[Syntax] Add mapping from spelled to expanded tokens for TokenBuffer

[hlopko]

Summary

Same restrictions apply as in the other direction: macro arguments are
not supported yet, only full macro expansions can be mapped.

Taking over from https://reviews.llvm.org/D72581.

[hlopko]

Adding comments.

[gribozavr2]

clang/lib/Tooling/Syntax/Tokens.cpp

239

It would be great to add an is_sorted assertion to the builder to check ordering of mappings based on both spelled and expanded token indices.

259

s/subarray/subrange/

Also no need to repeat "assert".

"Spelled must be a subrange of File.SpelledTokens."

261

I think we can improve the precision of this assertion. Comparing the sizes does not ensure that we really have a subrange. I think the second assert should be comparing end pointers instead of sizes. Something like:

assert(&Spelled.back() <= &File.SpelledTokens.back());

264

Or assert that SpelledFrontI is less than File.SpelledTokens.size().

[hlopko]

Adressing comments

[hlopko]

Adressing comments.

clang/lib/Tooling/Syntax/Tokens.cpp

264

I think the assert I added is good enough?

[gribozavr2]

clang/lib/Tooling/Syntax/Tokens.cpp

264

The assertion that I suggested is stronger, because it would prevent an out-of-bounds read from even happening, and would not rely on isBeforeInTranslationUnit returning false on garbage data.

586

I'd suggest moving this loop to the very bottom of this function. It is not related to the expanded token processing, which happens here (above we have assertions related to expanded tokens, and below we have the processExpandedToken call).

Also, fillGapsAtEndOfFiles below modifies mappings, it would be good if modifications were covered by the assertion as well.

587

We'd get an "unused variable" warning here when assertions are disabled. Please wrap the whole loop in #ifndef NDEBUG.

[hlopko]

Adressing comments

[hlopko]

Rebase

[hlopko]

clang/lib/Tooling/Syntax/Tokens.cpp

264

Great points! Done.

587

Done, added ifndef also above for T1/T2 variables.

[hlopko]

Assert the Spelled subrange more effectively

clang/include/clang/Tooling/Syntax/Tokens.h

public:
   ///         e f  => e f
   ///         g h  => can't map
   ///
   /// EXPECTS: \p Expanded is a subrange of expandedTokens().
   /// Complexity is logarithmic.
   llvm::Optional<llvm::ArrayRef<syntax::Token>>
   spelledForExpanded(llvm::ArrayRef<syntax::Token> Expanded) const;
 
+  /// Find the subranges of expanded tokens, corresponding to \p Spelled.
+  /// Some spelled tokens may not be present in the expanded token stream, so
+  /// this function can return an empty vector, e.g. for tokens of macro
+  /// directives or disabled preprocessor branches.
+  ///
+  /// Some spelled tokens can be duplicated in the expanded token stream
+  /// multiple times and this function will return multiple results in those
+  /// cases. This happens when \p Spelled is inside a macro argument.
+  ///
+  /// FIXME: return correct results on macro arguments. For now, we return an
+  ///        empty list.
+  ///
+  /// (!) will return empty vector on tokens from #define body:
+  /// E.g. for the following example:
+  ///
+  ///   #define FIRST(A) f1 A = A f2
+  ///   #define SECOND s
+  ///
+  ///   a FIRST(arg) b SECOND c // expanded tokens are: a f1 arg = arg f2 b s
+  /// The results would be
+  ///   spelled           => expanded
+  ///   ------------------------
+  ///   #define FIRST     => {}
+  ///   a FIRST(arg)      => {a f1 arg = arg f2}
+  ///   arg               => {arg, arg} // arg #1 is before `=` and arg #2 is
+  ///                                   // after `=` in the expanded tokens.
+  llvm::SmallVector<llvm::ArrayRef<syntax::Token>, 1>
+  expandedForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const;
+
   /// An expansion produced by the preprocessor, includes macro expansions and
   /// preprocessor directives. Preprocessor always maps a non-empty range of
   /// spelled tokens to a (possibly empty) range of expanded tokens. Here is a
   /// few examples of expansions:
   ///    #pragma once      // Expands to an empty range.
   ///    #define FOO 1 2 3 // Expands an empty range.
   ///    FOO               // Expands to "1 2 3".
   /// FIXME(ibiryukov): implement this, currently #include expansions are empty.
private:
 
   friend class TokenCollector;
 
   /// Maps a single expanded token to its spelled counterpart or a mapping that
   /// produced it.
   std::pair<const syntax::Token *, const Mapping *>
   spelledForExpandedToken(const syntax::Token *Expanded) const;
 
+  static const Mapping *
+  mappingStartingBeforeSpelled(const MarkedFile &F,
+                               const syntax::Token *Spelled);
+
   /// Token stream produced after preprocessing, conceputally this captures the
   /// same stream as 'clang -E' (excluding the preprocessor directives like
   /// #file, etc.).
   std::vector<syntax::Token> ExpandedTokens;
   llvm::DenseMap<FileID, MarkedFile> Files;
   // The value is never null, pointer instead of reference to avoid disabling
   // implicit assignment operator.
   const SourceManager *SourceMgr;

clang/lib/Tooling/Syntax/Tokens.cpp

TokenBuffer::spelledForExpandedToken(const syntax::Token *Expanded) const {
   unsigned ExpandedIndex = Expanded - ExpandedTokens.data();
   // Find the first mapping that produced tokens after \p Expanded.
   auto It = llvm::partition_point(File.Mappings, [&](const Mapping &M) {
     return M.BeginExpanded <= ExpandedIndex;
   });
   // Our token could only be produced by the previous mapping.
   if (It == File.Mappings.begin()) {
     // No previous mapping, no need to modify offsets.
-    return {&File.SpelledTokens[ExpandedIndex - File.BeginExpanded], nullptr};
+    return {&File.SpelledTokens[ExpandedIndex - File.BeginExpanded],
+            /*Mapping=*/nullptr};
   }
   --It; // 'It' now points to last mapping that started before our token.
 
   // Check if the token is part of the mapping.
   if (ExpandedIndex < It->EndExpanded)
-    return {&File.SpelledTokens[It->BeginSpelled], /*Mapping*/ &*It};
+    return {&File.SpelledTokens[It->BeginSpelled], /*Mapping=*/&*It};
 
   // Not part of the mapping, use the index from previous mapping to compute the
   // corresponding spelled token.
   return {
       &File.SpelledTokens[It->EndSpelled + (ExpandedIndex - It->EndExpanded)],
-      /*Mapping*/ nullptr};
+      /*Mapping=*/nullptr};
+}
+
+const TokenBuffer::Mapping *
+TokenBuffer::mappingStartingBeforeSpelled(const MarkedFile &F,
+                                          const syntax::Token *Spelled) {
+  assert(F.SpelledTokens.data() <= Spelled);
+  unsigned SpelledI = Spelled - F.SpelledTokens.data();
+  assert(SpelledI < F.SpelledTokens.size());
+
+  auto It = llvm::partition_point(F.Mappings, [SpelledI](const Mapping &M) {
+    return M.BeginSpelled <= SpelledI;
+  });
+  if (It == F.Mappings.begin())
+    return nullptr;
+  --It;
+  return &*It;
+}
+
+llvm::SmallVector<llvm::ArrayRef<syntax::Token>, 1>
+TokenBuffer::expandedForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const {
+  if (Spelled.empty())
+    return {};
+  assert(Spelled.front().location().isFileID());
+
+  auto FID = sourceManager().getFileID(Spelled.front().location());
+  auto It = Files.find(FID);
+  assert(It != Files.end());
+
+  const MarkedFile &File = It->second;
+  // assert that Spelled is a subarray of File.SpelledTokens.
+  assert(File.SpelledTokens.data() <= Spelled.data());
+  assert(Spelled.size() <= File.SpelledTokens.size());
+
+  auto *FrontMapping = mappingStartingBeforeSpelled(File, &Spelled.front());
+  unsigned SpelledFrontI = &Spelled.front() - File.SpelledTokens.data();
+  unsigned ExpandedBegin;
+  if (!FrontMapping) {
+    // Requested spelled tokens are mapped in the first mapping.
+    ExpandedBegin = File.BeginExpanded + SpelledFrontI;
+  } else if (SpelledFrontI < FrontMapping->EndSpelled) {
+    // Requested spelled tokens start in the FrontMapping.
+    if (SpelledFrontI != FrontMapping->BeginSpelled)
+      return {}; // FIXME: support macro arguments.
+    ExpandedBegin = FrontMapping->BeginExpanded;
+  } else {
+    ExpandedBegin =
+        FrontMapping->EndExpanded + (SpelledFrontI - FrontMapping->EndSpelled);
+  }
+
+  auto *BackMapping = mappingStartingBeforeSpelled(File, &Spelled.back());
+  unsigned SpelledBackI = &Spelled.back() - File.SpelledTokens.data();
+  unsigned ExpandedEnd;
+  if (!BackMapping) {
+    ExpandedEnd = File.BeginExpanded + SpelledBackI + 1;
+  } else if (SpelledBackI < BackMapping->EndSpelled) {
+    if (SpelledBackI + 1 != BackMapping->EndSpelled)
+      return {}; // FIXME: support macro arguments.
+    ExpandedEnd = BackMapping->EndExpanded;
+  } else {
+    ExpandedEnd =
+        BackMapping->EndExpanded + (SpelledBackI - BackMapping->EndSpelled) + 1;
+  }
+
+  assert(ExpandedBegin < ExpandedTokens.size());
+  assert(ExpandedEnd < ExpandedTokens.size());
+  // Avoid returning empty ranges.
+  if (ExpandedBegin == ExpandedEnd)
+    return {};
+  return {llvm::makeArrayRef(ExpandedTokens.data() + ExpandedBegin,
+                             ExpandedTokens.data() + ExpandedEnd)};
 }
 
 llvm::ArrayRef<syntax::Token> TokenBuffer::spelledTokens(FileID FID) const {
   auto It = Files.find(FID);
   assert(It != Files.end());
   return It->second.SpelledTokens;
 }
 
TokenBuffer build() && {
     // To correctly recover empty macro expansions, we also take locations
     // reported to PPCallbacks::MacroExpands into account as we do not have any
     // expanded tokens with source locations to guide us.
 
     // The 'eof' token is special, it is not part of spelled token stream. We
     // handle it separately at the end.
     assert(!Result.ExpandedTokens.empty());
     assert(Result.ExpandedTokens.back().kind() == tok::eof);
     for (unsigned I = 0; I < Result.ExpandedTokens.size() - 1; ++I) {
       // (!) I might be updated by the following call.
       processExpandedToken(I);
     }
 
     // 'eof' not handled in the loop, do it here.
     assert(SM.getMainFileID() ==
            SM.getFileID(Result.ExpandedTokens.back().location()));
     fillGapUntil(Result.Files[SM.getMainFileID()],
                  Result.ExpandedTokens.back().location(),

clang/unittests/Tooling/Syntax/TokensTest.cpp

 using namespace clang;
 using namespace clang::syntax;
 
 using llvm::ValueIs;
 using ::testing::AllOf;
 using ::testing::Contains;
 using ::testing::ElementsAre;
 using ::testing::Field;
+using ::testing::IsEmpty;
 using ::testing::Matcher;
 using ::testing::Not;
 using ::testing::Pointee;
 using ::testing::StartsWith;
 
 namespace {
 // Checks the passed ArrayRef<T> has the same begin() and end() iterators as the
 // argument.
Ls.second = SourceMgr->getLocForStartOfFile(SourceMgr->getMainFileID())
                     .getLocWithOffset(R.End);
     return ::RangeIs(Ls);
   }
 
   /// Finds a subrange in O(n * m).
   template <class T, class U, class Eq>
   llvm::ArrayRef<T> findSubrange(llvm::ArrayRef<U> Subrange,
                                  llvm::ArrayRef<T> Range, Eq F) {
-    for (auto Begin = Range.begin(); Begin < Range.end(); ++Begin) {
+    assert(Subrange.size() >= 1);
+    if (Range.size() < Subrange.size())
+      return llvm::makeArrayRef(Range.end(), Range.end());
+    for (auto Begin = Range.begin(), Last = Range.end() - Subrange.size();
+         Begin <= Last; ++Begin) {
       auto It = Begin;
-      for (auto ItSub = Subrange.begin();
-           ItSub != Subrange.end() && It != Range.end(); ++ItSub, ++It) {
+      for (auto ItSub = Subrange.begin(); ItSub != Subrange.end();
+           ++ItSub, ++It) {
         if (!F(*ItSub, *It))
           goto continue_outer;
       }
       return llvm::makeArrayRef(Begin, It);
     continue_outer:;
     }
     return llvm::makeArrayRef(Range.end(), Range.end());
   }
TEST_F(TokenBufferTest, Touching) {
   EXPECT_EQ(Identifier(6), "");
 
   EXPECT_THAT(Touching(7), SameRange(findSpelled(";")));
   EXPECT_EQ(Identifier(7), "");
 
   ASSERT_EQ(Code.points().size(), 8u);
 }
 
+TEST_F(TokenBufferTest, ExpandedBySpelled) {
+  recordTokens(R"cpp(
+    a1 a2 a3 b1 b2
+  )cpp");
+  // Sanity check: expanded and spelled tokens are stored separately.
+  EXPECT_THAT(findExpanded("a1 a2"), Not(SameRange(findSpelled("a1 a2"))));
+  // Searching for subranges of expanded tokens should give the corresponding
+  // spelled ones.
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("a1 a2 a3 b1 b2")),
+              ElementsAre(SameRange(findExpanded("a1 a2 a3 b1 b2"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("a1 a2 a3")),
+              ElementsAre(SameRange(findExpanded("a1 a2 a3"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("b1 b2")),
+              ElementsAre(SameRange(findExpanded("b1 b2"))));
+
+  // Test search on simple macro expansions.
+  recordTokens(R"cpp(
+    #define A a1 a2 a3
+    #define B b1 b2
+
+    A split B
+  )cpp");
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("A split B")),
+              ElementsAre(SameRange(findExpanded("a1 a2 a3 split b1 b2"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("A split").drop_back()),
+              ElementsAre(SameRange(findExpanded("a1 a2 a3"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("split B").drop_front()),
+              ElementsAre(SameRange(findExpanded("b1 b2"))));
+
+  // Recursive macro invocations.
+  recordTokens(R"cpp(
+    #define ID(x) x
+    #define B b1 b2
+
+    ID(ID(ID(a1) a2 a3)) split ID(B)
+  )cpp");
+
+  EXPECT_THAT(
+      Buffer.expandedForSpelled(findSpelled("ID ( ID ( ID ( a1 ) a2 a3 ) )")),
+      ElementsAre(SameRange(findExpanded("a1 a2 a3"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( B )")),
+              ElementsAre(SameRange(findExpanded("b1 b2"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(
+                  findSpelled("ID ( ID ( ID ( a1 ) a2 a3 ) ) split ID ( B )")),
+              ElementsAre(SameRange(findExpanded("a1 a2 a3 split b1 b2"))));
+  // FIXME: these should succeed, but we do not support macro arguments yet.
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("a1")), IsEmpty());
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( a1 ) a2")),
+              IsEmpty());
+
+  // Empty macro expansions.
+  recordTokens(R"cpp(
+    #define EMPTY
+    #define ID(X) X
+
+    EMPTY EMPTY ID(1 2 3) EMPTY EMPTY split1
+    EMPTY EMPTY ID(4 5 6) split2
+    ID(7 8 9) EMPTY EMPTY
+  )cpp");
+  // Covered by empty expansions on one of both of the sides.
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( 1 2 3 )")),
+              ElementsAre(SameRange(findExpanded("1 2 3"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( 4 5 6 )")),
+              ElementsAre(SameRange(findExpanded("4 5 6"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( 7 8 9 )")),
+              ElementsAre(SameRange(findExpanded("7 8 9"))));
+  // Including the empty macro expansions on the side.
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("EMPTY ID ( 1 2 3 )")),
+              ElementsAre(SameRange(findExpanded("1 2 3"))));
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("ID ( 1 2 3 ) EMPTY")),
+              ElementsAre(SameRange(findExpanded("1 2 3"))));
+  EXPECT_THAT(
+      Buffer.expandedForSpelled(findSpelled("EMPTY ID ( 1 2 3 ) EMPTY")),
+      ElementsAre(SameRange(findExpanded("1 2 3"))));
+
+  // Empty mappings coming from various directives.
+  recordTokens(R"cpp(
+    #define ID(X) X
+    ID(1)
+    #pragma lalala
+    not_mapped
+  )cpp");
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("# define ID ( X ) X")),
+              IsEmpty());
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("# pragma lalala")),
+              IsEmpty());
+
+  // Empty macro expansion.
+  recordTokens(R"cpp(
+    #define EMPTY
+    EMPTY int a = 100;
+  )cpp");
+  EXPECT_THAT(Buffer.expandedForSpelled(findSpelled("EMPTY int").drop_back()),
+              IsEmpty());
+}
+
 } // namespace