This commit replaces the `IncrementalDriver` abstraction which powered
incremental compilation in the compiler with a new `IncrementalCompilation`
design. Principally, it separates two concerns which were tied together in
the previous implementation:
1. Tracking the reusable state of a compilation at any given point that
could be reused in a subsequent future compilation.
2. Making use of a prior compilation's state to accelerate the current one.
The new abstraction adds explicit tracking and types to deal with both of
these concerns separately, which greatly reduces the complexity of the state
tracking that `IncrementalDriver` used to perform.
PR Close#41475
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
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
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
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
To avoid overwhelming a user with secondary diagnostics that derive from a
"root cause" error, the compiler has the notion of a "poisoned" NgModule.
An NgModule becomes poisoned when its declaration contains semantic errors:
declarations which are not components or pipes, imports which are not other
NgModules, etc. An NgModule also becomes poisoned if it imports or exports
another poisoned NgModule.
Previously, the compiler tracked this poisoned status as an alternate state
for each scope. Either a correct scope could be produced, or the entire
scope would be set to a sentinel error value. This meant that the compiler
would not track any information about a scope that was determined to be in
error.
This method presents several issues:
1. The compiler is unable to support the language service and return results
when a component or its module scope is poisoned.
This is fine for compilation, since diagnostics will be produced showing the
error(s), but the language service needs to still work for incorrect code.
2. `getComponentScopes()` does not return components with a poisoned scope,
which interferes with resource tracking of incremental builds.
If the component isn't included in that list, then the NgModule for it will
not have its dependencies properly tracked, and this can cause future
incremental build steps to produce incorrect results.
This commit changes the tracking of poisoned module scopes to use a flag on
the scope itself, rather than a sentinel value that replaces the scope. This
means that the scope itself will still be tracked, even if it contains
semantic errors. A test is added to the language service which verifies that
poisoned scopes can still be used in template type-checking.
PR Close#39923
This commit adds new language service testing infrastructure which allows
for in-memory testing. It solves a number of issues with the previous
testing infrastructure that relied on a single integration project across
all of the tests, and also provides for much faster builds by using
the compiler-cli's mock versions of @angular/core and @angular/common.
A new `LanguageServiceTestEnvironment` class (conceptually mirroring the
compiler-cli `NgtscTestEnvironment`) controls setup and execution of tests.
The `FileSystem` abstraction is used to drive a `ts.server.ServerHost`,
which backs the language service infrastructure.
Since many language service tests revolve around the template, the API is
currently optimized to spin up a "skeleton" project and then override its
template for each test.
The existing Quick Info tests (quick_info_spec.ts) were ported to the new
infrastructure for validation. The tests were cleaned up a bit to remove
unnecessary initializations as well as correct legitimate template errors
which did not affect the test outcome, but caused additional validation of
test correctness to fail. They still utilize a shared project with all
fields required for each individual unit test, which is an anti-pattern, but
new tests can now easily be written independently without relying on the
shared project, which was extremely difficult previously. Future cleanup
work might refactor these tests to be more independent.
PR Close#39594
This commit enables the Ivy Language Service to 'go to definition' of a
templateUrl or styleUrl, which would jump to the template/style file
itself.
PR Close#39202
This commit fixes a bug in which a new Ivy Compiler is created every time
language service receives a new request. This is not needed if the
`ts.Program` has not changed.
A new class `CompilerFactory` is created to manage Compiler lifecycle and
keep track of template changes so that it knows when to override them.
With this change, we no longer need the method `getModifiedResourceFile()`
on the adapter. Instead, we call `overrideComponentTemplate` on the
template type checker.
This commit also changes the incremental build strategy from
`PatchedIncrementalBuildStrategy` to `TrackedIncrementalBuildStrategy`.
PR Close#39231
Alex Rickabaugh
Keen Yee Liau
ayazhafiz
Andrew Scott