Previously we only compiled the typings files, in ngcc, if there was
an ES2015 formatted bundle avaiable. This turns out to be an artificial
constraint and we can also support typings compilation via ES5 formats
too.
This commit changes the ngcc compiler to attempt typings compilation
via ES5 if necessary. The order of the formats to consider is now:
FESM2015, FESM5, ESM2015, ESM5.
FW-1122
PR Close#29092
This fix is for a bug in the ngtsc PartialEvaluator, which statically
evaluates expressions.
Sometimes, evaluating a reference requires resolving a function which is
declared in another module, and thus no function body is available. To
support this case, the PartialEvaluator has the concept of a foreign
function resolver.
This allows the interpretation of expressions like:
const router = RouterModule.forRoot([]);
even though the definition of the 'forRoot' function has no body. In
ngtsc today, this will be resolved to a Reference to RouterModule itself,
via the ModuleWithProviders foreign function resolver.
However, the PartialEvaluator also associates any Identifiers in the path
of this resolution with the Reference. This is done so that if the user
writes
const x = imported.y;
'x' can be generated as a local identifier instead of adding an import for
'y'.
This was at the heart of a bug. In the above case with 'router', the
PartialEvaluator added the identifier 'router' to the Reference generated
(through FFR) to RouterModule.
This is not correct. References that result from FFR expressions may not
have the same value at runtime as they do at compile time (indeed, this is
not the case for ModuleWithProviders). The Reference generated via FFR is
"synthetic" in the sense that it's constructed based on a useful
interpretation of the code, not an accurate representation of the runtime
value. Therefore, it may not be legal to refer to the Reference via the
'router' identifier.
This commit adds the ability to mark such a Reference as 'synthetic', which
allows the PartialEvaluator to not add the 'router' identifier down the
line. Tests are included for both the PartialEvaluator itself as well as the
resultant buggy behavior in ngtsc overall.
PR Close#29387
The `resolve` phase (run after all handlers have analyzed) was
introduced in 7d954dffd, but `ngcc` was not updated to run the handlers'
`resolve()` methods. As a result, certain operations (such as listing
directives used in component templates) would not be performed by
`ngcc`.
This commit fixes it by running the `resolve()` methods once analysis
has been completed.
PR Close#28963
Currently if an Angular library has multiple unnamed module re-exports, NGC will
generate incorrect metdata if the project is using the flat-module bundle option.
e.g.
_public-api.ts_
```ts
export * from '@mypkg/secondary1';
export * from '@mypkg/secondary2';
```
There are clearly two unnamed re-exports in the `public-api.ts` file. NGC right now
accidentally overwrites all previous re-exports with the last one. Resulting in the
generated metadata only containing a reference to `@mypkg/secondary2`.
This is problematic as it is common for primary library entry-points to have
multiple re-exports (e.g. Material re-exporting all public symbols; or flex-layout
exporting all public symbols from their secondary entry-points).
Currently Angular Material works around this issue by manually creating
a metadata file that declares the re-exports from all unnamed re-exports.
(see: https://github.com/angular/material2/blob/master/tools/package-tools/build-release.ts#L78-L85)
This workaround works fine currently, but is no longer easily integrated when
building the package output with Bazel. In order to be able to build such
libraries with Bazel (Material/flex-layout), we need to make sure that NGC
generates the proper flat-module metadata bundle.
PR Close#29360
Previously, ngtsc would resolve forward references while evaluating the
bootstrap, declaration, imports, and exports fields of NgModule types.
However, when generating the resulting ngModuleDef, the forward nature of
these references was not taken into consideration, and so the generated JS
code would incorrectly reference types not yet declared.
This commit fixes this issue by introducing function closures in the
NgModuleDef type, similarly to how NgComponentDef uses them for forward
declarations of its directives and pipes arrays. ngtsc will then generate
closures when required, and the runtime will unwrap them if present.
PR Close#29198
This fixes an issue with commit b6f6b117. In this commit, default imports
processed in a type-to-value conversion were recorded as non-local imports
with a '*' name, and the ImportManager generated a new default import for
them. When transpiled to ES2015 modules, this resulted in the following
correct code:
import i3 from './module';
// somewhere in the file, a value reference of i3:
{type: i3}
However, when the AST with this synthetic import and reference was
transpiled to non-ES2015 modules (for example, to commonjs) an issue
appeared:
var module_1 = require('./module');
{type: i3}
TypeScript renames the imported identifier from i3 to module_1, but doesn't
substitute later references to i3. This is because the import and reference
are both synthetic, and never went through the TypeScript AST step of
"binding" which associates the reference to its import. This association is
important during emit when the identifiers might change.
Synthetic (transformer-added) imports will never be bound properly. The only
possible solution is to reuse the user's original import and the identifier
from it, which will be properly downleveled. The issue with this approach
(which prompted the fix in b6f6b117) is that if the import is only used in a
type position, TypeScript will mark it for deletion in the generated JS,
even though additional non-type usages are added in the transformer. This
again would leave a dangling import.
To work around this, it's necessary for the compiler to keep track of
identifiers that it emits which came from default imports, and tell TS not
to remove those imports during transpilation. A `DefaultImportTracker` class
is implemented to perform this tracking. It implements a
`DefaultImportRecorder` interface, which is used to record two significant
pieces of information:
* when a WrappedNodeExpr is generated which refers to a default imported
value, the ts.Identifier is associated to the ts.ImportDeclaration via
the recorder.
* when that WrappedNodeExpr is later emitted as part of the statement /
expression translators, the fact that the ts.Identifier was used is
also recorded.
Combined, this tracking gives the `DefaultImportTracker` enough information
to implement another TS transformer, which can recognize default imports
which were used in the output of the Ivy transform and can prevent them
from being elided. This is done by creating a new ts.ImportDeclaration for
the imports with the same ts.ImportClause. A test verifies that this works.
PR Close#29266
Prior to this change default selector for Components was not applied in case selector is missing or defined as an empty string. This update aligns this behavior between Ivy and VE: now default selector is used for Components when it's needed. Directives with empty selector are not allowed and trigger a compile-time error in both Ivy and VE.
PR Close#29239
Prior to this change the code didn't take into account the fact that decorators can be aliases while importing into a script. As a result, these decorators were not recognized by Angular and various failures happened because of that. Now we take aliases into account and resolve decorator name properly.
PR Close#29195
`getCurrentDirectory` directory doesn't return a posix separated normalized path. While `rootDir` and `rootDirs` should return posix separated paths, it's best to not assume as other paths within the compiler options can be returned not posix separated such as `basePath`
See: https://github.com/Microsoft/TypeScript/blob/master/src/compiler/sys.ts#L635
This partially fixes#29140, however there needs to be a change in the CLI as well to handle this, as at the moment we are leaking devkit paths which is not correct.
Fixes#29140
PR Close#29151
Sometimes declarations are not exported publicly but are exported under
a private name. In this case, rather than adding a completely new
export to the entry point, we should create an export that aliases the
private name back to the original public name.
This is important when the typings files have been rolled-up using a tool
such as the [API Extractor](https://api-extractor.com/). In this case
the internal type of an aliased private export will be removed completely
from the typings file, so there is no "original" type to re-export.
For example:
If there are the following TS files:
**entry-point.ts**
```ts
export {Internal as External} from './internal';
```
**internal.ts**
```ts
export class Internal {
foo(): void;
}
```
Then the API Extractor might roll up the .d.ts files into:
```ts
export declare class External {
foo(): void;
}
```
In this case ngcc should add an export so the file looks like:
```ts
export declare class External {
foo(): void;
}
export {External as Internal};
```
PR Close#28735
ngsummary files were generated with an export for each class declaration.
However, some Angular code declares classes (class Foo) and exports them
(export {Foo}) separately, which was causing incomplete summary files.
This commit expands the set of symbol names for which summary exports will
be generated, fixing this issue.
PR Close#29193
Previously, when the NgModule scope resolver discovered semantic errors
within a users NgModules, it would throw assertion errors. TODOs in the
codebase indicated these should become ts.Diagnostics eventually.
Besides producing better-looking errors, there is another reason to make
this change asap: these assertions were shadowing actual errors, via an
interesting mechanism:
1) a component would produce a ts.Diagnostic during its analyze() step
2) as a result, it wouldn't register component metadata with the scope
resolver
3) the NgModule for the component references it in exports, which was
detected as an invalid export (no metadata registering it as a
component).
4) the resulting assertion error would crash the compiler, hiding the
real cause of the problem (an invalid component).
This commit should mitigate this problem by converting scoping errors to
proper ts.Diagnostics. Additionally, we should consider registering some
marker indicating a class is a directive/component/pipe without actually
requiring full metadata to be produced for it, which would allow suppression
of errors like "invalid export" for such invalid types.
PR Close#29191
This commit refactors and expands ngtsc's support for generating imports of
values from imports of types (this is used for example when importing a
class referenced in a type annotation in a constructor).
Previously, this logic handled "import {Foo} from" and "import * as foo
from" style imports, but failed on imports of default values ("import
Foo from"). This commit moves the type-to-value logic to a separate file and
expands it to cover the default import case. Doing this also required
augmenting the ImportManager to track default as well as non-default import
generation. The APIs were made a little cleaner at the same time.
PR Close#29146
In the TypeScript compiler API, emit() can be performed either on a single
ts.SourceFile or on the entire ts.Program simultaneously.
ngtsc previously used whole-program emit, which was convenient to use while
spinning up the project but has a significant drawback: it causes a type
checking operation to occur for the whole program, including .d.ts files.
In large Bazel environments (such as Google's codebase), an ngtsc invocation
can have a few .ts files and thousands of .d.ts inputs. This unwanted type
checking is therefore a significant drain on performance.
This commit switches ngtsc to emit each .ts file individually, avoiding the
unwanted type checking.
PR Close#29147
When processing a JavaScript program, TS may come across a symbol that has
been imported from a TypeScript typings file.
In this case the compiler may pass the ReflectionHost a `prototype` symbol
as an export of the class.
This pseudo-member symbol has no declarations, which previously caused the
code in `Esm5ReflectionHost.reflectMembers()` to crash.
Now we just quietly ignore such a symbol and leave `Esm2015ReflectionHost`
to deal with it.
(As it happens `Esm2015ReflectionHost` also quietly ignores this symbol).
PR Close#29158
For the template type checking to work correctly, it needs to know
what attributes are bound to expressions or directives, which may
require expressions in the template to be evaluated in a different
scope.
In inline templates, there are attributes that are now marked as
"Template" attributes. We need to ensure that the template
type checking code looks at these "bound" attributes as well as the
"input" attributes.
PR Close#29041
ngtsc occasionally converts a type reference (such as the type of a
parameter in a constructor) to a value reference (argument to a
directiveInject call). TypeScript has a bad habit of sometimes removing
the import statement associated with this type reference, because it's a
type only import when it initially looks at the file.
A solution to this is to always add an import to refer to a type position
value that's imported, and not rely on the existing import.
PR Close#29111
When ngtsc generates a .ngfactory shim, it does so based on the contents of
an original file in the program. Occasionally these original files have
comments at the top which are load-bearing (e.g. they contain jsdoc
annotations which are significant to downstream bundling tools). The
generated factory shims should preserve this comment.
This commit adds a step to the ngfactory generator to preserve the top-level
comment from the original source file.
FW-1006 #resolve
FW-1095 #resolve
PR Close#29065
The ngtsc partial evaluator previously would not handle an enum reference
inside a template string expression correctly. Enums are resolved to an
`EnumValue` type, which has a `resolved` property with the actual value.
When effectively toString-ing a `ResolvedValue` as part of visiting a
template expression, the partial evaluator needs to translate `EnumValue`s
to their fully resolved value, which this commit does.
PR Close#29062
Currently, ngtsc has a bug where if you alias the name of a decorator when
importing it, it won't be detected properly. This is because the compiler
uses the aliased name and not the original, declared name of the decorator
for detection.
This commit fixes the compiler to compare against the declared name of
decorators when available, and adds a test to prevent regression.
PR Close#29061
ngtsc has cyclic import detection, to determine when adding an import to a
directive or pipe would create a cycle. However, this detection must also
account for already inserted imports, as it's possible for both directions
of a circular import to be inserted by Ivy (as opposed to at least one of
those edges existing in the user's program).
This commit fixes the circular import detection for components to take into
consideration already added edges. This is difficult for one critical
reason: only edges to files which will *actually* be imported should be
considered. However, that depends on which directives & pipes are used in
a given template, which is currently only known by running the
TemplateDefinitionBuilder during the 'compile' phase. This is too late; the
decision whether to use remote scoping (which consults the import graph) is
made during the 'resolve' phase, before any compilation has taken place.
Thus, the only way to correctly consider synthetic edges is for the compiler
to know exactly which directives & pipes are used in a template during
'resolve'. There are two ways to achieve this:
1) refactor `TemplateDefinitionBuilder` to do its work in two phases, with
directive matching occurring as a separate step which can be performed
earlier.
2) use the `R3TargetBinder` in the 'resolve' phase to independently bind the
template and get information about used directives.
Option 1 is ideal, but option 2 is currently used for practical reasons. The
cost of binding the template can be shared with template-typechecking.
PR Close#29040
In the @Component decorator, the 'host' field is an object which represents
host bindings. The type of this field is complex, but is generally of the
form {[key: string]: string}. Several different kinds of bindings can be
specified, depending on the structure of the key.
For example:
```
@Component({
host: {'[prop]': 'someExpr'}
})
```
will bind an expression 'someExpr' to the property 'prop'. This is known to
be a property binding because of the square brackets in the binding key.
If the binding key is a plain string (no brackets or parentheses), then it
is known as an attribute binding. In this case, the right-hand side is not
interpreted as an expression, but is instead a constant string.
There is no actual requirement that at build time, these constant strings
are known to the compiler, but this was previously enforced as a side effect
of requiring the binding expressions for property and event bindings to be
statically known (as they need to be parsed). This commit breaks that
relationship and allows the attribute bindings to be dynamic. In the case
that they are dynamic, the references to the dynamic values are reflected
into the Ivy instructions for attribute bindings.
PR Close#29033
DynamicValues are generated whenever a partially evaluated expression is
unable to be resolved statically. They contain a reference to the ts.Node
which wasn't resolvable.
They can also be nested. For example, the expression 'a + b' is resolvable
only if 'a' and 'b' are themselves resolvable. If either 'a' or 'b' resolve
to a DynamicValue, the whole expression must also resolve to a DynamicValue.
Previously, if 'a' resolved to a DynamicValue, the entire expression might
have been resolved to the same DynamicValue. This correctly indicated that
the expression wasn't resolvable, but didn't return a reference to the
shallow node that couldn't be resolved (the expression 'a + b'), only a
reference to the deep node that couldn't be resolved ('a').
In certain situations, it's very useful to know the shallow unresolvable
node (for example, to use it verbatim in the output). To support this,
the partial evaluator is updated to always wrap DynamicValue to point to
each unresolvable expression as it's processed, ensuring the receiver can
determine exactly which expression node failed to resolve.
PR Close#29033
This change helps highlight certain misoptimizations with Closure
compiler. It is also stylistically preferable to consistently use index
access on index sig types.
Roughly, when one sees '.foo' they know it is always checked for typos
in the prop name by the type system (unless 'any'), while "['foo']" is
always not.
Once all angular repos are conforming this will become a tsetse.info
check, enforced by bazel.
PR Close#28937
Prior to this change, TypeScript stripped out some imports in case we reference a type that can be represented as a value (for ex. classes). This fix ensures that we use correct symbol identifier, which makes TypeScript retain the necessary import statements.
PR Close#28941
Angular supports using <style> and <link> tags inline in component
templates, but previously such tags were not implemented within the ngtsc
compiler. This commit introduces that support.
FW-1069 #resolve
PR Close#28997
This commit introduces support for the windows paths in the new concrete types mechanism that was introduced in this PR https://github.com/angular/angular/pull/28523
Normalized posix paths that start with either a `/` or `C:/` are considered to be an absolute path.
Note: `C:/` is used as a reference, as other drive letters are also supported.
Fixes#28754
PR Close#28752
The partial evaluator in ngtsc can handle a shorthand property declaration
in the middle evaluation, but fails if evaluation starts at the shorthand
property itself. This is because evaluation starts at the ts.Identifier
of the property (the ts.Expression representing it), not the ts.Declaration
for the property.
The fix for this is to detect in TypeScriptReflectionHost when a ts.Symbol
refers to a shorthand property, and to use the ts.TypeChecker method
getShorthandAssignmentValueSymbol() to resolve the value of the assignment
instead.
FW-1089 #resolve
PR Close#28936
In certain configurations (such as the g3 repository) which have lots of
small compilation units as well as strict dependency checking on generated
code, ngtsc's default strategy of directly importing directives/pipes into
components will not work. To handle these cases, an additional mode is
introduced, and is enabled when using the FileToModuleHost provided by such
compilation environments.
In this mode, when ngtsc encounters an NgModule which re-exports another
from a different file, it will re-export all the directives it contains at
the ES2015 level. The exports will have a predictable name based on the
FileToModuleHost. For example, if the host says that a directive Foo is
from the 'root/external/foo' module, ngtsc will add:
```
export {Foo as ɵng$root$external$foo$$Foo} from 'root/external/foo';
```
Consumers of the re-exported directive will then import it via this path
instead of directly from root/external/foo, preserving strict dependency
semantics.
PR Close#28852
This commit splits apart selector_scope.ts in ngtsc and extracts the logic
into two separate classes, the LocalModuleScopeRegistry and the
DtsModuleScopeResolver. The logic is cleaned up significantly and new tests
are added to verify behavior.
LocalModuleScopeRegistry implements the NgModule semantics for compilation
scopes, and handles NgModules declared in the current compilation unit.
DtsModuleScopeResolver implements simpler logic for export scopes and
handles NgModules declared in .d.ts files.
This is done in preparation for the addition of re-export logic to solve
StrictDeps issues.
PR Close#28852
Prior to this change absolute file paths (like `/a/b/c/style.css`) were calculated taking current component file location into account. As a result, absolute file paths were calculated using current file as a root. This change updates this logic to ignore current file path in case of absolute paths.
PR Close#28789
Prior to this change, Ivy and VE CSS resource resolution was different: in addition to specified styleUrl (with .scss, .less and .styl extensions), VE also makes an attempt to resolve resource with .css extension. This change introduces similar logic for Ivy to make sure Ivy behavior is backwards compatible.
PR Close#28770
Prior to this change, the @fileoverview annotations added by users in source files or by tsickle during compilation might have change a location due to the fact that Ngtsc may prepend extra imports or constants. As a result, the output file is considered invalid by Closure (misplaced @fileoverview annotation). In order to resolve the problem we relocate @fileoverview annotation if we detect that its host node shifted.
PR Close#28723
This change is kind of similar to #27466, but instead of ensuring that
these shims can be generated, we also need to make sure that developers
are able to also use the factory shims like with `ngc`.
This issue is now surfacing because we have various old examples which
are now also built with `ngtsc` (due to the bazel migration). On case insensitive
platforms (e.g. windows) these examples cannot be built because ngtsc fails
the app imports a generated shim file (such as the factory shim files).
This is because the `GeneratedShimsHostWrapper` TypeScript host uses
the `getCanonicalFileName` method in order to check whether a given
file/module exists in the generator file maps. e.g.
```
// Generator Map:
'C:/users/paul/_bazel_paul/lm3s4mgv/execroot/angular/packages/core/index.ngfactory.ts' =>
'C:/users/paul/_bazel_paul/lm3s4mgv/execroot/angular/packages/core/index.ts',
// Path passed into `fileExists`
C:/users/paul/_bazel_paul/lm3s4mgv/execroot/angular/packages/core/index.ngfactory.ts
// After getCanonicalFileName (notice the **lower-case drive name**)
c:/users/paul/_bazel_paul/lm3s4mgv/execroot/angular/packages/core/index.ngfactory.ts
```
As seen above, the generator map does not use the canonical file names, as well as
TypeScript internally does not pass around canonical file names. We can fix this by removing
the manual call to `getCanonicalFileName` and just following TypeScript internal-semantics.
PR Close#28831
Pete Bacon Darwin
Alex Eagle
Alex Rickabaugh
George Kalpakas
Paul Gschwendtner
Kristiyan Kostadinov
Igor Minar
Andrew Kushnir
Alan
Greg Magolan
Rado Kirov
Wassim Chegham