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 syntax::Token::Token(SourceLocation Location, unsigned Length, 39 tok::TokenKind Kind) 40 : Location(Location), Length(Length), Kind(Kind) { 41 assert(Location.isValid()); 42 } 43 44 syntax::Token::Token(const clang::Token &T) 45 : Token(T.getLocation(), T.getLength(), T.getKind()) { 46 assert(!T.isAnnotation()); 47 } 48 49 llvm::StringRef syntax::Token::text(const SourceManager &SM) const { 50 bool Invalid = false; 51 const char *Start = SM.getCharacterData(location(), &Invalid); 52 assert(!Invalid); 53 return llvm::StringRef(Start, length()); 54 } 55 56 FileRange syntax::Token::range(const SourceManager &SM) const { 57 assert(location().isFileID() && "must be a spelled token"); 58 FileID File; 59 unsigned StartOffset; 60 std::tie(File, StartOffset) = SM.getDecomposedLoc(location()); 61 return FileRange(File, StartOffset, StartOffset + length()); 62 } 63 64 FileRange syntax::Token::range(const SourceManager &SM, 65 const syntax::Token &First, 66 const syntax::Token &Last) { 67 auto F = First.range(SM); 68 auto L = Last.range(SM); 69 assert(F.file() == L.file() && "tokens from different files"); 70 assert(F.endOffset() <= L.beginOffset() && "wrong order of tokens"); 71 return FileRange(F.file(), F.beginOffset(), L.endOffset()); 72 } 73 74 llvm::raw_ostream &syntax::operator<<(llvm::raw_ostream &OS, const Token &T) { 75 return OS << T.str(); 76 } 77 78 FileRange::FileRange(FileID File, unsigned BeginOffset, unsigned EndOffset) 79 : File(File), Begin(BeginOffset), End(EndOffset) { 80 assert(File.isValid()); 81 assert(BeginOffset <= EndOffset); 82 } 83 84 FileRange::FileRange(const SourceManager &SM, SourceLocation BeginLoc, 85 unsigned Length) { 86 assert(BeginLoc.isValid()); 87 assert(BeginLoc.isFileID()); 88 89 std::tie(File, Begin) = SM.getDecomposedLoc(BeginLoc); 90 End = Begin + Length; 91 } 92 FileRange::FileRange(const SourceManager &SM, SourceLocation BeginLoc, 93 SourceLocation EndLoc) { 94 assert(BeginLoc.isValid()); 95 assert(BeginLoc.isFileID()); 96 assert(EndLoc.isValid()); 97 assert(EndLoc.isFileID()); 98 assert(SM.getFileID(BeginLoc) == SM.getFileID(EndLoc)); 99 assert(SM.getFileOffset(BeginLoc) <= SM.getFileOffset(EndLoc)); 100 101 std::tie(File, Begin) = SM.getDecomposedLoc(BeginLoc); 102 End = SM.getFileOffset(EndLoc); 103 } 104 105 llvm::raw_ostream &syntax::operator<<(llvm::raw_ostream &OS, 106 const FileRange &R) { 107 return OS << llvm::formatv("FileRange(file = {0}, offsets = {1}-{2})", 108 R.file().getHashValue(), R.beginOffset(), 109 R.endOffset()); 110 } 111 112 llvm::StringRef FileRange::text(const SourceManager &SM) const { 113 bool Invalid = false; 114 StringRef Text = SM.getBufferData(File, &Invalid); 115 if (Invalid) 116 return ""; 117 assert(Begin <= Text.size()); 118 assert(End <= Text.size()); 119 return Text.substr(Begin, length()); 120 } 121 122 llvm::ArrayRef<syntax::Token> TokenBuffer::expandedTokens(SourceRange R) const { 123 if (R.isInvalid()) 124 return {}; 125 const Token *Begin = 126 llvm::partition_point(expandedTokens(), [&](const syntax::Token &T) { 127 return SourceMgr->isBeforeInTranslationUnit(T.location(), R.getBegin()); 128 }); 129 const Token *End = 130 llvm::partition_point(expandedTokens(), [&](const syntax::Token &T) { 131 return !SourceMgr->isBeforeInTranslationUnit(R.getEnd(), T.location()); 132 }); 133 if (Begin > End) 134 return {}; 135 return {Begin, End}; 136 } 137 138 std::pair<const syntax::Token *, const TokenBuffer::Mapping *> 139 TokenBuffer::spelledForExpandedToken(const syntax::Token *Expanded) const { 140 assert(Expanded); 141 assert(ExpandedTokens.data() <= Expanded && 142 Expanded < ExpandedTokens.data() + ExpandedTokens.size()); 143 144 auto FileIt = Files.find( 145 SourceMgr->getFileID(SourceMgr->getExpansionLoc(Expanded->location()))); 146 assert(FileIt != Files.end() && "no file for an expanded token"); 147 148 const MarkedFile &File = FileIt->second; 149 150 unsigned ExpandedIndex = Expanded - ExpandedTokens.data(); 151 // Find the first mapping that produced tokens after \p Expanded. 152 auto It = llvm::partition_point(File.Mappings, [&](const Mapping &M) { 153 return M.BeginExpanded <= ExpandedIndex; 154 }); 155 // Our token could only be produced by the previous mapping. 156 if (It == File.Mappings.begin()) { 157 // No previous mapping, no need to modify offsets. 158 return {&File.SpelledTokens[ExpandedIndex - File.BeginExpanded], nullptr}; 159 } 160 --It; // 'It' now points to last mapping that started before our token. 161 162 // Check if the token is part of the mapping. 163 if (ExpandedIndex < It->EndExpanded) 164 return {&File.SpelledTokens[It->BeginSpelled], /*Mapping*/ &*It}; 165 166 // Not part of the mapping, use the index from previous mapping to compute the 167 // corresponding spelled token. 168 return { 169 &File.SpelledTokens[It->EndSpelled + (ExpandedIndex - It->EndExpanded)], 170 /*Mapping*/ nullptr}; 171 } 172 173 llvm::ArrayRef<syntax::Token> TokenBuffer::spelledTokens(FileID FID) const { 174 auto It = Files.find(FID); 175 assert(It != Files.end()); 176 return It->second.SpelledTokens; 177 } 178 179 std::string TokenBuffer::Mapping::str() const { 180 return llvm::formatv("spelled tokens: [{0},{1}), expanded tokens: [{2},{3})", 181 BeginSpelled, EndSpelled, BeginExpanded, EndExpanded); 182 } 183 184 llvm::Optional<llvm::ArrayRef<syntax::Token>> 185 TokenBuffer::spelledForExpanded(llvm::ArrayRef<syntax::Token> Expanded) const { 186 // Mapping an empty range is ambiguous in case of empty mappings at either end 187 // of the range, bail out in that case. 188 if (Expanded.empty()) 189 return llvm::None; 190 191 // FIXME: also allow changes uniquely mapping to macro arguments. 192 193 const syntax::Token *BeginSpelled; 194 const Mapping *BeginMapping; 195 std::tie(BeginSpelled, BeginMapping) = 196 spelledForExpandedToken(&Expanded.front()); 197 198 const syntax::Token *LastSpelled; 199 const Mapping *LastMapping; 200 std::tie(LastSpelled, LastMapping) = 201 spelledForExpandedToken(&Expanded.back()); 202 203 FileID FID = SourceMgr->getFileID(BeginSpelled->location()); 204 // FIXME: Handle multi-file changes by trying to map onto a common root. 205 if (FID != SourceMgr->getFileID(LastSpelled->location())) 206 return llvm::None; 207 208 const MarkedFile &File = Files.find(FID)->second; 209 210 // Do not allow changes that cross macro expansion boundaries. 211 unsigned BeginExpanded = Expanded.begin() - ExpandedTokens.data(); 212 unsigned EndExpanded = Expanded.end() - ExpandedTokens.data(); 213 if (BeginMapping && BeginMapping->BeginExpanded < BeginExpanded) 214 return llvm::None; 215 if (LastMapping && EndExpanded < LastMapping->EndExpanded) 216 return llvm::None; 217 // All is good, return the result. 218 return llvm::makeArrayRef( 219 BeginMapping ? File.SpelledTokens.data() + BeginMapping->BeginSpelled 220 : BeginSpelled, 221 LastMapping ? File.SpelledTokens.data() + LastMapping->EndSpelled 222 : LastSpelled + 1); 223 } 224 225 llvm::Optional<TokenBuffer::Expansion> 226 TokenBuffer::expansionStartingAt(const syntax::Token *Spelled) const { 227 assert(Spelled); 228 assert(Spelled->location().isFileID() && "not a spelled token"); 229 auto FileIt = Files.find(SourceMgr->getFileID(Spelled->location())); 230 assert(FileIt != Files.end() && "file not tracked by token buffer"); 231 232 auto &File = FileIt->second; 233 assert(File.SpelledTokens.data() <= Spelled && 234 Spelled < (File.SpelledTokens.data() + File.SpelledTokens.size())); 235 236 unsigned SpelledIndex = Spelled - File.SpelledTokens.data(); 237 auto M = llvm::partition_point(File.Mappings, [&](const Mapping &M) { 238 return M.BeginSpelled < SpelledIndex; 239 }); 240 if (M == File.Mappings.end() || M->BeginSpelled != SpelledIndex) 241 return llvm::None; 242 243 Expansion E; 244 E.Spelled = llvm::makeArrayRef(File.SpelledTokens.data() + M->BeginSpelled, 245 File.SpelledTokens.data() + M->EndSpelled); 246 E.Expanded = llvm::makeArrayRef(ExpandedTokens.data() + M->BeginExpanded, 247 ExpandedTokens.data() + M->EndExpanded); 248 return E; 249 } 250 251 llvm::ArrayRef<syntax::Token> 252 syntax::spelledTokensTouching(SourceLocation Loc, 253 const syntax::TokenBuffer &Tokens) { 254 assert(Loc.isFileID()); 255 llvm::ArrayRef<syntax::Token> All = 256 Tokens.spelledTokens(Tokens.sourceManager().getFileID(Loc)); 257 auto *Right = llvm::partition_point( 258 All, [&](const syntax::Token &Tok) { return Tok.location() < Loc; }); 259 bool AcceptRight = Right != All.end() && Right->location() <= Loc; 260 bool AcceptLeft = Right != All.begin() && (Right - 1)->endLocation() >= Loc; 261 return llvm::makeArrayRef(Right - (AcceptLeft ? 1 : 0), 262 Right + (AcceptRight ? 1 : 0)); 263 } 264 265 const syntax::Token * 266 syntax::spelledIdentifierTouching(SourceLocation Loc, 267 const syntax::TokenBuffer &Tokens) { 268 for (const syntax::Token &Tok : spelledTokensTouching(Loc, Tokens)) { 269 if (Tok.kind() == tok::identifier) 270 return &Tok; 271 } 272 return nullptr; 273 } 274 275 std::vector<const syntax::Token *> 276 TokenBuffer::macroExpansions(FileID FID) const { 277 auto FileIt = Files.find(FID); 278 assert(FileIt != Files.end() && "file not tracked by token buffer"); 279 auto &File = FileIt->second; 280 std::vector<const syntax::Token *> Expansions; 281 auto &Spelled = File.SpelledTokens; 282 for (auto Mapping : File.Mappings) { 283 const syntax::Token *Token = &Spelled[Mapping.BeginSpelled]; 284 if (Token->kind() == tok::TokenKind::identifier) 285 Expansions.push_back(Token); 286 } 287 return Expansions; 288 } 289 290 std::vector<syntax::Token> syntax::tokenize(FileID FID, const SourceManager &SM, 291 const LangOptions &LO) { 292 std::vector<syntax::Token> Tokens; 293 IdentifierTable Identifiers(LO); 294 auto AddToken = [&](clang::Token T) { 295 // Fill the proper token kind for keywords, etc. 296 if (T.getKind() == tok::raw_identifier && !T.needsCleaning() && 297 !T.hasUCN()) { // FIXME: support needsCleaning and hasUCN cases. 298 clang::IdentifierInfo &II = Identifiers.get(T.getRawIdentifier()); 299 T.setIdentifierInfo(&II); 300 T.setKind(II.getTokenID()); 301 } 302 Tokens.push_back(syntax::Token(T)); 303 }; 304 305 Lexer L(FID, SM.getBuffer(FID), SM, LO); 306 307 clang::Token T; 308 while (!L.LexFromRawLexer(T)) 309 AddToken(T); 310 // 'eof' is only the last token if the input is null-terminated. Never store 311 // it, for consistency. 312 if (T.getKind() != tok::eof) 313 AddToken(T); 314 return Tokens; 315 } 316 317 /// Records information reqired to construct mappings for the token buffer that 318 /// we are collecting. 319 class TokenCollector::CollectPPExpansions : public PPCallbacks { 320 public: 321 CollectPPExpansions(TokenCollector &C) : Collector(&C) {} 322 323 /// Disabled instance will stop reporting anything to TokenCollector. 324 /// This ensures that uses of the preprocessor after TokenCollector::consume() 325 /// is called do not access the (possibly invalid) collector instance. 326 void disable() { Collector = nullptr; } 327 328 void MacroExpands(const clang::Token &MacroNameTok, const MacroDefinition &MD, 329 SourceRange Range, const MacroArgs *Args) override { 330 if (!Collector) 331 return; 332 // Only record top-level expansions, not those where: 333 // - the macro use is inside a macro body, 334 // - the macro appears in an argument to another macro. 335 if (!MacroNameTok.getLocation().isFileID() || 336 (LastExpansionEnd.isValid() && 337 Collector->PP.getSourceManager().isBeforeInTranslationUnit( 338 Range.getBegin(), LastExpansionEnd))) 339 return; 340 Collector->Expansions[Range.getBegin().getRawEncoding()] = Range.getEnd(); 341 LastExpansionEnd = Range.getEnd(); 342 } 343 // FIXME: handle directives like #pragma, #include, etc. 344 private: 345 TokenCollector *Collector; 346 /// Used to detect recursive macro expansions. 347 SourceLocation LastExpansionEnd; 348 }; 349 350 /// Fills in the TokenBuffer by tracing the run of a preprocessor. The 351 /// implementation tracks the tokens, macro expansions and directives coming 352 /// from the preprocessor and: 353 /// - for each token, figures out if it is a part of an expanded token stream, 354 /// spelled token stream or both. Stores the tokens appropriately. 355 /// - records mappings from the spelled to expanded token ranges, e.g. for macro 356 /// expansions. 357 /// FIXME: also properly record: 358 /// - #include directives, 359 /// - #pragma, #line and other PP directives, 360 /// - skipped pp regions, 361 /// - ... 362 363 TokenCollector::TokenCollector(Preprocessor &PP) : PP(PP) { 364 // Collect the expanded token stream during preprocessing. 365 PP.setTokenWatcher([this](const clang::Token &T) { 366 if (T.isAnnotation()) 367 return; 368 DEBUG_WITH_TYPE("collect-tokens", llvm::dbgs() 369 << "Token: " 370 << syntax::Token(T).dumpForTests( 371 this->PP.getSourceManager()) 372 << "\n" 373 374 ); 375 Expanded.push_back(syntax::Token(T)); 376 }); 377 // And locations of macro calls, to properly recover boundaries of those in 378 // case of empty expansions. 379 auto CB = std::make_unique<CollectPPExpansions>(*this); 380 this->Collector = CB.get(); 381 PP.addPPCallbacks(std::move(CB)); 382 } 383 384 /// Builds mappings and spelled tokens in the TokenBuffer based on the expanded 385 /// token stream. 386 class TokenCollector::Builder { 387 public: 388 Builder(std::vector<syntax::Token> Expanded, PPExpansions CollectedExpansions, 389 const SourceManager &SM, const LangOptions &LangOpts) 390 : Result(SM), CollectedExpansions(std::move(CollectedExpansions)), SM(SM), 391 LangOpts(LangOpts) { 392 Result.ExpandedTokens = std::move(Expanded); 393 } 394 395 TokenBuffer build() && { 396 buildSpelledTokens(); 397 398 // Walk over expanded tokens and spelled tokens in parallel, building the 399 // mappings between those using source locations. 400 // To correctly recover empty macro expansions, we also take locations 401 // reported to PPCallbacks::MacroExpands into account as we do not have any 402 // expanded tokens with source locations to guide us. 403 404 // The 'eof' token is special, it is not part of spelled token stream. We 405 // handle it separately at the end. 406 assert(!Result.ExpandedTokens.empty()); 407 assert(Result.ExpandedTokens.back().kind() == tok::eof); 408 for (unsigned I = 0; I < Result.ExpandedTokens.size() - 1; ++I) { 409 // (!) I might be updated by the following call. 410 processExpandedToken(I); 411 } 412 413 // 'eof' not handled in the loop, do it here. 414 assert(SM.getMainFileID() == 415 SM.getFileID(Result.ExpandedTokens.back().location())); 416 fillGapUntil(Result.Files[SM.getMainFileID()], 417 Result.ExpandedTokens.back().location(), 418 Result.ExpandedTokens.size() - 1); 419 Result.Files[SM.getMainFileID()].EndExpanded = Result.ExpandedTokens.size(); 420 421 // Some files might have unaccounted spelled tokens at the end, add an empty 422 // mapping for those as they did not have expanded counterparts. 423 fillGapsAtEndOfFiles(); 424 425 return std::move(Result); 426 } 427 428 private: 429 /// Process the next token in an expanded stream and move corresponding 430 /// spelled tokens, record any mapping if needed. 431 /// (!) \p I will be updated if this had to skip tokens, e.g. for macros. 432 void processExpandedToken(unsigned &I) { 433 auto L = Result.ExpandedTokens[I].location(); 434 if (L.isMacroID()) { 435 processMacroExpansion(SM.getExpansionRange(L), I); 436 return; 437 } 438 if (L.isFileID()) { 439 auto FID = SM.getFileID(L); 440 TokenBuffer::MarkedFile &File = Result.Files[FID]; 441 442 fillGapUntil(File, L, I); 443 444 // Skip the token. 445 assert(File.SpelledTokens[NextSpelled[FID]].location() == L && 446 "no corresponding token in the spelled stream"); 447 ++NextSpelled[FID]; 448 return; 449 } 450 } 451 452 /// Skipped expanded and spelled tokens of a macro expansion that covers \p 453 /// SpelledRange. Add a corresponding mapping. 454 /// (!) \p I will be the index of the last token in an expansion after this 455 /// function returns. 456 void processMacroExpansion(CharSourceRange SpelledRange, unsigned &I) { 457 auto FID = SM.getFileID(SpelledRange.getBegin()); 458 assert(FID == SM.getFileID(SpelledRange.getEnd())); 459 TokenBuffer::MarkedFile &File = Result.Files[FID]; 460 461 fillGapUntil(File, SpelledRange.getBegin(), I); 462 463 // Skip all expanded tokens from the same macro expansion. 464 unsigned BeginExpanded = I; 465 for (; I + 1 < Result.ExpandedTokens.size(); ++I) { 466 auto NextL = Result.ExpandedTokens[I + 1].location(); 467 if (!NextL.isMacroID() || 468 SM.getExpansionLoc(NextL) != SpelledRange.getBegin()) 469 break; 470 } 471 unsigned EndExpanded = I + 1; 472 consumeMapping(File, SM.getFileOffset(SpelledRange.getEnd()), BeginExpanded, 473 EndExpanded, NextSpelled[FID]); 474 } 475 476 /// Initializes TokenBuffer::Files and fills spelled tokens and expanded 477 /// ranges for each of the files. 478 void buildSpelledTokens() { 479 for (unsigned I = 0; I < Result.ExpandedTokens.size(); ++I) { 480 auto FID = 481 SM.getFileID(SM.getExpansionLoc(Result.ExpandedTokens[I].location())); 482 auto It = Result.Files.try_emplace(FID); 483 TokenBuffer::MarkedFile &File = It.first->second; 484 485 File.EndExpanded = I + 1; 486 if (!It.second) 487 continue; // we have seen this file before. 488 489 // This is the first time we see this file. 490 File.BeginExpanded = I; 491 File.SpelledTokens = tokenize(FID, SM, LangOpts); 492 } 493 } 494 495 void consumeEmptyMapping(TokenBuffer::MarkedFile &File, unsigned EndOffset, 496 unsigned ExpandedIndex, unsigned &SpelledIndex) { 497 consumeMapping(File, EndOffset, ExpandedIndex, ExpandedIndex, SpelledIndex); 498 } 499 500 /// Consumes spelled tokens that form a macro expansion and adds a entry to 501 /// the resulting token buffer. 502 /// (!) SpelledIndex is updated in-place. 503 void consumeMapping(TokenBuffer::MarkedFile &File, unsigned EndOffset, 504 unsigned BeginExpanded, unsigned EndExpanded, 505 unsigned &SpelledIndex) { 506 // We need to record this mapping before continuing. 507 unsigned MappingBegin = SpelledIndex; 508 ++SpelledIndex; 509 510 bool HitMapping = 511 tryConsumeSpelledUntil(File, EndOffset + 1, SpelledIndex).hasValue(); 512 (void)HitMapping; 513 assert(!HitMapping && "recursive macro expansion?"); 514 515 TokenBuffer::Mapping M; 516 M.BeginExpanded = BeginExpanded; 517 M.EndExpanded = EndExpanded; 518 M.BeginSpelled = MappingBegin; 519 M.EndSpelled = SpelledIndex; 520 521 File.Mappings.push_back(M); 522 } 523 524 /// Consumes spelled tokens until location \p L is reached and adds a mapping 525 /// covering the consumed tokens. The mapping will point to an empty expanded 526 /// range at position \p ExpandedIndex. 527 void fillGapUntil(TokenBuffer::MarkedFile &File, SourceLocation L, 528 unsigned ExpandedIndex) { 529 assert(L.isFileID()); 530 FileID FID; 531 unsigned Offset; 532 std::tie(FID, Offset) = SM.getDecomposedLoc(L); 533 534 unsigned &SpelledIndex = NextSpelled[FID]; 535 unsigned MappingBegin = SpelledIndex; 536 while (true) { 537 auto EndLoc = tryConsumeSpelledUntil(File, Offset, SpelledIndex); 538 if (SpelledIndex != MappingBegin) { 539 TokenBuffer::Mapping M; 540 M.BeginSpelled = MappingBegin; 541 M.EndSpelled = SpelledIndex; 542 M.BeginExpanded = M.EndExpanded = ExpandedIndex; 543 File.Mappings.push_back(M); 544 } 545 if (!EndLoc) 546 break; 547 consumeEmptyMapping(File, SM.getFileOffset(*EndLoc), ExpandedIndex, 548 SpelledIndex); 549 550 MappingBegin = SpelledIndex; 551 } 552 }; 553 554 /// Consumes spelled tokens until it reaches Offset or a mapping boundary, 555 /// i.e. a name of a macro expansion or the start '#' token of a PP directive. 556 /// (!) NextSpelled is updated in place. 557 /// 558 /// returns None if \p Offset was reached, otherwise returns the end location 559 /// of a mapping that starts at \p NextSpelled. 560 llvm::Optional<SourceLocation> 561 tryConsumeSpelledUntil(TokenBuffer::MarkedFile &File, unsigned Offset, 562 unsigned &NextSpelled) { 563 for (; NextSpelled < File.SpelledTokens.size(); ++NextSpelled) { 564 auto L = File.SpelledTokens[NextSpelled].location(); 565 if (Offset <= SM.getFileOffset(L)) 566 return llvm::None; // reached the offset we are looking for. 567 auto Mapping = CollectedExpansions.find(L.getRawEncoding()); 568 if (Mapping != CollectedExpansions.end()) 569 return Mapping->second; // found a mapping before the offset. 570 } 571 return llvm::None; // no more tokens, we "reached" the offset. 572 } 573 574 /// Adds empty mappings for unconsumed spelled tokens at the end of each file. 575 void fillGapsAtEndOfFiles() { 576 for (auto &F : Result.Files) { 577 if (F.second.SpelledTokens.empty()) 578 continue; 579 fillGapUntil(F.second, F.second.SpelledTokens.back().endLocation(), 580 F.second.EndExpanded); 581 } 582 } 583 584 TokenBuffer Result; 585 /// For each file, a position of the next spelled token we will consume. 586 llvm::DenseMap<FileID, unsigned> NextSpelled; 587 PPExpansions CollectedExpansions; 588 const SourceManager &SM; 589 const LangOptions &LangOpts; 590 }; 591 592 TokenBuffer TokenCollector::consume() && { 593 PP.setTokenWatcher(nullptr); 594 Collector->disable(); 595 return Builder(std::move(Expanded), std::move(Expansions), 596 PP.getSourceManager(), PP.getLangOpts()) 597 .build(); 598 } 599 600 std::string syntax::Token::str() const { 601 return llvm::formatv("Token({0}, length = {1})", tok::getTokenName(kind()), 602 length()); 603 } 604 605 std::string syntax::Token::dumpForTests(const SourceManager &SM) const { 606 return llvm::formatv("{0} {1}", tok::getTokenName(kind()), text(SM)); 607 } 608 609 std::string TokenBuffer::dumpForTests() const { 610 auto PrintToken = [this](const syntax::Token &T) -> std::string { 611 if (T.kind() == tok::eof) 612 return "<eof>"; 613 return T.text(*SourceMgr); 614 }; 615 616 auto DumpTokens = [this, &PrintToken](llvm::raw_ostream &OS, 617 llvm::ArrayRef<syntax::Token> Tokens) { 618 if (Tokens.empty()) { 619 OS << "<empty>"; 620 return; 621 } 622 OS << Tokens[0].text(*SourceMgr); 623 for (unsigned I = 1; I < Tokens.size(); ++I) { 624 if (Tokens[I].kind() == tok::eof) 625 continue; 626 OS << " " << PrintToken(Tokens[I]); 627 } 628 }; 629 630 std::string Dump; 631 llvm::raw_string_ostream OS(Dump); 632 633 OS << "expanded tokens:\n" 634 << " "; 635 // (!) we do not show '<eof>'. 636 DumpTokens(OS, llvm::makeArrayRef(ExpandedTokens).drop_back()); 637 OS << "\n"; 638 639 std::vector<FileID> Keys; 640 for (auto F : Files) 641 Keys.push_back(F.first); 642 llvm::sort(Keys); 643 644 for (FileID ID : Keys) { 645 const MarkedFile &File = Files.find(ID)->second; 646 auto *Entry = SourceMgr->getFileEntryForID(ID); 647 if (!Entry) 648 continue; // Skip builtin files. 649 OS << llvm::formatv("file '{0}'\n", Entry->getName()) 650 << " spelled tokens:\n" 651 << " "; 652 DumpTokens(OS, File.SpelledTokens); 653 OS << "\n"; 654 655 if (File.Mappings.empty()) { 656 OS << " no mappings.\n"; 657 continue; 658 } 659 OS << " mappings:\n"; 660 for (auto &M : File.Mappings) { 661 OS << llvm::formatv( 662 " ['{0}'_{1}, '{2}'_{3}) => ['{4}'_{5}, '{6}'_{7})\n", 663 PrintToken(File.SpelledTokens[M.BeginSpelled]), M.BeginSpelled, 664 M.EndSpelled == File.SpelledTokens.size() 665 ? "<eof>" 666 : PrintToken(File.SpelledTokens[M.EndSpelled]), 667 M.EndSpelled, PrintToken(ExpandedTokens[M.BeginExpanded]), 668 M.BeginExpanded, PrintToken(ExpandedTokens[M.EndExpanded]), 669 M.EndExpanded); 670 } 671 } 672 return OS.str(); 673 } 674