[clang][Sema] Suggest/Hint Standard Library Include File

Clang now tries to suggest and hint standard library include file by emitting
the following.

```
    4 |     std::unordered_map<std::string, std::string> map;
      |          ^~~~~~~~~~~~~
main.cpp:4:10: note: maybe try to include <unordered_map>; 'std::unordered_map' is defined in <unordered_map>
main.cpp:4:10: note: 'std::unordered_map' is a c++11 feature
main.cpp:4:42: note: maybe try to include <string>; 'std::string' is defined in <string>
```

This works for both C and C++.

fixes #120388

Author: Mr-Anyone

clang/docs/ReleaseNotes.rst

@@ -470,6 +470,7 @@ Improvements to Clang's diagnostics
   diagnostics. This fixes a bunch of `bool` being printed as `_Bool`, and also
   a bunch of HLSL types being printed as their C++ equivalents.
 - Clang now consistently quotes expressions in diagnostics.
+- Clang now suggest standard library include path and its associated C++ or C language version.
 - When printing types for diagnostics, clang now doesn't suppress the scopes of
   template arguments contained within nested names.
 - The ``-Wshift-bool`` warning has been added to warn about shifting a boolean. (#GH28334)

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -5978,6 +5978,10 @@ def err_template_expansion_into_fixed_list : Error<
   "template|concept}0">;
 def note_parameter_type : Note<
   "parameter of type %0 is declared here">;
+def note_standard_lib_include_suggestion : Note<
+  "maybe try to include %0; '%1' is defined in %0">;
+def note_standard_lib_version : Note<
+  "'%0' is a %1 feature">;
 
 // C++11 Variadic Templates
 def err_template_param_pack_default_arg : Error<

clang/include/clang/Sema/Sema.h

@@ -3611,6 +3611,16 @@ class Sema final : public SemaBase {
                                ParsedType &SuggestedType,
                                bool IsTemplateName = false);
 
+  // Try to suggest the missing standard library include files
+  //
+  // \param SymbolName the symbol name like 'cout'
+  // \param SourceLocation the location of the note being emitted
+  // \param Namespace the namespace that must end with ::, so like 'std::'
+  void NoteStandardIncludes(StringRef SymbolName, SourceLocation IILoc,
+                            StringRef Namespace);
+  void NoteStandardIncludes(StringRef SymbolName, SourceLocation IILoc,
+                            const CXXScopeSpec *SS);
+
   /// Attempt to behave like MSVC in situations where lookup of an unqualified
   /// type name has failed in a dependent context. In these situations, we
   /// automatically form a DependentTypeName that will retry lookup in a related

clang/include/clang/Tooling/Inclusions/StandardLibrary.h

@@ -29,6 +29,27 @@ class NamespaceDecl;
 class DeclContext;
 namespace tooling {
 namespace stdlib {
+enum Version {
+  Unknown,
+
+  // c++ versions
+  CPlusPlusStart,
+  CPlusPlus11,
+  CPlusPlus14,
+  CPlusPlus17,
+  CPlusPlus20,
+  CPlusPlus23,
+  CPlusPlus26,
+  CPlusPlusEnd,
+
+  // c version
+  CStart,
+  C99,
+  C11,
+  CEnd
+};
+
+llvm::StringRef GetAsString(Version Ver);
 
 class Symbol;
 enum class Lang { C = 0, CXX, LastValue = CXX };
@@ -85,6 +106,7 @@ class Symbol {
   std::optional<Header> header() const;
   // Some symbols may be provided by multiple headers.
   llvm::SmallVector<Header> headers() const;
+  Version version() const;
 
 private:
   Symbol(unsigned ID, Lang Language) : ID(ID), Language(Language) {}

clang/lib/Parse/ParseExprCXX.cpp

@@ -226,6 +226,11 @@ bool Parser::ParseOptionalCXXScopeSpecifier(
     HasScopeSpecifier = true;
   }
 
+  // If `FailedNestedNameBuilding` is true, attempt to suggest the standard
+  // include file corresponding to the current `FullNamespace` and symbol.
+  std::string FullNamespace = "";
+  bool FailedNesatedNameBuilding = false;
+
   // Preferred type might change when parsing qualifiers, we need the original.
   auto SavedType = PreferredType;
   while (true) {
@@ -454,6 +459,9 @@ bool Parser::ParseOptionalCXXScopeSpecifier(
       // We have an identifier followed by a '::'. Lookup this name
       // as the name in a nested-name-specifier.
       Token Identifier = Tok;
+      FullNamespace += Identifier.getIdentifierInfo()->getName();
+      FullNamespace += "::";
+
       SourceLocation IdLoc = ConsumeToken();
       assert(Tok.isOneOf(tok::coloncolon, tok::colon) &&
              "NextToken() not working properly!");
@@ -465,6 +473,7 @@ bool Parser::ParseOptionalCXXScopeSpecifier(
       if (Actions.ActOnCXXNestedNameSpecifier(
               getCurScope(), IdInfo, EnteringContext, SS, CorrectionFlagPtr,
               OnlyNamespace)) {
+        FailedNesatedNameBuilding = true;
         // Identifier is not recognized as a nested name, but we can have
         // mistyped '::' instead of ':'.
         if (CorrectionFlagPtr && IsCorrectedToColon) {
@@ -554,6 +563,11 @@ bool Parser::ParseOptionalCXXScopeSpecifier(
     break;
   }
 
+  if (FailedNesatedNameBuilding && Tok.getKind() == tok::identifier) {
+    Actions.NoteStandardIncludes(Tok.getIdentifierInfo()->getName(),
+                                 Tok.getLocation(), FullNamespace);
+  }
+
   // Even if we didn't see any pieces of a nested-name-specifier, we
   // still check whether there is a tilde in this position, which
   // indicates a potential pseudo-destructor.

clang/lib/Sema/CMakeLists.txt

@@ -115,4 +115,5 @@ add_clang_library(clangSema
   clangEdit
   clangLex
   clangSupport
+  clangToolingInclusionsStdlib
   )

clang/lib/Sema/SemaDecl.cpp

@@ -58,10 +58,12 @@
 #include "clang/Sema/SemaSwift.h"
 #include "clang/Sema/SemaWasm.h"
 #include "clang/Sema/Template.h"
+#include "clang/Tooling/Inclusions/StandardLibrary.h"
 #include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringSwitch.h"
 #include "llvm/Frontend/HLSL/HLSLRootSignature.h"
 #include "llvm/Support/SaveAndRestore.h"
 #include "llvm/TargetParser/Triple.h"
@@ -805,6 +807,52 @@ void Sema::DiagnoseUnknownTypeName(IdentifierInfo *&II,
     assert(SS && SS->isInvalid() &&
            "Invalid scope specifier has already been diagnosed");
   }
+
+  // don't note standard include files for OpenCL and Objective C
+  if ((getLangOpts().CPlusPlus || getLangOpts().C99) && !getLangOpts().OpenCL &&
+      !getLangOpts().ObjC)
+    NoteStandardIncludes(II->getName(), IILoc, SS);
+}
+
+void Sema::NoteStandardIncludes(StringRef SymbolName, SourceLocation IILoc,
+                                StringRef Namespace) {
+  using clang::tooling::stdlib::Lang;
+
+  llvm::StringRef HeaderName = "";
+  tooling::stdlib::Lang LangOption = tooling::stdlib::Lang::C;
+  if (getLangOpts().CPlusPlus)
+    LangOption = clang::tooling::stdlib::Lang::CXX;
+
+  if (auto StdSym =
+          tooling::stdlib::Symbol::named(Namespace, SymbolName, LangOption)) {
+    if (auto Header = StdSym->header()) {
+      HeaderName = Header->name();
+      Diag(IILoc, diag::note_standard_lib_include_suggestion)
+          << HeaderName << (Namespace + SymbolName).str();
+
+      // Noting the C/C++ version as well
+      if (StdSym->version() != tooling::stdlib::Unknown) {
+        llvm::StringRef CPlusPlusVersion =
+            tooling::stdlib::GetAsString(StdSym->version());
+
+        Diag(IILoc, diag::note_standard_lib_version)
+            << (Namespace + SymbolName).str() << CPlusPlusVersion;
+      }
+    }
+  }
+}
+
+void Sema::NoteStandardIncludes(StringRef SymbolName, SourceLocation IILoc,
+                                const CXXScopeSpec *SS) {
+  std::string Namespace = "";
+  if (SS) {
+    llvm::raw_string_ostream Stream(Namespace);
+    if (SS->isValid())
+      SS->getScopeRep()->dump(Stream);
+    Stream.flush();
+  }
+
+  NoteStandardIncludes(SymbolName, IILoc, Namespace);
 }
 
 /// Determine whether the given result set contains either a type name
@@ -16784,6 +16832,7 @@ NamedDecl *Sema::ImplicitlyDefineFunction(SourceLocation Loc,
   }
 
   Diag(Loc, diag_id) << &II;
+  NoteStandardIncludes(II.getName(), Loc, "");
   if (Corrected) {
     // If the correction is going to suggest an implicitly defined function,
     // skip the correction as not being a particularly good idea.

clang/lib/Sema/SemaExpr.cpp

@@ -2660,11 +2660,17 @@ bool Sema::DiagnoseEmptyLookup(Scope *S, CXXScopeSpec &SS, LookupResult &R,
   if (!SS.isEmpty()) {
     Diag(R.getNameLoc(), diag::err_no_member)
         << Name << computeDeclContext(SS, false) << NameRange;
+    NoteStandardIncludes(Name.getAsString(), R.getNameLoc(), &SS);
     return true;
   }
 
   // Give up, we can't recover.
   Diag(R.getNameLoc(), diagnostic) << Name << NameRange;
+
+  // don't note standard include files for OpenCL and Objective C
+  if ((getLangOpts().CPlusPlus || getLangOpts().C99) && !getLangOpts().OpenCL &&
+      !getLangOpts().ObjC)
+    NoteStandardIncludes(Name.getAsString(), R.getNameLoc(), /*Namespace=*/"");
   return true;
 }