[Clang][P1061] Add stuctured binding packs (#121417)

This is an implementation of P1061 Structure Bindings Introduce a Pack
without the ability to use packs outside of templates. There is a couple
of ways the AST could have been sliced so let me know what you think.
The only part of this change that I am unsure of is the
serialization/deserialization stuff. I followed the implementation of
other Exprs, but I do not really know how it is tested. Thank you for
your time considering this.

---------

Co-authored-by: Yanzuo Liu <[email protected]>

clang/include/clang/AST/Decl.h

@@ -698,6 +698,10 @@ class ValueDecl : public NamedDecl {
     return const_cast<ValueDecl *>(this)->getPotentiallyDecomposedVarDecl();
   }
 
+  /// Determine whether this value is actually a function parameter pack,
+  /// init-capture pack, or structured binding pack
+  bool isParameterPack() const;
+
   // Implement isa/cast/dyncast/etc.
   static bool classof(const Decl *D) { return classofKind(D->getKind()); }
   static bool classofKind(Kind K) { return K >= firstValue && K <= lastValue; }
@@ -1527,10 +1531,6 @@ class VarDecl : public DeclaratorDecl, public Redeclarable<VarDecl> {
     NonParmVarDeclBits.IsInitCapture = IC;
   }
 
-  /// Determine whether this variable is actually a function parameter pack or
-  /// init-capture pack.
-  bool isParameterPack() const;
-
   /// Whether this local extern variable declaration's previous declaration
   /// was declared in the same block scope. Only correct in C++.
   bool isPreviousDeclInSameBlockScope() const {

clang/include/clang/AST/DeclCXX.h

@@ -4175,31 +4175,32 @@ class BindingDecl : public ValueDecl {
   /// binding).
   Expr *Binding = nullptr;
 
-  BindingDecl(DeclContext *DC, SourceLocation IdLoc, IdentifierInfo *Id)
-      : ValueDecl(Decl::Binding, DC, IdLoc, Id, QualType()) {}
+  BindingDecl(DeclContext *DC, SourceLocation IdLoc, IdentifierInfo *Id,
+              QualType T)
+      : ValueDecl(Decl::Binding, DC, IdLoc, Id, T) {}
 
   void anchor() override;
 
 public:
   friend class ASTDeclReader;
 
   static BindingDecl *Create(ASTContext &C, DeclContext *DC,
-                             SourceLocation IdLoc, IdentifierInfo *Id);
+                             SourceLocation IdLoc, IdentifierInfo *Id,
+                             QualType T);
   static BindingDecl *CreateDeserialized(ASTContext &C, GlobalDeclID ID);
 
   /// Get the expression to which this declaration is bound. This may be null
   /// in two different cases: while parsing the initializer for the
   /// decomposition declaration, and when the initializer is type-dependent.
   Expr *getBinding() const { return Binding; }
 
+  // Get the array of Exprs when the binding represents a pack.
+  llvm::ArrayRef<Expr *> getBindingPackExprs() const;
+
   /// Get the decomposition declaration that this binding represents a
   /// decomposition of.
   ValueDecl *getDecomposedDecl() const { return Decomp; }
 
-  /// Get the variable (if any) that holds the value of evaluating the binding.
-  /// Only present for user-defined bindings for tuple-like types.
-  VarDecl *getHoldingVar() const;
-
   /// Set the binding for this BindingDecl, along with its declared type (which
   /// should be a possibly-cv-qualified form of the type of the binding, or a
   /// reference to such a type).
@@ -4211,6 +4212,10 @@ class BindingDecl : public ValueDecl {
   /// Set the decomposed variable for this BindingDecl.
   void setDecomposedDecl(ValueDecl *Decomposed) { Decomp = Decomposed; }
 
+  /// Get the variable (if any) that holds the value of evaluating the binding.
+  /// Only present for user-defined bindings for tuple-like types.
+  VarDecl *getHoldingVar() const;
+
   static bool classof(const Decl *D) { return classofKind(D->getKind()); }
   static bool classofKind(Kind K) { return K == Decl::Binding; }
 };
@@ -4238,8 +4243,16 @@ class DecompositionDecl final
         NumBindings(Bindings.size()) {
     std::uninitialized_copy(Bindings.begin(), Bindings.end(),
                             getTrailingObjects<BindingDecl *>());
-    for (auto *B : Bindings)
+    for (auto *B : Bindings) {
       B->setDecomposedDecl(this);
+      if (B->isParameterPack() && B->getBinding()) {
+        for (Expr *E : B->getBindingPackExprs()) {
+          auto *DRE = cast<DeclRefExpr>(E);
+          auto *NestedB = cast<BindingDecl>(DRE->getDecl());
+          NestedB->setDecomposedDecl(this);
+        }
+      }
+    }
   }
 
   void anchor() override;
@@ -4257,8 +4270,33 @@ class DecompositionDecl final
   static DecompositionDecl *CreateDeserialized(ASTContext &C, GlobalDeclID ID,
                                                unsigned NumBindings);
 
-  ArrayRef<BindingDecl *> bindings() const {
-    return llvm::ArrayRef(getTrailingObjects<BindingDecl *>(), NumBindings);
+  // Provide the range of bindings which may have a nested pack.
+  llvm::ArrayRef<BindingDecl *> bindings() const {
+    return {getTrailingObjects<BindingDecl *>(), NumBindings};
+  }
+
+  // Provide a flattened range to visit each binding.
+  auto flat_bindings() const {
+    llvm::ArrayRef<BindingDecl *> Bindings = bindings();
+    llvm::ArrayRef<Expr *> PackExprs;
+
+    // Split the bindings into subranges split by the pack.
+    auto S1 = Bindings.take_until(
+        [](BindingDecl *BD) { return BD->isParameterPack(); });
+
+    Bindings = Bindings.drop_front(S1.size());
+    if (!Bindings.empty()) {
+      PackExprs = Bindings.front()->getBindingPackExprs();
+      Bindings = Bindings.drop_front();
+    }
+
+    auto S2 = llvm::map_range(PackExprs, [](Expr *E) {
+      auto *DRE = cast<DeclRefExpr>(E);
+      return cast<BindingDecl>(DRE->getDecl());
+    });
+
+    return llvm::concat<BindingDecl *>(std::move(S1), std::move(S2),
+                                       std::move(Bindings));
   }
 
   void printName(raw_ostream &OS, const PrintingPolicy &Policy) const override;

clang/include/clang/AST/ExprCXX.h

@@ -5319,6 +5319,59 @@ class BuiltinBitCastExpr final
   }
 };
 
+// Represents an unexpanded pack where the list of expressions are
+// known. These are used when structured bindings introduce a pack.
+class ResolvedUnexpandedPackExpr final
+    : public Expr,
+      private llvm::TrailingObjects<ResolvedUnexpandedPackExpr, Expr *> {
+  friend class ASTStmtReader;
+  friend class ASTStmtWriter;
+  friend TrailingObjects;
+
+  SourceLocation BeginLoc;
+  unsigned NumExprs;
+
+  ResolvedUnexpandedPackExpr(SourceLocation BL, QualType QT, unsigned NumExprs);
+
+public:
+  static ResolvedUnexpandedPackExpr *CreateDeserialized(ASTContext &C,
+                                                        unsigned NumExprs);
+  static ResolvedUnexpandedPackExpr *
+  Create(ASTContext &C, SourceLocation BeginLoc, QualType T, unsigned NumExprs);
+  static ResolvedUnexpandedPackExpr *Create(ASTContext &C,
+                                            SourceLocation BeginLoc, QualType T,
+                                            llvm::ArrayRef<Expr *> Exprs);
+
+  unsigned getNumExprs() const { return NumExprs; }
+
+  llvm::MutableArrayRef<Expr *> getExprs() {
+    return {getTrailingObjects<Expr *>(), NumExprs};
+  }
+
+  llvm::ArrayRef<Expr *> getExprs() const {
+    return {getTrailingObjects<Expr *>(), NumExprs};
+  }
+
+  Expr *getExpansion(unsigned Idx) { return getExprs()[Idx]; }
+  Expr *getExpansion(unsigned Idx) const { return getExprs()[Idx]; }
+
+  // Iterators
+  child_range children() {
+    return child_range((Stmt **)getTrailingObjects<Expr *>(),
+                       (Stmt **)getTrailingObjects<Expr *>() + getNumExprs());
+  }
+
+  SourceLocation getBeginLoc() const LLVM_READONLY { return BeginLoc; }
+  SourceLocation getEndLoc() const LLVM_READONLY { return BeginLoc; }
+
+  // Returns the resolved pack of a decl or nullptr
+  static ResolvedUnexpandedPackExpr *getFromDecl(Decl *);
+
+  static bool classof(const Stmt *T) {
+    return T->getStmtClass() == ResolvedUnexpandedPackExprClass;
+  }
+};
+
 } // namespace clang
 
 #endif // LLVM_CLANG_AST_EXPRCXX_H

clang/include/clang/AST/RecursiveASTVisitor.h

@@ -2950,6 +2950,7 @@ DEF_TRAVERSE_STMT(FunctionParmPackExpr, {})
 DEF_TRAVERSE_STMT(CXXFoldExpr, {})
 DEF_TRAVERSE_STMT(AtomicExpr, {})
 DEF_TRAVERSE_STMT(CXXParenListInitExpr, {})
+DEF_TRAVERSE_STMT(ResolvedUnexpandedPackExpr, {})
 
 DEF_TRAVERSE_STMT(MaterializeTemporaryExpr, {
   if (S->getLifetimeExtendedTemporaryDecl()) {

clang/include/clang/Basic/DiagnosticParseKinds.td

@@ -1099,6 +1099,16 @@ def err_lambda_capture_misplaced_ellipsis : Error<
   "the name of the capture">;
 def err_lambda_capture_multiple_ellipses : Error<
   "multiple ellipses in pack capture">;
+def err_binding_multiple_ellipses : Error<
+  "multiple packs in structured binding declaration">;
+def note_previous_ellipsis : Note<
+  "previous binding pack specified here">;
+def ext_cxx_binding_pack : ExtWarn<
+  "structured binding packs are a C++2c extension ">,
+  InGroup<CXX26>;
+def warn_cxx23_compat_binding_pack : Warning<
+  "structured binding packs are incompatible with C++ standards before C++2c">,
+  InGroup<CXXPre26Compat>, DefaultIgnore;
 def err_capture_default_first : Error<
   "capture default must be first">;
 def ext_decl_attrs_on_lambda : ExtWarn<

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -5945,6 +5945,9 @@ def warn_cxx23_pack_indexing : Warning<
   "pack indexing is incompatible with C++ standards before C++2c">,
   DefaultIgnore, InGroup<CXXPre26Compat>;
 
+def err_pack_outside_template : Error<
+  "pack declaration outside of template">;
+
 def err_fold_expression_packs_both_sides : Error<
   "binary fold expression has unexpanded parameter packs in both operands">;
 def err_fold_expression_empty : Error<

clang/include/clang/Basic/StmtNodes.td

@@ -163,6 +163,7 @@ def MaterializeTemporaryExpr : StmtNode<Expr>;
 def LambdaExpr : StmtNode<Expr>;
 def CXXFoldExpr : StmtNode<Expr>;
 def CXXParenListInitExpr: StmtNode<Expr>;
+def ResolvedUnexpandedPackExpr : StmtNode<Expr>;
 
 // C++ Coroutines expressions
 def CoroutineSuspendExpr : StmtNode<Expr, 1>;

clang/include/clang/Sema/DeclSpec.h

@@ -1795,6 +1795,7 @@ class DecompositionDeclarator {
     IdentifierInfo *Name;
     SourceLocation NameLoc;
     std::optional<ParsedAttributes> Attrs;
+    SourceLocation EllipsisLoc;
   };
 
 private:

clang/include/clang/Sema/Sema.h

@@ -232,7 +232,8 @@ void threadSafetyCleanup(BeforeSet *Cache);
 
 // FIXME: No way to easily map from TemplateTypeParmTypes to
 // TemplateTypeParmDecls, so we have this horrible PointerUnion.
-typedef std::pair<llvm::PointerUnion<const TemplateTypeParmType *, NamedDecl *>,
+typedef std::pair<llvm::PointerUnion<const TemplateTypeParmType *, NamedDecl *,
+                                     ResolvedUnexpandedPackExpr *>,
                   SourceLocation>
     UnexpandedParameterPack;
 
@@ -6021,6 +6022,7 @@ class Sema final : public SemaBase {
                                          RecordDecl *ClassDecl,
                                          const IdentifierInfo *Name);
 
+  unsigned GetDecompositionElementCount(QualType DecompType);
   void CheckCompleteDecompositionDeclaration(DecompositionDecl *DD);
 
   /// Stack containing information needed when in C++2a an 'auto' is encountered

clang/include/clang/Serialization/ASTBitCodes.h

@@ -1908,6 +1908,7 @@ enum StmtCode {
   EXPR_PACK_EXPANSION,                    // PackExpansionExpr
   EXPR_PACK_INDEXING,                     // PackIndexingExpr
   EXPR_SIZEOF_PACK,                       // SizeOfPackExpr
+  EXPR_RESOLVED_UNEXPANDED_PACK,          // ResolvedUnexpandedPackExpr
   EXPR_SUBST_NON_TYPE_TEMPLATE_PARM,      // SubstNonTypeTemplateParmExpr
   EXPR_SUBST_NON_TYPE_TEMPLATE_PARM_PACK, // SubstNonTypeTemplateParmPackExpr
   EXPR_FUNCTION_PARM_PACK,                // FunctionParmPackExpr

clang/lib/AST/ASTContext.cpp

@@ -12841,11 +12841,12 @@ bool ASTContext::DeclMustBeEmitted(const Decl *D) {
 
   // Likewise, variables with tuple-like bindings are required if their
   // bindings have side-effects.
-  if (const auto *DD = dyn_cast<DecompositionDecl>(VD))
-    for (const auto *BD : DD->bindings())
+  if (const auto *DD = dyn_cast<DecompositionDecl>(VD)) {
+    for (const auto *BD : DD->flat_bindings())
       if (const auto *BindingVD = BD->getHoldingVar())
         if (DeclMustBeEmitted(BindingVD))
           return true;
+  }
 
   return false;
 }

clang/lib/AST/ASTImporter.cpp

@@ -2553,7 +2553,7 @@ ExpectedDecl ASTNodeImporter::VisitBindingDecl(BindingDecl *D) {
 
   BindingDecl *ToD;
   if (GetImportedOrCreateDecl(ToD, D, Importer.getToContext(), DC, Loc,
-                              Name.getAsIdentifierInfo()))
+                              Name.getAsIdentifierInfo(), D->getType()))
     return ToD;
 
   Error Err = Error::success();

clang/lib/AST/Decl.cpp

@@ -2659,10 +2659,6 @@ bool VarDecl::checkForConstantInitialization(
   return Eval->HasConstantInitialization;
 }
 
-bool VarDecl::isParameterPack() const {
-  return isa<PackExpansionType>(getType());
-}
-
 template<typename DeclT>
 static DeclT *getDefinitionOrSelf(DeclT *D) {
   assert(D);
@@ -5421,6 +5417,13 @@ bool ValueDecl::isInitCapture() const {
   return false;
 }
 
+bool ValueDecl::isParameterPack() const {
+  if (const auto *NTTP = dyn_cast<NonTypeTemplateParmDecl>(this))
+    return NTTP->isParameterPack();
+
+  return isa_and_nonnull<PackExpansionType>(getType().getTypePtrOrNull());
+}
+
 void ImplicitParamDecl::anchor() {}
 
 ImplicitParamDecl *ImplicitParamDecl::Create(ASTContext &C, DeclContext *DC,

clang/lib/AST/DeclBase.cpp

@@ -245,7 +245,7 @@ bool Decl::isTemplateParameterPack() const {
 }
 
 bool Decl::isParameterPack() const {
-  if (const auto *Var = dyn_cast<VarDecl>(this))
+  if (const auto *Var = dyn_cast<ValueDecl>(this))
     return Var->isParameterPack();
 
   return isTemplateParameterPack();

clang/lib/AST/DeclCXX.cpp

@@ -3462,19 +3462,21 @@ VarDecl *ValueDecl::getPotentiallyDecomposedVarDecl() {
   if (auto *Var = llvm::dyn_cast<VarDecl>(this))
     return Var;
   if (auto *BD = llvm::dyn_cast<BindingDecl>(this))
-    return llvm::dyn_cast<VarDecl>(BD->getDecomposedDecl());
+    return llvm::dyn_cast_if_present<VarDecl>(BD->getDecomposedDecl());
   return nullptr;
 }
 
 void BindingDecl::anchor() {}
 
 BindingDecl *BindingDecl::Create(ASTContext &C, DeclContext *DC,
-                                 SourceLocation IdLoc, IdentifierInfo *Id) {
-  return new (C, DC) BindingDecl(DC, IdLoc, Id);
+                                 SourceLocation IdLoc, IdentifierInfo *Id,
+                                 QualType T) {
+  return new (C, DC) BindingDecl(DC, IdLoc, Id, T);
 }
 
 BindingDecl *BindingDecl::CreateDeserialized(ASTContext &C, GlobalDeclID ID) {
-  return new (C, ID) BindingDecl(nullptr, SourceLocation(), nullptr);
+  return new (C, ID)
+      BindingDecl(nullptr, SourceLocation(), nullptr, QualType());
 }
 
 VarDecl *BindingDecl::getHoldingVar() const {
@@ -3490,6 +3492,12 @@ VarDecl *BindingDecl::getHoldingVar() const {
   return VD;
 }
 
+llvm::ArrayRef<Expr *> BindingDecl::getBindingPackExprs() const {
+  assert(Binding && "expecting a pack expr");
+  auto *RP = cast<ResolvedUnexpandedPackExpr>(Binding);
+  return RP->getExprs();
+}
+
 void DecompositionDecl::anchor() {}
 
 DecompositionDecl *DecompositionDecl::Create(ASTContext &C, DeclContext *DC,

clang/lib/AST/Expr.cpp

@@ -3659,6 +3659,7 @@ bool Expr::HasSideEffects(const ASTContext &Ctx,
   case PackIndexingExprClass:
   case HLSLOutArgExprClass:
   case OpenACCAsteriskSizeExprClass:
+  case ResolvedUnexpandedPackExprClass:
     // These never have a side-effect.
     return false;