1 //===- Tokens.cpp - collect tokens from preprocessing ---------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 #include "clang/Tooling/Syntax/Tokens.h"
9 
10 #include "clang/Basic/Diagnostic.h"
11 #include "clang/Basic/IdentifierTable.h"
12 #include "clang/Basic/LLVM.h"
13 #include "clang/Basic/LangOptions.h"
14 #include "clang/Basic/SourceLocation.h"
15 #include "clang/Basic/SourceManager.h"
16 #include "clang/Basic/TokenKinds.h"
17 #include "clang/Lex/PPCallbacks.h"
18 #include "clang/Lex/Preprocessor.h"
19 #include "clang/Lex/Token.h"
20 #include "llvm/ADT/ArrayRef.h"
21 #include "llvm/ADT/None.h"
22 #include "llvm/ADT/Optional.h"
23 #include "llvm/ADT/STLExtras.h"
24 #include "llvm/Support/Debug.h"
25 #include "llvm/Support/ErrorHandling.h"
26 #include "llvm/Support/FormatVariadic.h"
27 #include "llvm/Support/raw_ostream.h"
28 #include <algorithm>
29 #include <cassert>
30 #include <iterator>
31 #include <string>
32 #include <utility>
33 #include <vector>
34 
35 using namespace clang;
36 using namespace clang::syntax;
37 
38 namespace {
39 // Finds the smallest consecutive subsuquence of Toks that covers R.
40 llvm::ArrayRef<syntax::Token>
41 getTokensCovering(llvm::ArrayRef<syntax::Token> Toks, SourceRange R,
42                   const SourceManager &SM) {
43   if (R.isInvalid())
44     return {};
45   const syntax::Token *Begin =
46       llvm::partition_point(Toks, [&](const syntax::Token &T) {
47         return SM.isBeforeInTranslationUnit(T.location(), R.getBegin());
48       });
49   const syntax::Token *End =
50       llvm::partition_point(Toks, [&](const syntax::Token &T) {
51         return !SM.isBeforeInTranslationUnit(R.getEnd(), T.location());
52       });
53   if (Begin > End)
54     return {};
55   return {Begin, End};
56 }
57 
58 // Finds the range within FID corresponding to expanded tokens [First, Last].
59 // Prev precedes First and Next follows Last, these must *not* be included.
60 // If no range satisfies the criteria, returns an invalid range.
61 //
62 // #define ID(x) x
63 // ID(ID(ID(a1) a2))
64 //          ~~       -> a1
65 //              ~~   -> a2
66 //       ~~~~~~~~~   -> a1 a2
67 SourceRange spelledForExpandedSlow(SourceLocation First, SourceLocation Last,
68                                    SourceLocation Prev, SourceLocation Next,
69                                    FileID TargetFile,
70                                    const SourceManager &SM) {
71   // There are two main parts to this algorithm:
72   //  - identifying which spelled range covers the expanded tokens
73   //  - validating that this range doesn't cover any extra tokens (First/Last)
74   //
75   // We do these in order. However as we transform the expanded range into the
76   // spelled one, we adjust First/Last so the validation remains simple.
77 
78   assert(SM.getSLocEntry(TargetFile).isFile());
79   // In most cases, to select First and Last we must return their expansion
80   // range, i.e. the whole of any macros they are included in.
81   //
82   // When First and Last are part of the *same macro arg* of a macro written
83   // in TargetFile, we that slice of the arg, i.e. their spelling range.
84   //
85   // Unwrap such macro calls. If the target file has A(B(C)), the
86   // SourceLocation stack of a token inside C shows us the expansion of A first,
87   // then B, then any macros inside C's body, then C itself.
88   // (This is the reverse of the order the PP applies the expansions in).
89   while (First.isMacroID() && Last.isMacroID()) {
90     auto DecFirst = SM.getDecomposedLoc(First);
91     auto DecLast = SM.getDecomposedLoc(Last);
92     auto &ExpFirst = SM.getSLocEntry(DecFirst.first).getExpansion();
93     auto &ExpLast = SM.getSLocEntry(DecLast.first).getExpansion();
94 
95     if (!ExpFirst.isMacroArgExpansion() || !ExpLast.isMacroArgExpansion())
96       break;
97     // Locations are in the same macro arg if they expand to the same place.
98     // (They may still have different FileIDs - an arg can have >1 chunks!)
99     if (ExpFirst.getExpansionLocStart() != ExpLast.getExpansionLocStart())
100       break;
101     // Careful, given:
102     //   #define HIDE ID(ID(a))
103     //   ID(ID(HIDE))
104     // The token `a` is wrapped in 4 arg-expansions, we only want to unwrap 2.
105     // We distinguish them by whether the macro expands into the target file.
106     // Fortunately, the target file ones will always appear first.
107     auto &ExpMacro =
108         SM.getSLocEntry(SM.getFileID(ExpFirst.getExpansionLocStart()))
109             .getExpansion();
110     if (ExpMacro.getExpansionLocStart().isMacroID())
111       break;
112     // Replace each endpoint with its spelling inside the macro arg.
113     // (This is getImmediateSpellingLoc without repeating lookups).
114     First = ExpFirst.getSpellingLoc().getLocWithOffset(DecFirst.second);
115     Last = ExpLast.getSpellingLoc().getLocWithOffset(DecLast.second);
116 
117     // Now: how do we adjust the previous/next bounds? Three cases:
118     // A) If they are also part of the same macro arg, we translate them too.
119     //   This will ensure that we don't select any macros nested within the
120     //   macro arg that cover extra tokens. Critical case:
121     //      #define ID(X) X
122     //      ID(prev target) // selecting 'target' succeeds
123     //      #define LARGE ID(prev target)
124     //      LARGE // selecting 'target' fails.
125     // B) They are not in the macro at all, then their expansion range is a
126     //    sibling to it, and we can safely substitute that.
127     //      #define PREV prev
128     //      #define ID(X) X
129     //      PREV ID(target) // selecting 'target' succeeds.
130     //      #define LARGE PREV ID(target)
131     //      LARGE // selecting 'target' fails.
132     // C) They are in a different arg of this macro, or the macro body.
133     //    Now selecting the whole macro arg is fine, but the whole macro is not.
134     //    Model this by setting using the edge of the macro call as the bound.
135     //      #define ID2(X, Y) X Y
136     //      ID2(prev, target) // selecting 'target' succeeds
137     //      #define LARGE ID2(prev, target)
138     //      LARGE // selecting 'target' fails
139     auto AdjustBound = [&](SourceLocation &Bound) {
140       if (Bound.isInvalid() || !Bound.isMacroID()) // Non-macro must be case B.
141         return;
142       auto DecBound = SM.getDecomposedLoc(Bound);
143       auto &ExpBound = SM.getSLocEntry(DecBound.first).getExpansion();
144       if (ExpBound.isMacroArgExpansion() &&
145           ExpBound.getExpansionLocStart() == ExpFirst.getExpansionLocStart()) {
146         // Case A: translate to (spelling) loc within the macro arg.
147         Bound = ExpBound.getSpellingLoc().getLocWithOffset(DecBound.second);
148         return;
149       }
150       while (Bound.isMacroID()) {
151         SourceRange Exp = SM.getImmediateExpansionRange(Bound).getAsRange();
152         if (Exp.getBegin() == ExpMacro.getExpansionLocStart()) {
153           // Case B: bounds become the macro call itself.
154           Bound = (&Bound == &Prev) ? Exp.getBegin() : Exp.getEnd();
155           return;
156         }
157         // Either case C, or expansion location will later find case B.
158         // We choose the upper bound for Prev and the lower one for Next:
159         //   ID(prev) target ID(next)
160         //          ^        ^
161         //      new-prev  new-next
162         Bound = (&Bound == &Prev) ? Exp.getEnd() : Exp.getBegin();
163       }
164     };
165     AdjustBound(Prev);
166     AdjustBound(Next);
167   }
168 
169   // In all remaining cases we need the full containing macros.
170   // If this overlaps Prev or Next, then no range is possible.
171   SourceRange Candidate =
172       SM.getExpansionRange(SourceRange(First, Last)).getAsRange();
173   auto DecFirst = SM.getDecomposedExpansionLoc(Candidate.getBegin());
174   auto DecLast = SM.getDecomposedLoc(Candidate.getEnd());
175   // Can end up in the wrong file due to bad input or token-pasting shenanigans.
176   if (Candidate.isInvalid() || DecFirst.first != TargetFile || DecLast.first != TargetFile)
177     return SourceRange();
178   // Check bounds, which may still be inside macros.
179   if (Prev.isValid()) {
180     auto Dec = SM.getDecomposedLoc(SM.getExpansionRange(Prev).getBegin());
181     if (Dec.first != DecFirst.first || Dec.second >= DecFirst.second)
182       return SourceRange();
183   }
184   if (Next.isValid()) {
185     auto Dec = SM.getDecomposedLoc(SM.getExpansionRange(Next).getEnd());
186     if (Dec.first != DecLast.first || Dec.second <= DecLast.second)
187       return SourceRange();
188   }
189   // Now we know that Candidate is a file range that covers [First, Last]
190   // without encroaching on {Prev, Next}. Ship it!
191   return Candidate;
192 }
193 
194 } // namespace
195 
196 syntax::Token::Token(SourceLocation Location, unsigned Length,
197                      tok::TokenKind Kind)
198     : Location(Location), Length(Length), Kind(Kind) {
199   assert(Location.isValid());
200 }
201 
202 syntax::Token::Token(const clang::Token &T)
203     : Token(T.getLocation(), T.getLength(), T.getKind()) {
204   assert(!T.isAnnotation());
205 }
206 
207 llvm::StringRef syntax::Token::text(const SourceManager &SM) const {
208   bool Invalid = false;
209   const char *Start = SM.getCharacterData(location(), &Invalid);
210   assert(!Invalid);
211   return llvm::StringRef(Start, length());
212 }
213 
214 FileRange syntax::Token::range(const SourceManager &SM) const {
215   assert(location().isFileID() && "must be a spelled token");
216   FileID File;
217   unsigned StartOffset;
218   std::tie(File, StartOffset) = SM.getDecomposedLoc(location());
219   return FileRange(File, StartOffset, StartOffset + length());
220 }
221 
222 FileRange syntax::Token::range(const SourceManager &SM,
223                                const syntax::Token &First,
224                                const syntax::Token &Last) {
225   auto F = First.range(SM);
226   auto L = Last.range(SM);
227   assert(F.file() == L.file() && "tokens from different files");
228   assert((F == L || F.endOffset() <= L.beginOffset()) &&
229          "wrong order of tokens");
230   return FileRange(F.file(), F.beginOffset(), L.endOffset());
231 }
232 
233 llvm::raw_ostream &syntax::operator<<(llvm::raw_ostream &OS, const Token &T) {
234   return OS << T.str();
235 }
236 
237 FileRange::FileRange(FileID File, unsigned BeginOffset, unsigned EndOffset)
238     : File(File), Begin(BeginOffset), End(EndOffset) {
239   assert(File.isValid());
240   assert(BeginOffset <= EndOffset);
241 }
242 
243 FileRange::FileRange(const SourceManager &SM, SourceLocation BeginLoc,
244                      unsigned Length) {
245   assert(BeginLoc.isValid());
246   assert(BeginLoc.isFileID());
247 
248   std::tie(File, Begin) = SM.getDecomposedLoc(BeginLoc);
249   End = Begin + Length;
250 }
251 FileRange::FileRange(const SourceManager &SM, SourceLocation BeginLoc,
252                      SourceLocation EndLoc) {
253   assert(BeginLoc.isValid());
254   assert(BeginLoc.isFileID());
255   assert(EndLoc.isValid());
256   assert(EndLoc.isFileID());
257   assert(SM.getFileID(BeginLoc) == SM.getFileID(EndLoc));
258   assert(SM.getFileOffset(BeginLoc) <= SM.getFileOffset(EndLoc));
259 
260   std::tie(File, Begin) = SM.getDecomposedLoc(BeginLoc);
261   End = SM.getFileOffset(EndLoc);
262 }
263 
264 llvm::raw_ostream &syntax::operator<<(llvm::raw_ostream &OS,
265                                       const FileRange &R) {
266   return OS << llvm::formatv("FileRange(file = {0}, offsets = {1}-{2})",
267                              R.file().getHashValue(), R.beginOffset(),
268                              R.endOffset());
269 }
270 
271 llvm::StringRef FileRange::text(const SourceManager &SM) const {
272   bool Invalid = false;
273   StringRef Text = SM.getBufferData(File, &Invalid);
274   if (Invalid)
275     return "";
276   assert(Begin <= Text.size());
277   assert(End <= Text.size());
278   return Text.substr(Begin, length());
279 }
280 
281 void TokenBuffer::indexExpandedTokens() {
282   // No-op if the index is already created.
283   if (!ExpandedTokIndex.empty())
284     return;
285   ExpandedTokIndex.reserve(ExpandedTokens.size());
286   // Index ExpandedTokens for faster lookups by SourceLocation.
287   for (size_t I = 0, E = ExpandedTokens.size(); I != E; ++I) {
288     SourceLocation Loc = ExpandedTokens[I].location();
289     if (Loc.isValid())
290       ExpandedTokIndex[Loc] = I;
291   }
292 }
293 
294 llvm::ArrayRef<syntax::Token> TokenBuffer::expandedTokens(SourceRange R) const {
295   if (R.isInvalid())
296     return {};
297   if (!ExpandedTokIndex.empty()) {
298     // Quick lookup if `R` is a token range.
299     // This is a huge win since majority of the users use ranges provided by an
300     // AST. Ranges in AST are token ranges from expanded token stream.
301     const auto B = ExpandedTokIndex.find(R.getBegin());
302     const auto E = ExpandedTokIndex.find(R.getEnd());
303     if (B != ExpandedTokIndex.end() && E != ExpandedTokIndex.end()) {
304       const Token *L = ExpandedTokens.data() + B->getSecond();
305       // Add 1 to End to make a half-open range.
306       const Token *R = ExpandedTokens.data() + E->getSecond() + 1;
307       if (L > R)
308         return {};
309       return {L, R};
310     }
311   }
312   // Slow case. Use `isBeforeInTranslationUnit` to binary search for the
313   // required range.
314   return getTokensCovering(expandedTokens(), R, *SourceMgr);
315 }
316 
317 CharSourceRange FileRange::toCharRange(const SourceManager &SM) const {
318   return CharSourceRange(
319       SourceRange(SM.getComposedLoc(File, Begin), SM.getComposedLoc(File, End)),
320       /*IsTokenRange=*/false);
321 }
322 
323 std::pair<const syntax::Token *, const TokenBuffer::Mapping *>
324 TokenBuffer::spelledForExpandedToken(const syntax::Token *Expanded) const {
325   assert(Expanded);
326   assert(ExpandedTokens.data() <= Expanded &&
327          Expanded < ExpandedTokens.data() + ExpandedTokens.size());
328 
329   auto FileIt = Files.find(
330       SourceMgr->getFileID(SourceMgr->getExpansionLoc(Expanded->location())));
331   assert(FileIt != Files.end() && "no file for an expanded token");
332 
333   const MarkedFile &File = FileIt->second;
334 
335   unsigned ExpandedIndex = Expanded - ExpandedTokens.data();
336   // Find the first mapping that produced tokens after \p Expanded.
337   auto It = llvm::partition_point(File.Mappings, [&](const Mapping &M) {
338     return M.BeginExpanded <= ExpandedIndex;
339   });
340   // Our token could only be produced by the previous mapping.
341   if (It == File.Mappings.begin()) {
342     // No previous mapping, no need to modify offsets.
343     return {&File.SpelledTokens[ExpandedIndex - File.BeginExpanded],
344             /*Mapping=*/nullptr};
345   }
346   --It; // 'It' now points to last mapping that started before our token.
347 
348   // Check if the token is part of the mapping.
349   if (ExpandedIndex < It->EndExpanded)
350     return {&File.SpelledTokens[It->BeginSpelled], /*Mapping=*/&*It};
351 
352   // Not part of the mapping, use the index from previous mapping to compute the
353   // corresponding spelled token.
354   return {
355       &File.SpelledTokens[It->EndSpelled + (ExpandedIndex - It->EndExpanded)],
356       /*Mapping=*/nullptr};
357 }
358 
359 const TokenBuffer::Mapping *
360 TokenBuffer::mappingStartingBeforeSpelled(const MarkedFile &F,
361                                           const syntax::Token *Spelled) {
362   assert(F.SpelledTokens.data() <= Spelled);
363   unsigned SpelledI = Spelled - F.SpelledTokens.data();
364   assert(SpelledI < F.SpelledTokens.size());
365 
366   auto It = llvm::partition_point(F.Mappings, [SpelledI](const Mapping &M) {
367     return M.BeginSpelled <= SpelledI;
368   });
369   if (It == F.Mappings.begin())
370     return nullptr;
371   --It;
372   return &*It;
373 }
374 
375 llvm::SmallVector<llvm::ArrayRef<syntax::Token>, 1>
376 TokenBuffer::expandedForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const {
377   if (Spelled.empty())
378     return {};
379   const auto &File = fileForSpelled(Spelled);
380 
381   auto *FrontMapping = mappingStartingBeforeSpelled(File, &Spelled.front());
382   unsigned SpelledFrontI = &Spelled.front() - File.SpelledTokens.data();
383   assert(SpelledFrontI < File.SpelledTokens.size());
384   unsigned ExpandedBegin;
385   if (!FrontMapping) {
386     // No mapping that starts before the first token of Spelled, we don't have
387     // to modify offsets.
388     ExpandedBegin = File.BeginExpanded + SpelledFrontI;
389   } else if (SpelledFrontI < FrontMapping->EndSpelled) {
390     // This mapping applies to Spelled tokens.
391     if (SpelledFrontI != FrontMapping->BeginSpelled) {
392       // Spelled tokens don't cover the entire mapping, returning empty result.
393       return {}; // FIXME: support macro arguments.
394     }
395     // Spelled tokens start at the beginning of this mapping.
396     ExpandedBegin = FrontMapping->BeginExpanded;
397   } else {
398     // Spelled tokens start after the mapping ends (they start in the hole
399     // between 2 mappings, or between a mapping and end of the file).
400     ExpandedBegin =
401         FrontMapping->EndExpanded + (SpelledFrontI - FrontMapping->EndSpelled);
402   }
403 
404   auto *BackMapping = mappingStartingBeforeSpelled(File, &Spelled.back());
405   unsigned SpelledBackI = &Spelled.back() - File.SpelledTokens.data();
406   unsigned ExpandedEnd;
407   if (!BackMapping) {
408     // No mapping that starts before the last token of Spelled, we don't have to
409     // modify offsets.
410     ExpandedEnd = File.BeginExpanded + SpelledBackI + 1;
411   } else if (SpelledBackI < BackMapping->EndSpelled) {
412     // This mapping applies to Spelled tokens.
413     if (SpelledBackI + 1 != BackMapping->EndSpelled) {
414       // Spelled tokens don't cover the entire mapping, returning empty result.
415       return {}; // FIXME: support macro arguments.
416     }
417     ExpandedEnd = BackMapping->EndExpanded;
418   } else {
419     // Spelled tokens end after the mapping ends.
420     ExpandedEnd =
421         BackMapping->EndExpanded + (SpelledBackI - BackMapping->EndSpelled) + 1;
422   }
423 
424   assert(ExpandedBegin < ExpandedTokens.size());
425   assert(ExpandedEnd < ExpandedTokens.size());
426   // Avoid returning empty ranges.
427   if (ExpandedBegin == ExpandedEnd)
428     return {};
429   return {llvm::makeArrayRef(ExpandedTokens.data() + ExpandedBegin,
430                              ExpandedTokens.data() + ExpandedEnd)};
431 }
432 
433 llvm::ArrayRef<syntax::Token> TokenBuffer::spelledTokens(FileID FID) const {
434   auto It = Files.find(FID);
435   assert(It != Files.end());
436   return It->second.SpelledTokens;
437 }
438 
439 const syntax::Token *TokenBuffer::spelledTokenAt(SourceLocation Loc) const {
440   assert(Loc.isFileID());
441   const auto *Tok = llvm::partition_point(
442       spelledTokens(SourceMgr->getFileID(Loc)),
443       [&](const syntax::Token &Tok) { return Tok.location() < Loc; });
444   if (!Tok || Tok->location() != Loc)
445     return nullptr;
446   return Tok;
447 }
448 
449 std::string TokenBuffer::Mapping::str() const {
450   return std::string(
451       llvm::formatv("spelled tokens: [{0},{1}), expanded tokens: [{2},{3})",
452                     BeginSpelled, EndSpelled, BeginExpanded, EndExpanded));
453 }
454 
455 llvm::Optional<llvm::ArrayRef<syntax::Token>>
456 TokenBuffer::spelledForExpanded(llvm::ArrayRef<syntax::Token> Expanded) const {
457   // Mapping an empty range is ambiguous in case of empty mappings at either end
458   // of the range, bail out in that case.
459   if (Expanded.empty())
460     return llvm::None;
461   const syntax::Token *First = &Expanded.front();
462   const syntax::Token *Last = &Expanded.back();
463   const syntax::Token *FirstSpelled, *LastSpelled;
464   const TokenBuffer::Mapping *FirstMapping, *LastMapping;
465   std::tie(FirstSpelled, FirstMapping) = spelledForExpandedToken(First);
466   std::tie(LastSpelled, LastMapping) = spelledForExpandedToken(Last);
467 
468   FileID FID = SourceMgr->getFileID(FirstSpelled->location());
469   // FIXME: Handle multi-file changes by trying to map onto a common root.
470   if (FID != SourceMgr->getFileID(LastSpelled->location()))
471     return llvm::None;
472 
473   const MarkedFile &File = Files.find(FID)->second;
474 
475   // If the range is within one macro argument, the result may be only part of a
476   // Mapping. We must use the general (SourceManager-based) algorithm.
477   if (FirstMapping && FirstMapping == LastMapping &&
478       SourceMgr->isMacroArgExpansion(First->location()) &&
479       SourceMgr->isMacroArgExpansion(Last->location())) {
480     // We use excluded Prev/Next token for bounds checking.
481     SourceLocation Prev = (First == &ExpandedTokens.front())
482                               ? SourceLocation()
483                               : (First - 1)->location();
484     SourceLocation Next = (Last == &ExpandedTokens.back())
485                               ? SourceLocation()
486                               : (Last + 1)->location();
487     SourceRange Range = spelledForExpandedSlow(
488         First->location(), Last->location(), Prev, Next, FID, *SourceMgr);
489     if (Range.isInvalid())
490       return llvm::None;
491     return getTokensCovering(File.SpelledTokens, Range, *SourceMgr);
492   }
493 
494   // Otherwise, use the fast version based on Mappings.
495   // Do not allow changes that doesn't cover full expansion.
496   unsigned FirstExpanded = Expanded.begin() - ExpandedTokens.data();
497   unsigned LastExpanded = Expanded.end() - ExpandedTokens.data();
498   if (FirstMapping && FirstExpanded != FirstMapping->BeginExpanded)
499     return llvm::None;
500   if (LastMapping && LastMapping->EndExpanded != LastExpanded)
501     return llvm::None;
502   return llvm::makeArrayRef(
503       FirstMapping ? File.SpelledTokens.data() + FirstMapping->BeginSpelled
504                    : FirstSpelled,
505       LastMapping ? File.SpelledTokens.data() + LastMapping->EndSpelled
506                   : LastSpelled + 1);
507 }
508 
509 TokenBuffer::Expansion TokenBuffer::makeExpansion(const MarkedFile &F,
510                                                   const Mapping &M) const {
511   Expansion E;
512   E.Spelled = llvm::makeArrayRef(F.SpelledTokens.data() + M.BeginSpelled,
513                                  F.SpelledTokens.data() + M.EndSpelled);
514   E.Expanded = llvm::makeArrayRef(ExpandedTokens.data() + M.BeginExpanded,
515                                   ExpandedTokens.data() + M.EndExpanded);
516   return E;
517 }
518 
519 const TokenBuffer::MarkedFile &
520 TokenBuffer::fileForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const {
521   assert(!Spelled.empty());
522   assert(Spelled.front().location().isFileID() && "not a spelled token");
523   auto FileIt = Files.find(SourceMgr->getFileID(Spelled.front().location()));
524   assert(FileIt != Files.end() && "file not tracked by token buffer");
525   const auto &File = FileIt->second;
526   assert(File.SpelledTokens.data() <= Spelled.data() &&
527          Spelled.end() <=
528              (File.SpelledTokens.data() + File.SpelledTokens.size()) &&
529          "Tokens not in spelled range");
530 #ifndef NDEBUG
531   auto T1 = Spelled.back().location();
532   auto T2 = File.SpelledTokens.back().location();
533   assert(T1 == T2 || sourceManager().isBeforeInTranslationUnit(T1, T2));
534 #endif
535   return File;
536 }
537 
538 llvm::Optional<TokenBuffer::Expansion>
539 TokenBuffer::expansionStartingAt(const syntax::Token *Spelled) const {
540   assert(Spelled);
541   const auto &File = fileForSpelled(*Spelled);
542 
543   unsigned SpelledIndex = Spelled - File.SpelledTokens.data();
544   auto M = llvm::partition_point(File.Mappings, [&](const Mapping &M) {
545     return M.BeginSpelled < SpelledIndex;
546   });
547   if (M == File.Mappings.end() || M->BeginSpelled != SpelledIndex)
548     return llvm::None;
549   return makeExpansion(File, *M);
550 }
551 
552 std::vector<TokenBuffer::Expansion> TokenBuffer::expansionsOverlapping(
553     llvm::ArrayRef<syntax::Token> Spelled) const {
554   if (Spelled.empty())
555     return {};
556   const auto &File = fileForSpelled(Spelled);
557 
558   // Find the first overlapping range, and then copy until we stop overlapping.
559   unsigned SpelledBeginIndex = Spelled.begin() - File.SpelledTokens.data();
560   unsigned SpelledEndIndex = Spelled.end() - File.SpelledTokens.data();
561   auto M = llvm::partition_point(File.Mappings, [&](const Mapping &M) {
562     return M.EndSpelled <= SpelledBeginIndex;
563   });
564   std::vector<TokenBuffer::Expansion> Expansions;
565   for (; M != File.Mappings.end() && M->BeginSpelled < SpelledEndIndex; ++M)
566     Expansions.push_back(makeExpansion(File, *M));
567   return Expansions;
568 }
569 
570 llvm::ArrayRef<syntax::Token>
571 syntax::spelledTokensTouching(SourceLocation Loc,
572                               llvm::ArrayRef<syntax::Token> Tokens) {
573   assert(Loc.isFileID());
574 
575   auto *Right = llvm::partition_point(
576       Tokens, [&](const syntax::Token &Tok) { return Tok.location() < Loc; });
577   bool AcceptRight = Right != Tokens.end() && Right->location() <= Loc;
578   bool AcceptLeft =
579       Right != Tokens.begin() && (Right - 1)->endLocation() >= Loc;
580   return llvm::makeArrayRef(Right - (AcceptLeft ? 1 : 0),
581                             Right + (AcceptRight ? 1 : 0));
582 }
583 
584 llvm::ArrayRef<syntax::Token>
585 syntax::spelledTokensTouching(SourceLocation Loc,
586                               const syntax::TokenBuffer &Tokens) {
587   return spelledTokensTouching(
588       Loc, Tokens.spelledTokens(Tokens.sourceManager().getFileID(Loc)));
589 }
590 
591 const syntax::Token *
592 syntax::spelledIdentifierTouching(SourceLocation Loc,
593                                   llvm::ArrayRef<syntax::Token> Tokens) {
594   for (const syntax::Token &Tok : spelledTokensTouching(Loc, Tokens)) {
595     if (Tok.kind() == tok::identifier)
596       return &Tok;
597   }
598   return nullptr;
599 }
600 
601 const syntax::Token *
602 syntax::spelledIdentifierTouching(SourceLocation Loc,
603                                   const syntax::TokenBuffer &Tokens) {
604   return spelledIdentifierTouching(
605       Loc, Tokens.spelledTokens(Tokens.sourceManager().getFileID(Loc)));
606 }
607 
608 std::vector<const syntax::Token *>
609 TokenBuffer::macroExpansions(FileID FID) const {
610   auto FileIt = Files.find(FID);
611   assert(FileIt != Files.end() && "file not tracked by token buffer");
612   auto &File = FileIt->second;
613   std::vector<const syntax::Token *> Expansions;
614   auto &Spelled = File.SpelledTokens;
615   for (auto Mapping : File.Mappings) {
616     const syntax::Token *Token = &Spelled[Mapping.BeginSpelled];
617     if (Token->kind() == tok::TokenKind::identifier)
618       Expansions.push_back(Token);
619   }
620   return Expansions;
621 }
622 
623 std::vector<syntax::Token> syntax::tokenize(const FileRange &FR,
624                                             const SourceManager &SM,
625                                             const LangOptions &LO) {
626   std::vector<syntax::Token> Tokens;
627   IdentifierTable Identifiers(LO);
628   auto AddToken = [&](clang::Token T) {
629     // Fill the proper token kind for keywords, etc.
630     if (T.getKind() == tok::raw_identifier && !T.needsCleaning() &&
631         !T.hasUCN()) { // FIXME: support needsCleaning and hasUCN cases.
632       clang::IdentifierInfo &II = Identifiers.get(T.getRawIdentifier());
633       T.setIdentifierInfo(&II);
634       T.setKind(II.getTokenID());
635     }
636     Tokens.push_back(syntax::Token(T));
637   };
638 
639   auto SrcBuffer = SM.getBufferData(FR.file());
640   Lexer L(SM.getLocForStartOfFile(FR.file()), LO, SrcBuffer.data(),
641           SrcBuffer.data() + FR.beginOffset(),
642           // We can't make BufEnd point to FR.endOffset, as Lexer requires a
643           // null terminated buffer.
644           SrcBuffer.data() + SrcBuffer.size());
645 
646   clang::Token T;
647   while (!L.LexFromRawLexer(T) && L.getCurrentBufferOffset() < FR.endOffset())
648     AddToken(T);
649   // LexFromRawLexer returns true when it parses the last token of the file, add
650   // it iff it starts within the range we are interested in.
651   if (SM.getFileOffset(T.getLocation()) < FR.endOffset())
652     AddToken(T);
653   return Tokens;
654 }
655 
656 std::vector<syntax::Token> syntax::tokenize(FileID FID, const SourceManager &SM,
657                                             const LangOptions &LO) {
658   return tokenize(syntax::FileRange(FID, 0, SM.getFileIDSize(FID)), SM, LO);
659 }
660 
661 /// Records information reqired to construct mappings for the token buffer that
662 /// we are collecting.
663 class TokenCollector::CollectPPExpansions : public PPCallbacks {
664 public:
665   CollectPPExpansions(TokenCollector &C) : Collector(&C) {}
666 
667   /// Disabled instance will stop reporting anything to TokenCollector.
668   /// This ensures that uses of the preprocessor after TokenCollector::consume()
669   /// is called do not access the (possibly invalid) collector instance.
670   void disable() { Collector = nullptr; }
671 
672   void MacroExpands(const clang::Token &MacroNameTok, const MacroDefinition &MD,
673                     SourceRange Range, const MacroArgs *Args) override {
674     if (!Collector)
675       return;
676     const auto &SM = Collector->PP.getSourceManager();
677     // Only record top-level expansions that directly produce expanded tokens.
678     // This excludes those where:
679     //   - the macro use is inside a macro body,
680     //   - the macro appears in an argument to another macro.
681     // However macro expansion isn't really a tree, it's token rewrite rules,
682     // so there are other cases, e.g.
683     //   #define B(X) X
684     //   #define A 1 + B
685     //   A(2)
686     // Both A and B produce expanded tokens, though the macro name 'B' comes
687     // from an expansion. The best we can do is merge the mappings for both.
688 
689     // The *last* token of any top-level macro expansion must be in a file.
690     // (In the example above, see the closing paren of the expansion of B).
691     if (!Range.getEnd().isFileID())
692       return;
693     // If there's a current expansion that encloses this one, this one can't be
694     // top-level.
695     if (LastExpansionEnd.isValid() &&
696         !SM.isBeforeInTranslationUnit(LastExpansionEnd, Range.getEnd()))
697       return;
698 
699     // If the macro invocation (B) starts in a macro (A) but ends in a file,
700     // we'll create a merged mapping for A + B by overwriting the endpoint for
701     // A's startpoint.
702     if (!Range.getBegin().isFileID()) {
703       Range.setBegin(SM.getExpansionLoc(Range.getBegin()));
704       assert(Collector->Expansions.count(Range.getBegin()) &&
705              "Overlapping macros should have same expansion location");
706     }
707 
708     Collector->Expansions[Range.getBegin()] = Range.getEnd();
709     LastExpansionEnd = Range.getEnd();
710   }
711   // FIXME: handle directives like #pragma, #include, etc.
712 private:
713   TokenCollector *Collector;
714   /// Used to detect recursive macro expansions.
715   SourceLocation LastExpansionEnd;
716 };
717 
718 /// Fills in the TokenBuffer by tracing the run of a preprocessor. The
719 /// implementation tracks the tokens, macro expansions and directives coming
720 /// from the preprocessor and:
721 /// - for each token, figures out if it is a part of an expanded token stream,
722 ///   spelled token stream or both. Stores the tokens appropriately.
723 /// - records mappings from the spelled to expanded token ranges, e.g. for macro
724 ///   expansions.
725 /// FIXME: also properly record:
726 ///          - #include directives,
727 ///          - #pragma, #line and other PP directives,
728 ///          - skipped pp regions,
729 ///          - ...
730 
731 TokenCollector::TokenCollector(Preprocessor &PP) : PP(PP) {
732   // Collect the expanded token stream during preprocessing.
733   PP.setTokenWatcher([this](const clang::Token &T) {
734     if (T.isAnnotation())
735       return;
736     DEBUG_WITH_TYPE("collect-tokens", llvm::dbgs()
737                                           << "Token: "
738                                           << syntax::Token(T).dumpForTests(
739                                                  this->PP.getSourceManager())
740                                           << "\n"
741 
742     );
743     Expanded.push_back(syntax::Token(T));
744   });
745   // And locations of macro calls, to properly recover boundaries of those in
746   // case of empty expansions.
747   auto CB = std::make_unique<CollectPPExpansions>(*this);
748   this->Collector = CB.get();
749   PP.addPPCallbacks(std::move(CB));
750 }
751 
752 /// Builds mappings and spelled tokens in the TokenBuffer based on the expanded
753 /// token stream.
754 class TokenCollector::Builder {
755 public:
756   Builder(std::vector<syntax::Token> Expanded, PPExpansions CollectedExpansions,
757           const SourceManager &SM, const LangOptions &LangOpts)
758       : Result(SM), CollectedExpansions(std::move(CollectedExpansions)), SM(SM),
759         LangOpts(LangOpts) {
760     Result.ExpandedTokens = std::move(Expanded);
761   }
762 
763   TokenBuffer build() && {
764     assert(!Result.ExpandedTokens.empty());
765     assert(Result.ExpandedTokens.back().kind() == tok::eof);
766 
767     // Tokenize every file that contributed tokens to the expanded stream.
768     buildSpelledTokens();
769 
770     // The expanded token stream consists of runs of tokens that came from
771     // the same source (a macro expansion, part of a file etc).
772     // Between these runs are the logical positions of spelled tokens that
773     // didn't expand to anything.
774     while (NextExpanded < Result.ExpandedTokens.size() - 1 /* eof */) {
775       // Create empty mappings for spelled tokens that expanded to nothing here.
776       // May advance NextSpelled, but NextExpanded is unchanged.
777       discard();
778       // Create mapping for a contiguous run of expanded tokens.
779       // Advances NextExpanded past the run, and NextSpelled accordingly.
780       unsigned OldPosition = NextExpanded;
781       advance();
782       if (NextExpanded == OldPosition)
783         diagnoseAdvanceFailure();
784     }
785     // If any tokens remain in any of the files, they didn't expand to anything.
786     // Create empty mappings up until the end of the file.
787     for (const auto &File : Result.Files)
788       discard(File.first);
789 
790 #ifndef NDEBUG
791     for (auto &pair : Result.Files) {
792       auto &mappings = pair.second.Mappings;
793       assert(llvm::is_sorted(mappings, [](const TokenBuffer::Mapping &M1,
794                                           const TokenBuffer::Mapping &M2) {
795         return M1.BeginSpelled < M2.BeginSpelled &&
796                M1.EndSpelled < M2.EndSpelled &&
797                M1.BeginExpanded < M2.BeginExpanded &&
798                M1.EndExpanded < M2.EndExpanded;
799       }));
800     }
801 #endif
802 
803     return std::move(Result);
804   }
805 
806 private:
807   // Consume a sequence of spelled tokens that didn't expand to anything.
808   // In the simplest case, skips spelled tokens until finding one that produced
809   // the NextExpanded token, and creates an empty mapping for them.
810   // If Drain is provided, skips remaining tokens from that file instead.
811   void discard(llvm::Optional<FileID> Drain = llvm::None) {
812     SourceLocation Target =
813         Drain ? SM.getLocForEndOfFile(*Drain)
814               : SM.getExpansionLoc(
815                     Result.ExpandedTokens[NextExpanded].location());
816     FileID File = SM.getFileID(Target);
817     const auto &SpelledTokens = Result.Files[File].SpelledTokens;
818     auto &NextSpelled = this->NextSpelled[File];
819 
820     TokenBuffer::Mapping Mapping;
821     Mapping.BeginSpelled = NextSpelled;
822     // When dropping trailing tokens from a file, the empty mapping should
823     // be positioned within the file's expanded-token range (at the end).
824     Mapping.BeginExpanded = Mapping.EndExpanded =
825         Drain ? Result.Files[*Drain].EndExpanded : NextExpanded;
826     // We may want to split into several adjacent empty mappings.
827     // FlushMapping() emits the current mapping and starts a new one.
828     auto FlushMapping = [&, this] {
829       Mapping.EndSpelled = NextSpelled;
830       if (Mapping.BeginSpelled != Mapping.EndSpelled)
831         Result.Files[File].Mappings.push_back(Mapping);
832       Mapping.BeginSpelled = NextSpelled;
833     };
834 
835     while (NextSpelled < SpelledTokens.size() &&
836            SpelledTokens[NextSpelled].location() < Target) {
837       // If we know mapping bounds at [NextSpelled, KnownEnd] (macro expansion)
838       // then we want to partition our (empty) mapping.
839       //   [Start, NextSpelled) [NextSpelled, KnownEnd] (KnownEnd, Target)
840       SourceLocation KnownEnd =
841           CollectedExpansions.lookup(SpelledTokens[NextSpelled].location());
842       if (KnownEnd.isValid()) {
843         FlushMapping(); // Emits [Start, NextSpelled)
844         while (NextSpelled < SpelledTokens.size() &&
845                SpelledTokens[NextSpelled].location() <= KnownEnd)
846           ++NextSpelled;
847         FlushMapping(); // Emits [NextSpelled, KnownEnd]
848         // Now the loop contitues and will emit (KnownEnd, Target).
849       } else {
850         ++NextSpelled;
851       }
852     }
853     FlushMapping();
854   }
855 
856   // Consumes the NextExpanded token and others that are part of the same run.
857   // Increases NextExpanded and NextSpelled by at least one, and adds a mapping
858   // (unless this is a run of file tokens, which we represent with no mapping).
859   void advance() {
860     const syntax::Token &Tok = Result.ExpandedTokens[NextExpanded];
861     SourceLocation Expansion = SM.getExpansionLoc(Tok.location());
862     FileID File = SM.getFileID(Expansion);
863     const auto &SpelledTokens = Result.Files[File].SpelledTokens;
864     auto &NextSpelled = this->NextSpelled[File];
865 
866     if (Tok.location().isFileID()) {
867       // A run of file tokens continues while the expanded/spelled tokens match.
868       while (NextSpelled < SpelledTokens.size() &&
869              NextExpanded < Result.ExpandedTokens.size() &&
870              SpelledTokens[NextSpelled].location() ==
871                  Result.ExpandedTokens[NextExpanded].location()) {
872         ++NextSpelled;
873         ++NextExpanded;
874       }
875       // We need no mapping for file tokens copied to the expanded stream.
876     } else {
877       // We found a new macro expansion. We should have its spelling bounds.
878       auto End = CollectedExpansions.lookup(Expansion);
879       assert(End.isValid() && "Macro expansion wasn't captured?");
880 
881       // Mapping starts here...
882       TokenBuffer::Mapping Mapping;
883       Mapping.BeginExpanded = NextExpanded;
884       Mapping.BeginSpelled = NextSpelled;
885       // ... consumes spelled tokens within bounds we captured ...
886       while (NextSpelled < SpelledTokens.size() &&
887              SpelledTokens[NextSpelled].location() <= End)
888         ++NextSpelled;
889       // ... consumes expanded tokens rooted at the same expansion ...
890       while (NextExpanded < Result.ExpandedTokens.size() &&
891              SM.getExpansionLoc(
892                  Result.ExpandedTokens[NextExpanded].location()) == Expansion)
893         ++NextExpanded;
894       // ... and ends here.
895       Mapping.EndExpanded = NextExpanded;
896       Mapping.EndSpelled = NextSpelled;
897       Result.Files[File].Mappings.push_back(Mapping);
898     }
899   }
900 
901   // advance() is supposed to consume at least one token - if not, we crash.
902   void diagnoseAdvanceFailure() {
903 #ifndef NDEBUG
904     // Show the failed-to-map token in context.
905     for (unsigned I = (NextExpanded < 10) ? 0 : NextExpanded - 10;
906          I < NextExpanded + 5 && I < Result.ExpandedTokens.size(); ++I) {
907       const char *L =
908           (I == NextExpanded) ? "!! " : (I < NextExpanded) ? "ok " : "   ";
909       llvm::errs() << L << Result.ExpandedTokens[I].dumpForTests(SM) << "\n";
910     }
911 #endif
912     llvm_unreachable("Couldn't map expanded token to spelled tokens!");
913   }
914 
915   /// Initializes TokenBuffer::Files and fills spelled tokens and expanded
916   /// ranges for each of the files.
917   void buildSpelledTokens() {
918     for (unsigned I = 0; I < Result.ExpandedTokens.size(); ++I) {
919       const auto &Tok = Result.ExpandedTokens[I];
920       auto FID = SM.getFileID(SM.getExpansionLoc(Tok.location()));
921       auto It = Result.Files.try_emplace(FID);
922       TokenBuffer::MarkedFile &File = It.first->second;
923 
924       // The eof token should not be considered part of the main-file's range.
925       File.EndExpanded = Tok.kind() == tok::eof ? I : I + 1;
926 
927       if (!It.second)
928         continue; // we have seen this file before.
929       // This is the first time we see this file.
930       File.BeginExpanded = I;
931       File.SpelledTokens = tokenize(FID, SM, LangOpts);
932     }
933   }
934 
935   TokenBuffer Result;
936   unsigned NextExpanded = 0;                    // cursor in ExpandedTokens
937   llvm::DenseMap<FileID, unsigned> NextSpelled; // cursor in SpelledTokens
938   PPExpansions CollectedExpansions;
939   const SourceManager &SM;
940   const LangOptions &LangOpts;
941 };
942 
943 TokenBuffer TokenCollector::consume() && {
944   PP.setTokenWatcher(nullptr);
945   Collector->disable();
946   return Builder(std::move(Expanded), std::move(Expansions),
947                  PP.getSourceManager(), PP.getLangOpts())
948       .build();
949 }
950 
951 std::string syntax::Token::str() const {
952   return std::string(llvm::formatv("Token({0}, length = {1})",
953                                    tok::getTokenName(kind()), length()));
954 }
955 
956 std::string syntax::Token::dumpForTests(const SourceManager &SM) const {
957   return std::string(llvm::formatv("Token(`{0}`, {1}, length = {2})", text(SM),
958                                    tok::getTokenName(kind()), length()));
959 }
960 
961 std::string TokenBuffer::dumpForTests() const {
962   auto PrintToken = [this](const syntax::Token &T) -> std::string {
963     if (T.kind() == tok::eof)
964       return "<eof>";
965     return std::string(T.text(*SourceMgr));
966   };
967 
968   auto DumpTokens = [this, &PrintToken](llvm::raw_ostream &OS,
969                                         llvm::ArrayRef<syntax::Token> Tokens) {
970     if (Tokens.empty()) {
971       OS << "<empty>";
972       return;
973     }
974     OS << Tokens[0].text(*SourceMgr);
975     for (unsigned I = 1; I < Tokens.size(); ++I) {
976       if (Tokens[I].kind() == tok::eof)
977         continue;
978       OS << " " << PrintToken(Tokens[I]);
979     }
980   };
981 
982   std::string Dump;
983   llvm::raw_string_ostream OS(Dump);
984 
985   OS << "expanded tokens:\n"
986      << "  ";
987   // (!) we do not show '<eof>'.
988   DumpTokens(OS, llvm::makeArrayRef(ExpandedTokens).drop_back());
989   OS << "\n";
990 
991   std::vector<FileID> Keys;
992   for (auto F : Files)
993     Keys.push_back(F.first);
994   llvm::sort(Keys);
995 
996   for (FileID ID : Keys) {
997     const MarkedFile &File = Files.find(ID)->second;
998     auto *Entry = SourceMgr->getFileEntryForID(ID);
999     if (!Entry)
1000       continue; // Skip builtin files.
1001     OS << llvm::formatv("file '{0}'\n", Entry->getName())
1002        << "  spelled tokens:\n"
1003        << "    ";
1004     DumpTokens(OS, File.SpelledTokens);
1005     OS << "\n";
1006 
1007     if (File.Mappings.empty()) {
1008       OS << "  no mappings.\n";
1009       continue;
1010     }
1011     OS << "  mappings:\n";
1012     for (auto &M : File.Mappings) {
1013       OS << llvm::formatv(
1014           "    ['{0}'_{1}, '{2}'_{3}) => ['{4}'_{5}, '{6}'_{7})\n",
1015           PrintToken(File.SpelledTokens[M.BeginSpelled]), M.BeginSpelled,
1016           M.EndSpelled == File.SpelledTokens.size()
1017               ? "<eof>"
1018               : PrintToken(File.SpelledTokens[M.EndSpelled]),
1019           M.EndSpelled, PrintToken(ExpandedTokens[M.BeginExpanded]),
1020           M.BeginExpanded, PrintToken(ExpandedTokens[M.EndExpanded]),
1021           M.EndExpanded);
1022     }
1023   }
1024   return Dump;
1025 }
1026