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/Path.h" 30 #include "llvm/Support/YAMLTraits.h" 31 #include <queue> 32 #include <string> 33 34 using clang::format::FormatStyle; 35 36 namespace llvm { 37 namespace yaml { 38 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> { 39 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) { 40 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp); 41 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript); 42 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto); 43 } 44 }; 45 46 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> { 47 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) { 48 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03); 49 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03); 50 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11); 51 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11); 52 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto); 53 } 54 }; 55 56 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> { 57 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) { 58 IO.enumCase(Value, "Never", FormatStyle::UT_Never); 59 IO.enumCase(Value, "false", FormatStyle::UT_Never); 60 IO.enumCase(Value, "Always", FormatStyle::UT_Always); 61 IO.enumCase(Value, "true", FormatStyle::UT_Always); 62 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation); 63 } 64 }; 65 66 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> { 67 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) { 68 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach); 69 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux); 70 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup); 71 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman); 72 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU); 73 } 74 }; 75 76 template <> 77 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> { 78 static void enumeration(IO &IO, 79 FormatStyle::NamespaceIndentationKind &Value) { 80 IO.enumCase(Value, "None", FormatStyle::NI_None); 81 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner); 82 IO.enumCase(Value, "All", FormatStyle::NI_All); 83 } 84 }; 85 86 template <> 87 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> { 88 static void enumeration(IO &IO, 89 FormatStyle::SpaceBeforeParensOptions &Value) { 90 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never); 91 IO.enumCase(Value, "ControlStatements", 92 FormatStyle::SBPO_ControlStatements); 93 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always); 94 95 // For backward compatibility. 96 IO.enumCase(Value, "false", FormatStyle::SBPO_Never); 97 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements); 98 } 99 }; 100 101 template <> struct MappingTraits<FormatStyle> { 102 static void mapping(IO &IO, FormatStyle &Style) { 103 // When reading, read the language first, we need it for getPredefinedStyle. 104 IO.mapOptional("Language", Style.Language); 105 106 if (IO.outputting()) { 107 StringRef StylesArray[] = { "LLVM", "Google", "Chromium", 108 "Mozilla", "WebKit", "GNU" }; 109 ArrayRef<StringRef> Styles(StylesArray); 110 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 111 StringRef StyleName(Styles[i]); 112 FormatStyle PredefinedStyle; 113 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) && 114 Style == PredefinedStyle) { 115 IO.mapOptional("# BasedOnStyle", StyleName); 116 break; 117 } 118 } 119 } else { 120 StringRef BasedOnStyle; 121 IO.mapOptional("BasedOnStyle", BasedOnStyle); 122 if (!BasedOnStyle.empty()) { 123 FormatStyle::LanguageKind OldLanguage = Style.Language; 124 FormatStyle::LanguageKind Language = 125 ((FormatStyle *)IO.getContext())->Language; 126 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) { 127 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 128 return; 129 } 130 Style.Language = OldLanguage; 131 } 132 } 133 134 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 135 IO.mapOptional("ConstructorInitializerIndentWidth", 136 Style.ConstructorInitializerIndentWidth); 137 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 138 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 139 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 140 Style.AllowAllParametersOfDeclarationOnNextLine); 141 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 142 Style.AllowShortIfStatementsOnASingleLine); 143 IO.mapOptional("AllowShortLoopsOnASingleLine", 144 Style.AllowShortLoopsOnASingleLine); 145 IO.mapOptional("AllowShortFunctionsOnASingleLine", 146 Style.AllowShortFunctionsOnASingleLine); 147 IO.mapOptional("AlwaysBreakTemplateDeclarations", 148 Style.AlwaysBreakTemplateDeclarations); 149 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 150 Style.AlwaysBreakBeforeMultilineStrings); 151 IO.mapOptional("BreakBeforeBinaryOperators", 152 Style.BreakBeforeBinaryOperators); 153 IO.mapOptional("BreakBeforeTernaryOperators", 154 Style.BreakBeforeTernaryOperators); 155 IO.mapOptional("BreakConstructorInitializersBeforeComma", 156 Style.BreakConstructorInitializersBeforeComma); 157 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 158 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 159 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 160 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 161 IO.mapOptional("DerivePointerBinding", Style.DerivePointerBinding); 162 IO.mapOptional("ExperimentalAutoDetectBinPacking", 163 Style.ExperimentalAutoDetectBinPacking); 164 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 165 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 166 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 167 IO.mapOptional("ObjCSpaceBeforeProtocolList", 168 Style.ObjCSpaceBeforeProtocolList); 169 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 170 Style.PenaltyBreakBeforeFirstCallParameter); 171 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 172 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 173 IO.mapOptional("PenaltyBreakFirstLessLess", 174 Style.PenaltyBreakFirstLessLess); 175 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 176 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 177 Style.PenaltyReturnTypeOnItsOwnLine); 178 IO.mapOptional("PointerBindsToType", Style.PointerBindsToType); 179 IO.mapOptional("SpacesBeforeTrailingComments", 180 Style.SpacesBeforeTrailingComments); 181 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 182 IO.mapOptional("Standard", Style.Standard); 183 IO.mapOptional("IndentWidth", Style.IndentWidth); 184 IO.mapOptional("TabWidth", Style.TabWidth); 185 IO.mapOptional("UseTab", Style.UseTab); 186 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 187 IO.mapOptional("IndentFunctionDeclarationAfterType", 188 Style.IndentFunctionDeclarationAfterType); 189 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 190 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 191 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 192 IO.mapOptional("SpacesInCStyleCastParentheses", 193 Style.SpacesInCStyleCastParentheses); 194 IO.mapOptional("SpacesInContainerLiterals", 195 Style.SpacesInContainerLiterals); 196 IO.mapOptional("SpaceBeforeAssignmentOperators", 197 Style.SpaceBeforeAssignmentOperators); 198 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 199 IO.mapOptional("CommentPragmas", Style.CommentPragmas); 200 201 // For backward compatibility. 202 if (!IO.outputting()) { 203 IO.mapOptional("SpaceAfterControlStatementKeyword", 204 Style.SpaceBeforeParens); 205 } 206 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens); 207 } 208 }; 209 210 // Allows to read vector<FormatStyle> while keeping default values. 211 // IO.getContext() should contain a pointer to the FormatStyle structure, that 212 // will be used to get default values for missing keys. 213 // If the first element has no Language specified, it will be treated as the 214 // default one for the following elements. 215 template <> struct DocumentListTraits<std::vector<FormatStyle> > { 216 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) { 217 return Seq.size(); 218 } 219 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq, 220 size_t Index) { 221 if (Index >= Seq.size()) { 222 assert(Index == Seq.size()); 223 FormatStyle Template; 224 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) { 225 Template = Seq[0]; 226 } else { 227 Template = *((const FormatStyle*)IO.getContext()); 228 Template.Language = FormatStyle::LK_None; 229 } 230 Seq.resize(Index + 1, Template); 231 } 232 return Seq[Index]; 233 } 234 }; 235 } 236 } 237 238 namespace clang { 239 namespace format { 240 241 FormatStyle getLLVMStyle() { 242 FormatStyle LLVMStyle; 243 LLVMStyle.Language = FormatStyle::LK_Cpp; 244 LLVMStyle.AccessModifierOffset = -2; 245 LLVMStyle.AlignEscapedNewlinesLeft = false; 246 LLVMStyle.AlignTrailingComments = true; 247 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 248 LLVMStyle.AllowShortFunctionsOnASingleLine = true; 249 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 250 LLVMStyle.AllowShortLoopsOnASingleLine = false; 251 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 252 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 253 LLVMStyle.BinPackParameters = true; 254 LLVMStyle.BreakBeforeBinaryOperators = false; 255 LLVMStyle.BreakBeforeTernaryOperators = true; 256 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 257 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 258 LLVMStyle.ColumnLimit = 80; 259 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 260 LLVMStyle.ConstructorInitializerIndentWidth = 4; 261 LLVMStyle.Cpp11BracedListStyle = false; 262 LLVMStyle.DerivePointerBinding = false; 263 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 264 LLVMStyle.IndentCaseLabels = false; 265 LLVMStyle.IndentFunctionDeclarationAfterType = false; 266 LLVMStyle.IndentWidth = 2; 267 LLVMStyle.TabWidth = 8; 268 LLVMStyle.MaxEmptyLinesToKeep = 1; 269 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 270 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 271 LLVMStyle.PointerBindsToType = false; 272 LLVMStyle.SpacesBeforeTrailingComments = 1; 273 LLVMStyle.Standard = FormatStyle::LS_Cpp03; 274 LLVMStyle.UseTab = FormatStyle::UT_Never; 275 LLVMStyle.SpacesInParentheses = false; 276 LLVMStyle.SpaceInEmptyParentheses = false; 277 LLVMStyle.SpacesInContainerLiterals = true; 278 LLVMStyle.SpacesInCStyleCastParentheses = false; 279 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements; 280 LLVMStyle.SpaceBeforeAssignmentOperators = true; 281 LLVMStyle.ContinuationIndentWidth = 4; 282 LLVMStyle.SpacesInAngles = false; 283 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; 284 285 LLVMStyle.PenaltyBreakComment = 300; 286 LLVMStyle.PenaltyBreakFirstLessLess = 120; 287 LLVMStyle.PenaltyBreakString = 1000; 288 LLVMStyle.PenaltyExcessCharacter = 1000000; 289 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 290 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 291 292 return LLVMStyle; 293 } 294 295 FormatStyle getGoogleStyle() { 296 FormatStyle GoogleStyle = getLLVMStyle(); 297 GoogleStyle.AccessModifierOffset = -1; 298 GoogleStyle.AlignEscapedNewlinesLeft = true; 299 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 300 GoogleStyle.AllowShortLoopsOnASingleLine = true; 301 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 302 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 303 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 304 GoogleStyle.Cpp11BracedListStyle = true; 305 GoogleStyle.DerivePointerBinding = true; 306 GoogleStyle.IndentCaseLabels = true; 307 GoogleStyle.IndentFunctionDeclarationAfterType = true; 308 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 309 GoogleStyle.PointerBindsToType = true; 310 GoogleStyle.SpacesBeforeTrailingComments = 2; 311 GoogleStyle.Standard = FormatStyle::LS_Auto; 312 313 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 314 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 315 316 return GoogleStyle; 317 } 318 319 FormatStyle getGoogleJSStyle() { 320 FormatStyle GoogleJSStyle = getGoogleStyle(); 321 GoogleJSStyle.Language = FormatStyle::LK_JavaScript; 322 GoogleJSStyle.BreakBeforeTernaryOperators = false; 323 GoogleJSStyle.SpacesInContainerLiterals = false; 324 return GoogleJSStyle; 325 } 326 327 FormatStyle getGoogleProtoStyle() { 328 FormatStyle GoogleProtoStyle = getGoogleStyle(); 329 GoogleProtoStyle.Language = FormatStyle::LK_Proto; 330 GoogleProtoStyle.AllowShortFunctionsOnASingleLine = false; 331 return GoogleProtoStyle; 332 } 333 334 FormatStyle getChromiumStyle() { 335 FormatStyle ChromiumStyle = getGoogleStyle(); 336 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 337 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 338 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 339 ChromiumStyle.BinPackParameters = false; 340 ChromiumStyle.DerivePointerBinding = false; 341 ChromiumStyle.Standard = FormatStyle::LS_Cpp03; 342 return ChromiumStyle; 343 } 344 345 FormatStyle getMozillaStyle() { 346 FormatStyle MozillaStyle = getLLVMStyle(); 347 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 348 MozillaStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 349 MozillaStyle.DerivePointerBinding = true; 350 MozillaStyle.IndentCaseLabels = true; 351 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 352 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 353 MozillaStyle.PointerBindsToType = true; 354 return MozillaStyle; 355 } 356 357 FormatStyle getWebKitStyle() { 358 FormatStyle Style = getLLVMStyle(); 359 Style.AccessModifierOffset = -4; 360 Style.AlignTrailingComments = false; 361 Style.BreakBeforeBinaryOperators = true; 362 Style.BreakBeforeBraces = FormatStyle::BS_Stroustrup; 363 Style.BreakConstructorInitializersBeforeComma = true; 364 Style.ColumnLimit = 0; 365 Style.IndentWidth = 4; 366 Style.NamespaceIndentation = FormatStyle::NI_Inner; 367 Style.PointerBindsToType = true; 368 return Style; 369 } 370 371 FormatStyle getGNUStyle() { 372 FormatStyle Style = getLLVMStyle(); 373 Style.BreakBeforeBinaryOperators = true; 374 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 375 Style.BreakBeforeTernaryOperators = true; 376 Style.ColumnLimit = 79; 377 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 378 return Style; 379 } 380 381 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 382 FormatStyle *Style) { 383 if (Name.equals_lower("llvm")) { 384 *Style = getLLVMStyle(); 385 } else if (Name.equals_lower("chromium")) { 386 *Style = getChromiumStyle(); 387 } else if (Name.equals_lower("mozilla")) { 388 *Style = getMozillaStyle(); 389 } else if (Name.equals_lower("google")) { 390 switch (Language) { 391 case FormatStyle::LK_JavaScript: 392 *Style = getGoogleJSStyle(); 393 break; 394 case FormatStyle::LK_Proto: 395 *Style = getGoogleProtoStyle(); 396 break; 397 default: 398 *Style = getGoogleStyle(); 399 } 400 } else if (Name.equals_lower("webkit")) { 401 *Style = getWebKitStyle(); 402 } else if (Name.equals_lower("gnu")) { 403 *Style = getGNUStyle(); 404 } else { 405 return false; 406 } 407 408 Style->Language = Language; 409 return true; 410 } 411 412 llvm::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 413 assert(Style); 414 FormatStyle::LanguageKind Language = Style->Language; 415 assert(Language != FormatStyle::LK_None); 416 if (Text.trim().empty()) 417 return llvm::make_error_code(llvm::errc::invalid_argument); 418 419 std::vector<FormatStyle> Styles; 420 llvm::yaml::Input Input(Text); 421 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 422 // values for the fields, keys for which are missing from the configuration. 423 // Mapping also uses the context to get the language to find the correct 424 // base style. 425 Input.setContext(Style); 426 Input >> Styles; 427 if (Input.error()) 428 return Input.error(); 429 430 for (unsigned i = 0; i < Styles.size(); ++i) { 431 // Ensures that only the first configuration can skip the Language option. 432 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 433 return llvm::make_error_code(llvm::errc::invalid_argument); 434 // Ensure that each language is configured at most once. 435 for (unsigned j = 0; j < i; ++j) { 436 if (Styles[i].Language == Styles[j].Language) { 437 DEBUG(llvm::dbgs() 438 << "Duplicate languages in the config file on positions " << j 439 << " and " << i << "\n"); 440 return llvm::make_error_code(llvm::errc::invalid_argument); 441 } 442 } 443 } 444 // Look for a suitable configuration starting from the end, so we can 445 // find the configuration for the specific language first, and the default 446 // configuration (which can only be at slot 0) after it. 447 for (int i = Styles.size() - 1; i >= 0; --i) { 448 if (Styles[i].Language == Language || 449 Styles[i].Language == FormatStyle::LK_None) { 450 *Style = Styles[i]; 451 Style->Language = Language; 452 return llvm::make_error_code(llvm::errc::success); 453 } 454 } 455 return llvm::make_error_code(llvm::errc::not_supported); 456 } 457 458 std::string configurationAsText(const FormatStyle &Style) { 459 std::string Text; 460 llvm::raw_string_ostream Stream(Text); 461 llvm::yaml::Output Output(Stream); 462 // We use the same mapping method for input and output, so we need a non-const 463 // reference here. 464 FormatStyle NonConstStyle = Style; 465 Output << NonConstStyle; 466 return Stream.str(); 467 } 468 469 namespace { 470 471 class NoColumnLimitFormatter { 472 public: 473 NoColumnLimitFormatter(ContinuationIndenter *Indenter) : Indenter(Indenter) {} 474 475 /// \brief Formats the line starting at \p State, simply keeping all of the 476 /// input's line breaking decisions. 477 void format(unsigned FirstIndent, const AnnotatedLine *Line) { 478 LineState State = 479 Indenter->getInitialState(FirstIndent, Line, /*DryRun=*/false); 480 while (State.NextToken != NULL) { 481 bool Newline = 482 Indenter->mustBreak(State) || 483 (Indenter->canBreak(State) && State.NextToken->NewlinesBefore > 0); 484 Indenter->addTokenToState(State, Newline, /*DryRun=*/false); 485 } 486 } 487 488 private: 489 ContinuationIndenter *Indenter; 490 }; 491 492 class LineJoiner { 493 public: 494 LineJoiner(const FormatStyle &Style) : Style(Style) {} 495 496 /// \brief Calculates how many lines can be merged into 1 starting at \p I. 497 unsigned 498 tryFitMultipleLinesInOne(unsigned Indent, 499 SmallVectorImpl<AnnotatedLine *>::const_iterator I, 500 SmallVectorImpl<AnnotatedLine *>::const_iterator E) { 501 // We can never merge stuff if there are trailing line comments. 502 const AnnotatedLine *TheLine = *I; 503 if (TheLine->Last->Type == TT_LineComment) 504 return 0; 505 506 if (Style.ColumnLimit > 0 && Indent > Style.ColumnLimit) 507 return 0; 508 509 unsigned Limit = 510 Style.ColumnLimit == 0 ? UINT_MAX : Style.ColumnLimit - Indent; 511 // If we already exceed the column limit, we set 'Limit' to 0. The different 512 // tryMerge..() functions can then decide whether to still do merging. 513 Limit = TheLine->Last->TotalLength > Limit 514 ? 0 515 : Limit - TheLine->Last->TotalLength; 516 517 if (I + 1 == E || I[1]->Type == LT_Invalid) 518 return 0; 519 520 if (TheLine->Last->Type == TT_FunctionLBrace && 521 TheLine->First != TheLine->Last) { 522 return Style.AllowShortFunctionsOnASingleLine 523 ? tryMergeSimpleBlock(I, E, Limit) 524 : 0; 525 } 526 if (TheLine->Last->is(tok::l_brace)) { 527 return Style.BreakBeforeBraces == FormatStyle::BS_Attach 528 ? tryMergeSimpleBlock(I, E, Limit) 529 : 0; 530 } 531 if (I[1]->First->Type == TT_FunctionLBrace && 532 Style.BreakBeforeBraces != FormatStyle::BS_Attach) { 533 // Check for Limit <= 2 to account for the " {". 534 if (Limit <= 2 || (Style.ColumnLimit == 0 && containsMustBreak(TheLine))) 535 return 0; 536 Limit -= 2; 537 538 unsigned MergedLines = 0; 539 if (Style.AllowShortFunctionsOnASingleLine) { 540 MergedLines = tryMergeSimpleBlock(I + 1, E, Limit); 541 // If we managed to merge the block, count the function header, which is 542 // on a separate line. 543 if (MergedLines > 0) 544 ++MergedLines; 545 } 546 return MergedLines; 547 } 548 if (TheLine->First->is(tok::kw_if)) { 549 return Style.AllowShortIfStatementsOnASingleLine 550 ? tryMergeSimpleControlStatement(I, E, Limit) 551 : 0; 552 } 553 if (TheLine->First->isOneOf(tok::kw_for, tok::kw_while)) { 554 return Style.AllowShortLoopsOnASingleLine 555 ? tryMergeSimpleControlStatement(I, E, Limit) 556 : 0; 557 } 558 if (TheLine->InPPDirective && 559 (TheLine->First->HasUnescapedNewline || TheLine->First->IsFirst)) { 560 return tryMergeSimplePPDirective(I, E, Limit); 561 } 562 return 0; 563 } 564 565 private: 566 unsigned 567 tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 568 SmallVectorImpl<AnnotatedLine *>::const_iterator E, 569 unsigned Limit) { 570 if (Limit == 0) 571 return 0; 572 if (!I[1]->InPPDirective || I[1]->First->HasUnescapedNewline) 573 return 0; 574 if (I + 2 != E && I[2]->InPPDirective && !I[2]->First->HasUnescapedNewline) 575 return 0; 576 if (1 + I[1]->Last->TotalLength > Limit) 577 return 0; 578 return 1; 579 } 580 581 unsigned tryMergeSimpleControlStatement( 582 SmallVectorImpl<AnnotatedLine *>::const_iterator I, 583 SmallVectorImpl<AnnotatedLine *>::const_iterator E, unsigned Limit) { 584 if (Limit == 0) 585 return 0; 586 if ((Style.BreakBeforeBraces == FormatStyle::BS_Allman || 587 Style.BreakBeforeBraces == FormatStyle::BS_GNU) && 588 I[1]->First->is(tok::l_brace)) 589 return 0; 590 if (I[1]->InPPDirective != (*I)->InPPDirective || 591 (I[1]->InPPDirective && I[1]->First->HasUnescapedNewline)) 592 return 0; 593 AnnotatedLine &Line = **I; 594 if (Line.Last->isNot(tok::r_paren)) 595 return 0; 596 if (1 + I[1]->Last->TotalLength > Limit) 597 return 0; 598 if (I[1]->First->isOneOf(tok::semi, tok::kw_if, tok::kw_for, 599 tok::kw_while) || 600 I[1]->First->Type == TT_LineComment) 601 return 0; 602 // Only inline simple if's (no nested if or else). 603 if (I + 2 != E && Line.First->is(tok::kw_if) && 604 I[2]->First->is(tok::kw_else)) 605 return 0; 606 return 1; 607 } 608 609 unsigned 610 tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 611 SmallVectorImpl<AnnotatedLine *>::const_iterator E, 612 unsigned Limit) { 613 // First, check that the current line allows merging. This is the case if 614 // we're not in a control flow statement and the last token is an opening 615 // brace. 616 AnnotatedLine &Line = **I; 617 if (Line.First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_do, tok::r_brace, 618 tok::kw_else, tok::kw_try, tok::kw_catch, 619 tok::kw_for, 620 // This gets rid of all ObjC @ keywords and methods. 621 tok::at, tok::minus, tok::plus)) 622 return 0; 623 624 FormatToken *Tok = I[1]->First; 625 if (Tok->is(tok::r_brace) && !Tok->MustBreakBefore && 626 (Tok->getNextNonComment() == NULL || 627 Tok->getNextNonComment()->is(tok::semi))) { 628 // We merge empty blocks even if the line exceeds the column limit. 629 Tok->SpacesRequiredBefore = 0; 630 Tok->CanBreakBefore = true; 631 return 1; 632 } else if (Limit != 0 && Line.First->isNot(tok::kw_namespace)) { 633 // Check that we still have three lines and they fit into the limit. 634 if (I + 2 == E || I[2]->Type == LT_Invalid) 635 return 0; 636 637 if (!nextTwoLinesFitInto(I, Limit)) 638 return 0; 639 640 // Second, check that the next line does not contain any braces - if it 641 // does, readability declines when putting it into a single line. 642 if (I[1]->Last->Type == TT_LineComment || Tok->MustBreakBefore) 643 return 0; 644 do { 645 if (Tok->isOneOf(tok::l_brace, tok::r_brace)) 646 return 0; 647 Tok = Tok->Next; 648 } while (Tok != NULL); 649 650 // Last, check that the third line contains a single closing brace. 651 Tok = I[2]->First; 652 if (Tok->getNextNonComment() != NULL || Tok->isNot(tok::r_brace) || 653 Tok->MustBreakBefore) 654 return 0; 655 656 return 2; 657 } 658 return 0; 659 } 660 661 bool nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 662 unsigned Limit) { 663 return 1 + I[1]->Last->TotalLength + 1 + I[2]->Last->TotalLength <= Limit; 664 } 665 666 bool containsMustBreak(const AnnotatedLine *Line) { 667 for (const FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) { 668 if (Tok->MustBreakBefore) 669 return true; 670 } 671 return false; 672 } 673 674 const FormatStyle &Style; 675 }; 676 677 class UnwrappedLineFormatter { 678 public: 679 UnwrappedLineFormatter(ContinuationIndenter *Indenter, 680 WhitespaceManager *Whitespaces, 681 const FormatStyle &Style) 682 : Indenter(Indenter), Whitespaces(Whitespaces), Style(Style), 683 Joiner(Style) {} 684 685 unsigned format(const SmallVectorImpl<AnnotatedLine *> &Lines, bool DryRun, 686 int AdditionalIndent = 0, bool FixBadIndentation = false) { 687 assert(!Lines.empty()); 688 unsigned Penalty = 0; 689 std::vector<int> IndentForLevel; 690 for (unsigned i = 0, e = Lines[0]->Level; i != e; ++i) 691 IndentForLevel.push_back(Style.IndentWidth * i + AdditionalIndent); 692 const AnnotatedLine *PreviousLine = NULL; 693 for (SmallVectorImpl<AnnotatedLine *>::const_iterator I = Lines.begin(), 694 E = Lines.end(); 695 I != E; ++I) { 696 const AnnotatedLine &TheLine = **I; 697 const FormatToken *FirstTok = TheLine.First; 698 int Offset = getIndentOffset(*FirstTok); 699 700 // Determine indent and try to merge multiple unwrapped lines. 701 unsigned Indent; 702 if (TheLine.InPPDirective) { 703 Indent = TheLine.Level * Style.IndentWidth; 704 } else { 705 while (IndentForLevel.size() <= TheLine.Level) 706 IndentForLevel.push_back(-1); 707 IndentForLevel.resize(TheLine.Level + 1); 708 Indent = getIndent(IndentForLevel, TheLine.Level); 709 } 710 unsigned LevelIndent = Indent; 711 if (static_cast<int>(Indent) + Offset >= 0) 712 Indent += Offset; 713 714 // Merge multiple lines if possible. 715 unsigned MergedLines = Joiner.tryFitMultipleLinesInOne(Indent, I, E); 716 if (MergedLines > 0 && Style.ColumnLimit == 0) { 717 // Disallow line merging if there is a break at the start of one of the 718 // input lines. 719 for (unsigned i = 0; i < MergedLines; ++i) { 720 if (I[i + 1]->First->NewlinesBefore > 0) 721 MergedLines = 0; 722 } 723 } 724 if (!DryRun) { 725 for (unsigned i = 0; i < MergedLines; ++i) { 726 join(*I[i], *I[i + 1]); 727 } 728 } 729 I += MergedLines; 730 731 bool FixIndentation = 732 FixBadIndentation && (LevelIndent != FirstTok->OriginalColumn); 733 if (TheLine.First->is(tok::eof)) { 734 if (PreviousLine && PreviousLine->Affected && !DryRun) { 735 // Remove the file's trailing whitespace. 736 unsigned Newlines = std::min(FirstTok->NewlinesBefore, 1u); 737 Whitespaces->replaceWhitespace(*TheLine.First, Newlines, 738 /*IndentLevel=*/0, /*Spaces=*/0, 739 /*TargetColumn=*/0); 740 } 741 } else if (TheLine.Type != LT_Invalid && 742 (TheLine.Affected || FixIndentation)) { 743 if (FirstTok->WhitespaceRange.isValid()) { 744 if (!DryRun) 745 formatFirstToken(*TheLine.First, PreviousLine, TheLine.Level, 746 Indent, TheLine.InPPDirective); 747 } else { 748 Indent = LevelIndent = FirstTok->OriginalColumn; 749 } 750 751 // If everything fits on a single line, just put it there. 752 unsigned ColumnLimit = Style.ColumnLimit; 753 if (I + 1 != E) { 754 AnnotatedLine *NextLine = I[1]; 755 if (NextLine->InPPDirective && !NextLine->First->HasUnescapedNewline) 756 ColumnLimit = getColumnLimit(TheLine.InPPDirective); 757 } 758 759 if (TheLine.Last->TotalLength + Indent <= ColumnLimit) { 760 LineState State = Indenter->getInitialState(Indent, &TheLine, DryRun); 761 while (State.NextToken != NULL) 762 Indenter->addTokenToState(State, /*Newline=*/false, DryRun); 763 } else if (Style.ColumnLimit == 0) { 764 // FIXME: Implement nested blocks for ColumnLimit = 0. 765 NoColumnLimitFormatter Formatter(Indenter); 766 if (!DryRun) 767 Formatter.format(Indent, &TheLine); 768 } else { 769 Penalty += format(TheLine, Indent, DryRun); 770 } 771 772 if (!TheLine.InPPDirective) 773 IndentForLevel[TheLine.Level] = LevelIndent; 774 } else if (TheLine.ChildrenAffected) { 775 format(TheLine.Children, DryRun); 776 } else { 777 // Format the first token if necessary, and notify the WhitespaceManager 778 // about the unchanged whitespace. 779 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 780 if (Tok == TheLine.First && 781 (Tok->NewlinesBefore > 0 || Tok->IsFirst)) { 782 unsigned LevelIndent = Tok->OriginalColumn; 783 if (!DryRun) { 784 // Remove trailing whitespace of the previous line. 785 if ((PreviousLine && PreviousLine->Affected) || 786 TheLine.LeadingEmptyLinesAffected) { 787 formatFirstToken(*Tok, PreviousLine, TheLine.Level, LevelIndent, 788 TheLine.InPPDirective); 789 } else { 790 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective); 791 } 792 } 793 794 if (static_cast<int>(LevelIndent) - Offset >= 0) 795 LevelIndent -= Offset; 796 if (Tok->isNot(tok::comment) && !TheLine.InPPDirective) 797 IndentForLevel[TheLine.Level] = LevelIndent; 798 } else if (!DryRun) { 799 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective); 800 } 801 } 802 } 803 if (!DryRun) { 804 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 805 Tok->Finalized = true; 806 } 807 } 808 PreviousLine = *I; 809 } 810 return Penalty; 811 } 812 813 private: 814 /// \brief Formats an \c AnnotatedLine and returns the penalty. 815 /// 816 /// If \p DryRun is \c false, directly applies the changes. 817 unsigned format(const AnnotatedLine &Line, unsigned FirstIndent, 818 bool DryRun) { 819 LineState State = Indenter->getInitialState(FirstIndent, &Line, DryRun); 820 821 // If the ObjC method declaration does not fit on a line, we should format 822 // it with one arg per line. 823 if (State.Line->Type == LT_ObjCMethodDecl) 824 State.Stack.back().BreakBeforeParameter = true; 825 826 // Find best solution in solution space. 827 return analyzeSolutionSpace(State, DryRun); 828 } 829 830 /// \brief An edge in the solution space from \c Previous->State to \c State, 831 /// inserting a newline dependent on the \c NewLine. 832 struct StateNode { 833 StateNode(const LineState &State, bool NewLine, StateNode *Previous) 834 : State(State), NewLine(NewLine), Previous(Previous) {} 835 LineState State; 836 bool NewLine; 837 StateNode *Previous; 838 }; 839 840 /// \brief A pair of <penalty, count> that is used to prioritize the BFS on. 841 /// 842 /// In case of equal penalties, we want to prefer states that were inserted 843 /// first. During state generation we make sure that we insert states first 844 /// that break the line as late as possible. 845 typedef std::pair<unsigned, unsigned> OrderedPenalty; 846 847 /// \brief An item in the prioritized BFS search queue. The \c StateNode's 848 /// \c State has the given \c OrderedPenalty. 849 typedef std::pair<OrderedPenalty, StateNode *> QueueItem; 850 851 /// \brief The BFS queue type. 852 typedef std::priority_queue<QueueItem, std::vector<QueueItem>, 853 std::greater<QueueItem> > QueueType; 854 855 /// \brief Get the offset of the line relatively to the level. 856 /// 857 /// For example, 'public:' labels in classes are offset by 1 or 2 858 /// characters to the left from their level. 859 int getIndentOffset(const FormatToken &RootToken) { 860 if (RootToken.isAccessSpecifier(false) || RootToken.isObjCAccessSpecifier()) 861 return Style.AccessModifierOffset; 862 return 0; 863 } 864 865 /// \brief Add a new line and the required indent before the first Token 866 /// of the \c UnwrappedLine if there was no structural parsing error. 867 void formatFirstToken(FormatToken &RootToken, 868 const AnnotatedLine *PreviousLine, unsigned IndentLevel, 869 unsigned Indent, bool InPPDirective) { 870 unsigned Newlines = 871 std::min(RootToken.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1); 872 // Remove empty lines before "}" where applicable. 873 if (RootToken.is(tok::r_brace) && 874 (!RootToken.Next || 875 (RootToken.Next->is(tok::semi) && !RootToken.Next->Next))) 876 Newlines = std::min(Newlines, 1u); 877 if (Newlines == 0 && !RootToken.IsFirst) 878 Newlines = 1; 879 880 // Insert extra new line before access specifiers. 881 if (PreviousLine && PreviousLine->Last->isOneOf(tok::semi, tok::r_brace) && 882 RootToken.isAccessSpecifier() && RootToken.NewlinesBefore == 1) 883 ++Newlines; 884 885 // Remove empty lines after access specifiers. 886 if (PreviousLine && PreviousLine->First->isAccessSpecifier()) 887 Newlines = std::min(1u, Newlines); 888 889 Whitespaces->replaceWhitespace(RootToken, Newlines, IndentLevel, Indent, 890 Indent, InPPDirective && 891 !RootToken.HasUnescapedNewline); 892 } 893 894 /// \brief Get the indent of \p Level from \p IndentForLevel. 895 /// 896 /// \p IndentForLevel must contain the indent for the level \c l 897 /// at \p IndentForLevel[l], or a value < 0 if the indent for 898 /// that level is unknown. 899 unsigned getIndent(const std::vector<int> IndentForLevel, unsigned Level) { 900 if (IndentForLevel[Level] != -1) 901 return IndentForLevel[Level]; 902 if (Level == 0) 903 return 0; 904 return getIndent(IndentForLevel, Level - 1) + Style.IndentWidth; 905 } 906 907 void join(AnnotatedLine &A, const AnnotatedLine &B) { 908 assert(!A.Last->Next); 909 assert(!B.First->Previous); 910 if (B.Affected) 911 A.Affected = true; 912 A.Last->Next = B.First; 913 B.First->Previous = A.Last; 914 B.First->CanBreakBefore = true; 915 unsigned LengthA = A.Last->TotalLength + B.First->SpacesRequiredBefore; 916 for (FormatToken *Tok = B.First; Tok; Tok = Tok->Next) { 917 Tok->TotalLength += LengthA; 918 A.Last = Tok; 919 } 920 } 921 922 unsigned getColumnLimit(bool InPPDirective) const { 923 // In preprocessor directives reserve two chars for trailing " \" 924 return Style.ColumnLimit - (InPPDirective ? 2 : 0); 925 } 926 927 /// \brief Analyze the entire solution space starting from \p InitialState. 928 /// 929 /// This implements a variant of Dijkstra's algorithm on the graph that spans 930 /// the solution space (\c LineStates are the nodes). The algorithm tries to 931 /// find the shortest path (the one with lowest penalty) from \p InitialState 932 /// to a state where all tokens are placed. Returns the penalty. 933 /// 934 /// If \p DryRun is \c false, directly applies the changes. 935 unsigned analyzeSolutionSpace(LineState &InitialState, bool DryRun = false) { 936 std::set<LineState> Seen; 937 938 // Increasing count of \c StateNode items we have created. This is used to 939 // create a deterministic order independent of the container. 940 unsigned Count = 0; 941 QueueType Queue; 942 943 // Insert start element into queue. 944 StateNode *Node = 945 new (Allocator.Allocate()) StateNode(InitialState, false, NULL); 946 Queue.push(QueueItem(OrderedPenalty(0, Count), Node)); 947 ++Count; 948 949 unsigned Penalty = 0; 950 951 // While not empty, take first element and follow edges. 952 while (!Queue.empty()) { 953 Penalty = Queue.top().first.first; 954 StateNode *Node = Queue.top().second; 955 if (Node->State.NextToken == NULL) { 956 DEBUG(llvm::dbgs() << "\n---\nPenalty for line: " << Penalty << "\n"); 957 break; 958 } 959 Queue.pop(); 960 961 // Cut off the analysis of certain solutions if the analysis gets too 962 // complex. See description of IgnoreStackForComparison. 963 if (Count > 10000) 964 Node->State.IgnoreStackForComparison = true; 965 966 if (!Seen.insert(Node->State).second) 967 // State already examined with lower penalty. 968 continue; 969 970 FormatDecision LastFormat = Node->State.NextToken->Decision; 971 if (LastFormat == FD_Unformatted || LastFormat == FD_Continue) 972 addNextStateToQueue(Penalty, Node, /*NewLine=*/false, &Count, &Queue); 973 if (LastFormat == FD_Unformatted || LastFormat == FD_Break) 974 addNextStateToQueue(Penalty, Node, /*NewLine=*/true, &Count, &Queue); 975 } 976 977 if (Queue.empty()) { 978 // We were unable to find a solution, do nothing. 979 // FIXME: Add diagnostic? 980 DEBUG(llvm::dbgs() << "Could not find a solution.\n"); 981 return 0; 982 } 983 984 // Reconstruct the solution. 985 if (!DryRun) 986 reconstructPath(InitialState, Queue.top().second); 987 988 DEBUG(llvm::dbgs() << "Total number of analyzed states: " << Count << "\n"); 989 DEBUG(llvm::dbgs() << "---\n"); 990 991 return Penalty; 992 } 993 994 void reconstructPath(LineState &State, StateNode *Current) { 995 std::deque<StateNode *> Path; 996 // We do not need a break before the initial token. 997 while (Current->Previous) { 998 Path.push_front(Current); 999 Current = Current->Previous; 1000 } 1001 for (std::deque<StateNode *>::iterator I = Path.begin(), E = Path.end(); 1002 I != E; ++I) { 1003 unsigned Penalty = 0; 1004 formatChildren(State, (*I)->NewLine, /*DryRun=*/false, Penalty); 1005 Penalty += Indenter->addTokenToState(State, (*I)->NewLine, false); 1006 1007 DEBUG({ 1008 if ((*I)->NewLine) { 1009 llvm::dbgs() << "Penalty for placing " 1010 << (*I)->Previous->State.NextToken->Tok.getName() << ": " 1011 << Penalty << "\n"; 1012 } 1013 }); 1014 } 1015 } 1016 1017 /// \brief Add the following state to the analysis queue \c Queue. 1018 /// 1019 /// Assume the current state is \p PreviousNode and has been reached with a 1020 /// penalty of \p Penalty. Insert a line break if \p NewLine is \c true. 1021 void addNextStateToQueue(unsigned Penalty, StateNode *PreviousNode, 1022 bool NewLine, unsigned *Count, QueueType *Queue) { 1023 if (NewLine && !Indenter->canBreak(PreviousNode->State)) 1024 return; 1025 if (!NewLine && Indenter->mustBreak(PreviousNode->State)) 1026 return; 1027 1028 StateNode *Node = new (Allocator.Allocate()) 1029 StateNode(PreviousNode->State, NewLine, PreviousNode); 1030 if (!formatChildren(Node->State, NewLine, /*DryRun=*/true, Penalty)) 1031 return; 1032 1033 Penalty += Indenter->addTokenToState(Node->State, NewLine, true); 1034 1035 Queue->push(QueueItem(OrderedPenalty(Penalty, *Count), Node)); 1036 ++(*Count); 1037 } 1038 1039 /// \brief If the \p State's next token is an r_brace closing a nested block, 1040 /// format the nested block before it. 1041 /// 1042 /// Returns \c true if all children could be placed successfully and adapts 1043 /// \p Penalty as well as \p State. If \p DryRun is false, also directly 1044 /// creates changes using \c Whitespaces. 1045 /// 1046 /// The crucial idea here is that children always get formatted upon 1047 /// encountering the closing brace right after the nested block. Now, if we 1048 /// are currently trying to keep the "}" on the same line (i.e. \p NewLine is 1049 /// \c false), the entire block has to be kept on the same line (which is only 1050 /// possible if it fits on the line, only contains a single statement, etc. 1051 /// 1052 /// If \p NewLine is true, we format the nested block on separate lines, i.e. 1053 /// break after the "{", format all lines with correct indentation and the put 1054 /// the closing "}" on yet another new line. 1055 /// 1056 /// This enables us to keep the simple structure of the 1057 /// \c UnwrappedLineFormatter, where we only have two options for each token: 1058 /// break or don't break. 1059 bool formatChildren(LineState &State, bool NewLine, bool DryRun, 1060 unsigned &Penalty) { 1061 FormatToken &Previous = *State.NextToken->Previous; 1062 const FormatToken *LBrace = State.NextToken->getPreviousNonComment(); 1063 if (!LBrace || LBrace->isNot(tok::l_brace) || 1064 LBrace->BlockKind != BK_Block || Previous.Children.size() == 0) 1065 // The previous token does not open a block. Nothing to do. We don't 1066 // assert so that we can simply call this function for all tokens. 1067 return true; 1068 1069 if (NewLine) { 1070 int AdditionalIndent = State.Stack.back().Indent - 1071 Previous.Children[0]->Level * Style.IndentWidth; 1072 Penalty += format(Previous.Children, DryRun, AdditionalIndent, 1073 /*FixBadIndentation=*/true); 1074 return true; 1075 } 1076 1077 // Cannot merge multiple statements into a single line. 1078 if (Previous.Children.size() > 1) 1079 return false; 1080 1081 // We can't put the closing "}" on a line with a trailing comment. 1082 if (Previous.Children[0]->Last->isTrailingComment()) 1083 return false; 1084 1085 if (!DryRun) { 1086 Whitespaces->replaceWhitespace( 1087 *Previous.Children[0]->First, 1088 /*Newlines=*/0, /*IndentLevel=*/0, /*Spaces=*/1, 1089 /*StartOfTokenColumn=*/State.Column, State.Line->InPPDirective); 1090 } 1091 Penalty += format(*Previous.Children[0], State.Column + 1, DryRun); 1092 1093 State.Column += 1 + Previous.Children[0]->Last->TotalLength; 1094 return true; 1095 } 1096 1097 ContinuationIndenter *Indenter; 1098 WhitespaceManager *Whitespaces; 1099 FormatStyle Style; 1100 LineJoiner Joiner; 1101 1102 llvm::SpecificBumpPtrAllocator<StateNode> Allocator; 1103 }; 1104 1105 class FormatTokenLexer { 1106 public: 1107 FormatTokenLexer(Lexer &Lex, SourceManager &SourceMgr, FormatStyle &Style, 1108 encoding::Encoding Encoding) 1109 : FormatTok(NULL), IsFirstToken(true), GreaterStashed(false), Column(0), 1110 TrailingWhitespace(0), Lex(Lex), SourceMgr(SourceMgr), Style(Style), 1111 IdentTable(getFormattingLangOpts()), Encoding(Encoding) { 1112 Lex.SetKeepWhitespaceMode(true); 1113 } 1114 1115 ArrayRef<FormatToken *> lex() { 1116 assert(Tokens.empty()); 1117 do { 1118 Tokens.push_back(getNextToken()); 1119 tryMergePreviousTokens(); 1120 } while (Tokens.back()->Tok.isNot(tok::eof)); 1121 return Tokens; 1122 } 1123 1124 IdentifierTable &getIdentTable() { return IdentTable; } 1125 1126 private: 1127 void tryMergePreviousTokens() { 1128 if (tryMerge_TMacro()) 1129 return; 1130 1131 if (Style.Language == FormatStyle::LK_JavaScript) { 1132 static tok::TokenKind JSIdentity[] = { tok::equalequal, tok::equal }; 1133 static tok::TokenKind JSNotIdentity[] = { tok::exclaimequal, tok::equal }; 1134 static tok::TokenKind JSShiftEqual[] = { tok::greater, tok::greater, 1135 tok::greaterequal }; 1136 // FIXME: We probably need to change token type to mimic operator with the 1137 // correct priority. 1138 if (tryMergeTokens(JSIdentity)) 1139 return; 1140 if (tryMergeTokens(JSNotIdentity)) 1141 return; 1142 if (tryMergeTokens(JSShiftEqual)) 1143 return; 1144 } 1145 } 1146 1147 bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds) { 1148 if (Tokens.size() < Kinds.size()) 1149 return false; 1150 1151 SmallVectorImpl<FormatToken *>::const_iterator First = 1152 Tokens.end() - Kinds.size(); 1153 if (!First[0]->is(Kinds[0])) 1154 return false; 1155 unsigned AddLength = 0; 1156 for (unsigned i = 1; i < Kinds.size(); ++i) { 1157 if (!First[i]->is(Kinds[i]) || First[i]->WhitespaceRange.getBegin() != 1158 First[i]->WhitespaceRange.getEnd()) 1159 return false; 1160 AddLength += First[i]->TokenText.size(); 1161 } 1162 Tokens.resize(Tokens.size() - Kinds.size() + 1); 1163 First[0]->TokenText = StringRef(First[0]->TokenText.data(), 1164 First[0]->TokenText.size() + AddLength); 1165 First[0]->ColumnWidth += AddLength; 1166 return true; 1167 } 1168 1169 bool tryMerge_TMacro() { 1170 if (Tokens.size() < 4) 1171 return false; 1172 FormatToken *Last = Tokens.back(); 1173 if (!Last->is(tok::r_paren)) 1174 return false; 1175 1176 FormatToken *String = Tokens[Tokens.size() - 2]; 1177 if (!String->is(tok::string_literal) || String->IsMultiline) 1178 return false; 1179 1180 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) 1181 return false; 1182 1183 FormatToken *Macro = Tokens[Tokens.size() - 4]; 1184 if (Macro->TokenText != "_T") 1185 return false; 1186 1187 const char *Start = Macro->TokenText.data(); 1188 const char *End = Last->TokenText.data() + Last->TokenText.size(); 1189 String->TokenText = StringRef(Start, End - Start); 1190 String->IsFirst = Macro->IsFirst; 1191 String->LastNewlineOffset = Macro->LastNewlineOffset; 1192 String->WhitespaceRange = Macro->WhitespaceRange; 1193 String->OriginalColumn = Macro->OriginalColumn; 1194 String->ColumnWidth = encoding::columnWidthWithTabs( 1195 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); 1196 1197 Tokens.pop_back(); 1198 Tokens.pop_back(); 1199 Tokens.pop_back(); 1200 Tokens.back() = String; 1201 return true; 1202 } 1203 1204 FormatToken *getNextToken() { 1205 if (GreaterStashed) { 1206 // Create a synthesized second '>' token. 1207 // FIXME: Increment Column and set OriginalColumn. 1208 Token Greater = FormatTok->Tok; 1209 FormatTok = new (Allocator.Allocate()) FormatToken; 1210 FormatTok->Tok = Greater; 1211 SourceLocation GreaterLocation = 1212 FormatTok->Tok.getLocation().getLocWithOffset(1); 1213 FormatTok->WhitespaceRange = 1214 SourceRange(GreaterLocation, GreaterLocation); 1215 FormatTok->TokenText = ">"; 1216 FormatTok->ColumnWidth = 1; 1217 GreaterStashed = false; 1218 return FormatTok; 1219 } 1220 1221 FormatTok = new (Allocator.Allocate()) FormatToken; 1222 readRawToken(*FormatTok); 1223 SourceLocation WhitespaceStart = 1224 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); 1225 FormatTok->IsFirst = IsFirstToken; 1226 IsFirstToken = false; 1227 1228 // Consume and record whitespace until we find a significant token. 1229 unsigned WhitespaceLength = TrailingWhitespace; 1230 while (FormatTok->Tok.is(tok::unknown)) { 1231 for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) { 1232 switch (FormatTok->TokenText[i]) { 1233 case '\n': 1234 ++FormatTok->NewlinesBefore; 1235 // FIXME: This is technically incorrect, as it could also 1236 // be a literal backslash at the end of the line. 1237 if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' && 1238 (FormatTok->TokenText[i - 1] != '\r' || i == 1 || 1239 FormatTok->TokenText[i - 2] != '\\'))) 1240 FormatTok->HasUnescapedNewline = true; 1241 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 1242 Column = 0; 1243 break; 1244 case '\r': 1245 case '\f': 1246 case '\v': 1247 Column = 0; 1248 break; 1249 case ' ': 1250 ++Column; 1251 break; 1252 case '\t': 1253 Column += Style.TabWidth - Column % Style.TabWidth; 1254 break; 1255 case '\\': 1256 ++Column; 1257 if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' && 1258 FormatTok->TokenText[i + 1] != '\n')) 1259 FormatTok->Type = TT_ImplicitStringLiteral; 1260 break; 1261 default: 1262 FormatTok->Type = TT_ImplicitStringLiteral; 1263 ++Column; 1264 break; 1265 } 1266 } 1267 1268 if (FormatTok->Type == TT_ImplicitStringLiteral) 1269 break; 1270 WhitespaceLength += FormatTok->Tok.getLength(); 1271 1272 readRawToken(*FormatTok); 1273 } 1274 1275 // In case the token starts with escaped newlines, we want to 1276 // take them into account as whitespace - this pattern is quite frequent 1277 // in macro definitions. 1278 // FIXME: Add a more explicit test. 1279 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' && 1280 FormatTok->TokenText[1] == '\n') { 1281 // FIXME: ++FormatTok->NewlinesBefore is missing... 1282 WhitespaceLength += 2; 1283 Column = 0; 1284 FormatTok->TokenText = FormatTok->TokenText.substr(2); 1285 } 1286 1287 FormatTok->WhitespaceRange = SourceRange( 1288 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); 1289 1290 FormatTok->OriginalColumn = Column; 1291 1292 TrailingWhitespace = 0; 1293 if (FormatTok->Tok.is(tok::comment)) { 1294 // FIXME: Add the trimmed whitespace to Column. 1295 StringRef UntrimmedText = FormatTok->TokenText; 1296 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); 1297 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); 1298 } else if (FormatTok->Tok.is(tok::raw_identifier)) { 1299 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); 1300 FormatTok->Tok.setIdentifierInfo(&Info); 1301 FormatTok->Tok.setKind(Info.getTokenID()); 1302 } else if (FormatTok->Tok.is(tok::greatergreater)) { 1303 FormatTok->Tok.setKind(tok::greater); 1304 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1305 GreaterStashed = true; 1306 } 1307 1308 // Now FormatTok is the next non-whitespace token. 1309 1310 StringRef Text = FormatTok->TokenText; 1311 size_t FirstNewlinePos = Text.find('\n'); 1312 if (FirstNewlinePos == StringRef::npos) { 1313 // FIXME: ColumnWidth actually depends on the start column, we need to 1314 // take this into account when the token is moved. 1315 FormatTok->ColumnWidth = 1316 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); 1317 Column += FormatTok->ColumnWidth; 1318 } else { 1319 FormatTok->IsMultiline = true; 1320 // FIXME: ColumnWidth actually depends on the start column, we need to 1321 // take this into account when the token is moved. 1322 FormatTok->ColumnWidth = encoding::columnWidthWithTabs( 1323 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); 1324 1325 // The last line of the token always starts in column 0. 1326 // Thus, the length can be precomputed even in the presence of tabs. 1327 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( 1328 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, 1329 Encoding); 1330 Column = FormatTok->LastLineColumnWidth; 1331 } 1332 1333 return FormatTok; 1334 } 1335 1336 FormatToken *FormatTok; 1337 bool IsFirstToken; 1338 bool GreaterStashed; 1339 unsigned Column; 1340 unsigned TrailingWhitespace; 1341 Lexer &Lex; 1342 SourceManager &SourceMgr; 1343 FormatStyle &Style; 1344 IdentifierTable IdentTable; 1345 encoding::Encoding Encoding; 1346 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator; 1347 SmallVector<FormatToken *, 16> Tokens; 1348 1349 void readRawToken(FormatToken &Tok) { 1350 Lex.LexFromRawLexer(Tok.Tok); 1351 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), 1352 Tok.Tok.getLength()); 1353 // For formatting, treat unterminated string literals like normal string 1354 // literals. 1355 if (Tok.is(tok::unknown) && !Tok.TokenText.empty() && 1356 Tok.TokenText[0] == '"') { 1357 Tok.Tok.setKind(tok::string_literal); 1358 Tok.IsUnterminatedLiteral = true; 1359 } 1360 } 1361 }; 1362 1363 static StringRef getLanguageName(FormatStyle::LanguageKind Language) { 1364 switch (Language) { 1365 case FormatStyle::LK_Cpp: 1366 return "C++"; 1367 case FormatStyle::LK_JavaScript: 1368 return "JavaScript"; 1369 case FormatStyle::LK_Proto: 1370 return "Proto"; 1371 default: 1372 return "Unknown"; 1373 } 1374 } 1375 1376 class Formatter : public UnwrappedLineConsumer { 1377 public: 1378 Formatter(const FormatStyle &Style, Lexer &Lex, SourceManager &SourceMgr, 1379 const std::vector<CharSourceRange> &Ranges) 1380 : Style(Style), Lex(Lex), SourceMgr(SourceMgr), 1381 Whitespaces(SourceMgr, Style, inputUsesCRLF(Lex.getBuffer())), 1382 Ranges(Ranges.begin(), Ranges.end()), UnwrappedLines(1), 1383 Encoding(encoding::detectEncoding(Lex.getBuffer())) { 1384 DEBUG(llvm::dbgs() << "File encoding: " 1385 << (Encoding == encoding::Encoding_UTF8 ? "UTF8" 1386 : "unknown") 1387 << "\n"); 1388 DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language) 1389 << "\n"); 1390 } 1391 1392 tooling::Replacements format() { 1393 tooling::Replacements Result; 1394 FormatTokenLexer Tokens(Lex, SourceMgr, Style, Encoding); 1395 1396 UnwrappedLineParser Parser(Style, Tokens.lex(), *this); 1397 bool StructuralError = Parser.parse(); 1398 assert(UnwrappedLines.rbegin()->empty()); 1399 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE; 1400 ++Run) { 1401 DEBUG(llvm::dbgs() << "Run " << Run << "...\n"); 1402 SmallVector<AnnotatedLine *, 16> AnnotatedLines; 1403 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) { 1404 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i])); 1405 } 1406 tooling::Replacements RunResult = 1407 format(AnnotatedLines, StructuralError, Tokens); 1408 DEBUG({ 1409 llvm::dbgs() << "Replacements for run " << Run << ":\n"; 1410 for (tooling::Replacements::iterator I = RunResult.begin(), 1411 E = RunResult.end(); 1412 I != E; ++I) { 1413 llvm::dbgs() << I->toString() << "\n"; 1414 } 1415 }); 1416 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1417 delete AnnotatedLines[i]; 1418 } 1419 Result.insert(RunResult.begin(), RunResult.end()); 1420 Whitespaces.reset(); 1421 } 1422 return Result; 1423 } 1424 1425 tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1426 bool StructuralError, FormatTokenLexer &Tokens) { 1427 TokenAnnotator Annotator(Style, Tokens.getIdentTable().get("in")); 1428 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1429 Annotator.annotate(*AnnotatedLines[i]); 1430 } 1431 deriveLocalStyle(AnnotatedLines); 1432 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1433 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 1434 } 1435 computeAffectedLines(AnnotatedLines.begin(), AnnotatedLines.end()); 1436 1437 Annotator.setCommentLineLevels(AnnotatedLines); 1438 ContinuationIndenter Indenter(Style, SourceMgr, Whitespaces, Encoding, 1439 BinPackInconclusiveFunctions); 1440 UnwrappedLineFormatter Formatter(&Indenter, &Whitespaces, Style); 1441 Formatter.format(AnnotatedLines, /*DryRun=*/false); 1442 return Whitespaces.generateReplacements(); 1443 } 1444 1445 private: 1446 // Determines which lines are affected by the SourceRanges given as input. 1447 // Returns \c true if at least one line between I and E or one of their 1448 // children is affected. 1449 bool computeAffectedLines(SmallVectorImpl<AnnotatedLine *>::iterator I, 1450 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1451 bool SomeLineAffected = false; 1452 const AnnotatedLine *PreviousLine = NULL; 1453 while (I != E) { 1454 AnnotatedLine *Line = *I; 1455 Line->LeadingEmptyLinesAffected = affectsLeadingEmptyLines(*Line->First); 1456 1457 // If a line is part of a preprocessor directive, it needs to be formatted 1458 // if any token within the directive is affected. 1459 if (Line->InPPDirective) { 1460 FormatToken *Last = Line->Last; 1461 SmallVectorImpl<AnnotatedLine *>::iterator PPEnd = I + 1; 1462 while (PPEnd != E && !(*PPEnd)->First->HasUnescapedNewline) { 1463 Last = (*PPEnd)->Last; 1464 ++PPEnd; 1465 } 1466 1467 if (affectsTokenRange(*Line->First, *Last, 1468 /*IncludeLeadingNewlines=*/false)) { 1469 SomeLineAffected = true; 1470 markAllAsAffected(I, PPEnd); 1471 } 1472 I = PPEnd; 1473 continue; 1474 } 1475 1476 if (nonPPLineAffected(Line, PreviousLine)) 1477 SomeLineAffected = true; 1478 1479 PreviousLine = Line; 1480 ++I; 1481 } 1482 return SomeLineAffected; 1483 } 1484 1485 // Determines whether 'Line' is affected by the SourceRanges given as input. 1486 // Returns \c true if line or one if its children is affected. 1487 bool nonPPLineAffected(AnnotatedLine *Line, 1488 const AnnotatedLine *PreviousLine) { 1489 bool SomeLineAffected = false; 1490 Line->ChildrenAffected = 1491 computeAffectedLines(Line->Children.begin(), Line->Children.end()); 1492 if (Line->ChildrenAffected) 1493 SomeLineAffected = true; 1494 1495 // Stores whether one of the line's tokens is directly affected. 1496 bool SomeTokenAffected = false; 1497 // Stores whether we need to look at the leading newlines of the next token 1498 // in order to determine whether it was affected. 1499 bool IncludeLeadingNewlines = false; 1500 1501 // Stores whether the first child line of any of this line's tokens is 1502 // affected. 1503 bool SomeFirstChildAffected = false; 1504 1505 for (FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) { 1506 // Determine whether 'Tok' was affected. 1507 if (affectsTokenRange(*Tok, *Tok, IncludeLeadingNewlines)) 1508 SomeTokenAffected = true; 1509 1510 // Determine whether the first child of 'Tok' was affected. 1511 if (!Tok->Children.empty() && Tok->Children.front()->Affected) 1512 SomeFirstChildAffected = true; 1513 1514 IncludeLeadingNewlines = Tok->Children.empty(); 1515 } 1516 1517 // Was this line moved, i.e. has it previously been on the same line as an 1518 // affected line? 1519 bool LineMoved = PreviousLine && PreviousLine->Affected && 1520 Line->First->NewlinesBefore == 0; 1521 1522 bool IsContinuedComment = Line->First->is(tok::comment) && 1523 Line->First->Next == NULL && 1524 Line->First->NewlinesBefore < 2 && PreviousLine && 1525 PreviousLine->Affected && 1526 PreviousLine->Last->is(tok::comment); 1527 1528 if (SomeTokenAffected || SomeFirstChildAffected || LineMoved || 1529 IsContinuedComment) { 1530 Line->Affected = true; 1531 SomeLineAffected = true; 1532 } 1533 return SomeLineAffected; 1534 } 1535 1536 // Marks all lines between I and E as well as all their children as affected. 1537 void markAllAsAffected(SmallVectorImpl<AnnotatedLine *>::iterator I, 1538 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1539 while (I != E) { 1540 (*I)->Affected = true; 1541 markAllAsAffected((*I)->Children.begin(), (*I)->Children.end()); 1542 ++I; 1543 } 1544 } 1545 1546 // Returns true if the range from 'First' to 'Last' intersects with one of the 1547 // input ranges. 1548 bool affectsTokenRange(const FormatToken &First, const FormatToken &Last, 1549 bool IncludeLeadingNewlines) { 1550 SourceLocation Start = First.WhitespaceRange.getBegin(); 1551 if (!IncludeLeadingNewlines) 1552 Start = Start.getLocWithOffset(First.LastNewlineOffset); 1553 SourceLocation End = Last.getStartOfNonWhitespace(); 1554 if (Last.TokenText.size() > 0) 1555 End = End.getLocWithOffset(Last.TokenText.size() - 1); 1556 CharSourceRange Range = CharSourceRange::getCharRange(Start, End); 1557 return affectsCharSourceRange(Range); 1558 } 1559 1560 // Returns true if one of the input ranges intersect the leading empty lines 1561 // before 'Tok'. 1562 bool affectsLeadingEmptyLines(const FormatToken &Tok) { 1563 CharSourceRange EmptyLineRange = CharSourceRange::getCharRange( 1564 Tok.WhitespaceRange.getBegin(), 1565 Tok.WhitespaceRange.getBegin().getLocWithOffset(Tok.LastNewlineOffset)); 1566 return affectsCharSourceRange(EmptyLineRange); 1567 } 1568 1569 // Returns true if 'Range' intersects with one of the input ranges. 1570 bool affectsCharSourceRange(const CharSourceRange &Range) { 1571 for (SmallVectorImpl<CharSourceRange>::const_iterator I = Ranges.begin(), 1572 E = Ranges.end(); 1573 I != E; ++I) { 1574 if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), I->getBegin()) && 1575 !SourceMgr.isBeforeInTranslationUnit(I->getEnd(), Range.getBegin())) 1576 return true; 1577 } 1578 return false; 1579 } 1580 1581 static bool inputUsesCRLF(StringRef Text) { 1582 return Text.count('\r') * 2 > Text.count('\n'); 1583 } 1584 1585 void 1586 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1587 unsigned CountBoundToVariable = 0; 1588 unsigned CountBoundToType = 0; 1589 bool HasCpp03IncompatibleFormat = false; 1590 bool HasBinPackedFunction = false; 1591 bool HasOnePerLineFunction = false; 1592 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1593 if (!AnnotatedLines[i]->First->Next) 1594 continue; 1595 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1596 while (Tok->Next) { 1597 if (Tok->Type == TT_PointerOrReference) { 1598 bool SpacesBefore = 1599 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1600 bool SpacesAfter = Tok->Next->WhitespaceRange.getBegin() != 1601 Tok->Next->WhitespaceRange.getEnd(); 1602 if (SpacesBefore && !SpacesAfter) 1603 ++CountBoundToVariable; 1604 else if (!SpacesBefore && SpacesAfter) 1605 ++CountBoundToType; 1606 } 1607 1608 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1609 if (Tok->is(tok::coloncolon) && 1610 Tok->Previous->Type == TT_TemplateOpener) 1611 HasCpp03IncompatibleFormat = true; 1612 if (Tok->Type == TT_TemplateCloser && 1613 Tok->Previous->Type == TT_TemplateCloser) 1614 HasCpp03IncompatibleFormat = true; 1615 } 1616 1617 if (Tok->PackingKind == PPK_BinPacked) 1618 HasBinPackedFunction = true; 1619 if (Tok->PackingKind == PPK_OnePerLine) 1620 HasOnePerLineFunction = true; 1621 1622 Tok = Tok->Next; 1623 } 1624 } 1625 if (Style.DerivePointerBinding) { 1626 if (CountBoundToType > CountBoundToVariable) 1627 Style.PointerBindsToType = true; 1628 else if (CountBoundToType < CountBoundToVariable) 1629 Style.PointerBindsToType = false; 1630 } 1631 if (Style.Standard == FormatStyle::LS_Auto) { 1632 Style.Standard = HasCpp03IncompatibleFormat ? FormatStyle::LS_Cpp11 1633 : FormatStyle::LS_Cpp03; 1634 } 1635 BinPackInconclusiveFunctions = 1636 HasBinPackedFunction || !HasOnePerLineFunction; 1637 } 1638 1639 virtual void consumeUnwrappedLine(const UnwrappedLine &TheLine) { 1640 assert(!UnwrappedLines.empty()); 1641 UnwrappedLines.back().push_back(TheLine); 1642 } 1643 1644 virtual void finishRun() { 1645 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>()); 1646 } 1647 1648 FormatStyle Style; 1649 Lexer &Lex; 1650 SourceManager &SourceMgr; 1651 WhitespaceManager Whitespaces; 1652 SmallVector<CharSourceRange, 8> Ranges; 1653 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines; 1654 1655 encoding::Encoding Encoding; 1656 bool BinPackInconclusiveFunctions; 1657 }; 1658 1659 } // end anonymous namespace 1660 1661 tooling::Replacements reformat(const FormatStyle &Style, Lexer &Lex, 1662 SourceManager &SourceMgr, 1663 std::vector<CharSourceRange> Ranges) { 1664 Formatter formatter(Style, Lex, SourceMgr, Ranges); 1665 return formatter.format(); 1666 } 1667 1668 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1669 std::vector<tooling::Range> Ranges, 1670 StringRef FileName) { 1671 FileManager Files((FileSystemOptions())); 1672 DiagnosticsEngine Diagnostics( 1673 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), 1674 new DiagnosticOptions); 1675 SourceManager SourceMgr(Diagnostics, Files); 1676 llvm::MemoryBuffer *Buf = llvm::MemoryBuffer::getMemBuffer(Code, FileName); 1677 const clang::FileEntry *Entry = 1678 Files.getVirtualFile(FileName, Buf->getBufferSize(), 0); 1679 SourceMgr.overrideFileContents(Entry, Buf); 1680 FileID ID = 1681 SourceMgr.createFileID(Entry, SourceLocation(), clang::SrcMgr::C_User); 1682 Lexer Lex(ID, SourceMgr.getBuffer(ID), SourceMgr, 1683 getFormattingLangOpts(Style.Standard)); 1684 SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID); 1685 std::vector<CharSourceRange> CharRanges; 1686 for (unsigned i = 0, e = Ranges.size(); i != e; ++i) { 1687 SourceLocation Start = StartOfFile.getLocWithOffset(Ranges[i].getOffset()); 1688 SourceLocation End = Start.getLocWithOffset(Ranges[i].getLength()); 1689 CharRanges.push_back(CharSourceRange::getCharRange(Start, End)); 1690 } 1691 return reformat(Style, Lex, SourceMgr, CharRanges); 1692 } 1693 1694 LangOptions getFormattingLangOpts(FormatStyle::LanguageStandard Standard) { 1695 LangOptions LangOpts; 1696 LangOpts.CPlusPlus = 1; 1697 LangOpts.CPlusPlus11 = Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1698 LangOpts.LineComment = 1; 1699 LangOpts.Bool = 1; 1700 LangOpts.ObjC1 = 1; 1701 LangOpts.ObjC2 = 1; 1702 return LangOpts; 1703 } 1704 1705 const char *StyleOptionHelpDescription = 1706 "Coding style, currently supports:\n" 1707 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1708 "Use -style=file to load style configuration from\n" 1709 ".clang-format file located in one of the parent\n" 1710 "directories of the source file (or current\n" 1711 "directory for stdin).\n" 1712 "Use -style=\"{key: value, ...}\" to set specific\n" 1713 "parameters, e.g.:\n" 1714 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1715 1716 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 1717 if (FileName.endswith_lower(".js")) { 1718 return FormatStyle::LK_JavaScript; 1719 } else if (FileName.endswith_lower(".proto") || 1720 FileName.endswith_lower(".protodevel")) { 1721 return FormatStyle::LK_Proto; 1722 } 1723 return FormatStyle::LK_Cpp; 1724 } 1725 1726 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 1727 StringRef FallbackStyle) { 1728 FormatStyle Style = getLLVMStyle(); 1729 Style.Language = getLanguageByFileName(FileName); 1730 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) { 1731 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 1732 << "\" using LLVM style\n"; 1733 return Style; 1734 } 1735 1736 if (StyleName.startswith("{")) { 1737 // Parse YAML/JSON style from the command line. 1738 if (llvm::error_code ec = parseConfiguration(StyleName, &Style)) { 1739 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1740 << FallbackStyle << " style\n"; 1741 } 1742 return Style; 1743 } 1744 1745 if (!StyleName.equals_lower("file")) { 1746 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 1747 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1748 << " style\n"; 1749 return Style; 1750 } 1751 1752 // Look for .clang-format/_clang-format file in the file's parent directories. 1753 SmallString<128> UnsuitableConfigFiles; 1754 SmallString<128> Path(FileName); 1755 llvm::sys::fs::make_absolute(Path); 1756 for (StringRef Directory = Path; !Directory.empty(); 1757 Directory = llvm::sys::path::parent_path(Directory)) { 1758 if (!llvm::sys::fs::is_directory(Directory)) 1759 continue; 1760 SmallString<128> ConfigFile(Directory); 1761 1762 llvm::sys::path::append(ConfigFile, ".clang-format"); 1763 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1764 bool IsFile = false; 1765 // Ignore errors from is_regular_file: we only need to know if we can read 1766 // the file or not. 1767 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1768 1769 if (!IsFile) { 1770 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1771 ConfigFile = Directory; 1772 llvm::sys::path::append(ConfigFile, "_clang-format"); 1773 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1774 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1775 } 1776 1777 if (IsFile) { 1778 OwningPtr<llvm::MemoryBuffer> Text; 1779 if (llvm::error_code ec = 1780 llvm::MemoryBuffer::getFile(ConfigFile.c_str(), Text)) { 1781 llvm::errs() << ec.message() << "\n"; 1782 break; 1783 } 1784 if (llvm::error_code ec = parseConfiguration(Text->getBuffer(), &Style)) { 1785 if (ec == llvm::errc::not_supported) { 1786 if (!UnsuitableConfigFiles.empty()) 1787 UnsuitableConfigFiles.append(", "); 1788 UnsuitableConfigFiles.append(ConfigFile); 1789 continue; 1790 } 1791 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1792 << "\n"; 1793 break; 1794 } 1795 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1796 return Style; 1797 } 1798 } 1799 llvm::errs() << "Can't find usable .clang-format, using " << FallbackStyle 1800 << " style\n"; 1801 if (!UnsuitableConfigFiles.empty()) { 1802 llvm::errs() << "Configuration file(s) do(es) not support " 1803 << getLanguageName(Style.Language) << ": " 1804 << UnsuitableConfigFiles << "\n"; 1805 } 1806 return Style; 1807 } 1808 1809 } // namespace format 1810 } // namespace clang 1811