Computing the type relation for every item in the code completion cache is way to expensive (~4x slowdown for global completion that imports `SwiftUI`). Instead, compute a type’s supertypes (protocol conformances and superclasses) once and write their USRs to the cache. To compute a type relation we can then check if the contextual type is in the completion item’s supertypes.
This reduces the overhead of computing the type relations (again global completion that imports `SwiftUI`) to ~6% – measured by instructions executed.
Technically, we might miss some conversions like
- retroactive conformances inside another module (because we can’t cache them if that other module isn’t imported)
- complex generic conversions (just too complicated to model using USRs)
Because of this, we never report an `unrelated` type relation for global items but always default to `unknown`.
But I believe this change covers the most common cases and is a good tradeoff between accuracy and performance.
rdar://83846531
Custom attributes were not printed because they are marked
'UserInaccesible'.
* Make CustomAttr 'RejectByParser' instead of 'UserInaccessible'
* Remove special treatment for Result Builder attributes
* Load implicit modules in module/header interface gen requests
rdar://79927502
Computing the type relation for every item in the code completion cache is way to expensive (~4x slowdown for global completion that imports `SwiftUI`). Instead, compute a type’s supertypes (protocol conformances and superclasses) once and write their USRs to the cache. To compute a type relation we can then check if the contextual type is in the completion item’s supertypes.
This reduces the overhead of computing the type relations (again global completion that imports `SwiftUI`) to ~6% – measured by instructions executed.
Technically, we might miss some conversions like
- retroactive conformances inside another module (because we can’t cache them if that other module isn’t imported)
- complex generic conversions (just too complicated to model using USRs)
Because of this, we never report an `unrelated` type relation for global items but always default to `unknown`.
But I believe this change covers the most common cases and is a good tradeoff between accuracy and performance.
rdar://83846531
`OverlayFileSystem` is very simple - it just passes along each request
to each VFS it contains until one is successful (or none are). Use it
when wrapping the VFS to pass down into the Clang invocation creation,
instead of the much more complicated `RedirectingFileSystem`.
This has the side effect of also fixing a case where due to a bug in
`RedirectingFileSystem`, the originally passed in path will be returned
regardless of `use-external-name`. While that should also be fixed,
there is no reason to use this VFS here anyway.
Added a small cursor info test case that should catch issues like this
in the future.
There was a recent change to error early if the stdlib couldn't be
loaded when semantic functionality is required. Since
`ide::initCompilerInvocation` was ignoring the given `Action` and just
setting to `Typecheck` instead, this would cause an error even if
semantic functionality *wasn't* required.
Resolves rdar://88968608.
If a method has an `async` variant, the non-`async` variant will now mark its completion handler parameter `@Sendable`. This shouldn't be a breaking change in Swift 5 code since these declarations are automatically `@_predatesConcurrency`.
Also adds:
• Support for `@_nonSendable` on parameters, which can be used to override this implicit `@Sendable`
• Support for `@Sendable` on block typedefs; it's generally going to be a good idea to mark completion block typedefs `@Sendable`.
Fixes rdar://85569247.
Ultimately this is to support the disambiguation of protocol requirements when printing stubs. This allows us to disambiguate the case where two modules declare a nominal type, and when that type appears in a protocol requirement. In such a case, we now fully qualify the types involved.
Fixing this also appears to now be consistently printing module qualification in many more places, hence the updates to the IDE/SourceKit tests.
rdar://72830118
LLVM arguments processing is a process global state. At this point,
there's no way to localize it to a CompilerInstance. For now, ignore any
LLVM arguemnts, so options from another compilation doesn't affect other
compilations.
rdar://86809003
Only declarations in the same module as synthesized extension's target
are placed within a synthesized extension. We should thus not add
"::SYNTHESIZED::" to the USR if the given declaration is in a different
and.
As an example, `Foundation` adds a method `components(separatedBy:)` to
`String` through an extension on `StringProtocol`. But since it is
within `Foundation` and not `Swift` it will *not* be in a synthesized
extension of `String` or `StringProtocol`. So it should not have
"::SYNTHESIZED::" added and should also not being in the `String` group.
Resolves rdar://71355632.
When opening a file for the first time, we don’t store a snapshot for it. This could cause a crash when trying to consult its snapshot to see whether an AST can be reused for cursor info.
Instead of checking that the stdlib can be loaded in a variety of places, check it when setting up the compiler instance. This required a couple more checks to avoid loading the stdlib in cases where it’s not needed.
To be able to differentiate stdlib loading failures from other setup errors, make `CompilerInstance::setup` return an error message on failure via an inout parameter. Consume that error on the call side, replacing a previous, more generic error message, adding error handling where appropriate or ignoring the error message, depending on the context.
Essentially, just wire up cancellation tokens and cancellation flags for `CompletionInstance` and make sure to return `CancellableResult::cancelled()` when cancellation is detected.
rdar://83391488
Currently, we were building an AST on document open or edit even if
- `key_enablesyntaxmap` = 0
- `key_enablesubstructure` = 0
- `key_enablediagnostics` = 0 and
- syntax tree transfer mode is off
In those cases we were just ignoring the result.
If all of the options are 0, don’t build an AST.
rdar://85847659
Arguments in `SubscriptExpr` are visited since the recent `ArgumentList`
refactoring, but were being added to the containing `CallExpr`. Add a
node for the `SubscriptExpr` itself so that its argument is added there
instead of the `CallExpr`.
Also remove `key.nameoffset` and `key.namelength` from the response when
both are 0 to match the rest of the offsets and lengths.
Resolves rdar://85412164.
An explicit swift_attr("@_nonSendable") will override it (except for ns_error_domain where the type is embedded in another type that's forced to be Sendable), but swift_attr("@_nonSendable(_assumed)") will not.
An explicit swift_attr("@_nonSendable") will override it (except for ns_error_domain where the type is embedded in another type that's forced to be Sendable), but swift_attr("@_nonSendable(_assumed)") will not.
Previously, `SwiftASTManager` and `SlowRequestSimulator` maintained their own list of in-progress cancellation tokens. With code completion cancellation coming up, there would need to be yet another place to track in-progress requests, so let’s centralize it.
While at it, also support cancelling requests before they are scheduled, eliminating the need for a `sleep` in a test case.
The current implementaiton leaks tiny amounts of memory if a request is cancelled after if finishes. I think this is fine because it is a pretty nieche case and the leaked memory is pretty small (a `std::map` entry pointing to a `std::function` + `bool`). Alternatively, we could require the client to always dispose of the cancellation token manually.
A keypath using dynamic member lookup results in various `KeyPathExpr`
that have components with no location. Ignore these and any other
references that have a missing location.
Resolves rdar://85237365
