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 "clang/Format/Format.h" 17 #include "AffectedRangeManager.h" 18 #include "ContinuationIndenter.h" 19 #include "TokenAnnotator.h" 20 #include "UnwrappedLineFormatter.h" 21 #include "UnwrappedLineParser.h" 22 #include "WhitespaceManager.h" 23 #include "clang/Basic/Diagnostic.h" 24 #include "clang/Basic/DiagnosticOptions.h" 25 #include "clang/Basic/SourceManager.h" 26 #include "clang/Basic/VirtualFileSystem.h" 27 #include "clang/Lex/Lexer.h" 28 #include "llvm/ADT/STLExtras.h" 29 #include "llvm/Support/Allocator.h" 30 #include "llvm/Support/Debug.h" 31 #include "llvm/Support/Path.h" 32 #include "llvm/Support/Regex.h" 33 #include "llvm/Support/YAMLTraits.h" 34 #include <memory> 35 #include <queue> 36 #include <string> 37 38 #define DEBUG_TYPE "format-formatter" 39 40 using clang::format::FormatStyle; 41 42 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string) 43 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory) 44 45 namespace llvm { 46 namespace yaml { 47 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> { 48 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) { 49 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp); 50 IO.enumCase(Value, "Java", FormatStyle::LK_Java); 51 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript); 52 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto); 53 IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen); 54 } 55 }; 56 57 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> { 58 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) { 59 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03); 60 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03); 61 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11); 62 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11); 63 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto); 64 } 65 }; 66 67 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> { 68 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) { 69 IO.enumCase(Value, "Never", FormatStyle::UT_Never); 70 IO.enumCase(Value, "false", FormatStyle::UT_Never); 71 IO.enumCase(Value, "Always", FormatStyle::UT_Always); 72 IO.enumCase(Value, "true", FormatStyle::UT_Always); 73 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation); 74 IO.enumCase(Value, "ForContinuationAndIndentation", 75 FormatStyle::UT_ForContinuationAndIndentation); 76 } 77 }; 78 79 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> { 80 static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) { 81 IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave); 82 IO.enumCase(Value, "Single", FormatStyle::JSQS_Single); 83 IO.enumCase(Value, "Double", FormatStyle::JSQS_Double); 84 } 85 }; 86 87 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> { 88 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) { 89 IO.enumCase(Value, "None", FormatStyle::SFS_None); 90 IO.enumCase(Value, "false", FormatStyle::SFS_None); 91 IO.enumCase(Value, "All", FormatStyle::SFS_All); 92 IO.enumCase(Value, "true", FormatStyle::SFS_All); 93 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline); 94 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty); 95 } 96 }; 97 98 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> { 99 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) { 100 IO.enumCase(Value, "All", FormatStyle::BOS_All); 101 IO.enumCase(Value, "true", FormatStyle::BOS_All); 102 IO.enumCase(Value, "None", FormatStyle::BOS_None); 103 IO.enumCase(Value, "false", FormatStyle::BOS_None); 104 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment); 105 } 106 }; 107 108 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> { 109 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) { 110 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach); 111 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux); 112 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla); 113 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup); 114 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman); 115 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU); 116 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit); 117 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom); 118 } 119 }; 120 121 template <> 122 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> { 123 static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) { 124 IO.enumCase(Value, "None", FormatStyle::RTBS_None); 125 IO.enumCase(Value, "All", FormatStyle::RTBS_All); 126 IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel); 127 IO.enumCase(Value, "TopLevelDefinitions", 128 FormatStyle::RTBS_TopLevelDefinitions); 129 IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions); 130 } 131 }; 132 133 template <> 134 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> { 135 static void 136 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) { 137 IO.enumCase(Value, "None", FormatStyle::DRTBS_None); 138 IO.enumCase(Value, "All", FormatStyle::DRTBS_All); 139 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel); 140 141 // For backward compatibility. 142 IO.enumCase(Value, "false", FormatStyle::DRTBS_None); 143 IO.enumCase(Value, "true", FormatStyle::DRTBS_All); 144 } 145 }; 146 147 template <> 148 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> { 149 static void enumeration(IO &IO, 150 FormatStyle::NamespaceIndentationKind &Value) { 151 IO.enumCase(Value, "None", FormatStyle::NI_None); 152 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner); 153 IO.enumCase(Value, "All", FormatStyle::NI_All); 154 } 155 }; 156 157 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> { 158 static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) { 159 IO.enumCase(Value, "Align", FormatStyle::BAS_Align); 160 IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign); 161 IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak); 162 163 // For backward compatibility. 164 IO.enumCase(Value, "true", FormatStyle::BAS_Align); 165 IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign); 166 } 167 }; 168 169 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> { 170 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) { 171 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle); 172 IO.enumCase(Value, "Left", FormatStyle::PAS_Left); 173 IO.enumCase(Value, "Right", FormatStyle::PAS_Right); 174 175 // For backward compatibility. 176 IO.enumCase(Value, "true", FormatStyle::PAS_Left); 177 IO.enumCase(Value, "false", FormatStyle::PAS_Right); 178 } 179 }; 180 181 template <> 182 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> { 183 static void enumeration(IO &IO, 184 FormatStyle::SpaceBeforeParensOptions &Value) { 185 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never); 186 IO.enumCase(Value, "ControlStatements", 187 FormatStyle::SBPO_ControlStatements); 188 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always); 189 190 // For backward compatibility. 191 IO.enumCase(Value, "false", FormatStyle::SBPO_Never); 192 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements); 193 } 194 }; 195 196 template <> struct MappingTraits<FormatStyle> { 197 static void mapping(IO &IO, FormatStyle &Style) { 198 // When reading, read the language first, we need it for getPredefinedStyle. 199 IO.mapOptional("Language", Style.Language); 200 201 if (IO.outputting()) { 202 StringRef StylesArray[] = {"LLVM", "Google", "Chromium", 203 "Mozilla", "WebKit", "GNU"}; 204 ArrayRef<StringRef> Styles(StylesArray); 205 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 206 StringRef StyleName(Styles[i]); 207 FormatStyle PredefinedStyle; 208 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) && 209 Style == PredefinedStyle) { 210 IO.mapOptional("# BasedOnStyle", StyleName); 211 break; 212 } 213 } 214 } else { 215 StringRef BasedOnStyle; 216 IO.mapOptional("BasedOnStyle", BasedOnStyle); 217 if (!BasedOnStyle.empty()) { 218 FormatStyle::LanguageKind OldLanguage = Style.Language; 219 FormatStyle::LanguageKind Language = 220 ((FormatStyle *)IO.getContext())->Language; 221 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) { 222 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 223 return; 224 } 225 Style.Language = OldLanguage; 226 } 227 } 228 229 // For backward compatibility. 230 if (!IO.outputting()) { 231 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment); 232 IO.mapOptional("IndentFunctionDeclarationAfterType", 233 Style.IndentWrappedFunctionNames); 234 IO.mapOptional("PointerBindsToType", Style.PointerAlignment); 235 IO.mapOptional("SpaceAfterControlStatementKeyword", 236 Style.SpaceBeforeParens); 237 } 238 239 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 240 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket); 241 IO.mapOptional("AlignConsecutiveAssignments", 242 Style.AlignConsecutiveAssignments); 243 IO.mapOptional("AlignConsecutiveDeclarations", 244 Style.AlignConsecutiveDeclarations); 245 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 246 IO.mapOptional("AlignOperands", Style.AlignOperands); 247 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 248 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 249 Style.AllowAllParametersOfDeclarationOnNextLine); 250 IO.mapOptional("AllowShortBlocksOnASingleLine", 251 Style.AllowShortBlocksOnASingleLine); 252 IO.mapOptional("AllowShortCaseLabelsOnASingleLine", 253 Style.AllowShortCaseLabelsOnASingleLine); 254 IO.mapOptional("AllowShortFunctionsOnASingleLine", 255 Style.AllowShortFunctionsOnASingleLine); 256 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 257 Style.AllowShortIfStatementsOnASingleLine); 258 IO.mapOptional("AllowShortLoopsOnASingleLine", 259 Style.AllowShortLoopsOnASingleLine); 260 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType", 261 Style.AlwaysBreakAfterDefinitionReturnType); 262 IO.mapOptional("AlwaysBreakAfterReturnType", 263 Style.AlwaysBreakAfterReturnType); 264 // If AlwaysBreakAfterDefinitionReturnType was specified but 265 // AlwaysBreakAfterReturnType was not, initialize the latter from the 266 // former for backwards compatibility. 267 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None && 268 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) { 269 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All) 270 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 271 else if (Style.AlwaysBreakAfterDefinitionReturnType == 272 FormatStyle::DRTBS_TopLevel) 273 Style.AlwaysBreakAfterReturnType = 274 FormatStyle::RTBS_TopLevelDefinitions; 275 } 276 277 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 278 Style.AlwaysBreakBeforeMultilineStrings); 279 IO.mapOptional("AlwaysBreakTemplateDeclarations", 280 Style.AlwaysBreakTemplateDeclarations); 281 IO.mapOptional("BinPackArguments", Style.BinPackArguments); 282 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 283 IO.mapOptional("BraceWrapping", Style.BraceWrapping); 284 IO.mapOptional("BreakBeforeBinaryOperators", 285 Style.BreakBeforeBinaryOperators); 286 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 287 IO.mapOptional("BreakBeforeTernaryOperators", 288 Style.BreakBeforeTernaryOperators); 289 IO.mapOptional("BreakConstructorInitializersBeforeComma", 290 Style.BreakConstructorInitializersBeforeComma); 291 IO.mapOptional("BreakAfterJavaFieldAnnotations", 292 Style.BreakAfterJavaFieldAnnotations); 293 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals); 294 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 295 IO.mapOptional("CommentPragmas", Style.CommentPragmas); 296 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 297 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 298 IO.mapOptional("ConstructorInitializerIndentWidth", 299 Style.ConstructorInitializerIndentWidth); 300 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 301 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 302 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment); 303 IO.mapOptional("DisableFormat", Style.DisableFormat); 304 IO.mapOptional("ExperimentalAutoDetectBinPacking", 305 Style.ExperimentalAutoDetectBinPacking); 306 IO.mapOptional("ForEachMacros", Style.ForEachMacros); 307 IO.mapOptional("IncludeCategories", Style.IncludeCategories); 308 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex); 309 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 310 IO.mapOptional("IndentWidth", Style.IndentWidth); 311 IO.mapOptional("IndentWrappedFunctionNames", 312 Style.IndentWrappedFunctionNames); 313 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks", 314 Style.KeepEmptyLinesAtTheStartOfBlocks); 315 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin); 316 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd); 317 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 318 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 319 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth); 320 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty); 321 IO.mapOptional("ObjCSpaceBeforeProtocolList", 322 Style.ObjCSpaceBeforeProtocolList); 323 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 324 Style.PenaltyBreakBeforeFirstCallParameter); 325 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 326 IO.mapOptional("PenaltyBreakFirstLessLess", 327 Style.PenaltyBreakFirstLessLess); 328 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 329 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 330 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 331 Style.PenaltyReturnTypeOnItsOwnLine); 332 IO.mapOptional("PointerAlignment", Style.PointerAlignment); 333 IO.mapOptional("ReflowComments", Style.ReflowComments); 334 IO.mapOptional("SortIncludes", Style.SortIncludes); 335 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast); 336 IO.mapOptional("SpaceBeforeAssignmentOperators", 337 Style.SpaceBeforeAssignmentOperators); 338 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens); 339 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 340 IO.mapOptional("SpacesBeforeTrailingComments", 341 Style.SpacesBeforeTrailingComments); 342 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 343 IO.mapOptional("SpacesInContainerLiterals", 344 Style.SpacesInContainerLiterals); 345 IO.mapOptional("SpacesInCStyleCastParentheses", 346 Style.SpacesInCStyleCastParentheses); 347 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 348 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets); 349 IO.mapOptional("Standard", Style.Standard); 350 IO.mapOptional("TabWidth", Style.TabWidth); 351 IO.mapOptional("UseTab", Style.UseTab); 352 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes); 353 } 354 }; 355 356 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> { 357 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) { 358 IO.mapOptional("AfterClass", Wrapping.AfterClass); 359 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement); 360 IO.mapOptional("AfterEnum", Wrapping.AfterEnum); 361 IO.mapOptional("AfterFunction", Wrapping.AfterFunction); 362 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace); 363 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration); 364 IO.mapOptional("AfterStruct", Wrapping.AfterStruct); 365 IO.mapOptional("AfterUnion", Wrapping.AfterUnion); 366 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch); 367 IO.mapOptional("BeforeElse", Wrapping.BeforeElse); 368 IO.mapOptional("IndentBraces", Wrapping.IndentBraces); 369 } 370 }; 371 372 template <> struct MappingTraits<FormatStyle::IncludeCategory> { 373 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) { 374 IO.mapOptional("Regex", Category.Regex); 375 IO.mapOptional("Priority", Category.Priority); 376 } 377 }; 378 379 // Allows to read vector<FormatStyle> while keeping default values. 380 // IO.getContext() should contain a pointer to the FormatStyle structure, that 381 // will be used to get default values for missing keys. 382 // If the first element has no Language specified, it will be treated as the 383 // default one for the following elements. 384 template <> struct DocumentListTraits<std::vector<FormatStyle>> { 385 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) { 386 return Seq.size(); 387 } 388 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq, 389 size_t Index) { 390 if (Index >= Seq.size()) { 391 assert(Index == Seq.size()); 392 FormatStyle Template; 393 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) { 394 Template = Seq[0]; 395 } else { 396 Template = *((const FormatStyle *)IO.getContext()); 397 Template.Language = FormatStyle::LK_None; 398 } 399 Seq.resize(Index + 1, Template); 400 } 401 return Seq[Index]; 402 } 403 }; 404 } // namespace yaml 405 } // namespace llvm 406 407 namespace clang { 408 namespace format { 409 410 const std::error_category &getParseCategory() { 411 static ParseErrorCategory C; 412 return C; 413 } 414 std::error_code make_error_code(ParseError e) { 415 return std::error_code(static_cast<int>(e), getParseCategory()); 416 } 417 418 const char *ParseErrorCategory::name() const LLVM_NOEXCEPT { 419 return "clang-format.parse_error"; 420 } 421 422 std::string ParseErrorCategory::message(int EV) const { 423 switch (static_cast<ParseError>(EV)) { 424 case ParseError::Success: 425 return "Success"; 426 case ParseError::Error: 427 return "Invalid argument"; 428 case ParseError::Unsuitable: 429 return "Unsuitable"; 430 } 431 llvm_unreachable("unexpected parse error"); 432 } 433 434 static FormatStyle expandPresets(const FormatStyle &Style) { 435 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom) 436 return Style; 437 FormatStyle Expanded = Style; 438 Expanded.BraceWrapping = {false, false, false, false, false, false, 439 false, false, false, false, false}; 440 switch (Style.BreakBeforeBraces) { 441 case FormatStyle::BS_Linux: 442 Expanded.BraceWrapping.AfterClass = true; 443 Expanded.BraceWrapping.AfterFunction = true; 444 Expanded.BraceWrapping.AfterNamespace = true; 445 break; 446 case FormatStyle::BS_Mozilla: 447 Expanded.BraceWrapping.AfterClass = true; 448 Expanded.BraceWrapping.AfterEnum = true; 449 Expanded.BraceWrapping.AfterFunction = true; 450 Expanded.BraceWrapping.AfterStruct = true; 451 Expanded.BraceWrapping.AfterUnion = true; 452 break; 453 case FormatStyle::BS_Stroustrup: 454 Expanded.BraceWrapping.AfterFunction = true; 455 Expanded.BraceWrapping.BeforeCatch = true; 456 Expanded.BraceWrapping.BeforeElse = true; 457 break; 458 case FormatStyle::BS_Allman: 459 Expanded.BraceWrapping.AfterClass = true; 460 Expanded.BraceWrapping.AfterControlStatement = true; 461 Expanded.BraceWrapping.AfterEnum = true; 462 Expanded.BraceWrapping.AfterFunction = true; 463 Expanded.BraceWrapping.AfterNamespace = true; 464 Expanded.BraceWrapping.AfterObjCDeclaration = true; 465 Expanded.BraceWrapping.AfterStruct = true; 466 Expanded.BraceWrapping.BeforeCatch = true; 467 Expanded.BraceWrapping.BeforeElse = true; 468 break; 469 case FormatStyle::BS_GNU: 470 Expanded.BraceWrapping = {true, true, true, true, true, true, 471 true, true, true, true, true}; 472 break; 473 case FormatStyle::BS_WebKit: 474 Expanded.BraceWrapping.AfterFunction = true; 475 break; 476 default: 477 break; 478 } 479 return Expanded; 480 } 481 482 FormatStyle getLLVMStyle() { 483 FormatStyle LLVMStyle; 484 LLVMStyle.Language = FormatStyle::LK_Cpp; 485 LLVMStyle.AccessModifierOffset = -2; 486 LLVMStyle.AlignEscapedNewlinesLeft = false; 487 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align; 488 LLVMStyle.AlignOperands = true; 489 LLVMStyle.AlignTrailingComments = true; 490 LLVMStyle.AlignConsecutiveAssignments = false; 491 LLVMStyle.AlignConsecutiveDeclarations = false; 492 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 493 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All; 494 LLVMStyle.AllowShortBlocksOnASingleLine = false; 495 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false; 496 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 497 LLVMStyle.AllowShortLoopsOnASingleLine = false; 498 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None; 499 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None; 500 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 501 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 502 LLVMStyle.BinPackParameters = true; 503 LLVMStyle.BinPackArguments = true; 504 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None; 505 LLVMStyle.BreakBeforeTernaryOperators = true; 506 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 507 LLVMStyle.BraceWrapping = {false, false, false, false, false, false, 508 false, false, false, false, false}; 509 LLVMStyle.BreakAfterJavaFieldAnnotations = false; 510 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 511 LLVMStyle.BreakStringLiterals = true; 512 LLVMStyle.ColumnLimit = 80; 513 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; 514 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 515 LLVMStyle.ConstructorInitializerIndentWidth = 4; 516 LLVMStyle.ContinuationIndentWidth = 4; 517 LLVMStyle.Cpp11BracedListStyle = true; 518 LLVMStyle.DerivePointerAlignment = false; 519 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 520 LLVMStyle.ForEachMacros.push_back("foreach"); 521 LLVMStyle.ForEachMacros.push_back("Q_FOREACH"); 522 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH"); 523 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2}, 524 {"^(<|\"(gtest|isl|json)/)", 3}, 525 {".*", 1}}; 526 LLVMStyle.IncludeIsMainRegex = "$"; 527 LLVMStyle.IndentCaseLabels = false; 528 LLVMStyle.IndentWrappedFunctionNames = false; 529 LLVMStyle.IndentWidth = 2; 530 LLVMStyle.TabWidth = 8; 531 LLVMStyle.MaxEmptyLinesToKeep = 1; 532 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true; 533 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 534 LLVMStyle.ObjCBlockIndentWidth = 2; 535 LLVMStyle.ObjCSpaceAfterProperty = false; 536 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 537 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right; 538 LLVMStyle.SpacesBeforeTrailingComments = 1; 539 LLVMStyle.Standard = FormatStyle::LS_Cpp11; 540 LLVMStyle.UseTab = FormatStyle::UT_Never; 541 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave; 542 LLVMStyle.ReflowComments = true; 543 LLVMStyle.SpacesInParentheses = false; 544 LLVMStyle.SpacesInSquareBrackets = false; 545 LLVMStyle.SpaceInEmptyParentheses = false; 546 LLVMStyle.SpacesInContainerLiterals = true; 547 LLVMStyle.SpacesInCStyleCastParentheses = false; 548 LLVMStyle.SpaceAfterCStyleCast = false; 549 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements; 550 LLVMStyle.SpaceBeforeAssignmentOperators = true; 551 LLVMStyle.SpacesInAngles = false; 552 553 LLVMStyle.PenaltyBreakComment = 300; 554 LLVMStyle.PenaltyBreakFirstLessLess = 120; 555 LLVMStyle.PenaltyBreakString = 1000; 556 LLVMStyle.PenaltyExcessCharacter = 1000000; 557 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 558 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 559 560 LLVMStyle.DisableFormat = false; 561 LLVMStyle.SortIncludes = true; 562 563 return LLVMStyle; 564 } 565 566 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { 567 FormatStyle GoogleStyle = getLLVMStyle(); 568 GoogleStyle.Language = Language; 569 570 GoogleStyle.AccessModifierOffset = -1; 571 GoogleStyle.AlignEscapedNewlinesLeft = true; 572 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 573 GoogleStyle.AllowShortLoopsOnASingleLine = true; 574 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 575 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 576 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 577 GoogleStyle.DerivePointerAlignment = true; 578 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}}; 579 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$"; 580 GoogleStyle.IndentCaseLabels = true; 581 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false; 582 GoogleStyle.ObjCSpaceAfterProperty = false; 583 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 584 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left; 585 GoogleStyle.SpacesBeforeTrailingComments = 2; 586 GoogleStyle.Standard = FormatStyle::LS_Auto; 587 588 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 589 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 590 591 if (Language == FormatStyle::LK_Java) { 592 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 593 GoogleStyle.AlignOperands = false; 594 GoogleStyle.AlignTrailingComments = false; 595 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 596 GoogleStyle.AllowShortIfStatementsOnASingleLine = false; 597 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 598 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment; 599 GoogleStyle.ColumnLimit = 100; 600 GoogleStyle.SpaceAfterCStyleCast = true; 601 GoogleStyle.SpacesBeforeTrailingComments = 1; 602 } else if (Language == FormatStyle::LK_JavaScript) { 603 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak; 604 GoogleStyle.AlignOperands = false; 605 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 606 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 607 GoogleStyle.BreakBeforeTernaryOperators = false; 608 GoogleStyle.CommentPragmas = "@(export|return|see|visibility) "; 609 GoogleStyle.MaxEmptyLinesToKeep = 3; 610 GoogleStyle.SpacesInContainerLiterals = false; 611 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single; 612 } else if (Language == FormatStyle::LK_Proto) { 613 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None; 614 GoogleStyle.SpacesInContainerLiterals = false; 615 } 616 617 return GoogleStyle; 618 } 619 620 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) { 621 FormatStyle ChromiumStyle = getGoogleStyle(Language); 622 if (Language == FormatStyle::LK_Java) { 623 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true; 624 ChromiumStyle.BreakAfterJavaFieldAnnotations = true; 625 ChromiumStyle.ContinuationIndentWidth = 8; 626 ChromiumStyle.IndentWidth = 4; 627 } else { 628 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 629 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 630 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 631 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 632 ChromiumStyle.BinPackParameters = false; 633 ChromiumStyle.DerivePointerAlignment = false; 634 } 635 ChromiumStyle.SortIncludes = false; 636 return ChromiumStyle; 637 } 638 639 FormatStyle getMozillaStyle() { 640 FormatStyle MozillaStyle = getLLVMStyle(); 641 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 642 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 643 MozillaStyle.AlwaysBreakAfterReturnType = 644 FormatStyle::RTBS_TopLevelDefinitions; 645 MozillaStyle.AlwaysBreakAfterDefinitionReturnType = 646 FormatStyle::DRTBS_TopLevel; 647 MozillaStyle.AlwaysBreakTemplateDeclarations = true; 648 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla; 649 MozillaStyle.BreakConstructorInitializersBeforeComma = true; 650 MozillaStyle.ConstructorInitializerIndentWidth = 2; 651 MozillaStyle.ContinuationIndentWidth = 2; 652 MozillaStyle.Cpp11BracedListStyle = false; 653 MozillaStyle.IndentCaseLabels = true; 654 MozillaStyle.ObjCSpaceAfterProperty = true; 655 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 656 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 657 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left; 658 return MozillaStyle; 659 } 660 661 FormatStyle getWebKitStyle() { 662 FormatStyle Style = getLLVMStyle(); 663 Style.AccessModifierOffset = -4; 664 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 665 Style.AlignOperands = false; 666 Style.AlignTrailingComments = false; 667 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 668 Style.BreakBeforeBraces = FormatStyle::BS_WebKit; 669 Style.BreakConstructorInitializersBeforeComma = true; 670 Style.Cpp11BracedListStyle = false; 671 Style.ColumnLimit = 0; 672 Style.IndentWidth = 4; 673 Style.NamespaceIndentation = FormatStyle::NI_Inner; 674 Style.ObjCBlockIndentWidth = 4; 675 Style.ObjCSpaceAfterProperty = true; 676 Style.PointerAlignment = FormatStyle::PAS_Left; 677 Style.Standard = FormatStyle::LS_Cpp03; 678 return Style; 679 } 680 681 FormatStyle getGNUStyle() { 682 FormatStyle Style = getLLVMStyle(); 683 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All; 684 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 685 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 686 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 687 Style.BreakBeforeTernaryOperators = true; 688 Style.Cpp11BracedListStyle = false; 689 Style.ColumnLimit = 79; 690 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 691 Style.Standard = FormatStyle::LS_Cpp03; 692 return Style; 693 } 694 695 FormatStyle getNoStyle() { 696 FormatStyle NoStyle = getLLVMStyle(); 697 NoStyle.DisableFormat = true; 698 NoStyle.SortIncludes = false; 699 return NoStyle; 700 } 701 702 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 703 FormatStyle *Style) { 704 if (Name.equals_lower("llvm")) { 705 *Style = getLLVMStyle(); 706 } else if (Name.equals_lower("chromium")) { 707 *Style = getChromiumStyle(Language); 708 } else if (Name.equals_lower("mozilla")) { 709 *Style = getMozillaStyle(); 710 } else if (Name.equals_lower("google")) { 711 *Style = getGoogleStyle(Language); 712 } else if (Name.equals_lower("webkit")) { 713 *Style = getWebKitStyle(); 714 } else if (Name.equals_lower("gnu")) { 715 *Style = getGNUStyle(); 716 } else if (Name.equals_lower("none")) { 717 *Style = getNoStyle(); 718 } else { 719 return false; 720 } 721 722 Style->Language = Language; 723 return true; 724 } 725 726 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 727 assert(Style); 728 FormatStyle::LanguageKind Language = Style->Language; 729 assert(Language != FormatStyle::LK_None); 730 if (Text.trim().empty()) 731 return make_error_code(ParseError::Error); 732 733 std::vector<FormatStyle> Styles; 734 llvm::yaml::Input Input(Text); 735 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 736 // values for the fields, keys for which are missing from the configuration. 737 // Mapping also uses the context to get the language to find the correct 738 // base style. 739 Input.setContext(Style); 740 Input >> Styles; 741 if (Input.error()) 742 return Input.error(); 743 744 for (unsigned i = 0; i < Styles.size(); ++i) { 745 // Ensures that only the first configuration can skip the Language option. 746 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 747 return make_error_code(ParseError::Error); 748 // Ensure that each language is configured at most once. 749 for (unsigned j = 0; j < i; ++j) { 750 if (Styles[i].Language == Styles[j].Language) { 751 DEBUG(llvm::dbgs() 752 << "Duplicate languages in the config file on positions " << j 753 << " and " << i << "\n"); 754 return make_error_code(ParseError::Error); 755 } 756 } 757 } 758 // Look for a suitable configuration starting from the end, so we can 759 // find the configuration for the specific language first, and the default 760 // configuration (which can only be at slot 0) after it. 761 for (int i = Styles.size() - 1; i >= 0; --i) { 762 if (Styles[i].Language == Language || 763 Styles[i].Language == FormatStyle::LK_None) { 764 *Style = Styles[i]; 765 Style->Language = Language; 766 return make_error_code(ParseError::Success); 767 } 768 } 769 return make_error_code(ParseError::Unsuitable); 770 } 771 772 std::string configurationAsText(const FormatStyle &Style) { 773 std::string Text; 774 llvm::raw_string_ostream Stream(Text); 775 llvm::yaml::Output Output(Stream); 776 // We use the same mapping method for input and output, so we need a non-const 777 // reference here. 778 FormatStyle NonConstStyle = expandPresets(Style); 779 Output << NonConstStyle; 780 return Stream.str(); 781 } 782 783 namespace { 784 785 class FormatTokenLexer { 786 public: 787 FormatTokenLexer(const SourceManager &SourceMgr, FileID ID, 788 const FormatStyle &Style, encoding::Encoding Encoding) 789 : FormatTok(nullptr), IsFirstToken(true), GreaterStashed(false), 790 LessStashed(false), Column(0), TrailingWhitespace(0), 791 SourceMgr(SourceMgr), ID(ID), Style(Style), 792 IdentTable(getFormattingLangOpts(Style)), Keywords(IdentTable), 793 Encoding(Encoding), FirstInLineIndex(0), FormattingDisabled(false), 794 MacroBlockBeginRegex(Style.MacroBlockBegin), 795 MacroBlockEndRegex(Style.MacroBlockEnd) { 796 Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr, 797 getFormattingLangOpts(Style))); 798 Lex->SetKeepWhitespaceMode(true); 799 800 for (const std::string &ForEachMacro : Style.ForEachMacros) 801 ForEachMacros.push_back(&IdentTable.get(ForEachMacro)); 802 std::sort(ForEachMacros.begin(), ForEachMacros.end()); 803 } 804 805 ArrayRef<FormatToken *> lex() { 806 assert(Tokens.empty()); 807 assert(FirstInLineIndex == 0); 808 do { 809 Tokens.push_back(getNextToken()); 810 if (Style.Language == FormatStyle::LK_JavaScript) 811 tryParseJSRegexLiteral(); 812 tryMergePreviousTokens(); 813 if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline) 814 FirstInLineIndex = Tokens.size() - 1; 815 } while (Tokens.back()->Tok.isNot(tok::eof)); 816 return Tokens; 817 } 818 819 const AdditionalKeywords &getKeywords() { return Keywords; } 820 821 private: 822 void tryMergePreviousTokens() { 823 if (tryMerge_TMacro()) 824 return; 825 if (tryMergeConflictMarkers()) 826 return; 827 if (tryMergeLessLess()) 828 return; 829 830 if (Style.Language == FormatStyle::LK_JavaScript) { 831 if (tryMergeTemplateString()) 832 return; 833 834 static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal}; 835 static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal, 836 tok::equal}; 837 static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater, 838 tok::greaterequal}; 839 static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater}; 840 // FIXME: Investigate what token type gives the correct operator priority. 841 if (tryMergeTokens(JSIdentity, TT_BinaryOperator)) 842 return; 843 if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator)) 844 return; 845 if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator)) 846 return; 847 if (tryMergeTokens(JSRightArrow, TT_JsFatArrow)) 848 return; 849 } 850 } 851 852 bool tryMergeLessLess() { 853 // Merge X,less,less,Y into X,lessless,Y unless X or Y is less. 854 if (Tokens.size() < 3) 855 return false; 856 857 bool FourthTokenIsLess = false; 858 if (Tokens.size() > 3) 859 FourthTokenIsLess = (Tokens.end() - 4)[0]->is(tok::less); 860 861 auto First = Tokens.end() - 3; 862 if (First[2]->is(tok::less) || First[1]->isNot(tok::less) || 863 First[0]->isNot(tok::less) || FourthTokenIsLess) 864 return false; 865 866 // Only merge if there currently is no whitespace between the two "<". 867 if (First[1]->WhitespaceRange.getBegin() != 868 First[1]->WhitespaceRange.getEnd()) 869 return false; 870 871 First[0]->Tok.setKind(tok::lessless); 872 First[0]->TokenText = "<<"; 873 First[0]->ColumnWidth += 1; 874 Tokens.erase(Tokens.end() - 2); 875 return true; 876 } 877 878 bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds, TokenType NewType) { 879 if (Tokens.size() < Kinds.size()) 880 return false; 881 882 SmallVectorImpl<FormatToken *>::const_iterator First = 883 Tokens.end() - Kinds.size(); 884 if (!First[0]->is(Kinds[0])) 885 return false; 886 unsigned AddLength = 0; 887 for (unsigned i = 1; i < Kinds.size(); ++i) { 888 if (!First[i]->is(Kinds[i]) || 889 First[i]->WhitespaceRange.getBegin() != 890 First[i]->WhitespaceRange.getEnd()) 891 return false; 892 AddLength += First[i]->TokenText.size(); 893 } 894 Tokens.resize(Tokens.size() - Kinds.size() + 1); 895 First[0]->TokenText = StringRef(First[0]->TokenText.data(), 896 First[0]->TokenText.size() + AddLength); 897 First[0]->ColumnWidth += AddLength; 898 First[0]->Type = NewType; 899 return true; 900 } 901 902 // Returns \c true if \p Tok can only be followed by an operand in JavaScript. 903 bool precedesOperand(FormatToken *Tok) { 904 // NB: This is not entirely correct, as an r_paren can introduce an operand 905 // location in e.g. `if (foo) /bar/.exec(...);`. That is a rare enough 906 // corner case to not matter in practice, though. 907 return Tok->isOneOf(tok::period, tok::l_paren, tok::comma, tok::l_brace, 908 tok::r_brace, tok::l_square, tok::semi, tok::exclaim, 909 tok::colon, tok::question, tok::tilde) || 910 Tok->isOneOf(tok::kw_return, tok::kw_do, tok::kw_case, tok::kw_throw, 911 tok::kw_else, tok::kw_new, tok::kw_delete, tok::kw_void, 912 tok::kw_typeof, Keywords.kw_instanceof, 913 Keywords.kw_in) || 914 Tok->isBinaryOperator(); 915 } 916 917 bool canPrecedeRegexLiteral(FormatToken *Prev) { 918 if (!Prev) 919 return true; 920 921 // Regex literals can only follow after prefix unary operators, not after 922 // postfix unary operators. If the '++' is followed by a non-operand 923 // introducing token, the slash here is the operand and not the start of a 924 // regex. 925 if (Prev->isOneOf(tok::plusplus, tok::minusminus)) 926 return (Tokens.size() < 3 || precedesOperand(Tokens[Tokens.size() - 3])); 927 928 // The previous token must introduce an operand location where regex 929 // literals can occur. 930 if (!precedesOperand(Prev)) 931 return false; 932 933 return true; 934 } 935 936 // Tries to parse a JavaScript Regex literal starting at the current token, 937 // if that begins with a slash and is in a location where JavaScript allows 938 // regex literals. Changes the current token to a regex literal and updates 939 // its text if successful. 940 void tryParseJSRegexLiteral() { 941 FormatToken *RegexToken = Tokens.back(); 942 if (!RegexToken->isOneOf(tok::slash, tok::slashequal)) 943 return; 944 945 FormatToken *Prev = nullptr; 946 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) { 947 // NB: Because previous pointers are not initialized yet, this cannot use 948 // Token.getPreviousNonComment. 949 if ((*I)->isNot(tok::comment)) { 950 Prev = *I; 951 break; 952 } 953 } 954 955 if (!canPrecedeRegexLiteral(Prev)) 956 return; 957 958 // 'Manually' lex ahead in the current file buffer. 959 const char *Offset = Lex->getBufferLocation(); 960 const char *RegexBegin = Offset - RegexToken->TokenText.size(); 961 StringRef Buffer = Lex->getBuffer(); 962 bool InCharacterClass = false; 963 bool HaveClosingSlash = false; 964 for (; !HaveClosingSlash && Offset != Buffer.end(); ++Offset) { 965 // Regular expressions are terminated with a '/', which can only be 966 // escaped using '\' or a character class between '[' and ']'. 967 // See http://www.ecma-international.org/ecma-262/5.1/#sec-7.8.5. 968 switch (*Offset) { 969 case '\\': 970 // Skip the escaped character. 971 ++Offset; 972 break; 973 case '[': 974 InCharacterClass = true; 975 break; 976 case ']': 977 InCharacterClass = false; 978 break; 979 case '/': 980 if (!InCharacterClass) 981 HaveClosingSlash = true; 982 break; 983 } 984 } 985 986 RegexToken->Type = TT_RegexLiteral; 987 // Treat regex literals like other string_literals. 988 RegexToken->Tok.setKind(tok::string_literal); 989 RegexToken->TokenText = StringRef(RegexBegin, Offset - RegexBegin); 990 RegexToken->ColumnWidth = RegexToken->TokenText.size(); 991 992 resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset))); 993 } 994 995 bool tryMergeTemplateString() { 996 if (Tokens.size() < 2) 997 return false; 998 999 FormatToken *EndBacktick = Tokens.back(); 1000 // Backticks get lexed as tok::unknown tokens. If a template string contains 1001 // a comment start, it gets lexed as a tok::comment, or tok::unknown if 1002 // unterminated. 1003 if (!EndBacktick->isOneOf(tok::comment, tok::string_literal, 1004 tok::char_constant, tok::unknown)) 1005 return false; 1006 size_t CommentBacktickPos = EndBacktick->TokenText.find('`'); 1007 // Unknown token that's not actually a backtick, or a comment that doesn't 1008 // contain a backtick. 1009 if (CommentBacktickPos == StringRef::npos) 1010 return false; 1011 1012 unsigned TokenCount = 0; 1013 bool IsMultiline = false; 1014 unsigned EndColumnInFirstLine = 1015 EndBacktick->OriginalColumn + EndBacktick->ColumnWidth; 1016 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; I++) { 1017 ++TokenCount; 1018 if (I[0]->IsMultiline) 1019 IsMultiline = true; 1020 1021 // If there was a preceding template string, this must be the start of a 1022 // template string, not the end. 1023 if (I[0]->is(TT_TemplateString)) 1024 return false; 1025 1026 if (I[0]->isNot(tok::unknown) || I[0]->TokenText != "`") { 1027 // Keep track of the rhs offset of the last token to wrap across lines - 1028 // its the rhs offset of the first line of the template string, used to 1029 // determine its width. 1030 if (I[0]->IsMultiline) 1031 EndColumnInFirstLine = I[0]->OriginalColumn + I[0]->ColumnWidth; 1032 // If the token has newlines, the token before it (if it exists) is the 1033 // rhs end of the previous line. 1034 if (I[0]->NewlinesBefore > 0 && (I + 1 != E)) { 1035 EndColumnInFirstLine = I[1]->OriginalColumn + I[1]->ColumnWidth; 1036 IsMultiline = true; 1037 } 1038 continue; 1039 } 1040 1041 Tokens.resize(Tokens.size() - TokenCount); 1042 Tokens.back()->Type = TT_TemplateString; 1043 const char *EndOffset = 1044 EndBacktick->TokenText.data() + 1 + CommentBacktickPos; 1045 if (CommentBacktickPos != 0) { 1046 // If the backtick was not the first character (e.g. in a comment), 1047 // re-lex after the backtick position. 1048 SourceLocation Loc = EndBacktick->Tok.getLocation(); 1049 resetLexer(SourceMgr.getFileOffset(Loc) + CommentBacktickPos + 1); 1050 } 1051 Tokens.back()->TokenText = 1052 StringRef(Tokens.back()->TokenText.data(), 1053 EndOffset - Tokens.back()->TokenText.data()); 1054 1055 unsigned EndOriginalColumn = EndBacktick->OriginalColumn; 1056 if (EndOriginalColumn == 0) { 1057 SourceLocation Loc = EndBacktick->Tok.getLocation(); 1058 EndOriginalColumn = SourceMgr.getSpellingColumnNumber(Loc); 1059 } 1060 // If the ` is further down within the token (e.g. in a comment). 1061 EndOriginalColumn += CommentBacktickPos; 1062 1063 if (IsMultiline) { 1064 // ColumnWidth is from backtick to last token in line. 1065 // LastLineColumnWidth is 0 to backtick. 1066 // x = `some content 1067 // until here`; 1068 Tokens.back()->ColumnWidth = 1069 EndColumnInFirstLine - Tokens.back()->OriginalColumn; 1070 // +1 for the ` itself. 1071 Tokens.back()->LastLineColumnWidth = EndOriginalColumn + 1; 1072 Tokens.back()->IsMultiline = true; 1073 } else { 1074 // Token simply spans from start to end, +1 for the ` itself. 1075 Tokens.back()->ColumnWidth = 1076 EndOriginalColumn - Tokens.back()->OriginalColumn + 1; 1077 } 1078 return true; 1079 } 1080 return false; 1081 } 1082 1083 bool tryMerge_TMacro() { 1084 if (Tokens.size() < 4) 1085 return false; 1086 FormatToken *Last = Tokens.back(); 1087 if (!Last->is(tok::r_paren)) 1088 return false; 1089 1090 FormatToken *String = Tokens[Tokens.size() - 2]; 1091 if (!String->is(tok::string_literal) || String->IsMultiline) 1092 return false; 1093 1094 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) 1095 return false; 1096 1097 FormatToken *Macro = Tokens[Tokens.size() - 4]; 1098 if (Macro->TokenText != "_T") 1099 return false; 1100 1101 const char *Start = Macro->TokenText.data(); 1102 const char *End = Last->TokenText.data() + Last->TokenText.size(); 1103 String->TokenText = StringRef(Start, End - Start); 1104 String->IsFirst = Macro->IsFirst; 1105 String->LastNewlineOffset = Macro->LastNewlineOffset; 1106 String->WhitespaceRange = Macro->WhitespaceRange; 1107 String->OriginalColumn = Macro->OriginalColumn; 1108 String->ColumnWidth = encoding::columnWidthWithTabs( 1109 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); 1110 String->NewlinesBefore = Macro->NewlinesBefore; 1111 String->HasUnescapedNewline = Macro->HasUnescapedNewline; 1112 1113 Tokens.pop_back(); 1114 Tokens.pop_back(); 1115 Tokens.pop_back(); 1116 Tokens.back() = String; 1117 return true; 1118 } 1119 1120 bool tryMergeConflictMarkers() { 1121 if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof)) 1122 return false; 1123 1124 // Conflict lines look like: 1125 // <marker> <text from the vcs> 1126 // For example: 1127 // >>>>>>> /file/in/file/system at revision 1234 1128 // 1129 // We merge all tokens in a line that starts with a conflict marker 1130 // into a single token with a special token type that the unwrapped line 1131 // parser will use to correctly rebuild the underlying code. 1132 1133 FileID ID; 1134 // Get the position of the first token in the line. 1135 unsigned FirstInLineOffset; 1136 std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc( 1137 Tokens[FirstInLineIndex]->getStartOfNonWhitespace()); 1138 StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer(); 1139 // Calculate the offset of the start of the current line. 1140 auto LineOffset = Buffer.rfind('\n', FirstInLineOffset); 1141 if (LineOffset == StringRef::npos) { 1142 LineOffset = 0; 1143 } else { 1144 ++LineOffset; 1145 } 1146 1147 auto FirstSpace = Buffer.find_first_of(" \n", LineOffset); 1148 StringRef LineStart; 1149 if (FirstSpace == StringRef::npos) { 1150 LineStart = Buffer.substr(LineOffset); 1151 } else { 1152 LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset); 1153 } 1154 1155 TokenType Type = TT_Unknown; 1156 if (LineStart == "<<<<<<<" || LineStart == ">>>>") { 1157 Type = TT_ConflictStart; 1158 } else if (LineStart == "|||||||" || LineStart == "=======" || 1159 LineStart == "====") { 1160 Type = TT_ConflictAlternative; 1161 } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") { 1162 Type = TT_ConflictEnd; 1163 } 1164 1165 if (Type != TT_Unknown) { 1166 FormatToken *Next = Tokens.back(); 1167 1168 Tokens.resize(FirstInLineIndex + 1); 1169 // We do not need to build a complete token here, as we will skip it 1170 // during parsing anyway (as we must not touch whitespace around conflict 1171 // markers). 1172 Tokens.back()->Type = Type; 1173 Tokens.back()->Tok.setKind(tok::kw___unknown_anytype); 1174 1175 Tokens.push_back(Next); 1176 return true; 1177 } 1178 1179 return false; 1180 } 1181 1182 FormatToken *getStashedToken() { 1183 // Create a synthesized second '>' or '<' token. 1184 Token Tok = FormatTok->Tok; 1185 StringRef TokenText = FormatTok->TokenText; 1186 1187 unsigned OriginalColumn = FormatTok->OriginalColumn; 1188 FormatTok = new (Allocator.Allocate()) FormatToken; 1189 FormatTok->Tok = Tok; 1190 SourceLocation TokLocation = 1191 FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1); 1192 FormatTok->Tok.setLocation(TokLocation); 1193 FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation); 1194 FormatTok->TokenText = TokenText; 1195 FormatTok->ColumnWidth = 1; 1196 FormatTok->OriginalColumn = OriginalColumn + 1; 1197 1198 return FormatTok; 1199 } 1200 1201 FormatToken *getNextToken() { 1202 if (GreaterStashed) { 1203 GreaterStashed = false; 1204 return getStashedToken(); 1205 } 1206 if (LessStashed) { 1207 LessStashed = false; 1208 return getStashedToken(); 1209 } 1210 1211 FormatTok = new (Allocator.Allocate()) FormatToken; 1212 readRawToken(*FormatTok); 1213 SourceLocation WhitespaceStart = 1214 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); 1215 FormatTok->IsFirst = IsFirstToken; 1216 IsFirstToken = false; 1217 1218 // Consume and record whitespace until we find a significant token. 1219 unsigned WhitespaceLength = TrailingWhitespace; 1220 while (FormatTok->Tok.is(tok::unknown)) { 1221 StringRef Text = FormatTok->TokenText; 1222 auto EscapesNewline = [&](int pos) { 1223 // A '\r' here is just part of '\r\n'. Skip it. 1224 if (pos >= 0 && Text[pos] == '\r') 1225 --pos; 1226 // See whether there is an odd number of '\' before this. 1227 unsigned count = 0; 1228 for (; pos >= 0; --pos, ++count) 1229 if (Text[pos] != '\\') 1230 break; 1231 return count & 1; 1232 }; 1233 // FIXME: This miscounts tok:unknown tokens that are not just 1234 // whitespace, e.g. a '`' character. 1235 for (int i = 0, e = Text.size(); i != e; ++i) { 1236 switch (Text[i]) { 1237 case '\n': 1238 ++FormatTok->NewlinesBefore; 1239 FormatTok->HasUnescapedNewline = !EscapesNewline(i - 1); 1240 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 1241 Column = 0; 1242 break; 1243 case '\r': 1244 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 1245 Column = 0; 1246 break; 1247 case '\f': 1248 case '\v': 1249 Column = 0; 1250 break; 1251 case ' ': 1252 ++Column; 1253 break; 1254 case '\t': 1255 Column += Style.TabWidth - Column % Style.TabWidth; 1256 break; 1257 case '\\': 1258 if (i + 1 == e || (Text[i + 1] != '\r' && Text[i + 1] != '\n')) 1259 FormatTok->Type = TT_ImplicitStringLiteral; 1260 break; 1261 default: 1262 FormatTok->Type = TT_ImplicitStringLiteral; 1263 break; 1264 } 1265 if (FormatTok->Type == TT_ImplicitStringLiteral) 1266 break; 1267 } 1268 1269 if (FormatTok->is(TT_ImplicitStringLiteral)) 1270 break; 1271 WhitespaceLength += FormatTok->Tok.getLength(); 1272 1273 readRawToken(*FormatTok); 1274 } 1275 1276 // In case the token starts with escaped newlines, we want to 1277 // take them into account as whitespace - this pattern is quite frequent 1278 // in macro definitions. 1279 // FIXME: Add a more explicit test. 1280 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' && 1281 FormatTok->TokenText[1] == '\n') { 1282 ++FormatTok->NewlinesBefore; 1283 WhitespaceLength += 2; 1284 FormatTok->LastNewlineOffset = 2; 1285 Column = 0; 1286 FormatTok->TokenText = FormatTok->TokenText.substr(2); 1287 } 1288 1289 FormatTok->WhitespaceRange = SourceRange( 1290 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); 1291 1292 FormatTok->OriginalColumn = Column; 1293 1294 TrailingWhitespace = 0; 1295 if (FormatTok->Tok.is(tok::comment)) { 1296 // FIXME: Add the trimmed whitespace to Column. 1297 StringRef UntrimmedText = FormatTok->TokenText; 1298 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); 1299 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); 1300 } else if (FormatTok->Tok.is(tok::raw_identifier)) { 1301 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); 1302 FormatTok->Tok.setIdentifierInfo(&Info); 1303 FormatTok->Tok.setKind(Info.getTokenID()); 1304 if (Style.Language == FormatStyle::LK_Java && 1305 FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete, 1306 tok::kw_operator)) { 1307 FormatTok->Tok.setKind(tok::identifier); 1308 FormatTok->Tok.setIdentifierInfo(nullptr); 1309 } else if (Style.Language == FormatStyle::LK_JavaScript && 1310 FormatTok->isOneOf(tok::kw_struct, tok::kw_union, 1311 tok::kw_operator)) { 1312 FormatTok->Tok.setKind(tok::identifier); 1313 FormatTok->Tok.setIdentifierInfo(nullptr); 1314 } 1315 } else if (FormatTok->Tok.is(tok::greatergreater)) { 1316 FormatTok->Tok.setKind(tok::greater); 1317 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1318 GreaterStashed = true; 1319 } else if (FormatTok->Tok.is(tok::lessless)) { 1320 FormatTok->Tok.setKind(tok::less); 1321 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1322 LessStashed = true; 1323 } 1324 1325 // Now FormatTok is the next non-whitespace token. 1326 1327 StringRef Text = FormatTok->TokenText; 1328 size_t FirstNewlinePos = Text.find('\n'); 1329 if (FirstNewlinePos == StringRef::npos) { 1330 // FIXME: ColumnWidth actually depends on the start column, we need to 1331 // take this into account when the token is moved. 1332 FormatTok->ColumnWidth = 1333 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); 1334 Column += FormatTok->ColumnWidth; 1335 } else { 1336 FormatTok->IsMultiline = true; 1337 // FIXME: ColumnWidth actually depends on the start column, we need to 1338 // take this into account when the token is moved. 1339 FormatTok->ColumnWidth = encoding::columnWidthWithTabs( 1340 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); 1341 1342 // The last line of the token always starts in column 0. 1343 // Thus, the length can be precomputed even in the presence of tabs. 1344 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( 1345 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, 1346 Encoding); 1347 Column = FormatTok->LastLineColumnWidth; 1348 } 1349 1350 if (Style.Language == FormatStyle::LK_Cpp) { 1351 if (!(Tokens.size() > 0 && Tokens.back()->Tok.getIdentifierInfo() && 1352 Tokens.back()->Tok.getIdentifierInfo()->getPPKeywordID() == 1353 tok::pp_define) && 1354 std::find(ForEachMacros.begin(), ForEachMacros.end(), 1355 FormatTok->Tok.getIdentifierInfo()) != 1356 ForEachMacros.end()) { 1357 FormatTok->Type = TT_ForEachMacro; 1358 } else if (FormatTok->is(tok::identifier)) { 1359 if (MacroBlockBeginRegex.match(Text)) { 1360 FormatTok->Type = TT_MacroBlockBegin; 1361 } else if (MacroBlockEndRegex.match(Text)) { 1362 FormatTok->Type = TT_MacroBlockEnd; 1363 } 1364 } 1365 } 1366 1367 return FormatTok; 1368 } 1369 1370 FormatToken *FormatTok; 1371 bool IsFirstToken; 1372 bool GreaterStashed, LessStashed; 1373 unsigned Column; 1374 unsigned TrailingWhitespace; 1375 std::unique_ptr<Lexer> Lex; 1376 const SourceManager &SourceMgr; 1377 FileID ID; 1378 const FormatStyle &Style; 1379 IdentifierTable IdentTable; 1380 AdditionalKeywords Keywords; 1381 encoding::Encoding Encoding; 1382 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator; 1383 // Index (in 'Tokens') of the last token that starts a new line. 1384 unsigned FirstInLineIndex; 1385 SmallVector<FormatToken *, 16> Tokens; 1386 SmallVector<IdentifierInfo *, 8> ForEachMacros; 1387 1388 bool FormattingDisabled; 1389 1390 llvm::Regex MacroBlockBeginRegex; 1391 llvm::Regex MacroBlockEndRegex; 1392 1393 void readRawToken(FormatToken &Tok) { 1394 Lex->LexFromRawLexer(Tok.Tok); 1395 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), 1396 Tok.Tok.getLength()); 1397 // For formatting, treat unterminated string literals like normal string 1398 // literals. 1399 if (Tok.is(tok::unknown)) { 1400 if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') { 1401 Tok.Tok.setKind(tok::string_literal); 1402 Tok.IsUnterminatedLiteral = true; 1403 } else if (Style.Language == FormatStyle::LK_JavaScript && 1404 Tok.TokenText == "''") { 1405 Tok.Tok.setKind(tok::string_literal); 1406 } 1407 } 1408 1409 if (Style.Language == FormatStyle::LK_JavaScript && 1410 Tok.is(tok::char_constant)) { 1411 Tok.Tok.setKind(tok::string_literal); 1412 } 1413 1414 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" || 1415 Tok.TokenText == "/* clang-format on */")) { 1416 FormattingDisabled = false; 1417 } 1418 1419 Tok.Finalized = FormattingDisabled; 1420 1421 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" || 1422 Tok.TokenText == "/* clang-format off */")) { 1423 FormattingDisabled = true; 1424 } 1425 } 1426 1427 void resetLexer(unsigned Offset) { 1428 StringRef Buffer = SourceMgr.getBufferData(ID); 1429 Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID), 1430 getFormattingLangOpts(Style), Buffer.begin(), 1431 Buffer.begin() + Offset, Buffer.end())); 1432 Lex->SetKeepWhitespaceMode(true); 1433 TrailingWhitespace = 0; 1434 } 1435 }; 1436 1437 static StringRef getLanguageName(FormatStyle::LanguageKind Language) { 1438 switch (Language) { 1439 case FormatStyle::LK_Cpp: 1440 return "C++"; 1441 case FormatStyle::LK_Java: 1442 return "Java"; 1443 case FormatStyle::LK_JavaScript: 1444 return "JavaScript"; 1445 case FormatStyle::LK_Proto: 1446 return "Proto"; 1447 default: 1448 return "Unknown"; 1449 } 1450 } 1451 1452 class Environment { 1453 public: 1454 Environment(SourceManager &SM, FileID ID, ArrayRef<CharSourceRange> Ranges) 1455 : ID(ID), CharRanges(Ranges.begin(), Ranges.end()), SM(SM) {} 1456 1457 Environment(FileID ID, std::unique_ptr<FileManager> FileMgr, 1458 std::unique_ptr<SourceManager> VirtualSM, 1459 std::unique_ptr<DiagnosticsEngine> Diagnostics, 1460 const std::vector<CharSourceRange> &CharRanges) 1461 : ID(ID), CharRanges(CharRanges.begin(), CharRanges.end()), 1462 SM(*VirtualSM), FileMgr(std::move(FileMgr)), 1463 VirtualSM(std::move(VirtualSM)), Diagnostics(std::move(Diagnostics)) {} 1464 1465 // This sets up an virtual file system with file \p FileName containing \p 1466 // Code. 1467 static std::unique_ptr<Environment> 1468 CreateVirtualEnvironment(StringRef Code, StringRef FileName, 1469 ArrayRef<tooling::Range> Ranges) { 1470 // This is referenced by `FileMgr` and will be released by `FileMgr` when it 1471 // is deleted. 1472 IntrusiveRefCntPtr<vfs::InMemoryFileSystem> InMemoryFileSystem( 1473 new vfs::InMemoryFileSystem); 1474 // This is passed to `SM` as reference, so the pointer has to be referenced 1475 // in `Environment` so that `FileMgr` can out-live this function scope. 1476 std::unique_ptr<FileManager> FileMgr( 1477 new FileManager(FileSystemOptions(), InMemoryFileSystem)); 1478 // This is passed to `SM` as reference, so the pointer has to be referenced 1479 // by `Environment` due to the same reason above. 1480 std::unique_ptr<DiagnosticsEngine> Diagnostics(new DiagnosticsEngine( 1481 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), 1482 new DiagnosticOptions)); 1483 // This will be stored as reference, so the pointer has to be stored in 1484 // due to the same reason above. 1485 std::unique_ptr<SourceManager> VirtualSM( 1486 new SourceManager(*Diagnostics, *FileMgr)); 1487 InMemoryFileSystem->addFile( 1488 FileName, 0, llvm::MemoryBuffer::getMemBuffer( 1489 Code, FileName, /*RequiresNullTerminator=*/false)); 1490 FileID ID = VirtualSM->createFileID( 1491 FileMgr->getFile(FileName), SourceLocation(), clang::SrcMgr::C_User); 1492 assert(ID.isValid()); 1493 SourceLocation StartOfFile = VirtualSM->getLocForStartOfFile(ID); 1494 std::vector<CharSourceRange> CharRanges; 1495 for (const tooling::Range &Range : Ranges) { 1496 SourceLocation Start = StartOfFile.getLocWithOffset(Range.getOffset()); 1497 SourceLocation End = Start.getLocWithOffset(Range.getLength()); 1498 CharRanges.push_back(CharSourceRange::getCharRange(Start, End)); 1499 } 1500 return llvm::make_unique<Environment>(ID, std::move(FileMgr), 1501 std::move(VirtualSM), 1502 std::move(Diagnostics), CharRanges); 1503 } 1504 1505 FileID getFileID() const { return ID; } 1506 1507 StringRef getFileName() const { return FileName; } 1508 1509 ArrayRef<CharSourceRange> getCharRanges() const { return CharRanges; } 1510 1511 const SourceManager &getSourceManager() const { return SM; } 1512 1513 private: 1514 FileID ID; 1515 StringRef FileName; 1516 SmallVector<CharSourceRange, 8> CharRanges; 1517 SourceManager &SM; 1518 1519 // The order of these fields are important - they should be in the same order 1520 // as they are created in `CreateVirtualEnvironment` so that they can be 1521 // deleted in the reverse order as they are created. 1522 std::unique_ptr<FileManager> FileMgr; 1523 std::unique_ptr<SourceManager> VirtualSM; 1524 std::unique_ptr<DiagnosticsEngine> Diagnostics; 1525 }; 1526 1527 class TokenAnalyzer : public UnwrappedLineConsumer { 1528 public: 1529 TokenAnalyzer(const Environment &Env, const FormatStyle &Style) 1530 : Style(Style), Env(Env), 1531 AffectedRangeMgr(Env.getSourceManager(), Env.getCharRanges()), 1532 UnwrappedLines(1), 1533 Encoding(encoding::detectEncoding( 1534 Env.getSourceManager().getBufferData(Env.getFileID()))) { 1535 DEBUG(llvm::dbgs() << "File encoding: " 1536 << (Encoding == encoding::Encoding_UTF8 ? "UTF8" 1537 : "unknown") 1538 << "\n"); 1539 DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language) 1540 << "\n"); 1541 } 1542 1543 tooling::Replacements process() { 1544 tooling::Replacements Result; 1545 FormatTokenLexer Tokens(Env.getSourceManager(), Env.getFileID(), Style, 1546 Encoding); 1547 1548 UnwrappedLineParser Parser(Style, Tokens.getKeywords(), Tokens.lex(), 1549 *this); 1550 Parser.parse(); 1551 assert(UnwrappedLines.rbegin()->empty()); 1552 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE; 1553 ++Run) { 1554 DEBUG(llvm::dbgs() << "Run " << Run << "...\n"); 1555 SmallVector<AnnotatedLine *, 16> AnnotatedLines; 1556 1557 TokenAnnotator Annotator(Style, Tokens.getKeywords()); 1558 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) { 1559 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i])); 1560 Annotator.annotate(*AnnotatedLines.back()); 1561 } 1562 1563 tooling::Replacements RunResult = 1564 analyze(Annotator, AnnotatedLines, Tokens, Result); 1565 1566 DEBUG({ 1567 llvm::dbgs() << "Replacements for run " << Run << ":\n"; 1568 for (tooling::Replacements::iterator I = RunResult.begin(), 1569 E = RunResult.end(); 1570 I != E; ++I) { 1571 llvm::dbgs() << I->toString() << "\n"; 1572 } 1573 }); 1574 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1575 delete AnnotatedLines[i]; 1576 } 1577 Result.insert(RunResult.begin(), RunResult.end()); 1578 } 1579 return Result; 1580 } 1581 1582 protected: 1583 virtual tooling::Replacements 1584 analyze(TokenAnnotator &Annotator, 1585 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1586 FormatTokenLexer &Tokens, tooling::Replacements &Result) = 0; 1587 1588 void consumeUnwrappedLine(const UnwrappedLine &TheLine) override { 1589 assert(!UnwrappedLines.empty()); 1590 UnwrappedLines.back().push_back(TheLine); 1591 } 1592 1593 void finishRun() override { 1594 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>()); 1595 } 1596 1597 FormatStyle Style; 1598 // Stores Style, FileID and SourceManager etc. 1599 const Environment &Env; 1600 // AffectedRangeMgr stores ranges to be fixed. 1601 AffectedRangeManager AffectedRangeMgr; 1602 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines; 1603 encoding::Encoding Encoding; 1604 }; 1605 1606 class Formatter : public TokenAnalyzer { 1607 public: 1608 Formatter(const Environment &Env, const FormatStyle &Style, 1609 bool *IncompleteFormat) 1610 : TokenAnalyzer(Env, Style), IncompleteFormat(IncompleteFormat) {} 1611 1612 tooling::Replacements 1613 analyze(TokenAnnotator &Annotator, 1614 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1615 FormatTokenLexer &Tokens, tooling::Replacements &Result) override { 1616 deriveLocalStyle(AnnotatedLines); 1617 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1618 AnnotatedLines.end()); 1619 1620 if (Style.Language == FormatStyle::LK_JavaScript && 1621 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) 1622 requoteJSStringLiteral(AnnotatedLines, Result); 1623 1624 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1625 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 1626 } 1627 1628 Annotator.setCommentLineLevels(AnnotatedLines); 1629 1630 WhitespaceManager Whitespaces( 1631 Env.getSourceManager(), Style, 1632 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID()))); 1633 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), 1634 Env.getSourceManager(), Whitespaces, Encoding, 1635 BinPackInconclusiveFunctions); 1636 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(), 1637 IncompleteFormat) 1638 .format(AnnotatedLines); 1639 return Whitespaces.generateReplacements(); 1640 } 1641 1642 private: 1643 // If the last token is a double/single-quoted string literal, generates a 1644 // replacement with a single/double quoted string literal, re-escaping the 1645 // contents in the process. 1646 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines, 1647 tooling::Replacements &Result) { 1648 for (AnnotatedLine *Line : Lines) { 1649 requoteJSStringLiteral(Line->Children, Result); 1650 if (!Line->Affected) 1651 continue; 1652 for (FormatToken *FormatTok = Line->First; FormatTok; 1653 FormatTok = FormatTok->Next) { 1654 StringRef Input = FormatTok->TokenText; 1655 if (!FormatTok->isStringLiteral() || 1656 // NB: testing for not starting with a double quote to avoid 1657 // breaking 1658 // `template strings`. 1659 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single && 1660 !Input.startswith("\"")) || 1661 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double && 1662 !Input.startswith("\'"))) 1663 continue; 1664 1665 // Change start and end quote. 1666 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single; 1667 SourceLocation Start = FormatTok->Tok.getLocation(); 1668 auto Replace = [&](SourceLocation Start, unsigned Length, 1669 StringRef ReplacementText) { 1670 Result.insert(tooling::Replacement(Env.getSourceManager(), Start, 1671 Length, ReplacementText)); 1672 }; 1673 Replace(Start, 1, IsSingle ? "'" : "\""); 1674 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1, 1675 IsSingle ? "'" : "\""); 1676 1677 // Escape internal quotes. 1678 size_t ColumnWidth = FormatTok->TokenText.size(); 1679 bool Escaped = false; 1680 for (size_t i = 1; i < Input.size() - 1; i++) { 1681 switch (Input[i]) { 1682 case '\\': 1683 if (!Escaped && i + 1 < Input.size() && 1684 ((IsSingle && Input[i + 1] == '"') || 1685 (!IsSingle && Input[i + 1] == '\''))) { 1686 // Remove this \, it's escaping a " or ' that no longer needs 1687 // escaping 1688 ColumnWidth--; 1689 Replace(Start.getLocWithOffset(i), 1, ""); 1690 continue; 1691 } 1692 Escaped = !Escaped; 1693 break; 1694 case '\"': 1695 case '\'': 1696 if (!Escaped && IsSingle == (Input[i] == '\'')) { 1697 // Escape the quote. 1698 Replace(Start.getLocWithOffset(i), 0, "\\"); 1699 ColumnWidth++; 1700 } 1701 Escaped = false; 1702 break; 1703 default: 1704 Escaped = false; 1705 break; 1706 } 1707 } 1708 1709 // For formatting, count the number of non-escaped single quotes in them 1710 // and adjust ColumnWidth to take the added escapes into account. 1711 // FIXME(martinprobst): this might conflict with code breaking a long 1712 // string literal (which clang-format doesn't do, yet). For that to 1713 // work, this code would have to modify TokenText directly. 1714 FormatTok->ColumnWidth = ColumnWidth; 1715 } 1716 } 1717 } 1718 1719 static bool inputUsesCRLF(StringRef Text) { 1720 return Text.count('\r') * 2 > Text.count('\n'); 1721 } 1722 1723 bool 1724 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) { 1725 for (const AnnotatedLine *Line : Lines) { 1726 if (hasCpp03IncompatibleFormat(Line->Children)) 1727 return true; 1728 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) { 1729 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1730 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener)) 1731 return true; 1732 if (Tok->is(TT_TemplateCloser) && 1733 Tok->Previous->is(TT_TemplateCloser)) 1734 return true; 1735 } 1736 } 1737 } 1738 return false; 1739 } 1740 1741 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) { 1742 int AlignmentDiff = 0; 1743 for (const AnnotatedLine *Line : Lines) { 1744 AlignmentDiff += countVariableAlignments(Line->Children); 1745 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) { 1746 if (!Tok->is(TT_PointerOrReference)) 1747 continue; 1748 bool SpaceBefore = 1749 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1750 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() != 1751 Tok->Next->WhitespaceRange.getEnd(); 1752 if (SpaceBefore && !SpaceAfter) 1753 ++AlignmentDiff; 1754 if (!SpaceBefore && SpaceAfter) 1755 --AlignmentDiff; 1756 } 1757 } 1758 return AlignmentDiff; 1759 } 1760 1761 void 1762 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1763 bool HasBinPackedFunction = false; 1764 bool HasOnePerLineFunction = false; 1765 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1766 if (!AnnotatedLines[i]->First->Next) 1767 continue; 1768 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1769 while (Tok->Next) { 1770 if (Tok->PackingKind == PPK_BinPacked) 1771 HasBinPackedFunction = true; 1772 if (Tok->PackingKind == PPK_OnePerLine) 1773 HasOnePerLineFunction = true; 1774 1775 Tok = Tok->Next; 1776 } 1777 } 1778 if (Style.DerivePointerAlignment) 1779 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0 1780 ? FormatStyle::PAS_Left 1781 : FormatStyle::PAS_Right; 1782 if (Style.Standard == FormatStyle::LS_Auto) 1783 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines) 1784 ? FormatStyle::LS_Cpp11 1785 : FormatStyle::LS_Cpp03; 1786 BinPackInconclusiveFunctions = 1787 HasBinPackedFunction || !HasOnePerLineFunction; 1788 } 1789 1790 bool BinPackInconclusiveFunctions; 1791 bool *IncompleteFormat; 1792 }; 1793 1794 // This class clean up the erroneous/redundant code around the given ranges in 1795 // file. 1796 class Cleaner : public TokenAnalyzer { 1797 public: 1798 Cleaner(const Environment &Env, const FormatStyle &Style) 1799 : TokenAnalyzer(Env, Style), 1800 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {} 1801 1802 // FIXME: eliminate unused parameters. 1803 tooling::Replacements 1804 analyze(TokenAnnotator &Annotator, 1805 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1806 FormatTokenLexer &Tokens, tooling::Replacements &Result) override { 1807 // FIXME: in the current implementation the granularity of affected range 1808 // is an annotated line. However, this is not sufficient. Furthermore, 1809 // redundant code introduced by replacements does not necessarily 1810 // intercept with ranges of replacements that result in the redundancy. 1811 // To determine if some redundant code is actually introduced by 1812 // replacements(e.g. deletions), we need to come up with a more 1813 // sophisticated way of computing affected ranges. 1814 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1815 AnnotatedLines.end()); 1816 1817 checkEmptyNamespace(AnnotatedLines); 1818 1819 return generateFixes(); 1820 } 1821 1822 private: 1823 bool containsOnlyComments(const AnnotatedLine &Line) { 1824 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) { 1825 if (Tok->isNot(tok::comment)) 1826 return false; 1827 } 1828 return true; 1829 } 1830 1831 // Iterate through all lines and remove any empty (nested) namespaces. 1832 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1833 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1834 auto &Line = *AnnotatedLines[i]; 1835 if (Line.startsWith(tok::kw_namespace) || 1836 Line.startsWith(tok::kw_inline, tok::kw_namespace)) { 1837 checkEmptyNamespace(AnnotatedLines, i, i); 1838 } 1839 } 1840 1841 for (auto Line : DeletedLines) { 1842 FormatToken *Tok = AnnotatedLines[Line]->First; 1843 while (Tok) { 1844 deleteToken(Tok); 1845 Tok = Tok->Next; 1846 } 1847 } 1848 } 1849 1850 // The function checks if the namespace, which starts from \p CurrentLine, and 1851 // its nested namespaces are empty and delete them if they are empty. It also 1852 // sets \p NewLine to the last line checked. 1853 // Returns true if the current namespace is empty. 1854 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1855 unsigned CurrentLine, unsigned &NewLine) { 1856 unsigned InitLine = CurrentLine, End = AnnotatedLines.size(); 1857 if (Style.BraceWrapping.AfterNamespace) { 1858 // If the left brace is in a new line, we should consume it first so that 1859 // it does not make the namespace non-empty. 1860 // FIXME: error handling if there is no left brace. 1861 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) { 1862 NewLine = CurrentLine; 1863 return false; 1864 } 1865 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) { 1866 return false; 1867 } 1868 while (++CurrentLine < End) { 1869 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace)) 1870 break; 1871 1872 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) || 1873 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline, 1874 tok::kw_namespace)) { 1875 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine)) 1876 return false; 1877 CurrentLine = NewLine; 1878 continue; 1879 } 1880 1881 if (containsOnlyComments(*AnnotatedLines[CurrentLine])) 1882 continue; 1883 1884 // If there is anything other than comments or nested namespaces in the 1885 // current namespace, the namespace cannot be empty. 1886 NewLine = CurrentLine; 1887 return false; 1888 } 1889 1890 NewLine = CurrentLine; 1891 if (CurrentLine >= End) 1892 return false; 1893 1894 // Check if the empty namespace is actually affected by changed ranges. 1895 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange( 1896 AnnotatedLines[InitLine]->First->Tok.getLocation(), 1897 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc()))) 1898 return false; 1899 1900 for (unsigned i = InitLine; i <= CurrentLine; ++i) { 1901 DeletedLines.insert(i); 1902 } 1903 1904 return true; 1905 } 1906 1907 // Delete the given token. 1908 inline void deleteToken(FormatToken *Tok) { 1909 if (Tok) 1910 DeletedTokens.insert(Tok); 1911 } 1912 1913 tooling::Replacements generateFixes() { 1914 tooling::Replacements Fixes; 1915 std::vector<FormatToken *> Tokens; 1916 std::copy(DeletedTokens.begin(), DeletedTokens.end(), 1917 std::back_inserter(Tokens)); 1918 1919 // Merge multiple continuous token deletions into one big deletion so that 1920 // the number of replacements can be reduced. This makes computing affected 1921 // ranges more efficient when we run reformat on the changed code. 1922 unsigned Idx = 0; 1923 while (Idx < Tokens.size()) { 1924 unsigned St = Idx, End = Idx; 1925 while ((End + 1) < Tokens.size() && 1926 Tokens[End]->Next == Tokens[End + 1]) { 1927 End++; 1928 } 1929 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(), 1930 Tokens[End]->Tok.getEndLoc()); 1931 Fixes.insert(tooling::Replacement(Env.getSourceManager(), SR, "")); 1932 Idx = End + 1; 1933 } 1934 1935 return Fixes; 1936 } 1937 1938 // Class for less-than inequality comparason for the set `RedundantTokens`. 1939 // We store tokens in the order they appear in the translation unit so that 1940 // we do not need to sort them in `generateFixes()`. 1941 struct FormatTokenLess { 1942 FormatTokenLess(const SourceManager &SM) : SM(SM) {} 1943 1944 bool operator()(const FormatToken *LHS, const FormatToken *RHS) { 1945 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(), 1946 RHS->Tok.getLocation()); 1947 } 1948 const SourceManager &SM; 1949 }; 1950 1951 // Tokens to be deleted. 1952 std::set<FormatToken *, FormatTokenLess> DeletedTokens; 1953 // The line numbers of lines to be deleted. 1954 std::set<unsigned> DeletedLines; 1955 }; 1956 1957 struct IncludeDirective { 1958 StringRef Filename; 1959 StringRef Text; 1960 unsigned Offset; 1961 int Category; 1962 }; 1963 1964 } // end anonymous namespace 1965 1966 // Determines whether 'Ranges' intersects with ('Start', 'End'). 1967 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start, 1968 unsigned End) { 1969 for (auto Range : Ranges) { 1970 if (Range.getOffset() < End && 1971 Range.getOffset() + Range.getLength() > Start) 1972 return true; 1973 } 1974 return false; 1975 } 1976 1977 // Sorts a block of includes given by 'Includes' alphabetically adding the 1978 // necessary replacement to 'Replaces'. 'Includes' must be in strict source 1979 // order. 1980 static void sortIncludes(const FormatStyle &Style, 1981 const SmallVectorImpl<IncludeDirective> &Includes, 1982 ArrayRef<tooling::Range> Ranges, StringRef FileName, 1983 tooling::Replacements &Replaces, unsigned *Cursor) { 1984 if (!affectsRange(Ranges, Includes.front().Offset, 1985 Includes.back().Offset + Includes.back().Text.size())) 1986 return; 1987 SmallVector<unsigned, 16> Indices; 1988 for (unsigned i = 0, e = Includes.size(); i != e; ++i) 1989 Indices.push_back(i); 1990 std::stable_sort( 1991 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) { 1992 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) < 1993 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename); 1994 }); 1995 1996 // If the #includes are out of order, we generate a single replacement fixing 1997 // the entire block. Otherwise, no replacement is generated. 1998 bool OutOfOrder = false; 1999 for (unsigned i = 1, e = Indices.size(); i != e; ++i) { 2000 if (Indices[i] != i) { 2001 OutOfOrder = true; 2002 break; 2003 } 2004 } 2005 if (!OutOfOrder) 2006 return; 2007 2008 std::string result; 2009 bool CursorMoved = false; 2010 for (unsigned Index : Indices) { 2011 if (!result.empty()) 2012 result += "\n"; 2013 result += Includes[Index].Text; 2014 2015 if (Cursor && !CursorMoved) { 2016 unsigned Start = Includes[Index].Offset; 2017 unsigned End = Start + Includes[Index].Text.size(); 2018 if (*Cursor >= Start && *Cursor < End) { 2019 *Cursor = Includes.front().Offset + result.size() + *Cursor - End; 2020 CursorMoved = true; 2021 } 2022 } 2023 } 2024 2025 // Sorting #includes shouldn't change their total number of characters. 2026 // This would otherwise mess up 'Ranges'. 2027 assert(result.size() == 2028 Includes.back().Offset + Includes.back().Text.size() - 2029 Includes.front().Offset); 2030 2031 Replaces.insert(tooling::Replacement(FileName, Includes.front().Offset, 2032 result.size(), result)); 2033 } 2034 2035 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code, 2036 ArrayRef<tooling::Range> Ranges, 2037 StringRef FileName, unsigned *Cursor) { 2038 tooling::Replacements Replaces; 2039 if (!Style.SortIncludes) 2040 return Replaces; 2041 2042 unsigned Prev = 0; 2043 unsigned SearchFrom = 0; 2044 llvm::Regex IncludeRegex( 2045 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))"); 2046 SmallVector<StringRef, 4> Matches; 2047 SmallVector<IncludeDirective, 16> IncludesInBlock; 2048 2049 // In compiled files, consider the first #include to be the main #include of 2050 // the file if it is not a system #include. This ensures that the header 2051 // doesn't have hidden dependencies 2052 // (http://llvm.org/docs/CodingStandards.html#include-style). 2053 // 2054 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix 2055 // cases where the first #include is unlikely to be the main header. 2056 bool IsSource = FileName.endswith(".c") || FileName.endswith(".cc") || 2057 FileName.endswith(".cpp") || FileName.endswith(".c++") || 2058 FileName.endswith(".cxx") || FileName.endswith(".m") || 2059 FileName.endswith(".mm"); 2060 StringRef FileStem = llvm::sys::path::stem(FileName); 2061 bool FirstIncludeBlock = true; 2062 bool MainIncludeFound = false; 2063 2064 // Create pre-compiled regular expressions for the #include categories. 2065 SmallVector<llvm::Regex, 4> CategoryRegexs; 2066 for (const auto &Category : Style.IncludeCategories) 2067 CategoryRegexs.emplace_back(Category.Regex); 2068 2069 bool FormattingOff = false; 2070 2071 for (;;) { 2072 auto Pos = Code.find('\n', SearchFrom); 2073 StringRef Line = 2074 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev); 2075 2076 StringRef Trimmed = Line.trim(); 2077 if (Trimmed == "// clang-format off") 2078 FormattingOff = true; 2079 else if (Trimmed == "// clang-format on") 2080 FormattingOff = false; 2081 2082 if (!FormattingOff && !Line.endswith("\\")) { 2083 if (IncludeRegex.match(Line, &Matches)) { 2084 StringRef IncludeName = Matches[2]; 2085 int Category = INT_MAX; 2086 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i) { 2087 if (CategoryRegexs[i].match(IncludeName)) { 2088 Category = Style.IncludeCategories[i].Priority; 2089 break; 2090 } 2091 } 2092 if (IsSource && !MainIncludeFound && Category > 0 && 2093 FirstIncludeBlock && IncludeName.startswith("\"")) { 2094 StringRef HeaderStem = 2095 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1)); 2096 if (FileStem.startswith(HeaderStem)) { 2097 llvm::Regex MainIncludeRegex( 2098 (HeaderStem + Style.IncludeIsMainRegex).str()); 2099 if (MainIncludeRegex.match(FileStem)) { 2100 Category = 0; 2101 MainIncludeFound = true; 2102 } 2103 } 2104 } 2105 IncludesInBlock.push_back({IncludeName, Line, Prev, Category}); 2106 } else if (!IncludesInBlock.empty()) { 2107 sortIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, 2108 Cursor); 2109 IncludesInBlock.clear(); 2110 FirstIncludeBlock = false; 2111 } 2112 Prev = Pos + 1; 2113 } 2114 if (Pos == StringRef::npos || Pos + 1 == Code.size()) 2115 break; 2116 SearchFrom = Pos + 1; 2117 } 2118 if (!IncludesInBlock.empty()) 2119 sortIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor); 2120 return Replaces; 2121 } 2122 2123 template <typename T> 2124 static tooling::Replacements 2125 processReplacements(T ProcessFunc, StringRef Code, 2126 const tooling::Replacements &Replaces, 2127 const FormatStyle &Style) { 2128 if (Replaces.empty()) 2129 return tooling::Replacements(); 2130 2131 std::string NewCode = applyAllReplacements(Code, Replaces); 2132 std::vector<tooling::Range> ChangedRanges = 2133 tooling::calculateChangedRanges(Replaces); 2134 StringRef FileName = Replaces.begin()->getFilePath(); 2135 2136 tooling::Replacements FormatReplaces = 2137 ProcessFunc(Style, NewCode, ChangedRanges, FileName); 2138 2139 return mergeReplacements(Replaces, FormatReplaces); 2140 } 2141 2142 tooling::Replacements formatReplacements(StringRef Code, 2143 const tooling::Replacements &Replaces, 2144 const FormatStyle &Style) { 2145 // We need to use lambda function here since there are two versions of 2146 // `reformat`. 2147 auto Reformat = [](const FormatStyle &Style, StringRef Code, 2148 std::vector<tooling::Range> Ranges, 2149 StringRef FileName) -> tooling::Replacements { 2150 return reformat(Style, Code, Ranges, FileName); 2151 }; 2152 return processReplacements(Reformat, Code, Replaces, Style); 2153 } 2154 2155 tooling::Replacements 2156 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces, 2157 const FormatStyle &Style) { 2158 // We need to use lambda function here since there are two versions of 2159 // `cleanup`. 2160 auto Cleanup = [](const FormatStyle &Style, StringRef Code, 2161 std::vector<tooling::Range> Ranges, 2162 StringRef FileName) -> tooling::Replacements { 2163 return cleanup(Style, Code, Ranges, FileName); 2164 }; 2165 return processReplacements(Cleanup, Code, Replaces, Style); 2166 } 2167 2168 tooling::Replacements reformat(const FormatStyle &Style, SourceManager &SM, 2169 FileID ID, ArrayRef<CharSourceRange> Ranges, 2170 bool *IncompleteFormat) { 2171 FormatStyle Expanded = expandPresets(Style); 2172 if (Expanded.DisableFormat) 2173 return tooling::Replacements(); 2174 2175 Environment Env(SM, ID, Ranges); 2176 Formatter Format(Env, Expanded, IncompleteFormat); 2177 return Format.process(); 2178 } 2179 2180 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 2181 ArrayRef<tooling::Range> Ranges, 2182 StringRef FileName, bool *IncompleteFormat) { 2183 FormatStyle Expanded = expandPresets(Style); 2184 if (Expanded.DisableFormat) 2185 return tooling::Replacements(); 2186 2187 std::unique_ptr<Environment> Env = 2188 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 2189 Formatter Format(*Env, Expanded, IncompleteFormat); 2190 return Format.process(); 2191 } 2192 2193 tooling::Replacements cleanup(const FormatStyle &Style, SourceManager &SM, 2194 FileID ID, ArrayRef<CharSourceRange> Ranges) { 2195 Environment Env(SM, ID, Ranges); 2196 Cleaner Clean(Env, Style); 2197 return Clean.process(); 2198 } 2199 2200 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code, 2201 ArrayRef<tooling::Range> Ranges, 2202 StringRef FileName) { 2203 std::unique_ptr<Environment> Env = 2204 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 2205 Cleaner Clean(*Env, Style); 2206 return Clean.process(); 2207 } 2208 2209 LangOptions getFormattingLangOpts(const FormatStyle &Style) { 2210 LangOptions LangOpts; 2211 LangOpts.CPlusPlus = 1; 2212 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2213 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2214 LangOpts.LineComment = 1; 2215 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp; 2216 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0; 2217 LangOpts.Bool = 1; 2218 LangOpts.ObjC1 = 1; 2219 LangOpts.ObjC2 = 1; 2220 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally. 2221 LangOpts.DeclSpecKeyword = 1; // To get __declspec. 2222 return LangOpts; 2223 } 2224 2225 const char *StyleOptionHelpDescription = 2226 "Coding style, currently supports:\n" 2227 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 2228 "Use -style=file to load style configuration from\n" 2229 ".clang-format file located in one of the parent\n" 2230 "directories of the source file (or current\n" 2231 "directory for stdin).\n" 2232 "Use -style=\"{key: value, ...}\" to set specific\n" 2233 "parameters, e.g.:\n" 2234 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 2235 2236 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 2237 if (FileName.endswith(".java")) 2238 return FormatStyle::LK_Java; 2239 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) 2240 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript. 2241 if (FileName.endswith_lower(".proto") || 2242 FileName.endswith_lower(".protodevel")) 2243 return FormatStyle::LK_Proto; 2244 if (FileName.endswith_lower(".td")) 2245 return FormatStyle::LK_TableGen; 2246 return FormatStyle::LK_Cpp; 2247 } 2248 2249 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 2250 StringRef FallbackStyle, vfs::FileSystem *FS) { 2251 if (!FS) { 2252 FS = vfs::getRealFileSystem().get(); 2253 } 2254 FormatStyle Style = getLLVMStyle(); 2255 Style.Language = getLanguageByFileName(FileName); 2256 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) { 2257 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 2258 << "\" using LLVM style\n"; 2259 return Style; 2260 } 2261 2262 if (StyleName.startswith("{")) { 2263 // Parse YAML/JSON style from the command line. 2264 if (std::error_code ec = parseConfiguration(StyleName, &Style)) { 2265 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 2266 << FallbackStyle << " style\n"; 2267 } 2268 return Style; 2269 } 2270 2271 if (!StyleName.equals_lower("file")) { 2272 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 2273 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 2274 << " style\n"; 2275 return Style; 2276 } 2277 2278 // Look for .clang-format/_clang-format file in the file's parent directories. 2279 SmallString<128> UnsuitableConfigFiles; 2280 SmallString<128> Path(FileName); 2281 llvm::sys::fs::make_absolute(Path); 2282 for (StringRef Directory = Path; !Directory.empty(); 2283 Directory = llvm::sys::path::parent_path(Directory)) { 2284 2285 auto Status = FS->status(Directory); 2286 if (!Status || 2287 Status->getType() != llvm::sys::fs::file_type::directory_file) { 2288 continue; 2289 } 2290 2291 SmallString<128> ConfigFile(Directory); 2292 2293 llvm::sys::path::append(ConfigFile, ".clang-format"); 2294 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 2295 2296 Status = FS->status(ConfigFile.str()); 2297 bool IsFile = 2298 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file); 2299 if (!IsFile) { 2300 // Try _clang-format too, since dotfiles are not commonly used on Windows. 2301 ConfigFile = Directory; 2302 llvm::sys::path::append(ConfigFile, "_clang-format"); 2303 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 2304 Status = FS->status(ConfigFile.str()); 2305 IsFile = Status && 2306 (Status->getType() == llvm::sys::fs::file_type::regular_file); 2307 } 2308 2309 if (IsFile) { 2310 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text = 2311 FS->getBufferForFile(ConfigFile.str()); 2312 if (std::error_code EC = Text.getError()) { 2313 llvm::errs() << EC.message() << "\n"; 2314 break; 2315 } 2316 if (std::error_code ec = 2317 parseConfiguration(Text.get()->getBuffer(), &Style)) { 2318 if (ec == ParseError::Unsuitable) { 2319 if (!UnsuitableConfigFiles.empty()) 2320 UnsuitableConfigFiles.append(", "); 2321 UnsuitableConfigFiles.append(ConfigFile); 2322 continue; 2323 } 2324 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 2325 << "\n"; 2326 break; 2327 } 2328 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 2329 return Style; 2330 } 2331 } 2332 if (!UnsuitableConfigFiles.empty()) { 2333 llvm::errs() << "Configuration file(s) do(es) not support " 2334 << getLanguageName(Style.Language) << ": " 2335 << UnsuitableConfigFiles << "\n"; 2336 } 2337 return Style; 2338 } 2339 2340 } // namespace format 2341 } // namespace clang 2342