Previously, the compiler performed an incremental build by analyzing and
resolving all classes in the program (even unchanged ones) and then using
the dependency graph information to determine which .js files were stale and
needed to be re-emitted. This algorithm produced "correct" rebuilds, but the
cost of re-analyzing the entire program turned out to be higher than
anticipated, especially for component-heavy compilations.
To achieve performant rebuilds, it is necessary to reuse previous analysis
results if possible. Doing this safely requires knowing when prior work is
viable and when it is stale and needs to be re-done.
The new algorithm implemented by this commit is such:
1) Each incremental build starts with knowledge of the last known good
dependency graph and analysis results from the last successful build,
plus of course information about the set of files changed.
2) The previous dependency graph's information is used to determine the
set of source files which have "logically" changed. A source file is
considered logically changed if it or any of its dependencies have
physically changed (on disk) since the last successful compilation. Any
logically unchanged dependencies have their dependency information copied
over to the new dependency graph.
3) During the `TraitCompiler`'s loop to consider all source files in the
program, if a source file is logically unchanged then its previous
analyses are "adopted" (and their 'register' steps are run). If the file
is logically changed, then it is re-analyzed as usual.
4) Then, incremental build proceeds as before, with the new dependency graph
being used to determine the set of files which require re-emitting.
This analysis reuse avoids template parsing operations in many circumstances
and significantly reduces the time it takes ngtsc to rebuild a large
application.
Future work will increase performance even more, by tackling a variety of
other opportunities to reuse or avoid work.
PR Close#34288
Due to the fact that Tsickle runs between analyze and transform phases in Angular, Tsickle may transform nodes (add comments with type annotations for Closure) that we captured during the analyze phase. As a result, some patterns where a function is returned from another function may trigger automatic semicolon insertion, which breaks the code (makes functions return `undefined` instead of a function). In order to avoid the problem, this commit updates the code to wrap all functions in some expression ("privders" and "viewProviders") in parentheses. More info can be found in Tsickle source code here: d797426257/src/jsdoc_transformer.ts (L1021)
PR Close#33609
To improve cross platform support, all file access (and path manipulation)
is now done through a well known interface (`FileSystem`).
For testing a number of `MockFileSystem` implementations are provided.
These provide an in-memory file-system which emulates operating systems
like OS/X, Unix and Windows.
The current file system is always available via the static method,
`FileSystem.getFileSystem()`. This is also used by a number of static
methods on `AbsoluteFsPath` and `PathSegment`, to avoid having to pass
`FileSystem` objects around all the time. The result of this is that one
must be careful to ensure that the file-system has been initialized before
using any of these static methods. To prevent this happening accidentally
the current file system always starts out as an instance of `InvalidFileSystem`,
which will throw an error if any of its methods are called.
You can set the current file-system by calling `FileSystem.setFileSystem()`.
During testing you can call the helper function `initMockFileSystem(os)`
which takes a string name of the OS to emulate, and will also monkey-patch
aspects of the TypeScript library to ensure that TS is also using the
current file-system.
Finally there is the `NgtscCompilerHost` to be used for any TypeScript
compilation, which uses a given file-system.
All tests that interact with the file-system should be tested against each
of the mock file-systems. A series of helpers have been provided to support
such tests:
* `runInEachFileSystem()` - wrap your tests in this helper to run all the
wrapped tests in each of the mock file-systems.
* `addTestFilesToFileSystem()` - use this to add files and their contents
to the mock file system for testing.
* `loadTestFilesFromDisk()` - use this to load a mirror image of files on
disk into the in-memory mock file-system.
* `loadFakeCore()` - use this to load a fake version of `@angular/core`
into the mock file-system.
All ngcc and ngtsc source and tests now use this virtual file-system setup.
PR Close#30921
Previously, metadata registration (the recording of collected metadata
during analysis of directives, pipes, and NgModules) was only used to
produce the `LocalModuleScope`, and thus was handled by the
`LocalModuleScopeRegistry`.
However, the template type-checker also needs information about registered
directives, outside of the NgModule scope determinations. Rather than
reuse the scope registry for an unintended purpose, this commit introduces
new abstractions for metadata registration and lookups in a separate
'metadata' package, which the scope registry implements.
This paves the way for a future commit to make use of this metadata for the
template type-checking system.
Testing strategy: this commit is a refactoring which introduces no new
functionality, so existing tests are sufficient.
PR Close#29698
Previously, several `ngtsc` and `ngcc` APIs dealing with class
declaration nodes used inconsistent types. For example, some methods of
the `DecoratorHandler` interface expected a `ts.Declaration` argument,
but actual `DecoratorHandler` implementations specified a stricter
`ts.ClassDeclaration` type.
As a result, the stricter methods would operate under the incorrect
assumption that their arguments were of type `ts.ClassDeclaration`,
while the actual arguments might be of different types (e.g. `ngcc`
would call them with `ts.FunctionDeclaration` or
`ts.VariableDeclaration` arguments, when compiling ES5 code).
Additionally, since we need those class declarations to be referenced in
other parts of the program, `ngtsc`/`ngcc` had to either repeatedly
check for `ts.isIdentifier(node.name)` or assume there was a `name`
identifier and use `node.name!`. While this assumption happens to be
true in the current implementation, working around type-checking is
error-prone (e.g. the assumption might stop being true in the future).
This commit fixes this by introducing a new type to be used for such
class declarations (`ts.Declaration & {name: ts.Identifier}`) and using
it consistently throughput the code.
PR Close#29209
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
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
Previously, `ngtsc` detected class inheritance in a way that only worked
in TS or ES2015 code. As a result, inheritance would not be detected for
code in ES5 format, such as when running `ngtsc` through `ngcc` to
transform old-style Angular code to ivy format.
This commit fixes it by delegating class inheritance detection to the
current `ReflectionHost`, which is able to correctly interpret the used
code format.
PR Close#28773
