[clang][Index][USR][NFC] Allow customizing langopts for USR generation

[steakhal]

This helps to produce USRs for custom LangOpts - that differ from the one associated with the given Decl. This can unlock usecases in tooling opportunities that we have downstream.

This is NFC because existing calls will still result in the right overload, thus using the LangOpts associated with the ASTContext of the Decls and Types.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Balazs Benics (steakhal)

Changes

This helps to produce USRs for custom LangOpts - that differ from the one associated with the given Decl. This can unlock usecases in tooling opportunities that we have downstream.

This is NFC because existing calls will still result in the right overload, thus using the LangOpts associated with the ASTContext of the Decls and Types.

---

Full diff: https://github.com/llvm/llvm-project/pull/109574.diff

2 Files Affected:

• (modified) clang/include/clang/Index/USRGeneration.h (+7-1)
• (modified) clang/lib/Index/USRGeneration.cpp (+29-19)

diff --git a/clang/include/clang/Index/USRGeneration.h b/clang/include/clang/Index/USRGeneration.h
index f89fc5cf49302c..61d267f3545a70 100644
--- a/clang/include/clang/Index/USRGeneration.h
+++ b/clang/include/clang/Index/USRGeneration.h
@@ -15,6 +15,7 @@
 namespace clang {
 class ASTContext;
 class Decl;
+class LangOptions;
 class MacroDefinitionRecord;
 class Module;
 class SourceLocation;
@@ -30,6 +31,8 @@ static inline StringRef getUSRSpacePrefix() {
 /// Generate a USR for a Decl, including the USR prefix.
 /// \returns true if the results should be ignored, false otherwise.
 bool generateUSRForDecl(const Decl *D, SmallVectorImpl<char> &Buf);
+bool generateUSRForDecl(const Decl *D, SmallVectorImpl<char> &Buf,
+                        const LangOptions &LangOpts);
 
 /// Generate a USR fragment for an Objective-C class.
 void generateUSRForObjCClass(StringRef Cls, raw_ostream &OS,
@@ -75,7 +78,10 @@ bool generateUSRForMacro(StringRef MacroName, SourceLocation Loc,
 /// Generates a USR for a type.
 ///
 /// \return true on error, false on success.
-bool generateUSRForType(QualType T, ASTContext &Ctx, SmallVectorImpl<char> &Buf);
+bool generateUSRForType(QualType T, ASTContext &Ctx,
+                        SmallVectorImpl<char> &Buf);
+bool generateUSRForType(QualType T, ASTContext &Ctx, SmallVectorImpl<char> &Buf,
+                        const LangOptions &LangOpts);
 
 /// Generate a USR for a module, including the USR prefix.
 /// \returns true on error, false on success.
diff --git a/clang/lib/Index/USRGeneration.cpp b/clang/lib/Index/USRGeneration.cpp
index f00bc56429f1a7..1cfca8b6aaa368 100644
--- a/clang/lib/Index/USRGeneration.cpp
+++ b/clang/lib/Index/USRGeneration.cpp
@@ -62,20 +62,17 @@ namespace {
 class USRGenerator : public ConstDeclVisitor<USRGenerator> {
   SmallVectorImpl<char> &Buf;
   llvm::raw_svector_ostream Out;
-  bool IgnoreResults;
   ASTContext *Context;
-  bool generatedLoc;
+  const LangOptions &LangOpts;
+  bool IgnoreResults = false;
+  bool generatedLoc = false;
 
   llvm::DenseMap<const Type *, unsigned> TypeSubstitutions;
 
 public:
-  explicit USRGenerator(ASTContext *Ctx, SmallVectorImpl<char> &Buf)
-  : Buf(Buf),
-    Out(Buf),
-    IgnoreResults(false),
-    Context(Ctx),
-    generatedLoc(false)
-  {
+  USRGenerator(ASTContext *Ctx, SmallVectorImpl<char> &Buf,
+               const LangOptions &LangOpts)
+      : Buf(Buf), Out(Buf), Context(Ctx), LangOpts(LangOpts) {
     // Add the USR space prefix.
     Out << getUSRSpacePrefix();
   }
@@ -246,14 +243,13 @@ void USRGenerator::VisitFunctionDecl(const FunctionDecl *D) {
   } else
     Out << "@F@";
 
-  PrintingPolicy Policy(Context->getLangOpts());
+  PrintingPolicy Policy(LangOpts);
   // Forward references can have different template argument names. Suppress the
   // template argument names in constructors to make their USR more stable.
   Policy.SuppressTemplateArgsInCXXConstructors = true;
   D->getDeclName().print(Out, Policy);
 
-  ASTContext &Ctx = *Context;
-  if ((!Ctx.getLangOpts().CPlusPlus || D->isExternC()) &&
+  if ((!LangOpts.CPlusPlus || D->isExternC()) &&
       !D->hasAttr<OverloadableAttr>())
     return;
 
@@ -657,9 +653,10 @@ bool USRGenerator::GenLoc(const Decl *D, bool IncludeOffset) {
   return IgnoreResults;
 }
 
-static void printQualifier(llvm::raw_ostream &Out, ASTContext &Ctx, NestedNameSpecifier *NNS) {
+static void printQualifier(llvm::raw_ostream &Out, const LangOptions &LangOpts,
+                           NestedNameSpecifier *NNS) {
   // FIXME: Encode the qualifier, don't just print it.
-  PrintingPolicy PO(Ctx.getLangOpts());
+  PrintingPolicy PO(LangOpts);
   PO.SuppressTagKeyword = true;
   PO.SuppressUnwrittenScope = true;
   PO.ConstantArraySizeAsWritten = false;
@@ -948,7 +945,7 @@ void USRGenerator::VisitType(QualType T) {
     }
     if (const DependentNameType *DNT = T->getAs<DependentNameType>()) {
       Out << '^';
-      printQualifier(Out, Ctx, DNT->getQualifier());
+      printQualifier(Out, LangOpts, DNT->getQualifier());
       Out << ':' << DNT->getIdentifier()->getName();
       return;
     }
@@ -1090,7 +1087,7 @@ void USRGenerator::VisitUnresolvedUsingValueDecl(const UnresolvedUsingValueDecl
     return;
   VisitDeclContext(D->getDeclContext());
   Out << "@UUV@";
-  printQualifier(Out, D->getASTContext(), D->getQualifier());
+  printQualifier(Out, LangOpts, D->getQualifier());
   EmitDeclName(D);
 }
 
@@ -1099,7 +1096,7 @@ void USRGenerator::VisitUnresolvedUsingTypenameDecl(const UnresolvedUsingTypenam
     return;
   VisitDeclContext(D->getDeclContext());
   Out << "@UUT@";
-  printQualifier(Out, D->getASTContext(), D->getQualifier());
+  printQualifier(Out, LangOpts, D->getQualifier());
   Out << D->getName(); // Simple name.
 }
 
@@ -1190,6 +1187,13 @@ bool clang::index::generateUSRForDecl(const Decl *D,
                                       SmallVectorImpl<char> &Buf) {
   if (!D)
     return true;
+  return generateUSRForDecl(D, Buf, D->getASTContext().getLangOpts());
+}
+
+bool clang::index::generateUSRForDecl(const Decl *D, SmallVectorImpl<char> &Buf,
+                                      const LangOptions &LangOpts) {
+  if (!D)
+    return true;
   // We don't ignore decls with invalid source locations. Implicit decls, like
   // C++'s operator new function, can have invalid locations but it is fine to
   // create USRs that can identify them.
@@ -1203,7 +1207,7 @@ bool clang::index::generateUSRForDecl(const Decl *D,
       return false;
     }
   }
-  USRGenerator UG(&D->getASTContext(), Buf);
+  USRGenerator UG(&D->getASTContext(), Buf, LangOpts);
   UG.Visit(D);
   return UG.ignoreResults();
 }
@@ -1240,11 +1244,17 @@ bool clang::index::generateUSRForMacro(StringRef MacroName, SourceLocation Loc,
 
 bool clang::index::generateUSRForType(QualType T, ASTContext &Ctx,
                                       SmallVectorImpl<char> &Buf) {
+  return generateUSRForType(T, Ctx, Buf, Ctx.getLangOpts());
+}
+
+bool clang::index::generateUSRForType(QualType T, ASTContext &Ctx,
+                                      SmallVectorImpl<char> &Buf,
+                                      const LangOptions &LangOpts) {
   if (T.isNull())
     return true;
   T = T.getCanonicalType();
 
-  USRGenerator UG(&Ctx, Buf);
+  USRGenerator UG(&Ctx, Buf, LangOpts);
   UG.VisitType(T);
   return UG.ignoreResults();
 }

[steakhal]

For example the following code would get different USRs in C and C++:

static int f1(int* x);
// USR in C:   c:input.cc@F@f1
// USR in C++: c:input.cc@F@f1#*I#

So by having this patch, it would be possible to generate the C++ USR even if the AST (thus the ASTContext) was produced for a C compiler invocation.

[steakhal]

ping @sdkrystian

[steakhal]

Ping.

[steakhal]

Ping @AaronBallman

[steakhal]

@Endilll Could you please help me reaching out to the right set of reviewers?

[AaronBallman]

LGTM!

[Endilll]

@Endilll Could you please help me reaching out to the right set of reviewers?

Aaron noticed it faster 😄