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 #include "ContinuationIndenter.h" 17 #include "TokenAnnotator.h" 18 #include "UnwrappedLineFormatter.h" 19 #include "UnwrappedLineParser.h" 20 #include "WhitespaceManager.h" 21 #include "clang/Basic/Diagnostic.h" 22 #include "clang/Basic/DiagnosticOptions.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 #define DEBUG_TYPE "format-formatter" 35 36 using clang::format::FormatStyle; 37 38 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string) 39 40 namespace llvm { 41 namespace yaml { 42 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> { 43 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) { 44 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp); 45 IO.enumCase(Value, "Java", FormatStyle::LK_Java); 46 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript); 47 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto); 48 } 49 }; 50 51 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> { 52 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) { 53 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03); 54 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03); 55 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11); 56 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11); 57 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto); 58 } 59 }; 60 61 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> { 62 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) { 63 IO.enumCase(Value, "Never", FormatStyle::UT_Never); 64 IO.enumCase(Value, "false", FormatStyle::UT_Never); 65 IO.enumCase(Value, "Always", FormatStyle::UT_Always); 66 IO.enumCase(Value, "true", FormatStyle::UT_Always); 67 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation); 68 } 69 }; 70 71 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> { 72 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) { 73 IO.enumCase(Value, "None", FormatStyle::SFS_None); 74 IO.enumCase(Value, "false", FormatStyle::SFS_None); 75 IO.enumCase(Value, "All", FormatStyle::SFS_All); 76 IO.enumCase(Value, "true", FormatStyle::SFS_All); 77 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline); 78 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty); 79 } 80 }; 81 82 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> { 83 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) { 84 IO.enumCase(Value, "All", FormatStyle::BOS_All); 85 IO.enumCase(Value, "true", FormatStyle::BOS_All); 86 IO.enumCase(Value, "None", FormatStyle::BOS_None); 87 IO.enumCase(Value, "false", FormatStyle::BOS_None); 88 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment); 89 } 90 }; 91 92 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> { 93 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) { 94 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach); 95 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux); 96 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup); 97 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman); 98 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU); 99 } 100 }; 101 102 template <> 103 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> { 104 static void enumeration(IO &IO, 105 FormatStyle::NamespaceIndentationKind &Value) { 106 IO.enumCase(Value, "None", FormatStyle::NI_None); 107 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner); 108 IO.enumCase(Value, "All", FormatStyle::NI_All); 109 } 110 }; 111 112 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> { 113 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) { 114 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle); 115 IO.enumCase(Value, "Left", FormatStyle::PAS_Left); 116 IO.enumCase(Value, "Right", FormatStyle::PAS_Right); 117 118 // For backward compatibility. 119 IO.enumCase(Value, "true", FormatStyle::PAS_Left); 120 IO.enumCase(Value, "false", FormatStyle::PAS_Right); 121 } 122 }; 123 124 template <> 125 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> { 126 static void enumeration(IO &IO, 127 FormatStyle::SpaceBeforeParensOptions &Value) { 128 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never); 129 IO.enumCase(Value, "ControlStatements", 130 FormatStyle::SBPO_ControlStatements); 131 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always); 132 133 // For backward compatibility. 134 IO.enumCase(Value, "false", FormatStyle::SBPO_Never); 135 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements); 136 } 137 }; 138 139 template <> struct MappingTraits<FormatStyle> { 140 static void mapping(IO &IO, FormatStyle &Style) { 141 // When reading, read the language first, we need it for getPredefinedStyle. 142 IO.mapOptional("Language", Style.Language); 143 144 if (IO.outputting()) { 145 StringRef StylesArray[] = {"LLVM", "Google", "Chromium", 146 "Mozilla", "WebKit", "GNU"}; 147 ArrayRef<StringRef> Styles(StylesArray); 148 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 149 StringRef StyleName(Styles[i]); 150 FormatStyle PredefinedStyle; 151 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) && 152 Style == PredefinedStyle) { 153 IO.mapOptional("# BasedOnStyle", StyleName); 154 break; 155 } 156 } 157 } else { 158 StringRef BasedOnStyle; 159 IO.mapOptional("BasedOnStyle", BasedOnStyle); 160 if (!BasedOnStyle.empty()) { 161 FormatStyle::LanguageKind OldLanguage = Style.Language; 162 FormatStyle::LanguageKind Language = 163 ((FormatStyle *)IO.getContext())->Language; 164 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) { 165 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 166 return; 167 } 168 Style.Language = OldLanguage; 169 } 170 } 171 172 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 173 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket); 174 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 175 IO.mapOptional("AlignOperands", Style.AlignOperands); 176 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 177 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 178 Style.AllowAllParametersOfDeclarationOnNextLine); 179 IO.mapOptional("AllowShortBlocksOnASingleLine", 180 Style.AllowShortBlocksOnASingleLine); 181 IO.mapOptional("AllowShortCaseLabelsOnASingleLine", 182 Style.AllowShortCaseLabelsOnASingleLine); 183 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 184 Style.AllowShortIfStatementsOnASingleLine); 185 IO.mapOptional("AllowShortLoopsOnASingleLine", 186 Style.AllowShortLoopsOnASingleLine); 187 IO.mapOptional("AllowShortFunctionsOnASingleLine", 188 Style.AllowShortFunctionsOnASingleLine); 189 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType", 190 Style.AlwaysBreakAfterDefinitionReturnType); 191 IO.mapOptional("AlwaysBreakTemplateDeclarations", 192 Style.AlwaysBreakTemplateDeclarations); 193 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 194 Style.AlwaysBreakBeforeMultilineStrings); 195 IO.mapOptional("BreakBeforeBinaryOperators", 196 Style.BreakBeforeBinaryOperators); 197 IO.mapOptional("BreakBeforeTernaryOperators", 198 Style.BreakBeforeTernaryOperators); 199 IO.mapOptional("BreakConstructorInitializersBeforeComma", 200 Style.BreakConstructorInitializersBeforeComma); 201 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 202 IO.mapOptional("BinPackArguments", Style.BinPackArguments); 203 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 204 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 205 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 206 IO.mapOptional("ConstructorInitializerIndentWidth", 207 Style.ConstructorInitializerIndentWidth); 208 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment); 209 IO.mapOptional("ExperimentalAutoDetectBinPacking", 210 Style.ExperimentalAutoDetectBinPacking); 211 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 212 IO.mapOptional("IndentWrappedFunctionNames", 213 Style.IndentWrappedFunctionNames); 214 IO.mapOptional("IndentFunctionDeclarationAfterType", 215 Style.IndentWrappedFunctionNames); 216 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 217 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks", 218 Style.KeepEmptyLinesAtTheStartOfBlocks); 219 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 220 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth); 221 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty); 222 IO.mapOptional("ObjCSpaceBeforeProtocolList", 223 Style.ObjCSpaceBeforeProtocolList); 224 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 225 Style.PenaltyBreakBeforeFirstCallParameter); 226 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 227 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 228 IO.mapOptional("PenaltyBreakFirstLessLess", 229 Style.PenaltyBreakFirstLessLess); 230 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 231 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 232 Style.PenaltyReturnTypeOnItsOwnLine); 233 IO.mapOptional("PointerAlignment", Style.PointerAlignment); 234 IO.mapOptional("SpacesBeforeTrailingComments", 235 Style.SpacesBeforeTrailingComments); 236 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 237 IO.mapOptional("Standard", Style.Standard); 238 IO.mapOptional("IndentWidth", Style.IndentWidth); 239 IO.mapOptional("TabWidth", Style.TabWidth); 240 IO.mapOptional("UseTab", Style.UseTab); 241 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 242 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 243 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets); 244 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 245 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 246 IO.mapOptional("SpacesInCStyleCastParentheses", 247 Style.SpacesInCStyleCastParentheses); 248 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast); 249 IO.mapOptional("SpacesInContainerLiterals", 250 Style.SpacesInContainerLiterals); 251 IO.mapOptional("SpaceBeforeAssignmentOperators", 252 Style.SpaceBeforeAssignmentOperators); 253 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 254 IO.mapOptional("CommentPragmas", Style.CommentPragmas); 255 IO.mapOptional("ForEachMacros", Style.ForEachMacros); 256 257 // For backward compatibility. 258 if (!IO.outputting()) { 259 IO.mapOptional("SpaceAfterControlStatementKeyword", 260 Style.SpaceBeforeParens); 261 IO.mapOptional("PointerBindsToType", Style.PointerAlignment); 262 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment); 263 } 264 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens); 265 IO.mapOptional("DisableFormat", Style.DisableFormat); 266 } 267 }; 268 269 // Allows to read vector<FormatStyle> while keeping default values. 270 // IO.getContext() should contain a pointer to the FormatStyle structure, that 271 // will be used to get default values for missing keys. 272 // If the first element has no Language specified, it will be treated as the 273 // default one for the following elements. 274 template <> struct DocumentListTraits<std::vector<FormatStyle>> { 275 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) { 276 return Seq.size(); 277 } 278 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq, 279 size_t Index) { 280 if (Index >= Seq.size()) { 281 assert(Index == Seq.size()); 282 FormatStyle Template; 283 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) { 284 Template = Seq[0]; 285 } else { 286 Template = *((const FormatStyle *)IO.getContext()); 287 Template.Language = FormatStyle::LK_None; 288 } 289 Seq.resize(Index + 1, Template); 290 } 291 return Seq[Index]; 292 } 293 }; 294 } 295 } 296 297 namespace clang { 298 namespace format { 299 300 const std::error_category &getParseCategory() { 301 static ParseErrorCategory C; 302 return C; 303 } 304 std::error_code make_error_code(ParseError e) { 305 return std::error_code(static_cast<int>(e), getParseCategory()); 306 } 307 308 const char *ParseErrorCategory::name() const LLVM_NOEXCEPT { 309 return "clang-format.parse_error"; 310 } 311 312 std::string ParseErrorCategory::message(int EV) const { 313 switch (static_cast<ParseError>(EV)) { 314 case ParseError::Success: 315 return "Success"; 316 case ParseError::Error: 317 return "Invalid argument"; 318 case ParseError::Unsuitable: 319 return "Unsuitable"; 320 } 321 llvm_unreachable("unexpected parse error"); 322 } 323 324 FormatStyle getLLVMStyle() { 325 FormatStyle LLVMStyle; 326 LLVMStyle.Language = FormatStyle::LK_Cpp; 327 LLVMStyle.AccessModifierOffset = -2; 328 LLVMStyle.AlignEscapedNewlinesLeft = false; 329 LLVMStyle.AlignAfterOpenBracket = true; 330 LLVMStyle.AlignOperands = true; 331 LLVMStyle.AlignTrailingComments = true; 332 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 333 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All; 334 LLVMStyle.AllowShortBlocksOnASingleLine = false; 335 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false; 336 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 337 LLVMStyle.AllowShortLoopsOnASingleLine = false; 338 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = false; 339 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 340 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 341 LLVMStyle.BinPackParameters = true; 342 LLVMStyle.BinPackArguments = true; 343 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None; 344 LLVMStyle.BreakBeforeTernaryOperators = true; 345 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 346 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 347 LLVMStyle.ColumnLimit = 80; 348 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; 349 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 350 LLVMStyle.ConstructorInitializerIndentWidth = 4; 351 LLVMStyle.ContinuationIndentWidth = 4; 352 LLVMStyle.Cpp11BracedListStyle = true; 353 LLVMStyle.DerivePointerAlignment = false; 354 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 355 LLVMStyle.ForEachMacros.push_back("foreach"); 356 LLVMStyle.ForEachMacros.push_back("Q_FOREACH"); 357 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH"); 358 LLVMStyle.IndentCaseLabels = false; 359 LLVMStyle.IndentWrappedFunctionNames = false; 360 LLVMStyle.IndentWidth = 2; 361 LLVMStyle.TabWidth = 8; 362 LLVMStyle.MaxEmptyLinesToKeep = 1; 363 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true; 364 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 365 LLVMStyle.ObjCBlockIndentWidth = 2; 366 LLVMStyle.ObjCSpaceAfterProperty = false; 367 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 368 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right; 369 LLVMStyle.SpacesBeforeTrailingComments = 1; 370 LLVMStyle.Standard = FormatStyle::LS_Cpp11; 371 LLVMStyle.UseTab = FormatStyle::UT_Never; 372 LLVMStyle.SpacesInParentheses = false; 373 LLVMStyle.SpacesInSquareBrackets = false; 374 LLVMStyle.SpaceInEmptyParentheses = false; 375 LLVMStyle.SpacesInContainerLiterals = true; 376 LLVMStyle.SpacesInCStyleCastParentheses = false; 377 LLVMStyle.SpaceAfterCStyleCast = false; 378 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements; 379 LLVMStyle.SpaceBeforeAssignmentOperators = true; 380 LLVMStyle.SpacesInAngles = false; 381 382 LLVMStyle.PenaltyBreakComment = 300; 383 LLVMStyle.PenaltyBreakFirstLessLess = 120; 384 LLVMStyle.PenaltyBreakString = 1000; 385 LLVMStyle.PenaltyExcessCharacter = 1000000; 386 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 387 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 388 389 LLVMStyle.DisableFormat = false; 390 391 return LLVMStyle; 392 } 393 394 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { 395 FormatStyle GoogleStyle = getLLVMStyle(); 396 GoogleStyle.Language = Language; 397 398 GoogleStyle.AccessModifierOffset = -1; 399 GoogleStyle.AlignEscapedNewlinesLeft = true; 400 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 401 GoogleStyle.AllowShortLoopsOnASingleLine = true; 402 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 403 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 404 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 405 GoogleStyle.DerivePointerAlignment = true; 406 GoogleStyle.IndentCaseLabels = true; 407 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false; 408 GoogleStyle.ObjCSpaceAfterProperty = false; 409 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 410 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left; 411 GoogleStyle.SpacesBeforeTrailingComments = 2; 412 GoogleStyle.Standard = FormatStyle::LS_Auto; 413 414 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 415 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 416 417 if (Language == FormatStyle::LK_Java) { 418 GoogleStyle.AlignAfterOpenBracket = false; 419 GoogleStyle.AlignOperands = false; 420 GoogleStyle.AlignTrailingComments = false; 421 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 422 GoogleStyle.AllowShortIfStatementsOnASingleLine = false; 423 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 424 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment; 425 GoogleStyle.ColumnLimit = 100; 426 GoogleStyle.SpaceAfterCStyleCast = true; 427 GoogleStyle.SpacesBeforeTrailingComments = 1; 428 } else if (Language == FormatStyle::LK_JavaScript) { 429 GoogleStyle.BreakBeforeTernaryOperators = false; 430 GoogleStyle.MaxEmptyLinesToKeep = 3; 431 GoogleStyle.SpacesInContainerLiterals = false; 432 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 433 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 434 } else if (Language == FormatStyle::LK_Proto) { 435 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None; 436 GoogleStyle.SpacesInContainerLiterals = false; 437 } 438 439 return GoogleStyle; 440 } 441 442 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) { 443 FormatStyle ChromiumStyle = getGoogleStyle(Language); 444 if (Language == FormatStyle::LK_Java) { 445 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true; 446 ChromiumStyle.IndentWidth = 4; 447 ChromiumStyle.ContinuationIndentWidth = 8; 448 } else { 449 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 450 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 451 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 452 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 453 ChromiumStyle.BinPackParameters = false; 454 ChromiumStyle.DerivePointerAlignment = false; 455 } 456 return ChromiumStyle; 457 } 458 459 FormatStyle getMozillaStyle() { 460 FormatStyle MozillaStyle = getLLVMStyle(); 461 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 462 MozillaStyle.Cpp11BracedListStyle = false; 463 MozillaStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 464 MozillaStyle.DerivePointerAlignment = true; 465 MozillaStyle.IndentCaseLabels = true; 466 MozillaStyle.ObjCSpaceAfterProperty = true; 467 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 468 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 469 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left; 470 MozillaStyle.Standard = FormatStyle::LS_Cpp03; 471 return MozillaStyle; 472 } 473 474 FormatStyle getWebKitStyle() { 475 FormatStyle Style = getLLVMStyle(); 476 Style.AccessModifierOffset = -4; 477 Style.AlignAfterOpenBracket = false; 478 Style.AlignOperands = false; 479 Style.AlignTrailingComments = false; 480 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 481 Style.BreakBeforeBraces = FormatStyle::BS_Stroustrup; 482 Style.BreakConstructorInitializersBeforeComma = true; 483 Style.Cpp11BracedListStyle = false; 484 Style.ColumnLimit = 0; 485 Style.IndentWidth = 4; 486 Style.NamespaceIndentation = FormatStyle::NI_Inner; 487 Style.ObjCBlockIndentWidth = 4; 488 Style.ObjCSpaceAfterProperty = true; 489 Style.PointerAlignment = FormatStyle::PAS_Left; 490 Style.Standard = FormatStyle::LS_Cpp03; 491 return Style; 492 } 493 494 FormatStyle getGNUStyle() { 495 FormatStyle Style = getLLVMStyle(); 496 Style.AlwaysBreakAfterDefinitionReturnType = true; 497 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 498 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 499 Style.BreakBeforeTernaryOperators = true; 500 Style.Cpp11BracedListStyle = false; 501 Style.ColumnLimit = 79; 502 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 503 Style.Standard = FormatStyle::LS_Cpp03; 504 return Style; 505 } 506 507 FormatStyle getNoStyle() { 508 FormatStyle NoStyle = getLLVMStyle(); 509 NoStyle.DisableFormat = true; 510 return NoStyle; 511 } 512 513 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 514 FormatStyle *Style) { 515 if (Name.equals_lower("llvm")) { 516 *Style = getLLVMStyle(); 517 } else if (Name.equals_lower("chromium")) { 518 *Style = getChromiumStyle(Language); 519 } else if (Name.equals_lower("mozilla")) { 520 *Style = getMozillaStyle(); 521 } else if (Name.equals_lower("google")) { 522 *Style = getGoogleStyle(Language); 523 } else if (Name.equals_lower("webkit")) { 524 *Style = getWebKitStyle(); 525 } else if (Name.equals_lower("gnu")) { 526 *Style = getGNUStyle(); 527 } else if (Name.equals_lower("none")) { 528 *Style = getNoStyle(); 529 } else { 530 return false; 531 } 532 533 Style->Language = Language; 534 return true; 535 } 536 537 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 538 assert(Style); 539 FormatStyle::LanguageKind Language = Style->Language; 540 assert(Language != FormatStyle::LK_None); 541 if (Text.trim().empty()) 542 return make_error_code(ParseError::Error); 543 544 std::vector<FormatStyle> Styles; 545 llvm::yaml::Input Input(Text); 546 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 547 // values for the fields, keys for which are missing from the configuration. 548 // Mapping also uses the context to get the language to find the correct 549 // base style. 550 Input.setContext(Style); 551 Input >> Styles; 552 if (Input.error()) 553 return Input.error(); 554 555 for (unsigned i = 0; i < Styles.size(); ++i) { 556 // Ensures that only the first configuration can skip the Language option. 557 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 558 return make_error_code(ParseError::Error); 559 // Ensure that each language is configured at most once. 560 for (unsigned j = 0; j < i; ++j) { 561 if (Styles[i].Language == Styles[j].Language) { 562 DEBUG(llvm::dbgs() 563 << "Duplicate languages in the config file on positions " << j 564 << " and " << i << "\n"); 565 return make_error_code(ParseError::Error); 566 } 567 } 568 } 569 // Look for a suitable configuration starting from the end, so we can 570 // find the configuration for the specific language first, and the default 571 // configuration (which can only be at slot 0) after it. 572 for (int i = Styles.size() - 1; i >= 0; --i) { 573 if (Styles[i].Language == Language || 574 Styles[i].Language == FormatStyle::LK_None) { 575 *Style = Styles[i]; 576 Style->Language = Language; 577 return make_error_code(ParseError::Success); 578 } 579 } 580 return make_error_code(ParseError::Unsuitable); 581 } 582 583 std::string configurationAsText(const FormatStyle &Style) { 584 std::string Text; 585 llvm::raw_string_ostream Stream(Text); 586 llvm::yaml::Output Output(Stream); 587 // We use the same mapping method for input and output, so we need a non-const 588 // reference here. 589 FormatStyle NonConstStyle = Style; 590 Output << NonConstStyle; 591 return Stream.str(); 592 } 593 594 namespace { 595 596 class FormatTokenLexer { 597 public: 598 FormatTokenLexer(SourceManager &SourceMgr, FileID ID, FormatStyle &Style, 599 encoding::Encoding Encoding) 600 : FormatTok(nullptr), IsFirstToken(true), GreaterStashed(false), 601 LessStashed(false), Column(0), TrailingWhitespace(0), 602 SourceMgr(SourceMgr), ID(ID), Style(Style), 603 IdentTable(getFormattingLangOpts(Style)), Keywords(IdentTable), 604 Encoding(Encoding), FirstInLineIndex(0), FormattingDisabled(false) { 605 Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr, 606 getFormattingLangOpts(Style))); 607 Lex->SetKeepWhitespaceMode(true); 608 609 for (const std::string &ForEachMacro : Style.ForEachMacros) 610 ForEachMacros.push_back(&IdentTable.get(ForEachMacro)); 611 std::sort(ForEachMacros.begin(), ForEachMacros.end()); 612 } 613 614 ArrayRef<FormatToken *> lex() { 615 assert(Tokens.empty()); 616 assert(FirstInLineIndex == 0); 617 do { 618 Tokens.push_back(getNextToken()); 619 tryMergePreviousTokens(); 620 if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline) 621 FirstInLineIndex = Tokens.size() - 1; 622 } while (Tokens.back()->Tok.isNot(tok::eof)); 623 return Tokens; 624 } 625 626 const AdditionalKeywords &getKeywords() { return Keywords; } 627 628 private: 629 void tryMergePreviousTokens() { 630 if (tryMerge_TMacro()) 631 return; 632 if (tryMergeConflictMarkers()) 633 return; 634 if (tryMergeLessLess()) 635 return; 636 637 if (Style.Language == FormatStyle::LK_JavaScript) { 638 if (tryMergeJSRegexLiteral()) 639 return; 640 if (tryMergeEscapeSequence()) 641 return; 642 if (tryMergeTemplateString()) 643 return; 644 645 static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal}; 646 static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal, 647 tok::equal}; 648 static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater, 649 tok::greaterequal}; 650 static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater}; 651 // FIXME: We probably need to change token type to mimic operator with the 652 // correct priority. 653 if (tryMergeTokens(JSIdentity)) 654 return; 655 if (tryMergeTokens(JSNotIdentity)) 656 return; 657 if (tryMergeTokens(JSShiftEqual)) 658 return; 659 if (tryMergeTokens(JSRightArrow)) 660 return; 661 } 662 } 663 664 bool tryMergeLessLess() { 665 // Merge X,less,less,Y into X,lessless,Y unless X or Y is less. 666 if (Tokens.size() < 3) 667 return false; 668 669 bool FourthTokenIsLess = false; 670 if (Tokens.size() > 3) 671 FourthTokenIsLess = (Tokens.end() - 4)[0]->is(tok::less); 672 673 auto First = Tokens.end() - 3; 674 if (First[2]->is(tok::less) || First[1]->isNot(tok::less) || 675 First[0]->isNot(tok::less) || FourthTokenIsLess) 676 return false; 677 678 // Only merge if there currently is no whitespace between the two "<". 679 if (First[1]->WhitespaceRange.getBegin() != 680 First[1]->WhitespaceRange.getEnd()) 681 return false; 682 683 First[0]->Tok.setKind(tok::lessless); 684 First[0]->TokenText = "<<"; 685 First[0]->ColumnWidth += 1; 686 Tokens.erase(Tokens.end() - 2); 687 return true; 688 } 689 690 bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds) { 691 if (Tokens.size() < Kinds.size()) 692 return false; 693 694 SmallVectorImpl<FormatToken *>::const_iterator First = 695 Tokens.end() - Kinds.size(); 696 if (!First[0]->is(Kinds[0])) 697 return false; 698 unsigned AddLength = 0; 699 for (unsigned i = 1; i < Kinds.size(); ++i) { 700 if (!First[i]->is(Kinds[i]) || 701 First[i]->WhitespaceRange.getBegin() != 702 First[i]->WhitespaceRange.getEnd()) 703 return false; 704 AddLength += First[i]->TokenText.size(); 705 } 706 Tokens.resize(Tokens.size() - Kinds.size() + 1); 707 First[0]->TokenText = StringRef(First[0]->TokenText.data(), 708 First[0]->TokenText.size() + AddLength); 709 First[0]->ColumnWidth += AddLength; 710 return true; 711 } 712 713 // Tries to merge an escape sequence, i.e. a "\\" and the following 714 // character. Use e.g. inside JavaScript regex literals. 715 bool tryMergeEscapeSequence() { 716 if (Tokens.size() < 2) 717 return false; 718 FormatToken *Previous = Tokens[Tokens.size() - 2]; 719 if (Previous->isNot(tok::unknown) || Previous->TokenText != "\\") 720 return false; 721 ++Previous->ColumnWidth; 722 StringRef Text = Previous->TokenText; 723 Previous->TokenText = StringRef(Text.data(), Text.size() + 1); 724 resetLexer(SourceMgr.getFileOffset(Tokens.back()->Tok.getLocation()) + 1); 725 Tokens.resize(Tokens.size() - 1); 726 Column = Previous->OriginalColumn + Previous->ColumnWidth; 727 return true; 728 } 729 730 // Try to determine whether the current token ends a JavaScript regex literal. 731 // We heuristically assume that this is a regex literal if we find two 732 // unescaped slashes on a line and the token before the first slash is one of 733 // "(;,{}![:?", a binary operator or 'return', as those cannot be followed by 734 // a division. 735 bool tryMergeJSRegexLiteral() { 736 if (Tokens.size() < 2) 737 return false; 738 // If a regex literal ends in "\//", this gets represented by an unknown 739 // token "\" and a comment. 740 bool MightEndWithEscapedSlash = 741 Tokens.back()->is(tok::comment) && 742 Tokens.back()->TokenText.startswith("//") && 743 Tokens[Tokens.size() - 2]->TokenText == "\\"; 744 if (!MightEndWithEscapedSlash && 745 (Tokens.back()->isNot(tok::slash) || 746 (Tokens[Tokens.size() - 2]->is(tok::unknown) && 747 Tokens[Tokens.size() - 2]->TokenText == "\\"))) 748 return false; 749 unsigned TokenCount = 0; 750 unsigned LastColumn = Tokens.back()->OriginalColumn; 751 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) { 752 ++TokenCount; 753 if (I[0]->is(tok::slash) && I + 1 != E && 754 (I[1]->isOneOf(tok::l_paren, tok::semi, tok::l_brace, tok::r_brace, 755 tok::exclaim, tok::l_square, tok::colon, tok::comma, 756 tok::question, tok::kw_return) || 757 I[1]->isBinaryOperator())) { 758 if (MightEndWithEscapedSlash) { 759 // This regex literal ends in '\//'. Skip past the '//' of the last 760 // token and re-start lexing from there. 761 SourceLocation Loc = Tokens.back()->Tok.getLocation(); 762 resetLexer(SourceMgr.getFileOffset(Loc) + 2); 763 } 764 Tokens.resize(Tokens.size() - TokenCount); 765 Tokens.back()->Tok.setKind(tok::unknown); 766 Tokens.back()->Type = TT_RegexLiteral; 767 Tokens.back()->ColumnWidth += LastColumn - I[0]->OriginalColumn; 768 return true; 769 } 770 771 // There can't be a newline inside a regex literal. 772 if (I[0]->NewlinesBefore > 0) 773 return false; 774 } 775 return false; 776 } 777 778 bool tryMergeTemplateString() { 779 if (Tokens.size() < 2) 780 return false; 781 782 FormatToken *EndBacktick = Tokens.back(); 783 if (!(EndBacktick->is(tok::unknown) && EndBacktick->TokenText == "`")) 784 return false; 785 786 unsigned TokenCount = 0; 787 bool IsMultiline = false; 788 unsigned EndColumnInFirstLine = 0; 789 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; I++) { 790 ++TokenCount; 791 if (I[0]->NewlinesBefore > 0 || I[0]->IsMultiline) 792 IsMultiline = true; 793 794 // If there was a preceding template string, this must be the start of a 795 // template string, not the end. 796 if (I[0]->is(TT_TemplateString)) 797 return false; 798 799 if (I[0]->isNot(tok::unknown) || I[0]->TokenText != "`") { 800 // Keep track of the rhs offset of the last token to wrap across lines - 801 // its the rhs offset of the first line of the template string, used to 802 // determine its width. 803 if (I[0]->IsMultiline) 804 EndColumnInFirstLine = I[0]->OriginalColumn + I[0]->ColumnWidth; 805 // If the token has newlines, the token before it (if it exists) is the 806 // rhs end of the previous line. 807 if (I[0]->NewlinesBefore > 0 && (I + 1 != E)) 808 EndColumnInFirstLine = I[1]->OriginalColumn + I[1]->ColumnWidth; 809 810 continue; 811 } 812 813 Tokens.resize(Tokens.size() - TokenCount); 814 Tokens.back()->Type = TT_TemplateString; 815 const char *EndOffset = EndBacktick->TokenText.data() + 1; 816 Tokens.back()->TokenText = 817 StringRef(Tokens.back()->TokenText.data(), 818 EndOffset - Tokens.back()->TokenText.data()); 819 if (IsMultiline) { 820 // ColumnWidth is from backtick to last token in line. 821 // LastLineColumnWidth is 0 to backtick. 822 // x = `some content 823 // until here`; 824 Tokens.back()->ColumnWidth = 825 EndColumnInFirstLine - Tokens.back()->OriginalColumn; 826 Tokens.back()->LastLineColumnWidth = EndBacktick->OriginalColumn; 827 Tokens.back()->IsMultiline = true; 828 } else { 829 // Token simply spans from start to end, +1 for the ` itself. 830 Tokens.back()->ColumnWidth = 831 EndBacktick->OriginalColumn - Tokens.back()->OriginalColumn + 1; 832 } 833 return true; 834 } 835 return false; 836 } 837 838 bool tryMerge_TMacro() { 839 if (Tokens.size() < 4) 840 return false; 841 FormatToken *Last = Tokens.back(); 842 if (!Last->is(tok::r_paren)) 843 return false; 844 845 FormatToken *String = Tokens[Tokens.size() - 2]; 846 if (!String->is(tok::string_literal) || String->IsMultiline) 847 return false; 848 849 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) 850 return false; 851 852 FormatToken *Macro = Tokens[Tokens.size() - 4]; 853 if (Macro->TokenText != "_T") 854 return false; 855 856 const char *Start = Macro->TokenText.data(); 857 const char *End = Last->TokenText.data() + Last->TokenText.size(); 858 String->TokenText = StringRef(Start, End - Start); 859 String->IsFirst = Macro->IsFirst; 860 String->LastNewlineOffset = Macro->LastNewlineOffset; 861 String->WhitespaceRange = Macro->WhitespaceRange; 862 String->OriginalColumn = Macro->OriginalColumn; 863 String->ColumnWidth = encoding::columnWidthWithTabs( 864 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); 865 866 Tokens.pop_back(); 867 Tokens.pop_back(); 868 Tokens.pop_back(); 869 Tokens.back() = String; 870 return true; 871 } 872 873 bool tryMergeConflictMarkers() { 874 if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof)) 875 return false; 876 877 // Conflict lines look like: 878 // <marker> <text from the vcs> 879 // For example: 880 // >>>>>>> /file/in/file/system at revision 1234 881 // 882 // We merge all tokens in a line that starts with a conflict marker 883 // into a single token with a special token type that the unwrapped line 884 // parser will use to correctly rebuild the underlying code. 885 886 FileID ID; 887 // Get the position of the first token in the line. 888 unsigned FirstInLineOffset; 889 std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc( 890 Tokens[FirstInLineIndex]->getStartOfNonWhitespace()); 891 StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer(); 892 // Calculate the offset of the start of the current line. 893 auto LineOffset = Buffer.rfind('\n', FirstInLineOffset); 894 if (LineOffset == StringRef::npos) { 895 LineOffset = 0; 896 } else { 897 ++LineOffset; 898 } 899 900 auto FirstSpace = Buffer.find_first_of(" \n", LineOffset); 901 StringRef LineStart; 902 if (FirstSpace == StringRef::npos) { 903 LineStart = Buffer.substr(LineOffset); 904 } else { 905 LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset); 906 } 907 908 TokenType Type = TT_Unknown; 909 if (LineStart == "<<<<<<<" || LineStart == ">>>>") { 910 Type = TT_ConflictStart; 911 } else if (LineStart == "|||||||" || LineStart == "=======" || 912 LineStart == "====") { 913 Type = TT_ConflictAlternative; 914 } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") { 915 Type = TT_ConflictEnd; 916 } 917 918 if (Type != TT_Unknown) { 919 FormatToken *Next = Tokens.back(); 920 921 Tokens.resize(FirstInLineIndex + 1); 922 // We do not need to build a complete token here, as we will skip it 923 // during parsing anyway (as we must not touch whitespace around conflict 924 // markers). 925 Tokens.back()->Type = Type; 926 Tokens.back()->Tok.setKind(tok::kw___unknown_anytype); 927 928 Tokens.push_back(Next); 929 return true; 930 } 931 932 return false; 933 } 934 935 FormatToken *getStashedToken() { 936 // Create a synthesized second '>' or '<' token. 937 Token Tok = FormatTok->Tok; 938 StringRef TokenText = FormatTok->TokenText; 939 940 unsigned OriginalColumn = FormatTok->OriginalColumn; 941 FormatTok = new (Allocator.Allocate()) FormatToken; 942 FormatTok->Tok = Tok; 943 SourceLocation TokLocation = 944 FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1); 945 FormatTok->Tok.setLocation(TokLocation); 946 FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation); 947 FormatTok->TokenText = TokenText; 948 FormatTok->ColumnWidth = 1; 949 FormatTok->OriginalColumn = OriginalColumn + 1; 950 951 return FormatTok; 952 } 953 954 FormatToken *getNextToken() { 955 if (GreaterStashed) { 956 GreaterStashed = false; 957 return getStashedToken(); 958 } 959 if (LessStashed) { 960 LessStashed = false; 961 return getStashedToken(); 962 } 963 964 FormatTok = new (Allocator.Allocate()) FormatToken; 965 readRawToken(*FormatTok); 966 SourceLocation WhitespaceStart = 967 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); 968 FormatTok->IsFirst = IsFirstToken; 969 IsFirstToken = false; 970 971 // Consume and record whitespace until we find a significant token. 972 unsigned WhitespaceLength = TrailingWhitespace; 973 while (FormatTok->Tok.is(tok::unknown)) { 974 // FIXME: This miscounts tok:unknown tokens that are not just 975 // whitespace, e.g. a '`' character. 976 for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) { 977 switch (FormatTok->TokenText[i]) { 978 case '\n': 979 ++FormatTok->NewlinesBefore; 980 // FIXME: This is technically incorrect, as it could also 981 // be a literal backslash at the end of the line. 982 if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' && 983 (FormatTok->TokenText[i - 1] != '\r' || i == 1 || 984 FormatTok->TokenText[i - 2] != '\\'))) 985 FormatTok->HasUnescapedNewline = true; 986 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 987 Column = 0; 988 break; 989 case '\r': 990 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 991 Column = 0; 992 break; 993 case '\f': 994 case '\v': 995 Column = 0; 996 break; 997 case ' ': 998 ++Column; 999 break; 1000 case '\t': 1001 Column += Style.TabWidth - Column % Style.TabWidth; 1002 break; 1003 case '\\': 1004 if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' && 1005 FormatTok->TokenText[i + 1] != '\n')) 1006 FormatTok->Type = TT_ImplicitStringLiteral; 1007 break; 1008 default: 1009 FormatTok->Type = TT_ImplicitStringLiteral; 1010 ++Column; 1011 break; 1012 } 1013 } 1014 1015 if (FormatTok->is(TT_ImplicitStringLiteral)) 1016 break; 1017 WhitespaceLength += FormatTok->Tok.getLength(); 1018 1019 readRawToken(*FormatTok); 1020 } 1021 1022 // In case the token starts with escaped newlines, we want to 1023 // take them into account as whitespace - this pattern is quite frequent 1024 // in macro definitions. 1025 // FIXME: Add a more explicit test. 1026 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' && 1027 FormatTok->TokenText[1] == '\n') { 1028 ++FormatTok->NewlinesBefore; 1029 WhitespaceLength += 2; 1030 Column = 0; 1031 FormatTok->TokenText = FormatTok->TokenText.substr(2); 1032 } 1033 1034 FormatTok->WhitespaceRange = SourceRange( 1035 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); 1036 1037 FormatTok->OriginalColumn = Column; 1038 1039 TrailingWhitespace = 0; 1040 if (FormatTok->Tok.is(tok::comment)) { 1041 // FIXME: Add the trimmed whitespace to Column. 1042 StringRef UntrimmedText = FormatTok->TokenText; 1043 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); 1044 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); 1045 } else if (FormatTok->Tok.is(tok::raw_identifier)) { 1046 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); 1047 FormatTok->Tok.setIdentifierInfo(&Info); 1048 FormatTok->Tok.setKind(Info.getTokenID()); 1049 if (Style.Language == FormatStyle::LK_Java && 1050 FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete)) { 1051 FormatTok->Tok.setKind(tok::identifier); 1052 FormatTok->Tok.setIdentifierInfo(nullptr); 1053 } 1054 } else if (FormatTok->Tok.is(tok::greatergreater)) { 1055 FormatTok->Tok.setKind(tok::greater); 1056 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1057 GreaterStashed = true; 1058 } else if (FormatTok->Tok.is(tok::lessless)) { 1059 FormatTok->Tok.setKind(tok::less); 1060 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1061 LessStashed = true; 1062 } 1063 1064 // Now FormatTok is the next non-whitespace token. 1065 1066 StringRef Text = FormatTok->TokenText; 1067 size_t FirstNewlinePos = Text.find('\n'); 1068 if (FirstNewlinePos == StringRef::npos) { 1069 // FIXME: ColumnWidth actually depends on the start column, we need to 1070 // take this into account when the token is moved. 1071 FormatTok->ColumnWidth = 1072 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); 1073 Column += FormatTok->ColumnWidth; 1074 } else { 1075 FormatTok->IsMultiline = true; 1076 // FIXME: ColumnWidth actually depends on the start column, we need to 1077 // take this into account when the token is moved. 1078 FormatTok->ColumnWidth = encoding::columnWidthWithTabs( 1079 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); 1080 1081 // The last line of the token always starts in column 0. 1082 // Thus, the length can be precomputed even in the presence of tabs. 1083 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( 1084 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, 1085 Encoding); 1086 Column = FormatTok->LastLineColumnWidth; 1087 } 1088 1089 FormatTok->IsForEachMacro = 1090 std::binary_search(ForEachMacros.begin(), ForEachMacros.end(), 1091 FormatTok->Tok.getIdentifierInfo()); 1092 1093 return FormatTok; 1094 } 1095 1096 FormatToken *FormatTok; 1097 bool IsFirstToken; 1098 bool GreaterStashed, LessStashed; 1099 unsigned Column; 1100 unsigned TrailingWhitespace; 1101 std::unique_ptr<Lexer> Lex; 1102 SourceManager &SourceMgr; 1103 FileID ID; 1104 FormatStyle &Style; 1105 IdentifierTable IdentTable; 1106 AdditionalKeywords Keywords; 1107 encoding::Encoding Encoding; 1108 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator; 1109 // Index (in 'Tokens') of the last token that starts a new line. 1110 unsigned FirstInLineIndex; 1111 SmallVector<FormatToken *, 16> Tokens; 1112 SmallVector<IdentifierInfo *, 8> ForEachMacros; 1113 1114 bool FormattingDisabled; 1115 1116 void readRawToken(FormatToken &Tok) { 1117 Lex->LexFromRawLexer(Tok.Tok); 1118 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), 1119 Tok.Tok.getLength()); 1120 // For formatting, treat unterminated string literals like normal string 1121 // literals. 1122 if (Tok.is(tok::unknown)) { 1123 if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') { 1124 Tok.Tok.setKind(tok::string_literal); 1125 Tok.IsUnterminatedLiteral = true; 1126 } else if (Style.Language == FormatStyle::LK_JavaScript && 1127 Tok.TokenText == "''") { 1128 Tok.Tok.setKind(tok::char_constant); 1129 } 1130 } 1131 1132 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" || 1133 Tok.TokenText == "/* clang-format on */")) { 1134 FormattingDisabled = false; 1135 } 1136 1137 Tok.Finalized = FormattingDisabled; 1138 1139 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" || 1140 Tok.TokenText == "/* clang-format off */")) { 1141 FormattingDisabled = true; 1142 } 1143 } 1144 1145 void resetLexer(unsigned Offset) { 1146 StringRef Buffer = SourceMgr.getBufferData(ID); 1147 Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID), 1148 getFormattingLangOpts(Style), Buffer.begin(), 1149 Buffer.begin() + Offset, Buffer.end())); 1150 Lex->SetKeepWhitespaceMode(true); 1151 } 1152 }; 1153 1154 static StringRef getLanguageName(FormatStyle::LanguageKind Language) { 1155 switch (Language) { 1156 case FormatStyle::LK_Cpp: 1157 return "C++"; 1158 case FormatStyle::LK_Java: 1159 return "Java"; 1160 case FormatStyle::LK_JavaScript: 1161 return "JavaScript"; 1162 case FormatStyle::LK_Proto: 1163 return "Proto"; 1164 default: 1165 return "Unknown"; 1166 } 1167 } 1168 1169 class Formatter : public UnwrappedLineConsumer { 1170 public: 1171 Formatter(const FormatStyle &Style, SourceManager &SourceMgr, FileID ID, 1172 ArrayRef<CharSourceRange> Ranges) 1173 : Style(Style), ID(ID), SourceMgr(SourceMgr), 1174 Whitespaces(SourceMgr, Style, 1175 inputUsesCRLF(SourceMgr.getBufferData(ID))), 1176 Ranges(Ranges.begin(), Ranges.end()), UnwrappedLines(1), 1177 Encoding(encoding::detectEncoding(SourceMgr.getBufferData(ID))) { 1178 DEBUG(llvm::dbgs() << "File encoding: " 1179 << (Encoding == encoding::Encoding_UTF8 ? "UTF8" 1180 : "unknown") 1181 << "\n"); 1182 DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language) 1183 << "\n"); 1184 } 1185 1186 tooling::Replacements format() { 1187 tooling::Replacements Result; 1188 FormatTokenLexer Tokens(SourceMgr, ID, Style, Encoding); 1189 1190 UnwrappedLineParser Parser(Style, Tokens.getKeywords(), Tokens.lex(), 1191 *this); 1192 bool StructuralError = Parser.parse(); 1193 assert(UnwrappedLines.rbegin()->empty()); 1194 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE; 1195 ++Run) { 1196 DEBUG(llvm::dbgs() << "Run " << Run << "...\n"); 1197 SmallVector<AnnotatedLine *, 16> AnnotatedLines; 1198 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) { 1199 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i])); 1200 } 1201 tooling::Replacements RunResult = 1202 format(AnnotatedLines, StructuralError, Tokens); 1203 DEBUG({ 1204 llvm::dbgs() << "Replacements for run " << Run << ":\n"; 1205 for (tooling::Replacements::iterator I = RunResult.begin(), 1206 E = RunResult.end(); 1207 I != E; ++I) { 1208 llvm::dbgs() << I->toString() << "\n"; 1209 } 1210 }); 1211 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1212 delete AnnotatedLines[i]; 1213 } 1214 Result.insert(RunResult.begin(), RunResult.end()); 1215 Whitespaces.reset(); 1216 } 1217 return Result; 1218 } 1219 1220 tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1221 bool StructuralError, FormatTokenLexer &Tokens) { 1222 TokenAnnotator Annotator(Style, Tokens.getKeywords()); 1223 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1224 Annotator.annotate(*AnnotatedLines[i]); 1225 } 1226 deriveLocalStyle(AnnotatedLines); 1227 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1228 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 1229 } 1230 computeAffectedLines(AnnotatedLines.begin(), AnnotatedLines.end()); 1231 1232 Annotator.setCommentLineLevels(AnnotatedLines); 1233 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), SourceMgr, 1234 Whitespaces, Encoding, 1235 BinPackInconclusiveFunctions); 1236 UnwrappedLineFormatter Formatter(&Indenter, &Whitespaces, Style, 1237 Tokens.getKeywords()); 1238 Formatter.format(AnnotatedLines, /*DryRun=*/false); 1239 return Whitespaces.generateReplacements(); 1240 } 1241 1242 private: 1243 // Determines which lines are affected by the SourceRanges given as input. 1244 // Returns \c true if at least one line between I and E or one of their 1245 // children is affected. 1246 bool computeAffectedLines(SmallVectorImpl<AnnotatedLine *>::iterator I, 1247 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1248 bool SomeLineAffected = false; 1249 const AnnotatedLine *PreviousLine = nullptr; 1250 while (I != E) { 1251 AnnotatedLine *Line = *I; 1252 Line->LeadingEmptyLinesAffected = affectsLeadingEmptyLines(*Line->First); 1253 1254 // If a line is part of a preprocessor directive, it needs to be formatted 1255 // if any token within the directive is affected. 1256 if (Line->InPPDirective) { 1257 FormatToken *Last = Line->Last; 1258 SmallVectorImpl<AnnotatedLine *>::iterator PPEnd = I + 1; 1259 while (PPEnd != E && !(*PPEnd)->First->HasUnescapedNewline) { 1260 Last = (*PPEnd)->Last; 1261 ++PPEnd; 1262 } 1263 1264 if (affectsTokenRange(*Line->First, *Last, 1265 /*IncludeLeadingNewlines=*/false)) { 1266 SomeLineAffected = true; 1267 markAllAsAffected(I, PPEnd); 1268 } 1269 I = PPEnd; 1270 continue; 1271 } 1272 1273 if (nonPPLineAffected(Line, PreviousLine)) 1274 SomeLineAffected = true; 1275 1276 PreviousLine = Line; 1277 ++I; 1278 } 1279 return SomeLineAffected; 1280 } 1281 1282 // Determines whether 'Line' is affected by the SourceRanges given as input. 1283 // Returns \c true if line or one if its children is affected. 1284 bool nonPPLineAffected(AnnotatedLine *Line, 1285 const AnnotatedLine *PreviousLine) { 1286 bool SomeLineAffected = false; 1287 Line->ChildrenAffected = 1288 computeAffectedLines(Line->Children.begin(), Line->Children.end()); 1289 if (Line->ChildrenAffected) 1290 SomeLineAffected = true; 1291 1292 // Stores whether one of the line's tokens is directly affected. 1293 bool SomeTokenAffected = false; 1294 // Stores whether we need to look at the leading newlines of the next token 1295 // in order to determine whether it was affected. 1296 bool IncludeLeadingNewlines = false; 1297 1298 // Stores whether the first child line of any of this line's tokens is 1299 // affected. 1300 bool SomeFirstChildAffected = false; 1301 1302 for (FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) { 1303 // Determine whether 'Tok' was affected. 1304 if (affectsTokenRange(*Tok, *Tok, IncludeLeadingNewlines)) 1305 SomeTokenAffected = true; 1306 1307 // Determine whether the first child of 'Tok' was affected. 1308 if (!Tok->Children.empty() && Tok->Children.front()->Affected) 1309 SomeFirstChildAffected = true; 1310 1311 IncludeLeadingNewlines = Tok->Children.empty(); 1312 } 1313 1314 // Was this line moved, i.e. has it previously been on the same line as an 1315 // affected line? 1316 bool LineMoved = PreviousLine && PreviousLine->Affected && 1317 Line->First->NewlinesBefore == 0; 1318 1319 bool IsContinuedComment = 1320 Line->First->is(tok::comment) && Line->First->Next == nullptr && 1321 Line->First->NewlinesBefore < 2 && PreviousLine && 1322 PreviousLine->Affected && PreviousLine->Last->is(tok::comment); 1323 1324 if (SomeTokenAffected || SomeFirstChildAffected || LineMoved || 1325 IsContinuedComment) { 1326 Line->Affected = true; 1327 SomeLineAffected = true; 1328 } 1329 return SomeLineAffected; 1330 } 1331 1332 // Marks all lines between I and E as well as all their children as affected. 1333 void markAllAsAffected(SmallVectorImpl<AnnotatedLine *>::iterator I, 1334 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1335 while (I != E) { 1336 (*I)->Affected = true; 1337 markAllAsAffected((*I)->Children.begin(), (*I)->Children.end()); 1338 ++I; 1339 } 1340 } 1341 1342 // Returns true if the range from 'First' to 'Last' intersects with one of the 1343 // input ranges. 1344 bool affectsTokenRange(const FormatToken &First, const FormatToken &Last, 1345 bool IncludeLeadingNewlines) { 1346 SourceLocation Start = First.WhitespaceRange.getBegin(); 1347 if (!IncludeLeadingNewlines) 1348 Start = Start.getLocWithOffset(First.LastNewlineOffset); 1349 SourceLocation End = Last.getStartOfNonWhitespace(); 1350 End = End.getLocWithOffset(Last.TokenText.size()); 1351 CharSourceRange Range = CharSourceRange::getCharRange(Start, End); 1352 return affectsCharSourceRange(Range); 1353 } 1354 1355 // Returns true if one of the input ranges intersect the leading empty lines 1356 // before 'Tok'. 1357 bool affectsLeadingEmptyLines(const FormatToken &Tok) { 1358 CharSourceRange EmptyLineRange = CharSourceRange::getCharRange( 1359 Tok.WhitespaceRange.getBegin(), 1360 Tok.WhitespaceRange.getBegin().getLocWithOffset(Tok.LastNewlineOffset)); 1361 return affectsCharSourceRange(EmptyLineRange); 1362 } 1363 1364 // Returns true if 'Range' intersects with one of the input ranges. 1365 bool affectsCharSourceRange(const CharSourceRange &Range) { 1366 for (SmallVectorImpl<CharSourceRange>::const_iterator I = Ranges.begin(), 1367 E = Ranges.end(); 1368 I != E; ++I) { 1369 if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), I->getBegin()) && 1370 !SourceMgr.isBeforeInTranslationUnit(I->getEnd(), Range.getBegin())) 1371 return true; 1372 } 1373 return false; 1374 } 1375 1376 static bool inputUsesCRLF(StringRef Text) { 1377 return Text.count('\r') * 2 > Text.count('\n'); 1378 } 1379 1380 void 1381 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1382 unsigned CountBoundToVariable = 0; 1383 unsigned CountBoundToType = 0; 1384 bool HasCpp03IncompatibleFormat = false; 1385 bool HasBinPackedFunction = false; 1386 bool HasOnePerLineFunction = false; 1387 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1388 if (!AnnotatedLines[i]->First->Next) 1389 continue; 1390 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1391 while (Tok->Next) { 1392 if (Tok->is(TT_PointerOrReference)) { 1393 bool SpacesBefore = 1394 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1395 bool SpacesAfter = Tok->Next->WhitespaceRange.getBegin() != 1396 Tok->Next->WhitespaceRange.getEnd(); 1397 if (SpacesBefore && !SpacesAfter) 1398 ++CountBoundToVariable; 1399 else if (!SpacesBefore && SpacesAfter) 1400 ++CountBoundToType; 1401 } 1402 1403 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1404 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener)) 1405 HasCpp03IncompatibleFormat = true; 1406 if (Tok->is(TT_TemplateCloser) && 1407 Tok->Previous->is(TT_TemplateCloser)) 1408 HasCpp03IncompatibleFormat = true; 1409 } 1410 1411 if (Tok->PackingKind == PPK_BinPacked) 1412 HasBinPackedFunction = true; 1413 if (Tok->PackingKind == PPK_OnePerLine) 1414 HasOnePerLineFunction = true; 1415 1416 Tok = Tok->Next; 1417 } 1418 } 1419 if (Style.DerivePointerAlignment) { 1420 if (CountBoundToType > CountBoundToVariable) 1421 Style.PointerAlignment = FormatStyle::PAS_Left; 1422 else if (CountBoundToType < CountBoundToVariable) 1423 Style.PointerAlignment = FormatStyle::PAS_Right; 1424 } 1425 if (Style.Standard == FormatStyle::LS_Auto) { 1426 Style.Standard = HasCpp03IncompatibleFormat ? FormatStyle::LS_Cpp11 1427 : FormatStyle::LS_Cpp03; 1428 } 1429 BinPackInconclusiveFunctions = 1430 HasBinPackedFunction || !HasOnePerLineFunction; 1431 } 1432 1433 void consumeUnwrappedLine(const UnwrappedLine &TheLine) override { 1434 assert(!UnwrappedLines.empty()); 1435 UnwrappedLines.back().push_back(TheLine); 1436 } 1437 1438 void finishRun() override { 1439 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>()); 1440 } 1441 1442 FormatStyle Style; 1443 FileID ID; 1444 SourceManager &SourceMgr; 1445 WhitespaceManager Whitespaces; 1446 SmallVector<CharSourceRange, 8> Ranges; 1447 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines; 1448 1449 encoding::Encoding Encoding; 1450 bool BinPackInconclusiveFunctions; 1451 }; 1452 1453 } // end anonymous namespace 1454 1455 tooling::Replacements reformat(const FormatStyle &Style, Lexer &Lex, 1456 SourceManager &SourceMgr, 1457 ArrayRef<CharSourceRange> Ranges) { 1458 if (Style.DisableFormat) 1459 return tooling::Replacements(); 1460 return reformat(Style, SourceMgr, 1461 SourceMgr.getFileID(Lex.getSourceLocation()), Ranges); 1462 } 1463 1464 tooling::Replacements reformat(const FormatStyle &Style, 1465 SourceManager &SourceMgr, FileID ID, 1466 ArrayRef<CharSourceRange> Ranges) { 1467 if (Style.DisableFormat) 1468 return tooling::Replacements(); 1469 Formatter formatter(Style, SourceMgr, ID, Ranges); 1470 return formatter.format(); 1471 } 1472 1473 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1474 ArrayRef<tooling::Range> Ranges, 1475 StringRef FileName) { 1476 if (Style.DisableFormat) 1477 return tooling::Replacements(); 1478 1479 FileManager Files((FileSystemOptions())); 1480 DiagnosticsEngine Diagnostics( 1481 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), 1482 new DiagnosticOptions); 1483 SourceManager SourceMgr(Diagnostics, Files); 1484 std::unique_ptr<llvm::MemoryBuffer> Buf = 1485 llvm::MemoryBuffer::getMemBuffer(Code, FileName); 1486 const clang::FileEntry *Entry = 1487 Files.getVirtualFile(FileName, Buf->getBufferSize(), 0); 1488 SourceMgr.overrideFileContents(Entry, std::move(Buf)); 1489 FileID ID = 1490 SourceMgr.createFileID(Entry, SourceLocation(), clang::SrcMgr::C_User); 1491 SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID); 1492 std::vector<CharSourceRange> CharRanges; 1493 for (const tooling::Range &Range : Ranges) { 1494 SourceLocation Start = StartOfFile.getLocWithOffset(Range.getOffset()); 1495 SourceLocation End = Start.getLocWithOffset(Range.getLength()); 1496 CharRanges.push_back(CharSourceRange::getCharRange(Start, End)); 1497 } 1498 return reformat(Style, SourceMgr, ID, CharRanges); 1499 } 1500 1501 LangOptions getFormattingLangOpts(const FormatStyle &Style) { 1502 LangOptions LangOpts; 1503 LangOpts.CPlusPlus = 1; 1504 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1505 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1506 LangOpts.LineComment = 1; 1507 bool AlternativeOperators = Style.Language != FormatStyle::LK_JavaScript && 1508 Style.Language != FormatStyle::LK_Java; 1509 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0; 1510 LangOpts.Bool = 1; 1511 LangOpts.ObjC1 = 1; 1512 LangOpts.ObjC2 = 1; 1513 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally. 1514 return LangOpts; 1515 } 1516 1517 const char *StyleOptionHelpDescription = 1518 "Coding style, currently supports:\n" 1519 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1520 "Use -style=file to load style configuration from\n" 1521 ".clang-format file located in one of the parent\n" 1522 "directories of the source file (or current\n" 1523 "directory for stdin).\n" 1524 "Use -style=\"{key: value, ...}\" to set specific\n" 1525 "parameters, e.g.:\n" 1526 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1527 1528 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 1529 if (FileName.endswith(".java")) { 1530 return FormatStyle::LK_Java; 1531 } else if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) { 1532 // JavaScript or TypeScript. 1533 return FormatStyle::LK_JavaScript; 1534 } else if (FileName.endswith_lower(".proto") || 1535 FileName.endswith_lower(".protodevel")) { 1536 return FormatStyle::LK_Proto; 1537 } 1538 return FormatStyle::LK_Cpp; 1539 } 1540 1541 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 1542 StringRef FallbackStyle) { 1543 FormatStyle Style = getLLVMStyle(); 1544 Style.Language = getLanguageByFileName(FileName); 1545 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) { 1546 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 1547 << "\" using LLVM style\n"; 1548 return Style; 1549 } 1550 1551 if (StyleName.startswith("{")) { 1552 // Parse YAML/JSON style from the command line. 1553 if (std::error_code ec = parseConfiguration(StyleName, &Style)) { 1554 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1555 << FallbackStyle << " style\n"; 1556 } 1557 return Style; 1558 } 1559 1560 if (!StyleName.equals_lower("file")) { 1561 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 1562 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1563 << " style\n"; 1564 return Style; 1565 } 1566 1567 // Look for .clang-format/_clang-format file in the file's parent directories. 1568 SmallString<128> UnsuitableConfigFiles; 1569 SmallString<128> Path(FileName); 1570 llvm::sys::fs::make_absolute(Path); 1571 for (StringRef Directory = Path; !Directory.empty(); 1572 Directory = llvm::sys::path::parent_path(Directory)) { 1573 if (!llvm::sys::fs::is_directory(Directory)) 1574 continue; 1575 SmallString<128> ConfigFile(Directory); 1576 1577 llvm::sys::path::append(ConfigFile, ".clang-format"); 1578 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1579 bool IsFile = false; 1580 // Ignore errors from is_regular_file: we only need to know if we can read 1581 // the file or not. 1582 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1583 1584 if (!IsFile) { 1585 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1586 ConfigFile = Directory; 1587 llvm::sys::path::append(ConfigFile, "_clang-format"); 1588 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1589 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1590 } 1591 1592 if (IsFile) { 1593 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text = 1594 llvm::MemoryBuffer::getFile(ConfigFile.c_str()); 1595 if (std::error_code EC = Text.getError()) { 1596 llvm::errs() << EC.message() << "\n"; 1597 break; 1598 } 1599 if (std::error_code ec = 1600 parseConfiguration(Text.get()->getBuffer(), &Style)) { 1601 if (ec == ParseError::Unsuitable) { 1602 if (!UnsuitableConfigFiles.empty()) 1603 UnsuitableConfigFiles.append(", "); 1604 UnsuitableConfigFiles.append(ConfigFile); 1605 continue; 1606 } 1607 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1608 << "\n"; 1609 break; 1610 } 1611 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1612 return Style; 1613 } 1614 } 1615 llvm::errs() << "Can't find usable .clang-format, using " << FallbackStyle 1616 << " style\n"; 1617 if (!UnsuitableConfigFiles.empty()) { 1618 llvm::errs() << "Configuration file(s) do(es) not support " 1619 << getLanguageName(Style.Language) << ": " 1620 << UnsuitableConfigFiles << "\n"; 1621 } 1622 return Style; 1623 } 1624 1625 } // namespace format 1626 } // namespace clang 1627