1 //===--- Format.cpp - Format C++ code -------------------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 /// 10 /// \file 11 /// \brief This file implements functions declared in Format.h. This will be 12 /// split into separate files as we go. 13 /// 14 //===----------------------------------------------------------------------===// 15 16 #define DEBUG_TYPE "format-formatter" 17 18 #include "ContinuationIndenter.h" 19 #include "TokenAnnotator.h" 20 #include "UnwrappedLineParser.h" 21 #include "WhitespaceManager.h" 22 #include "clang/Basic/Diagnostic.h" 23 #include "clang/Basic/SourceManager.h" 24 #include "clang/Format/Format.h" 25 #include "clang/Lex/Lexer.h" 26 #include "llvm/ADT/STLExtras.h" 27 #include "llvm/Support/Allocator.h" 28 #include "llvm/Support/Debug.h" 29 #include "llvm/Support/YAMLTraits.h" 30 #include "llvm/Support/Path.h" 31 #include <queue> 32 #include <string> 33 34 namespace llvm { 35 namespace yaml { 36 template <> 37 struct ScalarEnumerationTraits<clang::format::FormatStyle::LanguageStandard> { 38 static void enumeration(IO &IO, 39 clang::format::FormatStyle::LanguageStandard &Value) { 40 IO.enumCase(Value, "Cpp03", clang::format::FormatStyle::LS_Cpp03); 41 IO.enumCase(Value, "C++03", clang::format::FormatStyle::LS_Cpp03); 42 IO.enumCase(Value, "Cpp11", clang::format::FormatStyle::LS_Cpp11); 43 IO.enumCase(Value, "C++11", clang::format::FormatStyle::LS_Cpp11); 44 IO.enumCase(Value, "Auto", clang::format::FormatStyle::LS_Auto); 45 } 46 }; 47 48 template <> 49 struct ScalarEnumerationTraits<clang::format::FormatStyle::UseTabStyle> { 50 static void enumeration(IO &IO, 51 clang::format::FormatStyle::UseTabStyle &Value) { 52 IO.enumCase(Value, "Never", clang::format::FormatStyle::UT_Never); 53 IO.enumCase(Value, "false", clang::format::FormatStyle::UT_Never); 54 IO.enumCase(Value, "Always", clang::format::FormatStyle::UT_Always); 55 IO.enumCase(Value, "true", clang::format::FormatStyle::UT_Always); 56 IO.enumCase(Value, "ForIndentation", 57 clang::format::FormatStyle::UT_ForIndentation); 58 } 59 }; 60 61 template <> 62 struct ScalarEnumerationTraits<clang::format::FormatStyle::BraceBreakingStyle> { 63 static void 64 enumeration(IO &IO, clang::format::FormatStyle::BraceBreakingStyle &Value) { 65 IO.enumCase(Value, "Attach", clang::format::FormatStyle::BS_Attach); 66 IO.enumCase(Value, "Linux", clang::format::FormatStyle::BS_Linux); 67 IO.enumCase(Value, "Stroustrup", clang::format::FormatStyle::BS_Stroustrup); 68 IO.enumCase(Value, "Allman", clang::format::FormatStyle::BS_Allman); 69 } 70 }; 71 72 template <> 73 struct ScalarEnumerationTraits< 74 clang::format::FormatStyle::NamespaceIndentationKind> { 75 static void 76 enumeration(IO &IO, 77 clang::format::FormatStyle::NamespaceIndentationKind &Value) { 78 IO.enumCase(Value, "None", clang::format::FormatStyle::NI_None); 79 IO.enumCase(Value, "Inner", clang::format::FormatStyle::NI_Inner); 80 IO.enumCase(Value, "All", clang::format::FormatStyle::NI_All); 81 } 82 }; 83 84 template <> struct MappingTraits<clang::format::FormatStyle> { 85 static void mapping(llvm::yaml::IO &IO, clang::format::FormatStyle &Style) { 86 if (IO.outputting()) { 87 StringRef StylesArray[] = { "LLVM", "Google", "Chromium", 88 "Mozilla", "WebKit" }; 89 ArrayRef<StringRef> Styles(StylesArray); 90 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 91 StringRef StyleName(Styles[i]); 92 clang::format::FormatStyle PredefinedStyle; 93 if (clang::format::getPredefinedStyle(StyleName, &PredefinedStyle) && 94 Style == PredefinedStyle) { 95 IO.mapOptional("# BasedOnStyle", StyleName); 96 break; 97 } 98 } 99 } else { 100 StringRef BasedOnStyle; 101 IO.mapOptional("BasedOnStyle", BasedOnStyle); 102 if (!BasedOnStyle.empty()) 103 if (!clang::format::getPredefinedStyle(BasedOnStyle, &Style)) { 104 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 105 return; 106 } 107 } 108 109 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 110 IO.mapOptional("ConstructorInitializerIndentWidth", 111 Style.ConstructorInitializerIndentWidth); 112 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 113 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 114 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 115 Style.AllowAllParametersOfDeclarationOnNextLine); 116 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 117 Style.AllowShortIfStatementsOnASingleLine); 118 IO.mapOptional("AllowShortLoopsOnASingleLine", 119 Style.AllowShortLoopsOnASingleLine); 120 IO.mapOptional("AlwaysBreakTemplateDeclarations", 121 Style.AlwaysBreakTemplateDeclarations); 122 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 123 Style.AlwaysBreakBeforeMultilineStrings); 124 IO.mapOptional("BreakBeforeBinaryOperators", 125 Style.BreakBeforeBinaryOperators); 126 IO.mapOptional("BreakConstructorInitializersBeforeComma", 127 Style.BreakConstructorInitializersBeforeComma); 128 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 129 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 130 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 131 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 132 IO.mapOptional("DerivePointerBinding", Style.DerivePointerBinding); 133 IO.mapOptional("ExperimentalAutoDetectBinPacking", 134 Style.ExperimentalAutoDetectBinPacking); 135 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 136 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 137 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 138 IO.mapOptional("ObjCSpaceBeforeProtocolList", 139 Style.ObjCSpaceBeforeProtocolList); 140 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 141 Style.PenaltyBreakBeforeFirstCallParameter); 142 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 143 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 144 IO.mapOptional("PenaltyBreakFirstLessLess", 145 Style.PenaltyBreakFirstLessLess); 146 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 147 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 148 Style.PenaltyReturnTypeOnItsOwnLine); 149 IO.mapOptional("PointerBindsToType", Style.PointerBindsToType); 150 IO.mapOptional("SpacesBeforeTrailingComments", 151 Style.SpacesBeforeTrailingComments); 152 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 153 IO.mapOptional("Standard", Style.Standard); 154 IO.mapOptional("IndentWidth", Style.IndentWidth); 155 IO.mapOptional("TabWidth", Style.TabWidth); 156 IO.mapOptional("UseTab", Style.UseTab); 157 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 158 IO.mapOptional("IndentFunctionDeclarationAfterType", 159 Style.IndentFunctionDeclarationAfterType); 160 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 161 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 162 IO.mapOptional("SpacesInCStyleCastParentheses", 163 Style.SpacesInCStyleCastParentheses); 164 IO.mapOptional("SpaceAfterControlStatementKeyword", 165 Style.SpaceAfterControlStatementKeyword); 166 IO.mapOptional("SpaceBeforeAssignmentOperators", 167 Style.SpaceBeforeAssignmentOperators); 168 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 169 } 170 }; 171 } 172 } 173 174 namespace clang { 175 namespace format { 176 177 void setDefaultPenalties(FormatStyle &Style) { 178 Style.PenaltyBreakComment = 60; 179 Style.PenaltyBreakFirstLessLess = 120; 180 Style.PenaltyBreakString = 1000; 181 Style.PenaltyExcessCharacter = 1000000; 182 } 183 184 FormatStyle getLLVMStyle() { 185 FormatStyle LLVMStyle; 186 LLVMStyle.AccessModifierOffset = -2; 187 LLVMStyle.AlignEscapedNewlinesLeft = false; 188 LLVMStyle.AlignTrailingComments = true; 189 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 190 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 191 LLVMStyle.AllowShortLoopsOnASingleLine = false; 192 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 193 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 194 LLVMStyle.BinPackParameters = true; 195 LLVMStyle.BreakBeforeBinaryOperators = false; 196 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 197 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 198 LLVMStyle.ColumnLimit = 80; 199 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 200 LLVMStyle.ConstructorInitializerIndentWidth = 4; 201 LLVMStyle.Cpp11BracedListStyle = false; 202 LLVMStyle.DerivePointerBinding = false; 203 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 204 LLVMStyle.IndentCaseLabels = false; 205 LLVMStyle.IndentFunctionDeclarationAfterType = false; 206 LLVMStyle.IndentWidth = 2; 207 LLVMStyle.TabWidth = 8; 208 LLVMStyle.MaxEmptyLinesToKeep = 1; 209 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 210 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 211 LLVMStyle.PointerBindsToType = false; 212 LLVMStyle.SpacesBeforeTrailingComments = 1; 213 LLVMStyle.Standard = FormatStyle::LS_Cpp03; 214 LLVMStyle.UseTab = FormatStyle::UT_Never; 215 LLVMStyle.SpacesInParentheses = false; 216 LLVMStyle.SpaceInEmptyParentheses = false; 217 LLVMStyle.SpacesInCStyleCastParentheses = false; 218 LLVMStyle.SpaceAfterControlStatementKeyword = true; 219 LLVMStyle.SpaceBeforeAssignmentOperators = true; 220 LLVMStyle.ContinuationIndentWidth = 4; 221 222 setDefaultPenalties(LLVMStyle); 223 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 224 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 225 226 return LLVMStyle; 227 } 228 229 FormatStyle getGoogleStyle() { 230 FormatStyle GoogleStyle; 231 GoogleStyle.AccessModifierOffset = -1; 232 GoogleStyle.AlignEscapedNewlinesLeft = true; 233 GoogleStyle.AlignTrailingComments = true; 234 GoogleStyle.AllowAllParametersOfDeclarationOnNextLine = true; 235 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 236 GoogleStyle.AllowShortLoopsOnASingleLine = true; 237 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 238 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 239 GoogleStyle.BinPackParameters = true; 240 GoogleStyle.BreakBeforeBinaryOperators = false; 241 GoogleStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 242 GoogleStyle.BreakConstructorInitializersBeforeComma = false; 243 GoogleStyle.ColumnLimit = 80; 244 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 245 GoogleStyle.ConstructorInitializerIndentWidth = 4; 246 GoogleStyle.Cpp11BracedListStyle = true; 247 GoogleStyle.DerivePointerBinding = true; 248 GoogleStyle.ExperimentalAutoDetectBinPacking = false; 249 GoogleStyle.IndentCaseLabels = true; 250 GoogleStyle.IndentFunctionDeclarationAfterType = true; 251 GoogleStyle.IndentWidth = 2; 252 GoogleStyle.TabWidth = 8; 253 GoogleStyle.MaxEmptyLinesToKeep = 1; 254 GoogleStyle.NamespaceIndentation = FormatStyle::NI_None; 255 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 256 GoogleStyle.PointerBindsToType = true; 257 GoogleStyle.SpacesBeforeTrailingComments = 2; 258 GoogleStyle.Standard = FormatStyle::LS_Auto; 259 GoogleStyle.UseTab = FormatStyle::UT_Never; 260 GoogleStyle.SpacesInParentheses = false; 261 GoogleStyle.SpaceInEmptyParentheses = false; 262 GoogleStyle.SpacesInCStyleCastParentheses = false; 263 GoogleStyle.SpaceAfterControlStatementKeyword = true; 264 GoogleStyle.SpaceBeforeAssignmentOperators = true; 265 GoogleStyle.ContinuationIndentWidth = 4; 266 267 setDefaultPenalties(GoogleStyle); 268 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 269 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 270 271 return GoogleStyle; 272 } 273 274 FormatStyle getChromiumStyle() { 275 FormatStyle ChromiumStyle = getGoogleStyle(); 276 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 277 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 278 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 279 ChromiumStyle.BinPackParameters = false; 280 ChromiumStyle.DerivePointerBinding = false; 281 ChromiumStyle.Standard = FormatStyle::LS_Cpp03; 282 return ChromiumStyle; 283 } 284 285 FormatStyle getMozillaStyle() { 286 FormatStyle MozillaStyle = getLLVMStyle(); 287 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 288 MozillaStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 289 MozillaStyle.DerivePointerBinding = true; 290 MozillaStyle.IndentCaseLabels = true; 291 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 292 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 293 MozillaStyle.PointerBindsToType = true; 294 return MozillaStyle; 295 } 296 297 FormatStyle getWebKitStyle() { 298 FormatStyle Style = getLLVMStyle(); 299 Style.AccessModifierOffset = -4; 300 Style.AlignTrailingComments = false; 301 Style.BreakBeforeBinaryOperators = true; 302 Style.BreakBeforeBraces = FormatStyle::BS_Stroustrup; 303 Style.BreakConstructorInitializersBeforeComma = true; 304 Style.ColumnLimit = 0; 305 Style.IndentWidth = 4; 306 Style.NamespaceIndentation = FormatStyle::NI_Inner; 307 Style.PointerBindsToType = true; 308 return Style; 309 } 310 311 bool getPredefinedStyle(StringRef Name, FormatStyle *Style) { 312 if (Name.equals_lower("llvm")) 313 *Style = getLLVMStyle(); 314 else if (Name.equals_lower("chromium")) 315 *Style = getChromiumStyle(); 316 else if (Name.equals_lower("mozilla")) 317 *Style = getMozillaStyle(); 318 else if (Name.equals_lower("google")) 319 *Style = getGoogleStyle(); 320 else if (Name.equals_lower("webkit")) 321 *Style = getWebKitStyle(); 322 else 323 return false; 324 325 return true; 326 } 327 328 llvm::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 329 if (Text.trim().empty()) 330 return llvm::make_error_code(llvm::errc::invalid_argument); 331 llvm::yaml::Input Input(Text); 332 Input >> *Style; 333 return Input.error(); 334 } 335 336 std::string configurationAsText(const FormatStyle &Style) { 337 std::string Text; 338 llvm::raw_string_ostream Stream(Text); 339 llvm::yaml::Output Output(Stream); 340 // We use the same mapping method for input and output, so we need a non-const 341 // reference here. 342 FormatStyle NonConstStyle = Style; 343 Output << NonConstStyle; 344 return Stream.str(); 345 } 346 347 namespace { 348 349 class NoColumnLimitFormatter { 350 public: 351 NoColumnLimitFormatter(ContinuationIndenter *Indenter) : Indenter(Indenter) {} 352 353 /// \brief Formats the line starting at \p State, simply keeping all of the 354 /// input's line breaking decisions. 355 void format(unsigned FirstIndent, const AnnotatedLine *Line) { 356 LineState State = 357 Indenter->getInitialState(FirstIndent, Line, /*DryRun=*/false); 358 while (State.NextToken != NULL) { 359 bool Newline = 360 Indenter->mustBreak(State) || 361 (Indenter->canBreak(State) && State.NextToken->NewlinesBefore > 0); 362 Indenter->addTokenToState(State, Newline, /*DryRun=*/false); 363 } 364 } 365 366 private: 367 ContinuationIndenter *Indenter; 368 }; 369 370 class UnwrappedLineFormatter { 371 public: 372 UnwrappedLineFormatter(ContinuationIndenter *Indenter, 373 WhitespaceManager *Whitespaces, 374 const FormatStyle &Style, const AnnotatedLine &Line) 375 : Indenter(Indenter), Whitespaces(Whitespaces), Style(Style), Line(Line), 376 Count(0) {} 377 378 /// \brief Formats an \c UnwrappedLine and returns the penalty. 379 /// 380 /// If \p DryRun is \c false, directly applies the changes. 381 unsigned format(unsigned FirstIndent, bool DryRun = false) { 382 LineState State = Indenter->getInitialState(FirstIndent, &Line, DryRun); 383 384 // If the ObjC method declaration does not fit on a line, we should format 385 // it with one arg per line. 386 if (Line.Type == LT_ObjCMethodDecl) 387 State.Stack.back().BreakBeforeParameter = true; 388 389 // Find best solution in solution space. 390 return analyzeSolutionSpace(State, DryRun); 391 } 392 393 private: 394 /// \brief An edge in the solution space from \c Previous->State to \c State, 395 /// inserting a newline dependent on the \c NewLine. 396 struct StateNode { 397 StateNode(const LineState &State, bool NewLine, StateNode *Previous) 398 : State(State), NewLine(NewLine), Previous(Previous) {} 399 LineState State; 400 bool NewLine; 401 StateNode *Previous; 402 }; 403 404 /// \brief A pair of <penalty, count> that is used to prioritize the BFS on. 405 /// 406 /// In case of equal penalties, we want to prefer states that were inserted 407 /// first. During state generation we make sure that we insert states first 408 /// that break the line as late as possible. 409 typedef std::pair<unsigned, unsigned> OrderedPenalty; 410 411 /// \brief An item in the prioritized BFS search queue. The \c StateNode's 412 /// \c State has the given \c OrderedPenalty. 413 typedef std::pair<OrderedPenalty, StateNode *> QueueItem; 414 415 /// \brief The BFS queue type. 416 typedef std::priority_queue<QueueItem, std::vector<QueueItem>, 417 std::greater<QueueItem> > QueueType; 418 419 /// \brief Analyze the entire solution space starting from \p InitialState. 420 /// 421 /// This implements a variant of Dijkstra's algorithm on the graph that spans 422 /// the solution space (\c LineStates are the nodes). The algorithm tries to 423 /// find the shortest path (the one with lowest penalty) from \p InitialState 424 /// to a state where all tokens are placed. Returns the penalty. 425 /// 426 /// If \p DryRun is \c false, directly applies the changes. 427 unsigned analyzeSolutionSpace(LineState &InitialState, bool DryRun = false) { 428 std::set<LineState> Seen; 429 430 // Insert start element into queue. 431 StateNode *Node = 432 new (Allocator.Allocate()) StateNode(InitialState, false, NULL); 433 Queue.push(QueueItem(OrderedPenalty(0, Count), Node)); 434 ++Count; 435 436 unsigned Penalty = 0; 437 438 // While not empty, take first element and follow edges. 439 while (!Queue.empty()) { 440 Penalty = Queue.top().first.first; 441 StateNode *Node = Queue.top().second; 442 if (Node->State.NextToken == NULL) { 443 DEBUG(llvm::dbgs() << "\n---\nPenalty for line: " << Penalty << "\n"); 444 break; 445 } 446 Queue.pop(); 447 448 // Cut off the analysis of certain solutions if the analysis gets too 449 // complex. See description of IgnoreStackForComparison. 450 if (Count > 10000) 451 Node->State.IgnoreStackForComparison = true; 452 453 if (!Seen.insert(Node->State).second) 454 // State already examined with lower penalty. 455 continue; 456 457 FormatDecision LastFormat = Node->State.NextToken->Decision; 458 if (LastFormat == FD_Unformatted || LastFormat == FD_Continue) 459 addNextStateToQueue(Penalty, Node, /*NewLine=*/false); 460 if (LastFormat == FD_Unformatted || LastFormat == FD_Break) 461 addNextStateToQueue(Penalty, Node, /*NewLine=*/true); 462 } 463 464 if (Queue.empty()) { 465 // We were unable to find a solution, do nothing. 466 // FIXME: Add diagnostic? 467 DEBUG(llvm::dbgs() << "Could not find a solution.\n"); 468 return 0; 469 } 470 471 // Reconstruct the solution. 472 if (!DryRun) 473 reconstructPath(InitialState, Queue.top().second); 474 475 DEBUG(llvm::dbgs() << "Total number of analyzed states: " << Count << "\n"); 476 DEBUG(llvm::dbgs() << "---\n"); 477 478 return Penalty; 479 } 480 481 void reconstructPath(LineState &State, StateNode *Current) { 482 std::deque<StateNode *> Path; 483 // We do not need a break before the initial token. 484 while (Current->Previous) { 485 Path.push_front(Current); 486 Current = Current->Previous; 487 } 488 for (std::deque<StateNode *>::iterator I = Path.begin(), E = Path.end(); 489 I != E; ++I) { 490 unsigned Penalty = 0; 491 formatChildren(State, (*I)->NewLine, /*DryRun=*/false, Penalty); 492 Penalty += Indenter->addTokenToState(State, (*I)->NewLine, false); 493 494 DEBUG({ 495 if ((*I)->NewLine) { 496 llvm::dbgs() << "Penalty for placing " 497 << (*I)->Previous->State.NextToken->Tok.getName() << ": " 498 << Penalty << "\n"; 499 } 500 }); 501 } 502 } 503 504 /// \brief Add the following state to the analysis queue \c Queue. 505 /// 506 /// Assume the current state is \p PreviousNode and has been reached with a 507 /// penalty of \p Penalty. Insert a line break if \p NewLine is \c true. 508 void addNextStateToQueue(unsigned Penalty, StateNode *PreviousNode, 509 bool NewLine) { 510 if (NewLine && !Indenter->canBreak(PreviousNode->State)) 511 return; 512 if (!NewLine && Indenter->mustBreak(PreviousNode->State)) 513 return; 514 515 StateNode *Node = new (Allocator.Allocate()) 516 StateNode(PreviousNode->State, NewLine, PreviousNode); 517 if (!formatChildren(Node->State, NewLine, /*DryRun=*/true, Penalty)) 518 return; 519 520 Penalty += Indenter->addTokenToState(Node->State, NewLine, true); 521 522 Queue.push(QueueItem(OrderedPenalty(Penalty, Count), Node)); 523 ++Count; 524 } 525 526 /// \brief If the \p State's next token is an r_brace closing a nested block, 527 /// format the nested block before it. 528 /// 529 /// Returns \c true if all children could be placed successfully and adapts 530 /// \p Penalty as well as \p State. If \p DryRun is false, also directly 531 /// creates changes using \c Whitespaces. 532 /// 533 /// The crucial idea here is that children always get formatted upon 534 /// encountering the closing brace right after the nested block. Now, if we 535 /// are currently trying to keep the "}" on the same line (i.e. \p NewLine is 536 /// \c false), the entire block has to be kept on the same line (which is only 537 /// possible if it fits on the line, only contains a single statement, etc. 538 /// 539 /// If \p NewLine is true, we format the nested block on separate lines, i.e. 540 /// break after the "{", format all lines with correct indentation and the put 541 /// the closing "}" on yet another new line. 542 /// 543 /// This enables us to keep the simple structure of the 544 /// \c UnwrappedLineFormatter, where we only have two options for each token: 545 /// break or don't break. 546 bool formatChildren(LineState &State, bool NewLine, bool DryRun, 547 unsigned &Penalty) { 548 const FormatToken &Previous = *State.NextToken->Previous; 549 const FormatToken *LBrace = State.NextToken->getPreviousNonComment(); 550 if (!LBrace || LBrace->isNot(tok::l_brace) || 551 LBrace->BlockKind != BK_Block || Previous.Children.size() == 0) 552 // The previous token does not open a block. Nothing to do. We don't 553 // assert so that we can simply call this function for all tokens. 554 return true; 555 556 if (NewLine) { 557 unsigned ParentIndent = State.Stack.back().Indent; 558 for (SmallVector<AnnotatedLine *, 1>::const_iterator 559 I = Previous.Children.begin(), 560 E = Previous.Children.end(); 561 I != E; ++I) { 562 unsigned Indent = 563 ParentIndent + ((*I)->Level - Line.Level - 1) * Style.IndentWidth; 564 if (!DryRun) { 565 unsigned Newlines = std::min((*I)->First->NewlinesBefore, 566 Style.MaxEmptyLinesToKeep + 1); 567 Newlines = std::max(1u, Newlines); 568 Whitespaces->replaceWhitespace( 569 *(*I)->First, Newlines, (*I)->Level, /*Spaces=*/Indent, 570 /*StartOfTokenColumn=*/Indent, Line.InPPDirective); 571 } 572 UnwrappedLineFormatter Formatter(Indenter, Whitespaces, Style, **I); 573 Penalty += Formatter.format(Indent, DryRun); 574 } 575 return true; 576 } 577 578 if (Previous.Children.size() > 1) 579 return false; // Cannot merge multiple statements into a single line. 580 581 // We can't put the closing "}" on a line with a trailing comment. 582 if (Previous.Children[0]->Last->isTrailingComment()) 583 return false; 584 585 if (!DryRun) { 586 Whitespaces->replaceWhitespace( 587 *Previous.Children[0]->First, 588 /*Newlines=*/0, /*IndentLevel=*/0, /*Spaces=*/1, 589 /*StartOfTokenColumn=*/State.Column, State.Line->InPPDirective); 590 UnwrappedLineFormatter Formatter(Indenter, Whitespaces, Style, 591 *Previous.Children[0]); 592 Penalty += Formatter.format(State.Column + 1, DryRun); 593 } 594 595 State.Column += 1 + Previous.Children[0]->Last->TotalLength; 596 return true; 597 } 598 599 ContinuationIndenter *Indenter; 600 WhitespaceManager *Whitespaces; 601 FormatStyle Style; 602 const AnnotatedLine &Line; 603 604 llvm::SpecificBumpPtrAllocator<StateNode> Allocator; 605 QueueType Queue; 606 // Increasing count of \c StateNode items we have created. This is used 607 // to create a deterministic order independent of the container. 608 unsigned Count; 609 }; 610 611 class FormatTokenLexer { 612 public: 613 FormatTokenLexer(Lexer &Lex, SourceManager &SourceMgr, FormatStyle &Style, 614 encoding::Encoding Encoding) 615 : FormatTok(NULL), GreaterStashed(false), Column(0), 616 TrailingWhitespace(0), Lex(Lex), SourceMgr(SourceMgr), Style(Style), 617 IdentTable(getFormattingLangOpts()), Encoding(Encoding) { 618 Lex.SetKeepWhitespaceMode(true); 619 } 620 621 ArrayRef<FormatToken *> lex() { 622 assert(Tokens.empty()); 623 do { 624 Tokens.push_back(getNextToken()); 625 maybeJoinPreviousTokens(); 626 } while (Tokens.back()->Tok.isNot(tok::eof)); 627 return Tokens; 628 } 629 630 IdentifierTable &getIdentTable() { return IdentTable; } 631 632 private: 633 void maybeJoinPreviousTokens() { 634 if (Tokens.size() < 4) 635 return; 636 FormatToken *Last = Tokens.back(); 637 if (!Last->is(tok::r_paren)) 638 return; 639 640 FormatToken *String = Tokens[Tokens.size() - 2]; 641 if (!String->is(tok::string_literal) || String->IsMultiline) 642 return; 643 644 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) 645 return; 646 647 FormatToken *Macro = Tokens[Tokens.size() - 4]; 648 if (Macro->TokenText != "_T") 649 return; 650 651 const char *Start = Macro->TokenText.data(); 652 const char *End = Last->TokenText.data() + Last->TokenText.size(); 653 String->TokenText = StringRef(Start, End - Start); 654 String->IsFirst = Macro->IsFirst; 655 String->LastNewlineOffset = Macro->LastNewlineOffset; 656 String->WhitespaceRange = Macro->WhitespaceRange; 657 String->OriginalColumn = Macro->OriginalColumn; 658 String->ColumnWidth = encoding::columnWidthWithTabs( 659 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); 660 661 Tokens.pop_back(); 662 Tokens.pop_back(); 663 Tokens.pop_back(); 664 Tokens.back() = String; 665 } 666 667 FormatToken *getNextToken() { 668 if (GreaterStashed) { 669 // Create a synthesized second '>' token. 670 // FIXME: Increment Column and set OriginalColumn. 671 Token Greater = FormatTok->Tok; 672 FormatTok = new (Allocator.Allocate()) FormatToken; 673 FormatTok->Tok = Greater; 674 SourceLocation GreaterLocation = 675 FormatTok->Tok.getLocation().getLocWithOffset(1); 676 FormatTok->WhitespaceRange = 677 SourceRange(GreaterLocation, GreaterLocation); 678 FormatTok->TokenText = ">"; 679 FormatTok->ColumnWidth = 1; 680 GreaterStashed = false; 681 return FormatTok; 682 } 683 684 FormatTok = new (Allocator.Allocate()) FormatToken; 685 readRawToken(*FormatTok); 686 SourceLocation WhitespaceStart = 687 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); 688 if (SourceMgr.getFileOffset(WhitespaceStart) == 0) 689 FormatTok->IsFirst = true; 690 691 // Consume and record whitespace until we find a significant token. 692 unsigned WhitespaceLength = TrailingWhitespace; 693 while (FormatTok->Tok.is(tok::unknown)) { 694 for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) { 695 switch (FormatTok->TokenText[i]) { 696 case '\n': 697 ++FormatTok->NewlinesBefore; 698 // FIXME: This is technically incorrect, as it could also 699 // be a literal backslash at the end of the line. 700 if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' && 701 (FormatTok->TokenText[i - 1] != '\r' || i == 1 || 702 FormatTok->TokenText[i - 2] != '\\'))) 703 FormatTok->HasUnescapedNewline = true; 704 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 705 Column = 0; 706 break; 707 case '\r': 708 case '\f': 709 case '\v': 710 Column = 0; 711 break; 712 case ' ': 713 ++Column; 714 break; 715 case '\t': 716 Column += Style.TabWidth - Column % Style.TabWidth; 717 break; 718 case '\\': 719 ++Column; 720 if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' && 721 FormatTok->TokenText[i + 1] != '\n')) 722 FormatTok->Type = TT_ImplicitStringLiteral; 723 break; 724 default: 725 FormatTok->Type = TT_ImplicitStringLiteral; 726 ++Column; 727 break; 728 } 729 } 730 731 if (FormatTok->Type == TT_ImplicitStringLiteral) 732 break; 733 WhitespaceLength += FormatTok->Tok.getLength(); 734 735 readRawToken(*FormatTok); 736 } 737 738 // In case the token starts with escaped newlines, we want to 739 // take them into account as whitespace - this pattern is quite frequent 740 // in macro definitions. 741 // FIXME: Add a more explicit test. 742 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' && 743 FormatTok->TokenText[1] == '\n') { 744 // FIXME: ++FormatTok->NewlinesBefore is missing... 745 WhitespaceLength += 2; 746 Column = 0; 747 FormatTok->TokenText = FormatTok->TokenText.substr(2); 748 } 749 750 FormatTok->WhitespaceRange = SourceRange( 751 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); 752 753 FormatTok->OriginalColumn = Column; 754 755 TrailingWhitespace = 0; 756 if (FormatTok->Tok.is(tok::comment)) { 757 // FIXME: Add the trimmed whitespace to Column. 758 StringRef UntrimmedText = FormatTok->TokenText; 759 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); 760 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); 761 } else if (FormatTok->Tok.is(tok::raw_identifier)) { 762 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); 763 FormatTok->Tok.setIdentifierInfo(&Info); 764 FormatTok->Tok.setKind(Info.getTokenID()); 765 } else if (FormatTok->Tok.is(tok::greatergreater)) { 766 FormatTok->Tok.setKind(tok::greater); 767 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 768 GreaterStashed = true; 769 } 770 771 // Now FormatTok is the next non-whitespace token. 772 773 StringRef Text = FormatTok->TokenText; 774 size_t FirstNewlinePos = Text.find('\n'); 775 if (FirstNewlinePos == StringRef::npos) { 776 // FIXME: ColumnWidth actually depends on the start column, we need to 777 // take this into account when the token is moved. 778 FormatTok->ColumnWidth = 779 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); 780 Column += FormatTok->ColumnWidth; 781 } else { 782 FormatTok->IsMultiline = true; 783 // FIXME: ColumnWidth actually depends on the start column, we need to 784 // take this into account when the token is moved. 785 FormatTok->ColumnWidth = encoding::columnWidthWithTabs( 786 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); 787 788 // The last line of the token always starts in column 0. 789 // Thus, the length can be precomputed even in the presence of tabs. 790 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( 791 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, 792 Encoding); 793 Column = FormatTok->LastLineColumnWidth; 794 } 795 796 return FormatTok; 797 } 798 799 FormatToken *FormatTok; 800 bool GreaterStashed; 801 unsigned Column; 802 unsigned TrailingWhitespace; 803 Lexer &Lex; 804 SourceManager &SourceMgr; 805 FormatStyle &Style; 806 IdentifierTable IdentTable; 807 encoding::Encoding Encoding; 808 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator; 809 SmallVector<FormatToken *, 16> Tokens; 810 811 void readRawToken(FormatToken &Tok) { 812 Lex.LexFromRawLexer(Tok.Tok); 813 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), 814 Tok.Tok.getLength()); 815 // For formatting, treat unterminated string literals like normal string 816 // literals. 817 if (Tok.is(tok::unknown) && !Tok.TokenText.empty() && 818 Tok.TokenText[0] == '"') { 819 Tok.Tok.setKind(tok::string_literal); 820 Tok.IsUnterminatedLiteral = true; 821 } 822 } 823 }; 824 825 class Formatter : public UnwrappedLineConsumer { 826 public: 827 Formatter(const FormatStyle &Style, Lexer &Lex, SourceManager &SourceMgr, 828 const std::vector<CharSourceRange> &Ranges) 829 : Style(Style), Lex(Lex), SourceMgr(SourceMgr), 830 Whitespaces(SourceMgr, Style, inputUsesCRLF(Lex.getBuffer())), 831 Ranges(Ranges), UnwrappedLines(1), 832 Encoding(encoding::detectEncoding(Lex.getBuffer())) { 833 DEBUG(llvm::dbgs() << "File encoding: " 834 << (Encoding == encoding::Encoding_UTF8 ? "UTF8" 835 : "unknown") 836 << "\n"); 837 } 838 839 tooling::Replacements format() { 840 tooling::Replacements Result; 841 FormatTokenLexer Tokens(Lex, SourceMgr, Style, Encoding); 842 843 UnwrappedLineParser Parser(Style, Tokens.lex(), *this); 844 bool StructuralError = Parser.parse(); 845 assert(UnwrappedLines.rbegin()->empty()); 846 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE; 847 ++Run) { 848 DEBUG(llvm::dbgs() << "Run " << Run << "...\n"); 849 SmallVector<AnnotatedLine *, 16> AnnotatedLines; 850 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) { 851 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i])); 852 } 853 tooling::Replacements RunResult = 854 format(AnnotatedLines, StructuralError, Tokens); 855 DEBUG({ 856 llvm::dbgs() << "Replacements for run " << Run << ":\n"; 857 for (tooling::Replacements::iterator I = RunResult.begin(), 858 E = RunResult.end(); 859 I != E; ++I) { 860 llvm::dbgs() << I->toString() << "\n"; 861 } 862 }); 863 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 864 delete AnnotatedLines[i]; 865 } 866 Result.insert(RunResult.begin(), RunResult.end()); 867 Whitespaces.reset(); 868 } 869 return Result; 870 } 871 872 tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 873 bool StructuralError, FormatTokenLexer &Tokens) { 874 TokenAnnotator Annotator(Style, Tokens.getIdentTable().get("in")); 875 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 876 Annotator.annotate(*AnnotatedLines[i]); 877 } 878 deriveLocalStyle(AnnotatedLines); 879 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 880 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 881 } 882 883 Annotator.setCommentLineLevels(AnnotatedLines); 884 885 std::vector<int> IndentForLevel; 886 bool PreviousLineWasTouched = false; 887 const AnnotatedLine *PreviousLine = NULL; 888 bool FormatPPDirective = false; 889 for (SmallVectorImpl<AnnotatedLine *>::iterator I = AnnotatedLines.begin(), 890 E = AnnotatedLines.end(); 891 I != E; ++I) { 892 const AnnotatedLine &TheLine = **I; 893 const FormatToken *FirstTok = TheLine.First; 894 int Offset = getIndentOffset(*TheLine.First); 895 896 // Check whether this line is part of a formatted preprocessor directive. 897 if (FirstTok->HasUnescapedNewline) 898 FormatPPDirective = false; 899 if (!FormatPPDirective && TheLine.InPPDirective && 900 (touchesLine(TheLine) || touchesPPDirective(I + 1, E))) 901 FormatPPDirective = true; 902 903 // Determine indent and try to merge multiple unwrapped lines. 904 while (IndentForLevel.size() <= TheLine.Level) 905 IndentForLevel.push_back(-1); 906 IndentForLevel.resize(TheLine.Level + 1); 907 unsigned Indent = getIndent(IndentForLevel, TheLine.Level); 908 if (static_cast<int>(Indent) + Offset >= 0) 909 Indent += Offset; 910 tryFitMultipleLinesInOne(Indent, I, E); 911 912 bool WasMoved = PreviousLineWasTouched && FirstTok->NewlinesBefore == 0; 913 if (TheLine.First->is(tok::eof)) { 914 if (PreviousLineWasTouched) { 915 unsigned Newlines = std::min(FirstTok->NewlinesBefore, 1u); 916 Whitespaces.replaceWhitespace(*TheLine.First, Newlines, 917 /*IndentLevel=*/0, /*Spaces=*/0, 918 /*TargetColumn=*/0); 919 } 920 } else if (TheLine.Type != LT_Invalid && 921 (WasMoved || FormatPPDirective || touchesLine(TheLine))) { 922 unsigned LevelIndent = getIndent(IndentForLevel, TheLine.Level); 923 if (FirstTok->WhitespaceRange.isValid()) { 924 formatFirstToken(*TheLine.First, PreviousLine, TheLine.Level, Indent, 925 TheLine.InPPDirective); 926 } else { 927 Indent = LevelIndent = FirstTok->OriginalColumn; 928 } 929 ContinuationIndenter Indenter(Style, SourceMgr, Whitespaces, Encoding, 930 BinPackInconclusiveFunctions); 931 932 // If everything fits on a single line, just put it there. 933 unsigned ColumnLimit = Style.ColumnLimit; 934 AnnotatedLine *NextLine = *(I + 1); 935 if ((I + 1) != E && NextLine->InPPDirective && 936 !NextLine->First->HasUnescapedNewline) 937 ColumnLimit = getColumnLimit(TheLine.InPPDirective); 938 939 if (TheLine.Last->TotalLength + Indent <= ColumnLimit) { 940 LineState State = 941 Indenter.getInitialState(Indent, &TheLine, /*DryRun=*/false); 942 while (State.NextToken != NULL) 943 Indenter.addTokenToState(State, false, false); 944 } else if (Style.ColumnLimit == 0) { 945 NoColumnLimitFormatter Formatter(&Indenter); 946 Formatter.format(Indent, &TheLine); 947 } else { 948 UnwrappedLineFormatter Formatter(&Indenter, &Whitespaces, Style, 949 TheLine); 950 Formatter.format(Indent); 951 } 952 953 IndentForLevel[TheLine.Level] = LevelIndent; 954 PreviousLineWasTouched = true; 955 } else { 956 // Format the first token if necessary, and notify the WhitespaceManager 957 // about the unchanged whitespace. 958 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 959 if (Tok == TheLine.First && 960 (Tok->NewlinesBefore > 0 || Tok->IsFirst)) { 961 unsigned LevelIndent = Tok->OriginalColumn; 962 // Remove trailing whitespace of the previous line if it was 963 // touched. 964 if (PreviousLineWasTouched || touchesEmptyLineBefore(TheLine)) { 965 formatFirstToken(*Tok, PreviousLine, TheLine.Level, LevelIndent, 966 TheLine.InPPDirective); 967 } else { 968 Whitespaces.addUntouchableToken(*Tok, TheLine.InPPDirective); 969 } 970 971 if (static_cast<int>(LevelIndent) - Offset >= 0) 972 LevelIndent -= Offset; 973 if (Tok->isNot(tok::comment)) 974 IndentForLevel[TheLine.Level] = LevelIndent; 975 } else { 976 Whitespaces.addUntouchableToken(*Tok, TheLine.InPPDirective); 977 } 978 } 979 // If we did not reformat this unwrapped line, the column at the end of 980 // the last token is unchanged - thus, we can calculate the end of the 981 // last token. 982 PreviousLineWasTouched = false; 983 } 984 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 985 Tok->Finalized = true; 986 } 987 PreviousLine = *I; 988 } 989 return Whitespaces.generateReplacements(); 990 } 991 992 private: 993 static bool inputUsesCRLF(StringRef Text) { 994 return Text.count('\r') * 2 > Text.count('\n'); 995 } 996 997 void 998 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 999 unsigned CountBoundToVariable = 0; 1000 unsigned CountBoundToType = 0; 1001 bool HasCpp03IncompatibleFormat = false; 1002 bool HasBinPackedFunction = false; 1003 bool HasOnePerLineFunction = false; 1004 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1005 if (!AnnotatedLines[i]->First->Next) 1006 continue; 1007 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1008 while (Tok->Next) { 1009 if (Tok->Type == TT_PointerOrReference) { 1010 bool SpacesBefore = 1011 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1012 bool SpacesAfter = Tok->Next->WhitespaceRange.getBegin() != 1013 Tok->Next->WhitespaceRange.getEnd(); 1014 if (SpacesBefore && !SpacesAfter) 1015 ++CountBoundToVariable; 1016 else if (!SpacesBefore && SpacesAfter) 1017 ++CountBoundToType; 1018 } 1019 1020 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1021 if (Tok->is(tok::coloncolon) && 1022 Tok->Previous->Type == TT_TemplateOpener) 1023 HasCpp03IncompatibleFormat = true; 1024 if (Tok->Type == TT_TemplateCloser && 1025 Tok->Previous->Type == TT_TemplateCloser) 1026 HasCpp03IncompatibleFormat = true; 1027 } 1028 1029 if (Tok->PackingKind == PPK_BinPacked) 1030 HasBinPackedFunction = true; 1031 if (Tok->PackingKind == PPK_OnePerLine) 1032 HasOnePerLineFunction = true; 1033 1034 Tok = Tok->Next; 1035 } 1036 } 1037 if (Style.DerivePointerBinding) { 1038 if (CountBoundToType > CountBoundToVariable) 1039 Style.PointerBindsToType = true; 1040 else if (CountBoundToType < CountBoundToVariable) 1041 Style.PointerBindsToType = false; 1042 } 1043 if (Style.Standard == FormatStyle::LS_Auto) { 1044 Style.Standard = HasCpp03IncompatibleFormat ? FormatStyle::LS_Cpp11 1045 : FormatStyle::LS_Cpp03; 1046 } 1047 BinPackInconclusiveFunctions = 1048 HasBinPackedFunction || !HasOnePerLineFunction; 1049 } 1050 1051 /// \brief Get the indent of \p Level from \p IndentForLevel. 1052 /// 1053 /// \p IndentForLevel must contain the indent for the level \c l 1054 /// at \p IndentForLevel[l], or a value < 0 if the indent for 1055 /// that level is unknown. 1056 unsigned getIndent(const std::vector<int> IndentForLevel, unsigned Level) { 1057 if (IndentForLevel[Level] != -1) 1058 return IndentForLevel[Level]; 1059 if (Level == 0) 1060 return 0; 1061 return getIndent(IndentForLevel, Level - 1) + Style.IndentWidth; 1062 } 1063 1064 /// \brief Get the offset of the line relatively to the level. 1065 /// 1066 /// For example, 'public:' labels in classes are offset by 1 or 2 1067 /// characters to the left from their level. 1068 int getIndentOffset(const FormatToken &RootToken) { 1069 if (RootToken.isAccessSpecifier(false) || RootToken.isObjCAccessSpecifier()) 1070 return Style.AccessModifierOffset; 1071 return 0; 1072 } 1073 1074 /// \brief Tries to merge lines into one. 1075 /// 1076 /// This will change \c Line and \c AnnotatedLine to contain the merged line, 1077 /// if possible; note that \c I will be incremented when lines are merged. 1078 void tryFitMultipleLinesInOne(unsigned Indent, 1079 SmallVectorImpl<AnnotatedLine *>::iterator &I, 1080 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1081 // We can never merge stuff if there are trailing line comments. 1082 AnnotatedLine *TheLine = *I; 1083 if (TheLine->Last->Type == TT_LineComment) 1084 return; 1085 1086 if (Indent > Style.ColumnLimit) 1087 return; 1088 1089 unsigned Limit = Style.ColumnLimit - Indent; 1090 // If we already exceed the column limit, we set 'Limit' to 0. The different 1091 // tryMerge..() functions can then decide whether to still do merging. 1092 Limit = TheLine->Last->TotalLength > Limit 1093 ? 0 1094 : Limit - TheLine->Last->TotalLength; 1095 1096 if (I + 1 == E || (*(I + 1))->Type == LT_Invalid) 1097 return; 1098 1099 if (TheLine->Last->is(tok::l_brace)) { 1100 tryMergeSimpleBlock(I, E, Limit); 1101 } else if (Style.AllowShortIfStatementsOnASingleLine && 1102 TheLine->First->is(tok::kw_if)) { 1103 tryMergeSimpleControlStatement(I, E, Limit); 1104 } else if (Style.AllowShortLoopsOnASingleLine && 1105 TheLine->First->isOneOf(tok::kw_for, tok::kw_while)) { 1106 tryMergeSimpleControlStatement(I, E, Limit); 1107 } else if (TheLine->InPPDirective && (TheLine->First->HasUnescapedNewline || 1108 TheLine->First->IsFirst)) { 1109 tryMergeSimplePPDirective(I, E, Limit); 1110 } 1111 } 1112 1113 void tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine *>::iterator &I, 1114 SmallVectorImpl<AnnotatedLine *>::iterator E, 1115 unsigned Limit) { 1116 if (Limit == 0) 1117 return; 1118 AnnotatedLine &Line = **I; 1119 if (!(*(I + 1))->InPPDirective || (*(I + 1))->First->HasUnescapedNewline) 1120 return; 1121 if (I + 2 != E && (*(I + 2))->InPPDirective && 1122 !(*(I + 2))->First->HasUnescapedNewline) 1123 return; 1124 if (1 + (*(I + 1))->Last->TotalLength > Limit) 1125 return; 1126 join(Line, **(++I)); 1127 } 1128 1129 void 1130 tryMergeSimpleControlStatement(SmallVectorImpl<AnnotatedLine *>::iterator &I, 1131 SmallVectorImpl<AnnotatedLine *>::iterator E, 1132 unsigned Limit) { 1133 if (Limit == 0) 1134 return; 1135 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman && 1136 (*(I + 1))->First->is(tok::l_brace)) 1137 return; 1138 if ((*(I + 1))->InPPDirective != (*I)->InPPDirective || 1139 ((*(I + 1))->InPPDirective && (*(I + 1))->First->HasUnescapedNewline)) 1140 return; 1141 AnnotatedLine &Line = **I; 1142 if (Line.Last->isNot(tok::r_paren)) 1143 return; 1144 if (1 + (*(I + 1))->Last->TotalLength > Limit) 1145 return; 1146 if ((*(I + 1))->First->isOneOf(tok::semi, tok::kw_if, tok::kw_for, 1147 tok::kw_while) || 1148 (*(I + 1))->First->Type == TT_LineComment) 1149 return; 1150 // Only inline simple if's (no nested if or else). 1151 if (I + 2 != E && Line.First->is(tok::kw_if) && 1152 (*(I + 2))->First->is(tok::kw_else)) 1153 return; 1154 join(Line, **(++I)); 1155 } 1156 1157 void tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine *>::iterator &I, 1158 SmallVectorImpl<AnnotatedLine *>::iterator E, 1159 unsigned Limit) { 1160 // No merging if the brace already is on the next line. 1161 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 1162 return; 1163 1164 // First, check that the current line allows merging. This is the case if 1165 // we're not in a control flow statement and the last token is an opening 1166 // brace. 1167 AnnotatedLine &Line = **I; 1168 if (Line.First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_do, tok::r_brace, 1169 tok::kw_else, tok::kw_try, tok::kw_catch, 1170 tok::kw_for, 1171 // This gets rid of all ObjC @ keywords and methods. 1172 tok::at, tok::minus, tok::plus)) 1173 return; 1174 1175 FormatToken *Tok = (*(I + 1))->First; 1176 if (Tok->is(tok::r_brace) && !Tok->MustBreakBefore && 1177 (Tok->getNextNonComment() == NULL || 1178 Tok->getNextNonComment()->is(tok::semi))) { 1179 // We merge empty blocks even if the line exceeds the column limit. 1180 Tok->SpacesRequiredBefore = 0; 1181 Tok->CanBreakBefore = true; 1182 join(Line, **(I + 1)); 1183 I += 1; 1184 } else if (Limit != 0 && Line.First->isNot(tok::kw_namespace)) { 1185 // Check that we still have three lines and they fit into the limit. 1186 if (I + 2 == E || (*(I + 2))->Type == LT_Invalid || 1187 !nextTwoLinesFitInto(I, Limit)) 1188 return; 1189 1190 // Second, check that the next line does not contain any braces - if it 1191 // does, readability declines when putting it into a single line. 1192 if ((*(I + 1))->Last->Type == TT_LineComment || Tok->MustBreakBefore) 1193 return; 1194 do { 1195 if (Tok->isOneOf(tok::l_brace, tok::r_brace)) 1196 return; 1197 Tok = Tok->Next; 1198 } while (Tok != NULL); 1199 1200 // Last, check that the third line contains a single closing brace. 1201 Tok = (*(I + 2))->First; 1202 if (Tok->getNextNonComment() != NULL || Tok->isNot(tok::r_brace) || 1203 Tok->MustBreakBefore) 1204 return; 1205 1206 join(Line, **(I + 1)); 1207 join(Line, **(I + 2)); 1208 I += 2; 1209 } 1210 } 1211 1212 bool nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine *>::iterator I, 1213 unsigned Limit) { 1214 return 1 + (*(I + 1))->Last->TotalLength + 1 + 1215 (*(I + 2))->Last->TotalLength <= 1216 Limit; 1217 } 1218 1219 void join(AnnotatedLine &A, const AnnotatedLine &B) { 1220 assert(!A.Last->Next); 1221 assert(!B.First->Previous); 1222 A.Last->Next = B.First; 1223 B.First->Previous = A.Last; 1224 unsigned LengthA = A.Last->TotalLength + B.First->SpacesRequiredBefore; 1225 for (FormatToken *Tok = B.First; Tok; Tok = Tok->Next) { 1226 Tok->TotalLength += LengthA; 1227 A.Last = Tok; 1228 } 1229 } 1230 1231 bool touchesRanges(const CharSourceRange &Range) { 1232 for (unsigned i = 0, e = Ranges.size(); i != e; ++i) { 1233 if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), 1234 Ranges[i].getBegin()) && 1235 !SourceMgr.isBeforeInTranslationUnit(Ranges[i].getEnd(), 1236 Range.getBegin())) 1237 return true; 1238 } 1239 return false; 1240 } 1241 1242 bool touchesLine(const AnnotatedLine &TheLine) { 1243 const FormatToken *First = TheLine.First; 1244 const FormatToken *Last = TheLine.Last; 1245 CharSourceRange LineRange = CharSourceRange::getCharRange( 1246 First->WhitespaceRange.getBegin().getLocWithOffset( 1247 First->LastNewlineOffset), 1248 Last->getStartOfNonWhitespace().getLocWithOffset( 1249 Last->TokenText.size() - 1)); 1250 return touchesRanges(LineRange); 1251 } 1252 1253 bool touchesPPDirective(SmallVectorImpl<AnnotatedLine *>::iterator I, 1254 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1255 for (; I != E; ++I) { 1256 if ((*I)->First->HasUnescapedNewline) 1257 return false; 1258 if (touchesLine(**I)) 1259 return true; 1260 } 1261 return false; 1262 } 1263 1264 bool touchesEmptyLineBefore(const AnnotatedLine &TheLine) { 1265 const FormatToken *First = TheLine.First; 1266 CharSourceRange LineRange = CharSourceRange::getCharRange( 1267 First->WhitespaceRange.getBegin(), 1268 First->WhitespaceRange.getBegin().getLocWithOffset( 1269 First->LastNewlineOffset)); 1270 return touchesRanges(LineRange); 1271 } 1272 1273 virtual void consumeUnwrappedLine(const UnwrappedLine &TheLine) { 1274 assert(!UnwrappedLines.empty()); 1275 UnwrappedLines.back().push_back(TheLine); 1276 } 1277 1278 virtual void finishRun() { 1279 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>()); 1280 } 1281 1282 /// \brief Add a new line and the required indent before the first Token 1283 /// of the \c UnwrappedLine if there was no structural parsing error. 1284 void formatFirstToken(FormatToken &RootToken, 1285 const AnnotatedLine *PreviousLine, unsigned IndentLevel, 1286 unsigned Indent, bool InPPDirective) { 1287 unsigned Newlines = 1288 std::min(RootToken.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1); 1289 // Remove empty lines before "}" where applicable. 1290 if (RootToken.is(tok::r_brace) && 1291 (!RootToken.Next || 1292 (RootToken.Next->is(tok::semi) && !RootToken.Next->Next))) 1293 Newlines = std::min(Newlines, 1u); 1294 if (Newlines == 0 && !RootToken.IsFirst) 1295 Newlines = 1; 1296 1297 // Insert extra new line before access specifiers. 1298 if (PreviousLine && PreviousLine->Last->isOneOf(tok::semi, tok::r_brace) && 1299 RootToken.isAccessSpecifier() && RootToken.NewlinesBefore == 1) 1300 ++Newlines; 1301 1302 // Remove empty lines after access specifiers. 1303 if (PreviousLine && PreviousLine->First->isAccessSpecifier()) 1304 Newlines = std::min(1u, Newlines); 1305 1306 Whitespaces.replaceWhitespace( 1307 RootToken, Newlines, IndentLevel, Indent, Indent, 1308 InPPDirective && !RootToken.HasUnescapedNewline); 1309 } 1310 1311 unsigned getColumnLimit(bool InPPDirective) const { 1312 // In preprocessor directives reserve two chars for trailing " \" 1313 return Style.ColumnLimit - (InPPDirective ? 2 : 0); 1314 } 1315 1316 FormatStyle Style; 1317 Lexer &Lex; 1318 SourceManager &SourceMgr; 1319 WhitespaceManager Whitespaces; 1320 std::vector<CharSourceRange> Ranges; 1321 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines; 1322 1323 encoding::Encoding Encoding; 1324 bool BinPackInconclusiveFunctions; 1325 }; 1326 1327 } // end anonymous namespace 1328 1329 tooling::Replacements reformat(const FormatStyle &Style, Lexer &Lex, 1330 SourceManager &SourceMgr, 1331 std::vector<CharSourceRange> Ranges) { 1332 Formatter formatter(Style, Lex, SourceMgr, Ranges); 1333 return formatter.format(); 1334 } 1335 1336 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1337 std::vector<tooling::Range> Ranges, 1338 StringRef FileName) { 1339 FileManager Files((FileSystemOptions())); 1340 DiagnosticsEngine Diagnostics( 1341 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), 1342 new DiagnosticOptions); 1343 SourceManager SourceMgr(Diagnostics, Files); 1344 llvm::MemoryBuffer *Buf = llvm::MemoryBuffer::getMemBuffer(Code, FileName); 1345 const clang::FileEntry *Entry = 1346 Files.getVirtualFile(FileName, Buf->getBufferSize(), 0); 1347 SourceMgr.overrideFileContents(Entry, Buf); 1348 FileID ID = 1349 SourceMgr.createFileID(Entry, SourceLocation(), clang::SrcMgr::C_User); 1350 Lexer Lex(ID, SourceMgr.getBuffer(ID), SourceMgr, 1351 getFormattingLangOpts(Style.Standard)); 1352 SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID); 1353 std::vector<CharSourceRange> CharRanges; 1354 for (unsigned i = 0, e = Ranges.size(); i != e; ++i) { 1355 SourceLocation Start = StartOfFile.getLocWithOffset(Ranges[i].getOffset()); 1356 SourceLocation End = Start.getLocWithOffset(Ranges[i].getLength()); 1357 CharRanges.push_back(CharSourceRange::getCharRange(Start, End)); 1358 } 1359 return reformat(Style, Lex, SourceMgr, CharRanges); 1360 } 1361 1362 LangOptions getFormattingLangOpts(FormatStyle::LanguageStandard Standard) { 1363 LangOptions LangOpts; 1364 LangOpts.CPlusPlus = 1; 1365 LangOpts.CPlusPlus11 = Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1366 LangOpts.LineComment = 1; 1367 LangOpts.Bool = 1; 1368 LangOpts.ObjC1 = 1; 1369 LangOpts.ObjC2 = 1; 1370 return LangOpts; 1371 } 1372 1373 const char *StyleOptionHelpDescription = 1374 "Coding style, currently supports:\n" 1375 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1376 "Use -style=file to load style configuration from\n" 1377 ".clang-format file located in one of the parent\n" 1378 "directories of the source file (or current\n" 1379 "directory for stdin).\n" 1380 "Use -style=\"{key: value, ...}\" to set specific\n" 1381 "parameters, e.g.:\n" 1382 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1383 1384 FormatStyle getStyle(StringRef StyleName, StringRef FileName) { 1385 // Fallback style in case the rest of this function can't determine a style. 1386 StringRef FallbackStyle = "LLVM"; 1387 FormatStyle Style; 1388 getPredefinedStyle(FallbackStyle, &Style); 1389 1390 if (StyleName.startswith("{")) { 1391 // Parse YAML/JSON style from the command line. 1392 if (llvm::error_code ec = parseConfiguration(StyleName, &Style)) { 1393 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1394 << FallbackStyle << " style\n"; 1395 } 1396 return Style; 1397 } 1398 1399 if (!StyleName.equals_lower("file")) { 1400 if (!getPredefinedStyle(StyleName, &Style)) 1401 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1402 << " style\n"; 1403 return Style; 1404 } 1405 1406 SmallString<128> Path(FileName); 1407 llvm::sys::fs::make_absolute(Path); 1408 for (StringRef Directory = Path; !Directory.empty(); 1409 Directory = llvm::sys::path::parent_path(Directory)) { 1410 if (!llvm::sys::fs::is_directory(Directory)) 1411 continue; 1412 SmallString<128> ConfigFile(Directory); 1413 1414 llvm::sys::path::append(ConfigFile, ".clang-format"); 1415 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1416 bool IsFile = false; 1417 // Ignore errors from is_regular_file: we only need to know if we can read 1418 // the file or not. 1419 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1420 1421 if (!IsFile) { 1422 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1423 ConfigFile = Directory; 1424 llvm::sys::path::append(ConfigFile, "_clang-format"); 1425 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1426 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1427 } 1428 1429 if (IsFile) { 1430 OwningPtr<llvm::MemoryBuffer> Text; 1431 if (llvm::error_code ec = 1432 llvm::MemoryBuffer::getFile(ConfigFile.c_str(), Text)) { 1433 llvm::errs() << ec.message() << "\n"; 1434 continue; 1435 } 1436 if (llvm::error_code ec = parseConfiguration(Text->getBuffer(), &Style)) { 1437 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1438 << "\n"; 1439 continue; 1440 } 1441 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1442 return Style; 1443 } 1444 } 1445 llvm::errs() << "Can't find usable .clang-format, using " << FallbackStyle 1446 << " style\n"; 1447 return Style; 1448 } 1449 1450 } // namespace format 1451 } // namespace clang 1452