41125
The Ivy Language Service uses the compiler's template type-checking engine,
which honors the configuration in the user's tsconfig.json. We recommend
that users upgrade to `strictTemplates` mode in their projects to take
advantage of the best possible type inference, and thus to have the best
experience in Language Service.
If a project is not using `strictTemplates`, then the compiler will not
leverage certain type inference options it has. One case where this is very
noticeable is the inference of let- variables for structural directives that
provide a template context guard (such as NgFor). Without `strictTemplates`,
these guards will not be applied and such variables will be inferred as
'any', degrading the user experience within Language Service.
This is working as designed, since the Language Service _should_ reflect
types exactly as the compiler sees them. However, the View Engine Language
Service used its own type system that _would_ infer these types even when
the compiler did not. As a result, it's confusing to some users why the
Ivy Language Service has "worse" type inference.
To address this confusion, this commit implements a suggestion diagnostic
which is shown in the Language Service for variables which could have been
narrowed via a context guard, but the type checking configuration didn't
allow it. This should make the reason why variables receive the 'any' type
as well as the action needed to improve the typings much more obvious,
improving the Language Service experience.
Fixes angular/vscode-ng-language-service#1155
Closes#41042
PR Close#41072
This commit fixes the behavior when creating a type constructor for a directive when the following
conditions are met.
1. The directive has bound generic parameters.
2. Inlining is not available. (This happens for language service compiles).
Previously, we would throw an error saying 'Inlining is not supported in this environment.' The
compiler would stop type checking, and the developer could lose out on getting errors after the
compiler gives up.
This commit adds a useInlineTypeConstructors to the type check config. When set to false, we use
`any` type for bound generic parameters to avoid crashing. When set to true, we inline the type
constructor when inlining is required.
Addresses #40963
PR Close#41043
Tsserver expects `@angular/language-service` to provide a factory function
as the default export (commonjs-style) of the package.
The current implementation side steps TypeScript's import syntax by using
`module.exports = factory`.
This allows the code to incorrectly re-export other symbols:
```ts
export * from './api';
```
which transpiles to:
```js
var tslib_1 = require("tslib");
tslib_1.__exportStar(require("@angular/language-service/api"), exports);
```
Doing this meant that the package now has a runtime dependency on `tslib`,
which is totally unnecessary.
With the proper `export =` syntax, `tslib` is removed, and no other exports
are allowed.
Output:
```js
(function (factory) {
if (typeof module === "object" && typeof module.exports === "object") {
var v = factory(require, exports);
if (v !== undefined) module.exports = v;
}
else if (typeof define === "function" && define.amd) {
define("@angular/language-service", ["require", "exports"], factory);
}
})(function (require, exports) {
"use strict";
return function factory(tsModule) {
var plugin;
return {
create: function (info) {
var config = info.config;
var bundleName = config.ivy ? 'ivy.js' : 'language-service.js';
plugin = require("./bundles/" + bundleName)(tsModule);
return plugin.create(info);
},
getExternalFiles: function (project) {
var _a, _b;
return (_b = (_a = plugin === null || plugin === void 0 ? void 0 : plugin.getExternalFiles) === null || _a === void 0 ? void 0 : _a.call(plugin, project)) !== null && _b !== void 0 ? _b : [];
},
onConfigurationChanged: function (config) {
var _a;
(_a = plugin === null || plugin === void 0 ? void 0 : plugin.onConfigurationChanged) === null || _a === void 0 ? void 0 : _a.call(plugin, config);
},
};
};
});
```
PR Close#41165
We currently provide completions for DOM elements in the schema as well
as attributes when we are in the context of an external template.
However, these completions are already provided by other extensions for
HTML contexts (like Emmet). To avoid duplication of results, this commit
updates the language service to exclude DOM completions for external
templates. They are still provided for inline templates because those
are not handled by the HTML language extensions.
PR Close#41078
The compiler considers template diagnostics to "belong" to the source file
of the component using the template. This means that when diagnostics for
a source file are reported, it returns diagnostics of TS structures in the
actual source file, diagnostics for any inline templates, and diagnostics of
any external templates.
The Language Service uses a different model, and wants to show template
diagnostics in the actual .html file. Thus, it's not necessary (and in fact
incorrect) to include such diagnostics for the actual .ts file as well.
Doing this currently causes a bug where external diagnostics appear in the
TS file with "random" source spans.
This commit changes the Language Service to filter the set of diagnostics
returned by the compiler and only include those diagnostics with spans
actually within the .ts file itself.
Fixes#41032
PR Close#41070
The current logic in the compiler is to bail when there are errors when
parsing a template into an HTML AST or when there are errors in the i18n
metadata. As a result, a template with these types of parse errors
_will not have any information for the language service_. This is because we
never attempt to conver the HTML AST to a template AST in these
scenarios, so there are no template AST nodes for the language service
to look at for information. In addition, this also means that the errors
are never displayed in the template to the user because there are no
nodes to map the error to.
This commit adds an option to the template parser to temporarily ignore
the html parse and i18n meta errors and always perform the template AST
conversion. At the end, the i18n and HTML parse errors are appended to
the returned errors list. While this seems risky, it at least provides
us with more information than we had before (which was 0) and it's only
done in the context of the language service, when the compiler is
configured to use poisoned data (HTML parse and i18n meta errors can be
interpreted as a "poisoned" template).
fixes angular/vscode-ng-language-service#1140
PR Close#41068
An opening tag `<` without any characters after it is interperted as a
text node (just a "less than" character) rather than the start of an
element in the template AST. This commit adjusts the autocomplete engine
to provide element autocompletions when the nearest character to the
left of the cursor is `<`.
Part of the fix for angular/vscode-ng-language-service#1140
PR Close#41068
This commit adds a new configuration option, `forceStrictTemplates` to the
language service plugin to allow users to force enable `strictTemplates`.
This is needed so that the Angular extension can be used inside Google without
changing the underlying compiler options in the `ng_module` build rule.
PR Close#41062
VSCode only de-duplicates references results for "go to references" requests
but does not de-duplicate them for "find all references" requests. The
result is that users see duplicate references for results in TypeScript
files - one from the built-in TS extension and one from us.
While this is an issue in VSCode (see https://github.com/microsoft/vscode/issues/117095)
this commit provides a quick workaround on our end until it can be addressed there.
This commit should be reverted when microsoft/vscode/issues/117095 is resolved.
fixes https://github.com/angular/vscode-ng-language-service/issues/1124
PR Close#41041
Now the language service always uses the name of the JavaScript property on the
component or directive instance for this input or output. This PR will use the right
binding property name.
PR Close#41005
Currently there are two entry points for the `@angular/language-service`
package:
- `@angular/language-service`
This default entry point is for View Engine LS. Through the redirection
of `main` field in `package.json`, it resolves to
`./bundles/language-service.js`.
- `@angular/language-service/bundles/ivy.js`
This secondary entry point is for Ivy LS.
TypeScript recently changed the behavior of tsserver to allow only package
names as plugin names [1] for security reasons. This means the secondary
entry point for Ivy LS can no longer be used.
We implemented a quick hack in the module resolver (in the extension repo)
to fix this, but the long term fix should be in `@angular/language-service`.
Here, the `main` field in `package.json` is changed to `index.js`, and in the
index file we conditionally load View Engine or Ivy based on the input config.
This eliminates the need for multiple entry points.
As part of this PR, I also removed all source code for View Engine and Ivy
included in the NPM package. Consumers of this package should run the bundled
output and nothing else. This would help us prevent an accidental import that
results in execution of unbundled code.
[1]: https://github.com/microsoft/TypeScript/pull/42713
PR Close#40967
This commit updates compiler_spec.ts in the Ivy LS suite to utilize the new
testing environment which was introduced in the previous commit. Eventually
all specs should be converted, but converting one right now helps ensure
that the new testing env is working properly and able to support real tests.
PR Close#40966
The Angular LS does not provide quick info when the given position is not
inside a template. As an optimization, we can quickly look at the
file and determine if we are at a position that is part of an Angular
template. If not, we bail before asking the compiler for any more
information. Note that the Angular LS _already_ provides no quick info
when outside a template file, but currently asks the compiler to analyze
the program before it determines that information.
PR Close#40956
When certain information is requested from the Angular Language Service, we
know that there will be no additional Angular information if the requested
position is not in an inline template, template url, or style url. To avoid
unnecessary compiler compilations, we short circuit and return `undefined`
before asking the compiler for any type of answer which would trigger a
partial compilation, at the very least.
fixes https://github.com/angular/vscode-ng-language-service/issues/1104
PR Close#40946
This commit updates compiler_spec.ts in the Ivy LS suite to utilize the new
testing environment which was introduced in the previous commit. Eventually
all specs should be converted, but converting one right now helps ensure
that the new testing env is working properly and able to support real tests.
PR Close#40679
The Ivy Language Service codebase testing suite contains a few testing
utilities which allow for assertions of Language Service operations against
an in-memory project. However, this existing utility lacks the flexibility
to test more complex scenarios, such as those involving multiple TS projects
with dependencies between them.
This commit introduces a new 'testing' package for the Ivy LS which attempts
to more faithfully represent the possible states of an IDE, and allows for
testing of more advanced scenarios. The new utility borrows from the prior
version and is geared towards more ergonomic testing. Only basic
functionality is present in this initial implementation, but this will grow
over time.
PR Close#40679
In 5c547675b11a24b16c20df1718583a0e7ed49cbd the `EventEmitter.subscribe`
API was extended with a new signature that allows the emitter's generic
type `T` to flow into the subscribe callback. This new signature removes
the need for the special `_outputHelper` function that used to be
emitted into TCBs when `strictOutputEventTypes`/`strictTemplates` is
enabled.
PR Close#40738
Two motivations behind this change:
1. We would like to expose the types of the Language Service to external
users (like the VSCode extension) via the npm package, on the top
level of the package
2. We would like the View Engine and Ivy LS to share a common interface
(notably after the inclusion of `getTcb`, the Ivy LS upholds a
strict superset of `ts.LanguageService`; previously both VE and Ivy
LS were aligned on `ts.LanguageService`.)
To this end, this commit refactors the exports on the toplevel of the
`language-service/` package to just be types common to both the VE and
Ivy language services. The VE and Ivy build targets then import and use
these types accordingly, and the expectation is that an external user
will just import the relevant typings from the toplevel package without
diving into either the VE or Ivy sources.
Follow up on #40607
PR Close#40621
The `TemplateTypeChecker.overrideComponentTemplate` operation was originally
conceived as a "fast path" for the Language Service to react to a template
change without needing to go through a full incremental compilation step. It
served this purpose until the previous commit, which switches the LS to use
the new resource-only incremental change operation provided by `NgCompiler`.
`overrideComponentTemplate` is now no longer utilized, and is known to have
several hard-to-overcome issues that prevent it from being useful in any
other situations. As such, this commit removes it entirely.
PR Close#40585
This commit changes the Language Service's "compiler factory" mechanism to
leverage the new resource-only update path for `NgCompiler`. When an
incoming change only affects a resource file like a component template or
stylesheet, going through the new API allows the Language Service to avoid
unnecessary incremental steps of the `NgCompiler` and return answers more
efficiently.
PR Close#40585
When parsing interpolations, if we encounter an empty interpolation
(`{{}}`), the current code uses a "pretend" value of `$implicit` for the
name as if the interplotion were really `{{$implicit}}`. This is
problematic because the spans are then incorrect downstream since they
are based off of the `$implicit` text.
This commit changes the interpretation of empty interpolations so that
the text is simply an empty string.
Fixes https://github.com/angular/vscode-ng-language-service/issues/1077
Fixes https://github.com/angular/vscode-ng-language-service/issues/1078
PR Close#40583
If the template parse option `leadingTriviaChars` is configured to
consider whitespace as trivia, any trailing whitespace of an element
would be considered as leading trivia of the subsequent element, such
that its `start` span would start _after_ the whitespace. This means
that the start span cannot be used to mark the end of the current
element, as its trailing whitespace would then be included in its span.
Instead, the full start of the subsequent element should be used.
To harden the tests that for the Ivy parser, the test utility `parseR3`
has been adjusted to use the same configuration for `leadingTriviaChars`
as would be the case in its production counterpart `parseTemplate`. This
uncovered another bug in offset handling of the interpolation parser,
where the absolute offset was computed from the start source span
(which excludes leading trivia) whereas the interpolation expression
would include the leading trivia. As such, the absolute offset now also
uses the full start span.
Fixes#39148
PR Close#40513
Previously, the incremental flow for NgCompiler was simple: when creating a
new NgCompiler instance, the consumer could pass state from a previous
compilation, which would cause the new compilation to be performed
incrementally. "Local" information about TypeScript files which had not
changed would be passed from the old compilation to the new and reused,
while "global" information would always be recalculated.
However, this flow could be made more efficient in certain cases, such as
when no TypeScript files are changed in a new compilation. In this case,
_all_ information extracted during the first compilation is reusable. Doing
this involves reusing the previous `NgCompiler` instance (the container for
such global information) and updating it, instead of creating a new one for
the next compilation. This approach works cleanly, but complicates the
lifecycle of `NgCompiler`.
To prevent consumers from having to deal with the mechanics of reuse vs
incremental steps of `NgCompiler`, a new `CompilationTicket` mechanism is
added in this commit. Consumers obtain a `CompilationTicket` via one of
several code paths depending on the nature of the incoming compilation, and
use the `CompilationTicket` to obtain an `NgCompiler` instance. This
instance may be a fresh compilation, a new `NgCompiler` for an incremental
compilation, or an existing `NgCompiler` that's been updated to optimally
process a resource-only change. Consumers can use the new `NgCompiler`
without knowledge of its provenance.
PR Close#40561
The Language Service uses the source span of AST nodes to recognize which
node a user has selected, given their cursor position in a template. This is
used to trigger autocompletion.
The previous source span of BindingPipe nodes created a problem when:
1) the pipe binding had no identifier (incomplete or in-progress expression)
2) the user typed trailing whitespace after the pipe character ('|')
For example, the expression `{{foo | }}`. If the cursor preceded the '}' in
that expression, the Language Service was unable to detect that the user was
autocompleting the BindingPipe expression, since the span of the BindingPipe
ended after the '|'.
This commit changes the expression parser to expand the span of BindingPipe
expressions with a missing identifier, to include any trailing whitespace.
This allows the Language Service to correctly recognize this case as
targeting the BindingPipe and complete it successfully. The `nameSpan` of
the BindingPipe is also moved to be right-aligned with the end of any
whitespace present in the pipe binding expression.
This change allows for the disabled test in the Language Service for pipe
completion in this case to be re-enabled.
PR Close#40346
The `LanguageServiceAdapter` must implement `realpath` in order to resolve
symlinks in `node_modules`.
Local libraries are often symlinked in `node_modules` by adding a local
dependency in `package.json`.
Fix https://github.com/angular/vscode-ng-language-service/issues/1083
PR Close#40593
This PR adds a way for the language server to retrieve compiler options
diagnostics via `languageService.getCompilerOptionsDiagnostics()`.
This will be used by the language server to show a prompt in the editor if
users don't have `strict` or `fullTemplateTypeCheck` turned on.
Ref https://github.com/angular/vscode-ng-language-service/issues/1053
PR Close#40423
This patch adds an API to retrieve the template typecheck block for a
template (if any) at a file location, and a selection of the TS node
in the TCB corresponding to the template node at which the request for
a TCB was made (if any).
Probably not something we want to land soon, but a useful debugging tool
for folks working with TCBs.
PR Close#39974
Rather than mutating the span on the template when renaming literal strings,
this commit updates the logic to mutate the `TextSpan` equivalent that
is used by the Language Service.
PR Close#40484
Keen Yee Liau
Alex Rickabaugh
Andrew Scott
Zach Arend
Kristiyan Kostadinov
Kapunahele Wong
ivanwonder
JoostK
ayazhafiz