[SourceKit] Add a request to get the semantic tokens of a file #69294
Conversation
ahoppen
requested review from
bnbarham,
rintaro and
hamishknight
as code owners
|
@swift-ci Please smoke test |
|
@swift-ci Please build toolchain macOS |
ahoppen
force-pushed
the
ahoppen/semantic-tokens
branch
2 times, most recently
from
ff55f00 to
8441396
Compare
|
@swift-ci Please smoke test |
ahoppen
force-pushed
the
ahoppen/semantic-tokens
branch
from
8441396 to
e543961
Compare
We need this request for semantic highlighting in LSP. Previously, we were getting the semantic tokens using a 0,0 edit after a document update notification but that will no longer be possible if we open the documents in syntactic only mode.
ahoppen
force-pushed
the
ahoppen/semantic-tokens
branch
from
e543961 to
4bc03f8
Compare
|
@swift-ci Please smoke test |
| @@ -0,0 +1,17 @@ | |||
| // RUN: %sourcekitd-test -req=semantic-tokens %s -- %s | %FileCheck %s | |||
|
|
|||
There was a problem hiding this comment.
Do we get semantic tokens of buffers within a primary file?
There was a problem hiding this comment.
We do now. Added 1776e37
|
@swift-ci Please smoke test |
| // IN_BUFFER: source.lang.swift.ref.struct | ||
| // IN_BUFFER: key.kind: source.lang.swift.ref.class | ||
|
|
||
| // Check that we don't get semantic tokens inside the buffer when requesting semantic tokens for the outer type |
There was a problem hiding this comment.
| // Check that we don't get semantic tokens inside the buffer when requesting semantic tokens for the outer type | |
| // Check that we don't get semantic tokens inside the buffer when requesting semantic tokens for the outer file |
| if (!InvocationError.empty()) { | ||
| LOG_WARN_FUNC("error creating ASTInvocation: " << InvocationError); | ||
| } |
There was a problem hiding this comment.
Where does this actually show up? Is it ever more obvious than just the actual error response?
There was a problem hiding this comment.
This is copied from
and
Not sure if it’s more obvious but it’s also not less obvious so 🤷🏽
lib/IDE/SourceEntityWalker.cpp
Outdated
| } else if (auto ME = dyn_cast<MacroExpansionExpr>(E)) { | ||
| // Add a reference to the macro | ||
| auto macroRef = ME->getMacroRef(); | ||
| if (auto *macroDecl = dyn_cast_or_null<MacroDecl>(macroRef.getDecl())) { | ||
| auto macroRefType = macroDecl->getDeclaredInterfaceType(); | ||
| if (!passReference( | ||
| macroDecl, macroRefType, ME->getMacroNameLoc(), | ||
| ReferenceMetaData(SemaReferenceKind::DeclRef, llvm::None))) | ||
| return Action::Stop(); | ||
| } |
There was a problem hiding this comment.
Out of curiosity, why remove this? Wouldn't this impact things like cursor info?
There was a problem hiding this comment.
My assumption was that we end up passing the reference through some other means. But based on the failing test, seems like that isn't the case.
There was a problem hiding this comment.
We were passing a reference twice and I thought removing this fixed it. But looks like I didn’t rebuild before running tests or something along those lines. I need to look into it again.
There was a problem hiding this comment.
OK, we needed to check that the MacroExpansionExpr didn’t have a SubstituteDecl. Could you double-check that the comment I wrote in my new version of this PR makes sense?
ahoppen
force-pushed
the
ahoppen/semantic-tokens
branch
from
1776e37 to
2507e0e
Compare
|
@swift-ci Please smoke test |
We need this request for semantic highlighting in LSP. Previously, we were getting the semantic tokens using a 0,0 edit after a document update notification but that will no longer be possible if we open the documents in syntactic only mode.