1 //===- Tokens.h - collect tokens from preprocessing --------------*- C++-*-===// 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 // Record tokens that a preprocessor emits and define operations to map between 9 // the tokens written in a file and tokens produced by the preprocessor. 10 // 11 // When running the compiler, there are two token streams we are interested in: 12 // - "spelled" tokens directly correspond to a substring written in some 13 // source file. 14 // - "expanded" tokens represent the result of preprocessing, parses consumes 15 // this token stream to produce the AST. 16 // 17 // Expanded tokens correspond directly to locations found in the AST, allowing 18 // to find subranges of the token stream covered by various AST nodes. Spelled 19 // tokens correspond directly to the source code written by the user. 20 // 21 // To allow composing these two use-cases, we also define operations that map 22 // between expanded and spelled tokens that produced them (macro calls, 23 // directives, etc). 24 // 25 //===----------------------------------------------------------------------===// 26 27 #ifndef LLVM_CLANG_TOOLING_SYNTAX_TOKENS_H 28 #define LLVM_CLANG_TOOLING_SYNTAX_TOKENS_H 29 30 #include "clang/Basic/LangOptions.h" 31 #include "clang/Basic/SourceLocation.h" 32 #include "clang/Basic/SourceManager.h" 33 #include "clang/Basic/TokenKinds.h" 34 #include "clang/Lex/Token.h" 35 #include "llvm/ADT/ArrayRef.h" 36 #include "llvm/ADT/DenseMap.h" 37 #include "llvm/ADT/Optional.h" 38 #include "llvm/ADT/StringRef.h" 39 #include "llvm/Support/Compiler.h" 40 #include "llvm/Support/raw_ostream.h" 41 #include <cstdint> 42 #include <tuple> 43 44 namespace clang { 45 class Preprocessor; 46 47 namespace syntax { 48 49 /// A half-open character range inside a particular file, the start offset is 50 /// included and the end offset is excluded from the range. 51 struct FileRange { 52 /// EXPECTS: File.isValid() && Begin <= End. 53 FileRange(FileID File, unsigned BeginOffset, unsigned EndOffset); 54 /// EXPECTS: BeginLoc.isValid() && BeginLoc.isFileID(). 55 FileRange(const SourceManager &SM, SourceLocation BeginLoc, unsigned Length); 56 /// EXPECTS: BeginLoc.isValid() && BeginLoc.isFileID(), Begin <= End and files 57 /// are the same. 58 FileRange(const SourceManager &SM, SourceLocation BeginLoc, 59 SourceLocation EndLoc); 60 61 FileID file() const { return File; } 62 /// Start is a start offset (inclusive) in the corresponding file. 63 unsigned beginOffset() const { return Begin; } 64 /// End offset (exclusive) in the corresponding file. 65 unsigned endOffset() const { return End; } 66 67 unsigned length() const { return End - Begin; } 68 69 /// Check if \p Offset is inside the range. 70 bool contains(unsigned Offset) const { 71 return Begin <= Offset && Offset < End; 72 } 73 /// Check \p Offset is inside the range or equal to its endpoint. 74 bool touches(unsigned Offset) const { 75 return Begin <= Offset && Offset <= End; 76 } 77 78 /// Gets the substring that this FileRange refers to. 79 llvm::StringRef text(const SourceManager &SM) const; 80 81 /// Convert to the clang range. The returned range is always a char range, 82 /// never a token range. 83 CharSourceRange toCharRange(const SourceManager &SM) const; 84 85 friend bool operator==(const FileRange &L, const FileRange &R) { 86 return std::tie(L.File, L.Begin, L.End) == std::tie(R.File, R.Begin, R.End); 87 } 88 friend bool operator!=(const FileRange &L, const FileRange &R) { 89 return !(L == R); 90 } 91 92 private: 93 FileID File; 94 unsigned Begin; 95 unsigned End; 96 }; 97 98 /// For debugging purposes. 99 llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, const FileRange &R); 100 101 /// A token coming directly from a file or from a macro invocation. Has just 102 /// enough information to locate the token in the source code. 103 /// Can represent both expanded and spelled tokens. 104 class Token { 105 public: 106 Token(SourceLocation Location, unsigned Length, tok::TokenKind Kind); 107 /// EXPECTS: clang::Token is not an annotation token. 108 explicit Token(const clang::Token &T); 109 110 tok::TokenKind kind() const { return Kind; } 111 /// Location of the first character of a token. 112 SourceLocation location() const { return Location; } 113 /// Location right after the last character of a token. 114 SourceLocation endLocation() const { 115 return Location.getLocWithOffset(Length); 116 } 117 unsigned length() const { return Length; } 118 119 /// Get the substring covered by the token. Note that will include all 120 /// digraphs, newline continuations, etc. E.g. tokens for 'int' and 121 /// in\ 122 /// t 123 /// both have the same kind tok::kw_int, but results of text() are different. 124 llvm::StringRef text(const SourceManager &SM) const; 125 126 /// Gets a range of this token. 127 /// EXPECTS: token comes from a file, not from a macro expansion. 128 FileRange range(const SourceManager &SM) const; 129 130 /// Given two tokens inside the same file, returns a file range that starts at 131 /// \p First and ends at \p Last. 132 /// EXPECTS: First and Last are file tokens from the same file, Last starts 133 /// after First. 134 static FileRange range(const SourceManager &SM, const syntax::Token &First, 135 const syntax::Token &Last); 136 137 std::string dumpForTests(const SourceManager &SM) const; 138 /// For debugging purposes. 139 std::string str() const; 140 141 private: 142 SourceLocation Location; 143 unsigned Length; 144 tok::TokenKind Kind; 145 }; 146 /// For debugging purposes. Equivalent to a call to Token::str(). 147 llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, const Token &T); 148 149 /// A list of tokens obtained by preprocessing a text buffer and operations to 150 /// map between the expanded and spelled tokens, i.e. TokenBuffer has 151 /// information about two token streams: 152 /// 1. Expanded tokens: tokens produced by the preprocessor after all macro 153 /// replacements, 154 /// 2. Spelled tokens: corresponding directly to the source code of a file 155 /// before any macro replacements occurred. 156 /// Here's an example to illustrate a difference between those two: 157 /// #define FOO 10 158 /// int a = FOO; 159 /// 160 /// Spelled tokens are {'#','define','FOO','10','int','a','=','FOO',';'}. 161 /// Expanded tokens are {'int','a','=','10',';','eof'}. 162 /// 163 /// Note that the expanded token stream has a tok::eof token at the end, the 164 /// spelled tokens never store a 'eof' token. 165 /// 166 /// The full list expanded tokens can be obtained with expandedTokens(). Spelled 167 /// tokens for each of the files can be obtained via spelledTokens(FileID). 168 /// 169 /// To map between the expanded and spelled tokens use findSpelledByExpanded(). 170 /// 171 /// To build a token buffer use the TokenCollector class. You can also compute 172 /// the spelled tokens of a file using the tokenize() helper. 173 /// 174 /// FIXME: allow mappings into macro arguments. 175 class TokenBuffer { 176 public: 177 TokenBuffer(const SourceManager &SourceMgr) : SourceMgr(&SourceMgr) {} 178 179 TokenBuffer(TokenBuffer &&) = default; 180 TokenBuffer(const TokenBuffer &) = delete; 181 TokenBuffer &operator=(TokenBuffer &&) = default; 182 TokenBuffer &operator=(const TokenBuffer &) = delete; 183 184 /// All tokens produced by the preprocessor after all macro replacements, 185 /// directives, etc. Source locations found in the clang AST will always 186 /// point to one of these tokens. 187 /// Tokens are in TU order (per SourceManager::isBeforeInTranslationUnit()). 188 /// FIXME: figure out how to handle token splitting, e.g. '>>' can be split 189 /// into two '>' tokens by the parser. However, TokenBuffer currently 190 /// keeps it as a single '>>' token. 191 llvm::ArrayRef<syntax::Token> expandedTokens() const { 192 return ExpandedTokens; 193 } 194 195 /// Builds a cache to make future calls to expandedToken(SourceRange) faster. 196 /// Creates an index only once. Further calls to it will be no-op. 197 void indexExpandedTokens(); 198 199 /// Returns the subrange of expandedTokens() corresponding to the closed 200 /// token range R. 201 /// Consider calling indexExpandedTokens() before for faster lookups. 202 llvm::ArrayRef<syntax::Token> expandedTokens(SourceRange R) const; 203 204 /// Returns the subrange of spelled tokens corresponding to AST node spanning 205 /// \p Expanded. This is the text that should be replaced if a refactoring 206 /// were to rewrite the node. If \p Expanded is empty, the returned value is 207 /// llvm::None. 208 /// 209 /// Will fail if the expanded tokens do not correspond to a sequence of 210 /// spelled tokens. E.g. for the following example: 211 /// 212 /// #define FIRST f1 f2 f3 213 /// #define SECOND s1 s2 s3 214 /// #define ID2(X, Y) X Y 215 /// 216 /// a FIRST b SECOND c // expanded tokens are: a f1 f2 f3 b s1 s2 s3 c 217 /// d ID2(e f g, h) i // expanded tokens are: d e f g h i 218 /// 219 /// the results would be: 220 /// expanded => spelled 221 /// ------------------------ 222 /// a => a 223 /// s1 s2 s3 => SECOND 224 /// a f1 f2 f3 => a FIRST 225 /// a f1 => can't map 226 /// s1 s2 => can't map 227 /// e f => e f 228 /// g h => can't map 229 /// 230 /// EXPECTS: \p Expanded is a subrange of expandedTokens(). 231 /// Complexity is logarithmic. 232 llvm::Optional<llvm::ArrayRef<syntax::Token>> 233 spelledForExpanded(llvm::ArrayRef<syntax::Token> Expanded) const; 234 235 /// Find the subranges of expanded tokens, corresponding to \p Spelled. 236 /// 237 /// Some spelled tokens may not be present in the expanded token stream, so 238 /// this function can return an empty vector, e.g. for tokens of macro 239 /// directives or disabled preprocessor branches. 240 /// 241 /// Some spelled tokens can be duplicated in the expanded token stream 242 /// multiple times and this function will return multiple results in those 243 /// cases. This happens when \p Spelled is inside a macro argument. 244 /// 245 /// FIXME: return correct results on macro arguments. For now, we return an 246 /// empty list. 247 /// 248 /// (!) will return empty vector on tokens from #define body: 249 /// E.g. for the following example: 250 /// 251 /// #define FIRST(A) f1 A = A f2 252 /// #define SECOND s 253 /// 254 /// a FIRST(arg) b SECOND c // expanded tokens are: a f1 arg = arg f2 b s 255 /// The results would be 256 /// spelled => expanded 257 /// ------------------------ 258 /// #define FIRST => {} 259 /// a FIRST(arg) => {a f1 arg = arg f2} 260 /// arg => {arg, arg} // arg #1 is before `=` and arg #2 is 261 /// // after `=` in the expanded tokens. 262 llvm::SmallVector<llvm::ArrayRef<syntax::Token>, 1> 263 expandedForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const; 264 265 /// An expansion produced by the preprocessor, includes macro expansions and 266 /// preprocessor directives. Preprocessor always maps a non-empty range of 267 /// spelled tokens to a (possibly empty) range of expanded tokens. Here is a 268 /// few examples of expansions: 269 /// #pragma once // Expands to an empty range. 270 /// #define FOO 1 2 3 // Expands an empty range. 271 /// FOO // Expands to "1 2 3". 272 /// FIXME(ibiryukov): implement this, currently #include expansions are empty. 273 /// #include <vector> // Expands to tokens produced by the include. 274 struct Expansion { 275 llvm::ArrayRef<syntax::Token> Spelled; 276 llvm::ArrayRef<syntax::Token> Expanded; 277 }; 278 /// If \p Spelled starts a mapping (e.g. if it's a macro name or '#' starting 279 /// a preprocessor directive) return the subrange of expanded tokens that the 280 /// macro expands to. 281 llvm::Optional<Expansion> 282 expansionStartingAt(const syntax::Token *Spelled) const; 283 /// Returns all expansions (partially) expanded from the specified tokens. 284 /// This is the expansions whose Spelled range intersects \p Spelled. 285 std::vector<Expansion> 286 expansionsOverlapping(llvm::ArrayRef<syntax::Token> Spelled) const; 287 288 /// Lexed tokens of a file before preprocessing. E.g. for the following input 289 /// #define DECL(name) int name = 10 290 /// DECL(a); 291 /// spelledTokens() returns 292 /// {"#", "define", "DECL", "(", "name", ")", "int", "name", "=", "10", 293 /// "DECL", "(", "a", ")", ";"} 294 llvm::ArrayRef<syntax::Token> spelledTokens(FileID FID) const; 295 296 /// Returns the spelled Token starting at Loc, if there are no such tokens 297 /// returns nullptr. 298 const syntax::Token *spelledTokenAt(SourceLocation Loc) const; 299 300 /// Get all tokens that expand a macro in \p FID. For the following input 301 /// #define FOO B 302 /// #define FOO2(X) int X 303 /// FOO2(XY) 304 /// int B; 305 /// FOO; 306 /// macroExpansions() returns {"FOO2", "FOO"} (from line 3 and 5 307 /// respecitvely). 308 std::vector<const syntax::Token *> macroExpansions(FileID FID) const; 309 310 const SourceManager &sourceManager() const { return *SourceMgr; } 311 312 std::string dumpForTests() const; 313 314 private: 315 /// Describes a mapping between a continuous subrange of spelled tokens and 316 /// expanded tokens. Represents macro expansions, preprocessor directives, 317 /// conditionally disabled pp regions, etc. 318 /// #define FOO 1+2 319 /// #define BAR(a) a + 1 320 /// FOO // invocation #1, tokens = {'1','+','2'}, macroTokens = {'FOO'}. 321 /// BAR(1) // invocation #2, tokens = {'a', '+', '1'}, 322 /// macroTokens = {'BAR', '(', '1', ')'}. 323 struct Mapping { 324 // Positions in the corresponding spelled token stream. The corresponding 325 // range is never empty. 326 unsigned BeginSpelled = 0; 327 unsigned EndSpelled = 0; 328 // Positions in the expanded token stream. The corresponding range can be 329 // empty. 330 unsigned BeginExpanded = 0; 331 unsigned EndExpanded = 0; 332 333 /// For debugging purposes. 334 std::string str() const; 335 }; 336 /// Spelled tokens of the file with information about the subranges. 337 struct MarkedFile { 338 /// Lexed, but not preprocessed, tokens of the file. These map directly to 339 /// text in the corresponding files and include tokens of all preprocessor 340 /// directives. 341 /// FIXME: spelled tokens don't change across FileID that map to the same 342 /// FileEntry. We could consider deduplicating them to save memory. 343 std::vector<syntax::Token> SpelledTokens; 344 /// A sorted list to convert between the spelled and expanded token streams. 345 std::vector<Mapping> Mappings; 346 /// The first expanded token produced for this FileID. 347 unsigned BeginExpanded = 0; 348 unsigned EndExpanded = 0; 349 }; 350 351 friend class TokenCollector; 352 353 /// Maps a single expanded token to its spelled counterpart or a mapping that 354 /// produced it. 355 std::pair<const syntax::Token *, const Mapping *> 356 spelledForExpandedToken(const syntax::Token *Expanded) const; 357 358 /// Returns a mapping starting before \p Spelled token, or nullptr if no 359 /// such mapping exists. 360 static const Mapping * 361 mappingStartingBeforeSpelled(const MarkedFile &F, 362 const syntax::Token *Spelled); 363 364 /// Convert a private Mapping to a public Expansion. 365 Expansion makeExpansion(const MarkedFile &, const Mapping &) const; 366 /// Returns the file that the Spelled tokens are taken from. 367 /// Asserts that they are non-empty, from a tracked file, and in-bounds. 368 const MarkedFile &fileForSpelled(llvm::ArrayRef<syntax::Token> Spelled) const; 369 370 /// Token stream produced after preprocessing, conceputally this captures the 371 /// same stream as 'clang -E' (excluding the preprocessor directives like 372 /// #file, etc.). 373 std::vector<syntax::Token> ExpandedTokens; 374 // Index of ExpandedTokens for faster lookups by SourceLocation. 375 llvm::DenseMap<SourceLocation, unsigned> ExpandedTokIndex; 376 llvm::DenseMap<FileID, MarkedFile> Files; 377 // The value is never null, pointer instead of reference to avoid disabling 378 // implicit assignment operator. 379 const SourceManager *SourceMgr; 380 }; 381 382 /// The spelled tokens that overlap or touch a spelling location Loc. 383 /// This always returns 0-2 tokens. 384 llvm::ArrayRef<syntax::Token> 385 spelledTokensTouching(SourceLocation Loc, const syntax::TokenBuffer &Tokens); 386 llvm::ArrayRef<syntax::Token> 387 spelledTokensTouching(SourceLocation Loc, llvm::ArrayRef<syntax::Token> Tokens); 388 389 /// The identifier token that overlaps or touches a spelling location Loc. 390 /// If there is none, returns nullptr. 391 const syntax::Token * 392 spelledIdentifierTouching(SourceLocation Loc, 393 llvm::ArrayRef<syntax::Token> Tokens); 394 const syntax::Token * 395 spelledIdentifierTouching(SourceLocation Loc, 396 const syntax::TokenBuffer &Tokens); 397 398 /// Lex the text buffer, corresponding to \p FID, in raw mode and record the 399 /// resulting spelled tokens. Does minimal post-processing on raw identifiers, 400 /// setting the appropriate token kind (instead of the raw_identifier reported 401 /// by lexer in raw mode). This is a very low-level function, most users should 402 /// prefer to use TokenCollector. Lexing in raw mode produces wildly different 403 /// results from what one might expect when running a C++ frontend, e.g. 404 /// preprocessor does not run at all. 405 /// The result will *not* have a 'eof' token at the end. 406 std::vector<syntax::Token> tokenize(FileID FID, const SourceManager &SM, 407 const LangOptions &LO); 408 /// Similar to one above, instead of whole file tokenizes a part of it. Note 409 /// that, the first token might be incomplete if FR.startOffset is not at the 410 /// beginning of a token, and the last token returned will start before the 411 /// FR.endOffset but might end after it. 412 std::vector<syntax::Token> 413 tokenize(const FileRange &FR, const SourceManager &SM, const LangOptions &LO); 414 415 /// Collects tokens for the main file while running the frontend action. An 416 /// instance of this object should be created on 417 /// FrontendAction::BeginSourceFile() and the results should be consumed after 418 /// FrontendAction::Execute() finishes. 419 class TokenCollector { 420 public: 421 /// Adds the hooks to collect the tokens. Should be called before the 422 /// preprocessing starts, i.e. as a part of BeginSourceFile() or 423 /// CreateASTConsumer(). 424 TokenCollector(Preprocessor &P); 425 426 /// Finalizes token collection. Should be called after preprocessing is 427 /// finished, i.e. after running Execute(). 428 LLVM_NODISCARD TokenBuffer consume() &&; 429 430 private: 431 /// Maps from a start to an end spelling location of transformations 432 /// performed by the preprocessor. These include: 433 /// 1. range from '#' to the last token in the line for PP directives, 434 /// 2. macro name and arguments for macro expansions. 435 /// Note that we record only top-level macro expansions, intermediate 436 /// expansions (e.g. inside macro arguments) are ignored. 437 /// 438 /// Used to find correct boundaries of macro calls and directives when 439 /// building mappings from spelled to expanded tokens. 440 /// 441 /// Logically, at each point of the preprocessor execution there is a stack of 442 /// macro expansions being processed and we could use it to recover the 443 /// location information we need. However, the public preprocessor API only 444 /// exposes the points when macro expansions start (when we push a macro onto 445 /// the stack) and not when they end (when we pop a macro from the stack). 446 /// To workaround this limitation, we rely on source location information 447 /// stored in this map. 448 using PPExpansions = llvm::DenseMap<SourceLocation, SourceLocation>; 449 class Builder; 450 class CollectPPExpansions; 451 452 std::vector<syntax::Token> Expanded; 453 // FIXME: we only store macro expansions, also add directives(#pragma, etc.) 454 PPExpansions Expansions; 455 Preprocessor &PP; 456 CollectPPExpansions *Collector; 457 }; 458 459 } // namespace syntax 460 } // namespace clang 461 462 #endif 463