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