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 "FormatTokenLexer.h" 20 #include "SortJavaScriptImports.h" 21 #include "TokenAnalyzer.h" 22 #include "TokenAnnotator.h" 23 #include "UnwrappedLineFormatter.h" 24 #include "UnwrappedLineParser.h" 25 #include "WhitespaceManager.h" 26 #include "clang/Basic/Diagnostic.h" 27 #include "clang/Basic/DiagnosticOptions.h" 28 #include "clang/Basic/SourceManager.h" 29 #include "clang/Basic/VirtualFileSystem.h" 30 #include "clang/Lex/Lexer.h" 31 #include "llvm/ADT/STLExtras.h" 32 #include "llvm/Support/Allocator.h" 33 #include "llvm/Support/Debug.h" 34 #include "llvm/Support/Path.h" 35 #include "llvm/Support/Regex.h" 36 #include "llvm/Support/YAMLTraits.h" 37 #include <algorithm> 38 #include <memory> 39 #include <string> 40 41 #define DEBUG_TYPE "format-formatter" 42 43 using clang::format::FormatStyle; 44 45 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string) 46 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory) 47 48 namespace llvm { 49 namespace yaml { 50 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> { 51 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) { 52 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp); 53 IO.enumCase(Value, "Java", FormatStyle::LK_Java); 54 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript); 55 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto); 56 IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen); 57 } 58 }; 59 60 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> { 61 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) { 62 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03); 63 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03); 64 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11); 65 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11); 66 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto); 67 } 68 }; 69 70 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> { 71 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) { 72 IO.enumCase(Value, "Never", FormatStyle::UT_Never); 73 IO.enumCase(Value, "false", FormatStyle::UT_Never); 74 IO.enumCase(Value, "Always", FormatStyle::UT_Always); 75 IO.enumCase(Value, "true", FormatStyle::UT_Always); 76 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation); 77 IO.enumCase(Value, "ForContinuationAndIndentation", 78 FormatStyle::UT_ForContinuationAndIndentation); 79 } 80 }; 81 82 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> { 83 static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) { 84 IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave); 85 IO.enumCase(Value, "Single", FormatStyle::JSQS_Single); 86 IO.enumCase(Value, "Double", FormatStyle::JSQS_Double); 87 } 88 }; 89 90 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> { 91 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) { 92 IO.enumCase(Value, "None", FormatStyle::SFS_None); 93 IO.enumCase(Value, "false", FormatStyle::SFS_None); 94 IO.enumCase(Value, "All", FormatStyle::SFS_All); 95 IO.enumCase(Value, "true", FormatStyle::SFS_All); 96 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline); 97 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty); 98 } 99 }; 100 101 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> { 102 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) { 103 IO.enumCase(Value, "All", FormatStyle::BOS_All); 104 IO.enumCase(Value, "true", FormatStyle::BOS_All); 105 IO.enumCase(Value, "None", FormatStyle::BOS_None); 106 IO.enumCase(Value, "false", FormatStyle::BOS_None); 107 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment); 108 } 109 }; 110 111 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> { 112 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) { 113 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach); 114 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux); 115 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla); 116 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup); 117 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman); 118 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU); 119 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit); 120 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom); 121 } 122 }; 123 124 template <> 125 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> { 126 static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) { 127 IO.enumCase(Value, "None", FormatStyle::RTBS_None); 128 IO.enumCase(Value, "All", FormatStyle::RTBS_All); 129 IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel); 130 IO.enumCase(Value, "TopLevelDefinitions", 131 FormatStyle::RTBS_TopLevelDefinitions); 132 IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions); 133 } 134 }; 135 136 template <> 137 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> { 138 static void 139 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) { 140 IO.enumCase(Value, "None", FormatStyle::DRTBS_None); 141 IO.enumCase(Value, "All", FormatStyle::DRTBS_All); 142 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel); 143 144 // For backward compatibility. 145 IO.enumCase(Value, "false", FormatStyle::DRTBS_None); 146 IO.enumCase(Value, "true", FormatStyle::DRTBS_All); 147 } 148 }; 149 150 template <> 151 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> { 152 static void enumeration(IO &IO, 153 FormatStyle::NamespaceIndentationKind &Value) { 154 IO.enumCase(Value, "None", FormatStyle::NI_None); 155 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner); 156 IO.enumCase(Value, "All", FormatStyle::NI_All); 157 } 158 }; 159 160 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> { 161 static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) { 162 IO.enumCase(Value, "Align", FormatStyle::BAS_Align); 163 IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign); 164 IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak); 165 166 // For backward compatibility. 167 IO.enumCase(Value, "true", FormatStyle::BAS_Align); 168 IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign); 169 } 170 }; 171 172 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> { 173 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) { 174 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle); 175 IO.enumCase(Value, "Left", FormatStyle::PAS_Left); 176 IO.enumCase(Value, "Right", FormatStyle::PAS_Right); 177 178 // For backward compatibility. 179 IO.enumCase(Value, "true", FormatStyle::PAS_Left); 180 IO.enumCase(Value, "false", FormatStyle::PAS_Right); 181 } 182 }; 183 184 template <> 185 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> { 186 static void enumeration(IO &IO, 187 FormatStyle::SpaceBeforeParensOptions &Value) { 188 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never); 189 IO.enumCase(Value, "ControlStatements", 190 FormatStyle::SBPO_ControlStatements); 191 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always); 192 193 // For backward compatibility. 194 IO.enumCase(Value, "false", FormatStyle::SBPO_Never); 195 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements); 196 } 197 }; 198 199 template <> struct MappingTraits<FormatStyle> { 200 static void mapping(IO &IO, FormatStyle &Style) { 201 // When reading, read the language first, we need it for getPredefinedStyle. 202 IO.mapOptional("Language", Style.Language); 203 204 if (IO.outputting()) { 205 StringRef StylesArray[] = {"LLVM", "Google", "Chromium", 206 "Mozilla", "WebKit", "GNU"}; 207 ArrayRef<StringRef> Styles(StylesArray); 208 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 209 StringRef StyleName(Styles[i]); 210 FormatStyle PredefinedStyle; 211 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) && 212 Style == PredefinedStyle) { 213 IO.mapOptional("# BasedOnStyle", StyleName); 214 break; 215 } 216 } 217 } else { 218 StringRef BasedOnStyle; 219 IO.mapOptional("BasedOnStyle", BasedOnStyle); 220 if (!BasedOnStyle.empty()) { 221 FormatStyle::LanguageKind OldLanguage = Style.Language; 222 FormatStyle::LanguageKind Language = 223 ((FormatStyle *)IO.getContext())->Language; 224 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) { 225 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 226 return; 227 } 228 Style.Language = OldLanguage; 229 } 230 } 231 232 // For backward compatibility. 233 if (!IO.outputting()) { 234 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment); 235 IO.mapOptional("IndentFunctionDeclarationAfterType", 236 Style.IndentWrappedFunctionNames); 237 IO.mapOptional("PointerBindsToType", Style.PointerAlignment); 238 IO.mapOptional("SpaceAfterControlStatementKeyword", 239 Style.SpaceBeforeParens); 240 } 241 242 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 243 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket); 244 IO.mapOptional("AlignConsecutiveAssignments", 245 Style.AlignConsecutiveAssignments); 246 IO.mapOptional("AlignConsecutiveDeclarations", 247 Style.AlignConsecutiveDeclarations); 248 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 249 IO.mapOptional("AlignOperands", Style.AlignOperands); 250 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 251 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 252 Style.AllowAllParametersOfDeclarationOnNextLine); 253 IO.mapOptional("AllowShortBlocksOnASingleLine", 254 Style.AllowShortBlocksOnASingleLine); 255 IO.mapOptional("AllowShortCaseLabelsOnASingleLine", 256 Style.AllowShortCaseLabelsOnASingleLine); 257 IO.mapOptional("AllowShortFunctionsOnASingleLine", 258 Style.AllowShortFunctionsOnASingleLine); 259 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 260 Style.AllowShortIfStatementsOnASingleLine); 261 IO.mapOptional("AllowShortLoopsOnASingleLine", 262 Style.AllowShortLoopsOnASingleLine); 263 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType", 264 Style.AlwaysBreakAfterDefinitionReturnType); 265 IO.mapOptional("AlwaysBreakAfterReturnType", 266 Style.AlwaysBreakAfterReturnType); 267 // If AlwaysBreakAfterDefinitionReturnType was specified but 268 // AlwaysBreakAfterReturnType was not, initialize the latter from the 269 // former for backwards compatibility. 270 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None && 271 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) { 272 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All) 273 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 274 else if (Style.AlwaysBreakAfterDefinitionReturnType == 275 FormatStyle::DRTBS_TopLevel) 276 Style.AlwaysBreakAfterReturnType = 277 FormatStyle::RTBS_TopLevelDefinitions; 278 } 279 280 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 281 Style.AlwaysBreakBeforeMultilineStrings); 282 IO.mapOptional("AlwaysBreakTemplateDeclarations", 283 Style.AlwaysBreakTemplateDeclarations); 284 IO.mapOptional("BinPackArguments", Style.BinPackArguments); 285 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 286 IO.mapOptional("BraceWrapping", Style.BraceWrapping); 287 IO.mapOptional("BreakBeforeBinaryOperators", 288 Style.BreakBeforeBinaryOperators); 289 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 290 IO.mapOptional("BreakBeforeTernaryOperators", 291 Style.BreakBeforeTernaryOperators); 292 IO.mapOptional("BreakConstructorInitializersBeforeComma", 293 Style.BreakConstructorInitializersBeforeComma); 294 IO.mapOptional("BreakAfterJavaFieldAnnotations", 295 Style.BreakAfterJavaFieldAnnotations); 296 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals); 297 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 298 IO.mapOptional("CommentPragmas", Style.CommentPragmas); 299 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 300 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 301 IO.mapOptional("ConstructorInitializerIndentWidth", 302 Style.ConstructorInitializerIndentWidth); 303 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 304 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 305 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment); 306 IO.mapOptional("DisableFormat", Style.DisableFormat); 307 IO.mapOptional("ExperimentalAutoDetectBinPacking", 308 Style.ExperimentalAutoDetectBinPacking); 309 IO.mapOptional("ForEachMacros", Style.ForEachMacros); 310 IO.mapOptional("IncludeCategories", Style.IncludeCategories); 311 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex); 312 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 313 IO.mapOptional("IndentWidth", Style.IndentWidth); 314 IO.mapOptional("IndentWrappedFunctionNames", 315 Style.IndentWrappedFunctionNames); 316 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes); 317 IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports); 318 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks", 319 Style.KeepEmptyLinesAtTheStartOfBlocks); 320 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin); 321 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd); 322 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 323 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 324 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth); 325 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty); 326 IO.mapOptional("ObjCSpaceBeforeProtocolList", 327 Style.ObjCSpaceBeforeProtocolList); 328 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 329 Style.PenaltyBreakBeforeFirstCallParameter); 330 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 331 IO.mapOptional("PenaltyBreakFirstLessLess", 332 Style.PenaltyBreakFirstLessLess); 333 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 334 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 335 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 336 Style.PenaltyReturnTypeOnItsOwnLine); 337 IO.mapOptional("PointerAlignment", Style.PointerAlignment); 338 IO.mapOptional("ReflowComments", Style.ReflowComments); 339 IO.mapOptional("SortIncludes", Style.SortIncludes); 340 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast); 341 IO.mapOptional("SpaceAfterTemplateKeyword", Style.SpaceAfterTemplateKeyword); 342 IO.mapOptional("SpaceBeforeAssignmentOperators", 343 Style.SpaceBeforeAssignmentOperators); 344 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens); 345 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 346 IO.mapOptional("SpacesBeforeTrailingComments", 347 Style.SpacesBeforeTrailingComments); 348 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 349 IO.mapOptional("SpacesInContainerLiterals", 350 Style.SpacesInContainerLiterals); 351 IO.mapOptional("SpacesInCStyleCastParentheses", 352 Style.SpacesInCStyleCastParentheses); 353 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 354 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets); 355 IO.mapOptional("Standard", Style.Standard); 356 IO.mapOptional("TabWidth", Style.TabWidth); 357 IO.mapOptional("UseTab", Style.UseTab); 358 } 359 }; 360 361 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> { 362 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) { 363 IO.mapOptional("AfterClass", Wrapping.AfterClass); 364 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement); 365 IO.mapOptional("AfterEnum", Wrapping.AfterEnum); 366 IO.mapOptional("AfterFunction", Wrapping.AfterFunction); 367 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace); 368 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration); 369 IO.mapOptional("AfterStruct", Wrapping.AfterStruct); 370 IO.mapOptional("AfterUnion", Wrapping.AfterUnion); 371 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch); 372 IO.mapOptional("BeforeElse", Wrapping.BeforeElse); 373 IO.mapOptional("IndentBraces", Wrapping.IndentBraces); 374 } 375 }; 376 377 template <> struct MappingTraits<FormatStyle::IncludeCategory> { 378 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) { 379 IO.mapOptional("Regex", Category.Regex); 380 IO.mapOptional("Priority", Category.Priority); 381 } 382 }; 383 384 // Allows to read vector<FormatStyle> while keeping default values. 385 // IO.getContext() should contain a pointer to the FormatStyle structure, that 386 // will be used to get default values for missing keys. 387 // If the first element has no Language specified, it will be treated as the 388 // default one for the following elements. 389 template <> struct DocumentListTraits<std::vector<FormatStyle>> { 390 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) { 391 return Seq.size(); 392 } 393 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq, 394 size_t Index) { 395 if (Index >= Seq.size()) { 396 assert(Index == Seq.size()); 397 FormatStyle Template; 398 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) { 399 Template = Seq[0]; 400 } else { 401 Template = *((const FormatStyle *)IO.getContext()); 402 Template.Language = FormatStyle::LK_None; 403 } 404 Seq.resize(Index + 1, Template); 405 } 406 return Seq[Index]; 407 } 408 }; 409 } // namespace yaml 410 } // namespace llvm 411 412 namespace clang { 413 namespace format { 414 415 const std::error_category &getParseCategory() { 416 static ParseErrorCategory C; 417 return C; 418 } 419 std::error_code make_error_code(ParseError e) { 420 return std::error_code(static_cast<int>(e), getParseCategory()); 421 } 422 423 const char *ParseErrorCategory::name() const noexcept { 424 return "clang-format.parse_error"; 425 } 426 427 std::string ParseErrorCategory::message(int EV) const { 428 switch (static_cast<ParseError>(EV)) { 429 case ParseError::Success: 430 return "Success"; 431 case ParseError::Error: 432 return "Invalid argument"; 433 case ParseError::Unsuitable: 434 return "Unsuitable"; 435 } 436 llvm_unreachable("unexpected parse error"); 437 } 438 439 static FormatStyle expandPresets(const FormatStyle &Style) { 440 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom) 441 return Style; 442 FormatStyle Expanded = Style; 443 Expanded.BraceWrapping = {false, false, false, false, false, false, 444 false, false, false, false, false}; 445 switch (Style.BreakBeforeBraces) { 446 case FormatStyle::BS_Linux: 447 Expanded.BraceWrapping.AfterClass = true; 448 Expanded.BraceWrapping.AfterFunction = true; 449 Expanded.BraceWrapping.AfterNamespace = true; 450 break; 451 case FormatStyle::BS_Mozilla: 452 Expanded.BraceWrapping.AfterClass = true; 453 Expanded.BraceWrapping.AfterEnum = true; 454 Expanded.BraceWrapping.AfterFunction = true; 455 Expanded.BraceWrapping.AfterStruct = true; 456 Expanded.BraceWrapping.AfterUnion = true; 457 break; 458 case FormatStyle::BS_Stroustrup: 459 Expanded.BraceWrapping.AfterFunction = true; 460 Expanded.BraceWrapping.BeforeCatch = true; 461 Expanded.BraceWrapping.BeforeElse = true; 462 break; 463 case FormatStyle::BS_Allman: 464 Expanded.BraceWrapping.AfterClass = true; 465 Expanded.BraceWrapping.AfterControlStatement = true; 466 Expanded.BraceWrapping.AfterEnum = true; 467 Expanded.BraceWrapping.AfterFunction = true; 468 Expanded.BraceWrapping.AfterNamespace = true; 469 Expanded.BraceWrapping.AfterObjCDeclaration = true; 470 Expanded.BraceWrapping.AfterStruct = true; 471 Expanded.BraceWrapping.BeforeCatch = true; 472 Expanded.BraceWrapping.BeforeElse = true; 473 break; 474 case FormatStyle::BS_GNU: 475 Expanded.BraceWrapping = {true, true, true, true, true, true, 476 true, true, true, true, true}; 477 break; 478 case FormatStyle::BS_WebKit: 479 Expanded.BraceWrapping.AfterFunction = true; 480 break; 481 default: 482 break; 483 } 484 return Expanded; 485 } 486 487 FormatStyle getLLVMStyle() { 488 FormatStyle LLVMStyle; 489 LLVMStyle.Language = FormatStyle::LK_Cpp; 490 LLVMStyle.AccessModifierOffset = -2; 491 LLVMStyle.AlignEscapedNewlinesLeft = false; 492 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align; 493 LLVMStyle.AlignOperands = true; 494 LLVMStyle.AlignTrailingComments = true; 495 LLVMStyle.AlignConsecutiveAssignments = false; 496 LLVMStyle.AlignConsecutiveDeclarations = false; 497 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 498 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All; 499 LLVMStyle.AllowShortBlocksOnASingleLine = false; 500 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false; 501 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 502 LLVMStyle.AllowShortLoopsOnASingleLine = false; 503 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None; 504 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None; 505 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 506 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 507 LLVMStyle.BinPackParameters = true; 508 LLVMStyle.BinPackArguments = true; 509 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None; 510 LLVMStyle.BreakBeforeTernaryOperators = true; 511 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 512 LLVMStyle.BraceWrapping = {false, false, false, false, false, false, 513 false, false, false, false, false}; 514 LLVMStyle.BreakAfterJavaFieldAnnotations = false; 515 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 516 LLVMStyle.BreakStringLiterals = true; 517 LLVMStyle.ColumnLimit = 80; 518 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; 519 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 520 LLVMStyle.ConstructorInitializerIndentWidth = 4; 521 LLVMStyle.ContinuationIndentWidth = 4; 522 LLVMStyle.Cpp11BracedListStyle = true; 523 LLVMStyle.DerivePointerAlignment = false; 524 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 525 LLVMStyle.ForEachMacros.push_back("foreach"); 526 LLVMStyle.ForEachMacros.push_back("Q_FOREACH"); 527 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH"); 528 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2}, 529 {"^(<|\"(gtest|isl|json)/)", 3}, 530 {".*", 1}}; 531 LLVMStyle.IncludeIsMainRegex = "$"; 532 LLVMStyle.IndentCaseLabels = false; 533 LLVMStyle.IndentWrappedFunctionNames = false; 534 LLVMStyle.IndentWidth = 2; 535 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave; 536 LLVMStyle.JavaScriptWrapImports = true; 537 LLVMStyle.TabWidth = 8; 538 LLVMStyle.MaxEmptyLinesToKeep = 1; 539 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true; 540 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 541 LLVMStyle.ObjCBlockIndentWidth = 2; 542 LLVMStyle.ObjCSpaceAfterProperty = false; 543 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 544 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right; 545 LLVMStyle.SpacesBeforeTrailingComments = 1; 546 LLVMStyle.Standard = FormatStyle::LS_Cpp11; 547 LLVMStyle.UseTab = FormatStyle::UT_Never; 548 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave; 549 LLVMStyle.ReflowComments = true; 550 LLVMStyle.SpacesInParentheses = false; 551 LLVMStyle.SpacesInSquareBrackets = false; 552 LLVMStyle.SpaceInEmptyParentheses = false; 553 LLVMStyle.SpacesInContainerLiterals = true; 554 LLVMStyle.SpacesInCStyleCastParentheses = false; 555 LLVMStyle.SpaceAfterCStyleCast = false; 556 LLVMStyle.SpaceAfterTemplateKeyword = true; 557 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements; 558 LLVMStyle.SpaceBeforeAssignmentOperators = true; 559 LLVMStyle.SpacesInAngles = false; 560 561 LLVMStyle.PenaltyBreakComment = 300; 562 LLVMStyle.PenaltyBreakFirstLessLess = 120; 563 LLVMStyle.PenaltyBreakString = 1000; 564 LLVMStyle.PenaltyExcessCharacter = 1000000; 565 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 566 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 567 568 LLVMStyle.DisableFormat = false; 569 LLVMStyle.SortIncludes = true; 570 571 return LLVMStyle; 572 } 573 574 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { 575 FormatStyle GoogleStyle = getLLVMStyle(); 576 GoogleStyle.Language = Language; 577 578 GoogleStyle.AccessModifierOffset = -1; 579 GoogleStyle.AlignEscapedNewlinesLeft = true; 580 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 581 GoogleStyle.AllowShortLoopsOnASingleLine = true; 582 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 583 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 584 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 585 GoogleStyle.DerivePointerAlignment = true; 586 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}}; 587 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$"; 588 GoogleStyle.IndentCaseLabels = true; 589 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false; 590 GoogleStyle.ObjCSpaceAfterProperty = false; 591 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 592 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left; 593 GoogleStyle.SpacesBeforeTrailingComments = 2; 594 GoogleStyle.Standard = FormatStyle::LS_Auto; 595 596 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 597 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 598 599 if (Language == FormatStyle::LK_Java) { 600 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 601 GoogleStyle.AlignOperands = false; 602 GoogleStyle.AlignTrailingComments = false; 603 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 604 GoogleStyle.AllowShortIfStatementsOnASingleLine = false; 605 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 606 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment; 607 GoogleStyle.ColumnLimit = 100; 608 GoogleStyle.SpaceAfterCStyleCast = true; 609 GoogleStyle.SpacesBeforeTrailingComments = 1; 610 } else if (Language == FormatStyle::LK_JavaScript) { 611 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak; 612 GoogleStyle.AlignOperands = false; 613 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 614 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 615 GoogleStyle.BreakBeforeTernaryOperators = false; 616 GoogleStyle.CommentPragmas = 617 "(taze:|@(export|requirecss|return|returns|see|visibility)) "; 618 GoogleStyle.MaxEmptyLinesToKeep = 3; 619 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All; 620 GoogleStyle.SpacesInContainerLiterals = false; 621 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single; 622 GoogleStyle.JavaScriptWrapImports = false; 623 } else if (Language == FormatStyle::LK_Proto) { 624 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None; 625 GoogleStyle.SpacesInContainerLiterals = false; 626 } 627 628 return GoogleStyle; 629 } 630 631 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) { 632 FormatStyle ChromiumStyle = getGoogleStyle(Language); 633 if (Language == FormatStyle::LK_Java) { 634 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true; 635 ChromiumStyle.BreakAfterJavaFieldAnnotations = true; 636 ChromiumStyle.ContinuationIndentWidth = 8; 637 ChromiumStyle.IndentWidth = 4; 638 } else { 639 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 640 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 641 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 642 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 643 ChromiumStyle.BinPackParameters = false; 644 ChromiumStyle.DerivePointerAlignment = false; 645 } 646 ChromiumStyle.SortIncludes = false; 647 return ChromiumStyle; 648 } 649 650 FormatStyle getMozillaStyle() { 651 FormatStyle MozillaStyle = getLLVMStyle(); 652 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 653 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 654 MozillaStyle.AlwaysBreakAfterReturnType = 655 FormatStyle::RTBS_TopLevelDefinitions; 656 MozillaStyle.AlwaysBreakAfterDefinitionReturnType = 657 FormatStyle::DRTBS_TopLevel; 658 MozillaStyle.AlwaysBreakTemplateDeclarations = true; 659 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla; 660 MozillaStyle.BreakConstructorInitializersBeforeComma = true; 661 MozillaStyle.ConstructorInitializerIndentWidth = 2; 662 MozillaStyle.ContinuationIndentWidth = 2; 663 MozillaStyle.Cpp11BracedListStyle = false; 664 MozillaStyle.IndentCaseLabels = true; 665 MozillaStyle.ObjCSpaceAfterProperty = true; 666 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 667 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 668 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left; 669 MozillaStyle.SpaceAfterTemplateKeyword = false; 670 return MozillaStyle; 671 } 672 673 FormatStyle getWebKitStyle() { 674 FormatStyle Style = getLLVMStyle(); 675 Style.AccessModifierOffset = -4; 676 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 677 Style.AlignOperands = false; 678 Style.AlignTrailingComments = false; 679 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 680 Style.BreakBeforeBraces = FormatStyle::BS_WebKit; 681 Style.BreakConstructorInitializersBeforeComma = true; 682 Style.Cpp11BracedListStyle = false; 683 Style.ColumnLimit = 0; 684 Style.IndentWidth = 4; 685 Style.NamespaceIndentation = FormatStyle::NI_Inner; 686 Style.ObjCBlockIndentWidth = 4; 687 Style.ObjCSpaceAfterProperty = true; 688 Style.PointerAlignment = FormatStyle::PAS_Left; 689 return Style; 690 } 691 692 FormatStyle getGNUStyle() { 693 FormatStyle Style = getLLVMStyle(); 694 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All; 695 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 696 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 697 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 698 Style.BreakBeforeTernaryOperators = true; 699 Style.Cpp11BracedListStyle = false; 700 Style.ColumnLimit = 79; 701 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 702 Style.Standard = FormatStyle::LS_Cpp03; 703 return Style; 704 } 705 706 FormatStyle getNoStyle() { 707 FormatStyle NoStyle = getLLVMStyle(); 708 NoStyle.DisableFormat = true; 709 NoStyle.SortIncludes = false; 710 return NoStyle; 711 } 712 713 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 714 FormatStyle *Style) { 715 if (Name.equals_lower("llvm")) { 716 *Style = getLLVMStyle(); 717 } else if (Name.equals_lower("chromium")) { 718 *Style = getChromiumStyle(Language); 719 } else if (Name.equals_lower("mozilla")) { 720 *Style = getMozillaStyle(); 721 } else if (Name.equals_lower("google")) { 722 *Style = getGoogleStyle(Language); 723 } else if (Name.equals_lower("webkit")) { 724 *Style = getWebKitStyle(); 725 } else if (Name.equals_lower("gnu")) { 726 *Style = getGNUStyle(); 727 } else if (Name.equals_lower("none")) { 728 *Style = getNoStyle(); 729 } else { 730 return false; 731 } 732 733 Style->Language = Language; 734 return true; 735 } 736 737 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 738 assert(Style); 739 FormatStyle::LanguageKind Language = Style->Language; 740 assert(Language != FormatStyle::LK_None); 741 if (Text.trim().empty()) 742 return make_error_code(ParseError::Error); 743 744 std::vector<FormatStyle> Styles; 745 llvm::yaml::Input Input(Text); 746 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 747 // values for the fields, keys for which are missing from the configuration. 748 // Mapping also uses the context to get the language to find the correct 749 // base style. 750 Input.setContext(Style); 751 Input >> Styles; 752 if (Input.error()) 753 return Input.error(); 754 755 for (unsigned i = 0; i < Styles.size(); ++i) { 756 // Ensures that only the first configuration can skip the Language option. 757 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 758 return make_error_code(ParseError::Error); 759 // Ensure that each language is configured at most once. 760 for (unsigned j = 0; j < i; ++j) { 761 if (Styles[i].Language == Styles[j].Language) { 762 DEBUG(llvm::dbgs() 763 << "Duplicate languages in the config file on positions " << j 764 << " and " << i << "\n"); 765 return make_error_code(ParseError::Error); 766 } 767 } 768 } 769 // Look for a suitable configuration starting from the end, so we can 770 // find the configuration for the specific language first, and the default 771 // configuration (which can only be at slot 0) after it. 772 for (int i = Styles.size() - 1; i >= 0; --i) { 773 if (Styles[i].Language == Language || 774 Styles[i].Language == FormatStyle::LK_None) { 775 *Style = Styles[i]; 776 Style->Language = Language; 777 return make_error_code(ParseError::Success); 778 } 779 } 780 return make_error_code(ParseError::Unsuitable); 781 } 782 783 std::string configurationAsText(const FormatStyle &Style) { 784 std::string Text; 785 llvm::raw_string_ostream Stream(Text); 786 llvm::yaml::Output Output(Stream); 787 // We use the same mapping method for input and output, so we need a non-const 788 // reference here. 789 FormatStyle NonConstStyle = expandPresets(Style); 790 Output << NonConstStyle; 791 return Stream.str(); 792 } 793 794 namespace { 795 796 class JavaScriptRequoter : public TokenAnalyzer { 797 public: 798 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style) 799 : TokenAnalyzer(Env, Style) {} 800 801 tooling::Replacements 802 analyze(TokenAnnotator &Annotator, 803 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 804 FormatTokenLexer &Tokens) override { 805 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 806 AnnotatedLines.end()); 807 tooling::Replacements Result; 808 requoteJSStringLiteral(AnnotatedLines, Result); 809 return Result; 810 } 811 812 private: 813 // Replaces double/single-quoted string literal as appropriate, re-escaping 814 // the contents in the process. 815 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines, 816 tooling::Replacements &Result) { 817 for (AnnotatedLine *Line : Lines) { 818 requoteJSStringLiteral(Line->Children, Result); 819 if (!Line->Affected) 820 continue; 821 for (FormatToken *FormatTok = Line->First; FormatTok; 822 FormatTok = FormatTok->Next) { 823 StringRef Input = FormatTok->TokenText; 824 if (FormatTok->Finalized || !FormatTok->isStringLiteral() || 825 // NB: testing for not starting with a double quote to avoid 826 // breaking `template strings`. 827 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single && 828 !Input.startswith("\"")) || 829 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double && 830 !Input.startswith("\'"))) 831 continue; 832 833 // Change start and end quote. 834 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single; 835 SourceLocation Start = FormatTok->Tok.getLocation(); 836 auto Replace = [&](SourceLocation Start, unsigned Length, 837 StringRef ReplacementText) { 838 auto Err = Result.add(tooling::Replacement( 839 Env.getSourceManager(), Start, Length, ReplacementText)); 840 // FIXME: handle error. For now, print error message and skip the 841 // replacement for release version. 842 if (Err) 843 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 844 assert(!Err); 845 }; 846 Replace(Start, 1, IsSingle ? "'" : "\""); 847 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1, 848 IsSingle ? "'" : "\""); 849 850 // Escape internal quotes. 851 bool Escaped = false; 852 for (size_t i = 1; i < Input.size() - 1; i++) { 853 switch (Input[i]) { 854 case '\\': 855 if (!Escaped && i + 1 < Input.size() && 856 ((IsSingle && Input[i + 1] == '"') || 857 (!IsSingle && Input[i + 1] == '\''))) { 858 // Remove this \, it's escaping a " or ' that no longer needs 859 // escaping 860 Replace(Start.getLocWithOffset(i), 1, ""); 861 continue; 862 } 863 Escaped = !Escaped; 864 break; 865 case '\"': 866 case '\'': 867 if (!Escaped && IsSingle == (Input[i] == '\'')) { 868 // Escape the quote. 869 Replace(Start.getLocWithOffset(i), 0, "\\"); 870 } 871 Escaped = false; 872 break; 873 default: 874 Escaped = false; 875 break; 876 } 877 } 878 } 879 } 880 } 881 }; 882 883 class Formatter : public TokenAnalyzer { 884 public: 885 Formatter(const Environment &Env, const FormatStyle &Style, 886 bool *IncompleteFormat) 887 : TokenAnalyzer(Env, Style), IncompleteFormat(IncompleteFormat) {} 888 889 tooling::Replacements 890 analyze(TokenAnnotator &Annotator, 891 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 892 FormatTokenLexer &Tokens) override { 893 tooling::Replacements Result; 894 deriveLocalStyle(AnnotatedLines); 895 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 896 AnnotatedLines.end()); 897 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 898 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 899 } 900 Annotator.setCommentLineLevels(AnnotatedLines); 901 902 WhitespaceManager Whitespaces( 903 Env.getSourceManager(), Style, 904 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID()))); 905 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), 906 Env.getSourceManager(), Whitespaces, Encoding, 907 BinPackInconclusiveFunctions); 908 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(), 909 IncompleteFormat) 910 .format(AnnotatedLines); 911 for (const auto &R : Whitespaces.generateReplacements()) 912 if (Result.add(R)) 913 return Result; 914 return Result; 915 } 916 917 private: 918 919 static bool inputUsesCRLF(StringRef Text) { 920 return Text.count('\r') * 2 > Text.count('\n'); 921 } 922 923 bool 924 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) { 925 for (const AnnotatedLine *Line : Lines) { 926 if (hasCpp03IncompatibleFormat(Line->Children)) 927 return true; 928 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) { 929 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 930 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener)) 931 return true; 932 if (Tok->is(TT_TemplateCloser) && 933 Tok->Previous->is(TT_TemplateCloser)) 934 return true; 935 } 936 } 937 } 938 return false; 939 } 940 941 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) { 942 int AlignmentDiff = 0; 943 for (const AnnotatedLine *Line : Lines) { 944 AlignmentDiff += countVariableAlignments(Line->Children); 945 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) { 946 if (!Tok->is(TT_PointerOrReference)) 947 continue; 948 bool SpaceBefore = 949 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 950 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() != 951 Tok->Next->WhitespaceRange.getEnd(); 952 if (SpaceBefore && !SpaceAfter) 953 ++AlignmentDiff; 954 if (!SpaceBefore && SpaceAfter) 955 --AlignmentDiff; 956 } 957 } 958 return AlignmentDiff; 959 } 960 961 void 962 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 963 bool HasBinPackedFunction = false; 964 bool HasOnePerLineFunction = false; 965 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 966 if (!AnnotatedLines[i]->First->Next) 967 continue; 968 FormatToken *Tok = AnnotatedLines[i]->First->Next; 969 while (Tok->Next) { 970 if (Tok->PackingKind == PPK_BinPacked) 971 HasBinPackedFunction = true; 972 if (Tok->PackingKind == PPK_OnePerLine) 973 HasOnePerLineFunction = true; 974 975 Tok = Tok->Next; 976 } 977 } 978 if (Style.DerivePointerAlignment) 979 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0 980 ? FormatStyle::PAS_Left 981 : FormatStyle::PAS_Right; 982 if (Style.Standard == FormatStyle::LS_Auto) 983 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines) 984 ? FormatStyle::LS_Cpp11 985 : FormatStyle::LS_Cpp03; 986 BinPackInconclusiveFunctions = 987 HasBinPackedFunction || !HasOnePerLineFunction; 988 } 989 990 bool BinPackInconclusiveFunctions; 991 bool *IncompleteFormat; 992 }; 993 994 // This class clean up the erroneous/redundant code around the given ranges in 995 // file. 996 class Cleaner : public TokenAnalyzer { 997 public: 998 Cleaner(const Environment &Env, const FormatStyle &Style) 999 : TokenAnalyzer(Env, Style), 1000 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {} 1001 1002 // FIXME: eliminate unused parameters. 1003 tooling::Replacements 1004 analyze(TokenAnnotator &Annotator, 1005 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1006 FormatTokenLexer &Tokens) override { 1007 // FIXME: in the current implementation the granularity of affected range 1008 // is an annotated line. However, this is not sufficient. Furthermore, 1009 // redundant code introduced by replacements does not necessarily 1010 // intercept with ranges of replacements that result in the redundancy. 1011 // To determine if some redundant code is actually introduced by 1012 // replacements(e.g. deletions), we need to come up with a more 1013 // sophisticated way of computing affected ranges. 1014 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1015 AnnotatedLines.end()); 1016 1017 checkEmptyNamespace(AnnotatedLines); 1018 1019 for (auto &Line : AnnotatedLines) { 1020 if (Line->Affected) { 1021 cleanupRight(Line->First, tok::comma, tok::comma); 1022 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma); 1023 cleanupRight(Line->First, tok::l_paren, tok::comma); 1024 cleanupLeft(Line->First, tok::comma, tok::r_paren); 1025 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace); 1026 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace); 1027 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal); 1028 } 1029 } 1030 1031 return generateFixes(); 1032 } 1033 1034 private: 1035 bool containsOnlyComments(const AnnotatedLine &Line) { 1036 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) { 1037 if (Tok->isNot(tok::comment)) 1038 return false; 1039 } 1040 return true; 1041 } 1042 1043 // Iterate through all lines and remove any empty (nested) namespaces. 1044 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1045 std::set<unsigned> DeletedLines; 1046 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1047 auto &Line = *AnnotatedLines[i]; 1048 if (Line.startsWith(tok::kw_namespace) || 1049 Line.startsWith(tok::kw_inline, tok::kw_namespace)) { 1050 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines); 1051 } 1052 } 1053 1054 for (auto Line : DeletedLines) { 1055 FormatToken *Tok = AnnotatedLines[Line]->First; 1056 while (Tok) { 1057 deleteToken(Tok); 1058 Tok = Tok->Next; 1059 } 1060 } 1061 } 1062 1063 // The function checks if the namespace, which starts from \p CurrentLine, and 1064 // its nested namespaces are empty and delete them if they are empty. It also 1065 // sets \p NewLine to the last line checked. 1066 // Returns true if the current namespace is empty. 1067 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1068 unsigned CurrentLine, unsigned &NewLine, 1069 std::set<unsigned> &DeletedLines) { 1070 unsigned InitLine = CurrentLine, End = AnnotatedLines.size(); 1071 if (Style.BraceWrapping.AfterNamespace) { 1072 // If the left brace is in a new line, we should consume it first so that 1073 // it does not make the namespace non-empty. 1074 // FIXME: error handling if there is no left brace. 1075 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) { 1076 NewLine = CurrentLine; 1077 return false; 1078 } 1079 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) { 1080 return false; 1081 } 1082 while (++CurrentLine < End) { 1083 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace)) 1084 break; 1085 1086 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) || 1087 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline, 1088 tok::kw_namespace)) { 1089 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine, 1090 DeletedLines)) 1091 return false; 1092 CurrentLine = NewLine; 1093 continue; 1094 } 1095 1096 if (containsOnlyComments(*AnnotatedLines[CurrentLine])) 1097 continue; 1098 1099 // If there is anything other than comments or nested namespaces in the 1100 // current namespace, the namespace cannot be empty. 1101 NewLine = CurrentLine; 1102 return false; 1103 } 1104 1105 NewLine = CurrentLine; 1106 if (CurrentLine >= End) 1107 return false; 1108 1109 // Check if the empty namespace is actually affected by changed ranges. 1110 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange( 1111 AnnotatedLines[InitLine]->First->Tok.getLocation(), 1112 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc()))) 1113 return false; 1114 1115 for (unsigned i = InitLine; i <= CurrentLine; ++i) { 1116 DeletedLines.insert(i); 1117 } 1118 1119 return true; 1120 } 1121 1122 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one 1123 // of the token in the pair if the left token has \p LK token kind and the 1124 // right token has \p RK token kind. If \p DeleteLeft is true, the left token 1125 // is deleted on match; otherwise, the right token is deleted. 1126 template <typename LeftKind, typename RightKind> 1127 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK, 1128 bool DeleteLeft) { 1129 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * { 1130 for (auto *Res = Tok.Next; Res; Res = Res->Next) 1131 if (!Res->is(tok::comment) && 1132 DeletedTokens.find(Res) == DeletedTokens.end()) 1133 return Res; 1134 return nullptr; 1135 }; 1136 for (auto *Left = Start; Left;) { 1137 auto *Right = NextNotDeleted(*Left); 1138 if (!Right) 1139 break; 1140 if (Left->is(LK) && Right->is(RK)) { 1141 deleteToken(DeleteLeft ? Left : Right); 1142 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next) 1143 deleteToken(Tok); 1144 // If the right token is deleted, we should keep the left token 1145 // unchanged and pair it with the new right token. 1146 if (!DeleteLeft) 1147 continue; 1148 } 1149 Left = Right; 1150 } 1151 } 1152 1153 template <typename LeftKind, typename RightKind> 1154 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) { 1155 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true); 1156 } 1157 1158 template <typename LeftKind, typename RightKind> 1159 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) { 1160 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false); 1161 } 1162 1163 // Delete the given token. 1164 inline void deleteToken(FormatToken *Tok) { 1165 if (Tok) 1166 DeletedTokens.insert(Tok); 1167 } 1168 1169 tooling::Replacements generateFixes() { 1170 tooling::Replacements Fixes; 1171 std::vector<FormatToken *> Tokens; 1172 std::copy(DeletedTokens.begin(), DeletedTokens.end(), 1173 std::back_inserter(Tokens)); 1174 1175 // Merge multiple continuous token deletions into one big deletion so that 1176 // the number of replacements can be reduced. This makes computing affected 1177 // ranges more efficient when we run reformat on the changed code. 1178 unsigned Idx = 0; 1179 while (Idx < Tokens.size()) { 1180 unsigned St = Idx, End = Idx; 1181 while ((End + 1) < Tokens.size() && 1182 Tokens[End]->Next == Tokens[End + 1]) { 1183 End++; 1184 } 1185 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(), 1186 Tokens[End]->Tok.getEndLoc()); 1187 auto Err = 1188 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, "")); 1189 // FIXME: better error handling. for now just print error message and skip 1190 // for the release version. 1191 if (Err) 1192 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1193 assert(!Err && "Fixes must not conflict!"); 1194 Idx = End + 1; 1195 } 1196 1197 return Fixes; 1198 } 1199 1200 // Class for less-than inequality comparason for the set `RedundantTokens`. 1201 // We store tokens in the order they appear in the translation unit so that 1202 // we do not need to sort them in `generateFixes()`. 1203 struct FormatTokenLess { 1204 FormatTokenLess(const SourceManager &SM) : SM(SM) {} 1205 1206 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const { 1207 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(), 1208 RHS->Tok.getLocation()); 1209 } 1210 const SourceManager &SM; 1211 }; 1212 1213 // Tokens to be deleted. 1214 std::set<FormatToken *, FormatTokenLess> DeletedTokens; 1215 }; 1216 1217 struct IncludeDirective { 1218 StringRef Filename; 1219 StringRef Text; 1220 unsigned Offset; 1221 int Category; 1222 }; 1223 1224 } // end anonymous namespace 1225 1226 // Determines whether 'Ranges' intersects with ('Start', 'End'). 1227 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start, 1228 unsigned End) { 1229 for (auto Range : Ranges) { 1230 if (Range.getOffset() < End && 1231 Range.getOffset() + Range.getLength() > Start) 1232 return true; 1233 } 1234 return false; 1235 } 1236 1237 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor 1238 // before sorting/deduplicating. Index is the index of the include under the 1239 // cursor in the original set of includes. If this include has duplicates, it is 1240 // the index of the first of the duplicates as the others are going to be 1241 // removed. OffsetToEOL describes the cursor's position relative to the end of 1242 // its current line. 1243 // If `Cursor` is not on any #include, `Index` will be UINT_MAX. 1244 static std::pair<unsigned, unsigned> 1245 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes, 1246 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) { 1247 unsigned CursorIndex = UINT_MAX; 1248 unsigned OffsetToEOL = 0; 1249 for (int i = 0, e = Includes.size(); i != e; ++i) { 1250 unsigned Start = Includes[Indices[i]].Offset; 1251 unsigned End = Start + Includes[Indices[i]].Text.size(); 1252 if (!(Cursor >= Start && Cursor < End)) 1253 continue; 1254 CursorIndex = Indices[i]; 1255 OffsetToEOL = End - Cursor; 1256 // Put the cursor on the only remaining #include among the duplicate 1257 // #includes. 1258 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text) 1259 CursorIndex = i; 1260 break; 1261 } 1262 return std::make_pair(CursorIndex, OffsetToEOL); 1263 } 1264 1265 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically 1266 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict 1267 // source order. 1268 // #include directives with the same text will be deduplicated, and only the 1269 // first #include in the duplicate #includes remains. If the `Cursor` is 1270 // provided and put on a deleted #include, it will be moved to the remaining 1271 // #include in the duplicate #includes. 1272 static void sortCppIncludes(const FormatStyle &Style, 1273 const SmallVectorImpl<IncludeDirective> &Includes, 1274 ArrayRef<tooling::Range> Ranges, StringRef FileName, 1275 tooling::Replacements &Replaces, unsigned *Cursor) { 1276 unsigned IncludesBeginOffset = Includes.front().Offset; 1277 unsigned IncludesEndOffset = 1278 Includes.back().Offset + Includes.back().Text.size(); 1279 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset; 1280 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset)) 1281 return; 1282 SmallVector<unsigned, 16> Indices; 1283 for (unsigned i = 0, e = Includes.size(); i != e; ++i) 1284 Indices.push_back(i); 1285 std::stable_sort( 1286 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) { 1287 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) < 1288 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename); 1289 }); 1290 // The index of the include on which the cursor will be put after 1291 // sorting/deduplicating. 1292 unsigned CursorIndex; 1293 // The offset from cursor to the end of line. 1294 unsigned CursorToEOLOffset; 1295 if (Cursor) 1296 std::tie(CursorIndex, CursorToEOLOffset) = 1297 FindCursorIndex(Includes, Indices, *Cursor); 1298 1299 // Deduplicate #includes. 1300 Indices.erase(std::unique(Indices.begin(), Indices.end(), 1301 [&](unsigned LHSI, unsigned RHSI) { 1302 return Includes[LHSI].Text == Includes[RHSI].Text; 1303 }), 1304 Indices.end()); 1305 1306 // If the #includes are out of order, we generate a single replacement fixing 1307 // the entire block. Otherwise, no replacement is generated. 1308 if (Indices.size() == Includes.size() && 1309 std::is_sorted(Indices.begin(), Indices.end())) 1310 return; 1311 1312 std::string result; 1313 for (unsigned Index : Indices) { 1314 if (!result.empty()) 1315 result += "\n"; 1316 result += Includes[Index].Text; 1317 if (Cursor && CursorIndex == Index) 1318 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset; 1319 } 1320 1321 auto Err = Replaces.add(tooling::Replacement( 1322 FileName, Includes.front().Offset, IncludesBlockSize, result)); 1323 // FIXME: better error handling. For now, just skip the replacement for the 1324 // release version. 1325 if (Err) 1326 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1327 assert(!Err); 1328 } 1329 1330 namespace { 1331 1332 // This class manages priorities of #include categories and calculates 1333 // priorities for headers. 1334 class IncludeCategoryManager { 1335 public: 1336 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName) 1337 : Style(Style), FileName(FileName) { 1338 FileStem = llvm::sys::path::stem(FileName); 1339 for (const auto &Category : Style.IncludeCategories) 1340 CategoryRegexs.emplace_back(Category.Regex); 1341 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") || 1342 FileName.endswith(".cpp") || FileName.endswith(".c++") || 1343 FileName.endswith(".cxx") || FileName.endswith(".m") || 1344 FileName.endswith(".mm"); 1345 } 1346 1347 // Returns the priority of the category which \p IncludeName belongs to. 1348 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns 1349 // 0. Otherwise, returns the priority of the matching category or INT_MAX. 1350 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) { 1351 int Ret = INT_MAX; 1352 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i) 1353 if (CategoryRegexs[i].match(IncludeName)) { 1354 Ret = Style.IncludeCategories[i].Priority; 1355 break; 1356 } 1357 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName)) 1358 Ret = 0; 1359 return Ret; 1360 } 1361 1362 private: 1363 bool isMainHeader(StringRef IncludeName) const { 1364 if (!IncludeName.startswith("\"")) 1365 return false; 1366 StringRef HeaderStem = 1367 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1)); 1368 if (FileStem.startswith(HeaderStem)) { 1369 llvm::Regex MainIncludeRegex( 1370 (HeaderStem + Style.IncludeIsMainRegex).str()); 1371 if (MainIncludeRegex.match(FileStem)) 1372 return true; 1373 } 1374 return false; 1375 } 1376 1377 const FormatStyle &Style; 1378 bool IsMainFile; 1379 StringRef FileName; 1380 StringRef FileStem; 1381 SmallVector<llvm::Regex, 4> CategoryRegexs; 1382 }; 1383 1384 const char IncludeRegexPattern[] = 1385 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))"; 1386 1387 } // anonymous namespace 1388 1389 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code, 1390 ArrayRef<tooling::Range> Ranges, 1391 StringRef FileName, 1392 tooling::Replacements &Replaces, 1393 unsigned *Cursor) { 1394 unsigned Prev = 0; 1395 unsigned SearchFrom = 0; 1396 llvm::Regex IncludeRegex(IncludeRegexPattern); 1397 SmallVector<StringRef, 4> Matches; 1398 SmallVector<IncludeDirective, 16> IncludesInBlock; 1399 1400 // In compiled files, consider the first #include to be the main #include of 1401 // the file if it is not a system #include. This ensures that the header 1402 // doesn't have hidden dependencies 1403 // (http://llvm.org/docs/CodingStandards.html#include-style). 1404 // 1405 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix 1406 // cases where the first #include is unlikely to be the main header. 1407 IncludeCategoryManager Categories(Style, FileName); 1408 bool FirstIncludeBlock = true; 1409 bool MainIncludeFound = false; 1410 bool FormattingOff = false; 1411 1412 for (;;) { 1413 auto Pos = Code.find('\n', SearchFrom); 1414 StringRef Line = 1415 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev); 1416 1417 StringRef Trimmed = Line.trim(); 1418 if (Trimmed == "// clang-format off") 1419 FormattingOff = true; 1420 else if (Trimmed == "// clang-format on") 1421 FormattingOff = false; 1422 1423 if (!FormattingOff && !Line.endswith("\\")) { 1424 if (IncludeRegex.match(Line, &Matches)) { 1425 StringRef IncludeName = Matches[2]; 1426 int Category = Categories.getIncludePriority( 1427 IncludeName, 1428 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock); 1429 if (Category == 0) 1430 MainIncludeFound = true; 1431 IncludesInBlock.push_back({IncludeName, Line, Prev, Category}); 1432 } else if (!IncludesInBlock.empty()) { 1433 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, 1434 Cursor); 1435 IncludesInBlock.clear(); 1436 FirstIncludeBlock = false; 1437 } 1438 Prev = Pos + 1; 1439 } 1440 if (Pos == StringRef::npos || Pos + 1 == Code.size()) 1441 break; 1442 SearchFrom = Pos + 1; 1443 } 1444 if (!IncludesInBlock.empty()) 1445 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor); 1446 return Replaces; 1447 } 1448 1449 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code, 1450 ArrayRef<tooling::Range> Ranges, 1451 StringRef FileName, unsigned *Cursor) { 1452 tooling::Replacements Replaces; 1453 if (!Style.SortIncludes) 1454 return Replaces; 1455 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript) 1456 return sortJavaScriptImports(Style, Code, Ranges, FileName); 1457 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor); 1458 return Replaces; 1459 } 1460 1461 template <typename T> 1462 static llvm::Expected<tooling::Replacements> 1463 processReplacements(T ProcessFunc, StringRef Code, 1464 const tooling::Replacements &Replaces, 1465 const FormatStyle &Style) { 1466 if (Replaces.empty()) 1467 return tooling::Replacements(); 1468 1469 auto NewCode = applyAllReplacements(Code, Replaces); 1470 if (!NewCode) 1471 return NewCode.takeError(); 1472 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges(); 1473 StringRef FileName = Replaces.begin()->getFilePath(); 1474 1475 tooling::Replacements FormatReplaces = 1476 ProcessFunc(Style, *NewCode, ChangedRanges, FileName); 1477 1478 return Replaces.merge(FormatReplaces); 1479 } 1480 1481 llvm::Expected<tooling::Replacements> 1482 formatReplacements(StringRef Code, const tooling::Replacements &Replaces, 1483 const FormatStyle &Style) { 1484 // We need to use lambda function here since there are two versions of 1485 // `sortIncludes`. 1486 auto SortIncludes = [](const FormatStyle &Style, StringRef Code, 1487 std::vector<tooling::Range> Ranges, 1488 StringRef FileName) -> tooling::Replacements { 1489 return sortIncludes(Style, Code, Ranges, FileName); 1490 }; 1491 auto SortedReplaces = 1492 processReplacements(SortIncludes, Code, Replaces, Style); 1493 if (!SortedReplaces) 1494 return SortedReplaces.takeError(); 1495 1496 // We need to use lambda function here since there are two versions of 1497 // `reformat`. 1498 auto Reformat = [](const FormatStyle &Style, StringRef Code, 1499 std::vector<tooling::Range> Ranges, 1500 StringRef FileName) -> tooling::Replacements { 1501 return reformat(Style, Code, Ranges, FileName); 1502 }; 1503 return processReplacements(Reformat, Code, *SortedReplaces, Style); 1504 } 1505 1506 namespace { 1507 1508 inline bool isHeaderInsertion(const tooling::Replacement &Replace) { 1509 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 && 1510 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText()); 1511 } 1512 1513 inline bool isHeaderDeletion(const tooling::Replacement &Replace) { 1514 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1; 1515 } 1516 1517 void skipComments(Lexer &Lex, Token &Tok) { 1518 while (Tok.is(tok::comment)) 1519 if (Lex.LexFromRawLexer(Tok)) 1520 return; 1521 } 1522 1523 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is, 1524 // \p Tok will be the token after this directive; otherwise, it can be any token 1525 // after the given \p Tok (including \p Tok). 1526 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) { 1527 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) && 1528 Tok.is(tok::raw_identifier) && 1529 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) && 1530 Tok.is(tok::raw_identifier); 1531 if (Matched) 1532 Lex.LexFromRawLexer(Tok); 1533 return Matched; 1534 } 1535 1536 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName, 1537 StringRef Code, 1538 const FormatStyle &Style) { 1539 std::unique_ptr<Environment> Env = 1540 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{}); 1541 const SourceManager &SourceMgr = Env->getSourceManager(); 1542 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr, 1543 getFormattingLangOpts(Style)); 1544 Token Tok; 1545 // Get the first token. 1546 Lex.LexFromRawLexer(Tok); 1547 skipComments(Lex, Tok); 1548 unsigned AfterComments = SourceMgr.getFileOffset(Tok.getLocation()); 1549 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) { 1550 skipComments(Lex, Tok); 1551 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok)) 1552 return SourceMgr.getFileOffset(Tok.getLocation()); 1553 } 1554 return AfterComments; 1555 } 1556 1557 bool isDeletedHeader(llvm::StringRef HeaderName, 1558 const std::set<llvm::StringRef> HeadersToDelete) { 1559 return HeadersToDelete.find(HeaderName) != HeadersToDelete.end() || 1560 HeadersToDelete.find(HeaderName.trim("\"<>")) != HeadersToDelete.end(); 1561 } 1562 1563 // FIXME: we also need to insert a '\n' at the end of the code if we have an 1564 // insertion with offset Code.size(), and there is no '\n' at the end of the 1565 // code. 1566 // FIXME: do not insert headers into conditional #include blocks, e.g. #includes 1567 // surrounded by compile condition "#if...". 1568 // FIXME: insert empty lines between newly created blocks. 1569 tooling::Replacements 1570 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces, 1571 const FormatStyle &Style) { 1572 if (Style.Language != FormatStyle::LanguageKind::LK_Cpp) 1573 return Replaces; 1574 1575 tooling::Replacements HeaderInsertions; 1576 std::set<llvm::StringRef> HeadersToDelete; 1577 tooling::Replacements Result; 1578 for (const auto &R : Replaces) { 1579 if (isHeaderInsertion(R)) { 1580 // Replacements from \p Replaces must be conflict-free already, so we can 1581 // simply consume the error. 1582 llvm::consumeError(HeaderInsertions.add(R)); 1583 } else if (isHeaderDeletion(R)) { 1584 HeadersToDelete.insert(R.getReplacementText()); 1585 } else if (R.getOffset() == UINT_MAX) { 1586 llvm::errs() << "Insertions other than header #include insertion are " 1587 "not supported! " 1588 << R.getReplacementText() << "\n"; 1589 } else { 1590 llvm::consumeError(Result.add(R)); 1591 } 1592 } 1593 if (HeaderInsertions.empty() && HeadersToDelete.empty()) 1594 return Replaces; 1595 1596 llvm::Regex IncludeRegex(IncludeRegexPattern); 1597 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)"); 1598 SmallVector<StringRef, 4> Matches; 1599 1600 StringRef FileName = Replaces.begin()->getFilePath(); 1601 IncludeCategoryManager Categories(Style, FileName); 1602 1603 // Record the offset of the end of the last include in each category. 1604 std::map<int, int> CategoryEndOffsets; 1605 // All possible priorities. 1606 // Add 0 for main header and INT_MAX for headers that are not in any category. 1607 std::set<int> Priorities = {0, INT_MAX}; 1608 for (const auto &Category : Style.IncludeCategories) 1609 Priorities.insert(Category.Priority); 1610 int FirstIncludeOffset = -1; 1611 // All new headers should be inserted after this offset. 1612 unsigned MinInsertOffset = 1613 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style); 1614 StringRef TrimmedCode = Code.drop_front(MinInsertOffset); 1615 SmallVector<StringRef, 32> Lines; 1616 TrimmedCode.split(Lines, '\n'); 1617 unsigned Offset = MinInsertOffset; 1618 unsigned NextLineOffset; 1619 std::set<StringRef> ExistingIncludes; 1620 for (auto Line : Lines) { 1621 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1); 1622 if (IncludeRegex.match(Line, &Matches)) { 1623 // The header name with quotes or angle brackets. 1624 StringRef IncludeName = Matches[2]; 1625 ExistingIncludes.insert(IncludeName); 1626 int Category = Categories.getIncludePriority( 1627 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0); 1628 CategoryEndOffsets[Category] = NextLineOffset; 1629 if (FirstIncludeOffset < 0) 1630 FirstIncludeOffset = Offset; 1631 if (isDeletedHeader(IncludeName, HeadersToDelete)) { 1632 // If this is the last line without trailing newline, we need to make 1633 // sure we don't delete across the file boundary. 1634 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset); 1635 llvm::Error Err = 1636 Result.add(tooling::Replacement(FileName, Offset, Length, "")); 1637 if (Err) { 1638 // Ignore the deletion on conflict. 1639 llvm::errs() << "Failed to add header deletion replacement for " 1640 << IncludeName << ": " << llvm::toString(std::move(Err)) 1641 << "\n"; 1642 } 1643 } 1644 } 1645 Offset = NextLineOffset; 1646 } 1647 1648 // Populate CategoryEndOfssets: 1649 // - Ensure that CategoryEndOffset[Highest] is always populated. 1650 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that 1651 // is set, up to CategoryEndOffset[Highest]. 1652 auto Highest = Priorities.begin(); 1653 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) { 1654 if (FirstIncludeOffset >= 0) 1655 CategoryEndOffsets[*Highest] = FirstIncludeOffset; 1656 else 1657 CategoryEndOffsets[*Highest] = MinInsertOffset; 1658 } 1659 // By this point, CategoryEndOffset[Highest] is always set appropriately: 1660 // - to an appropriate location before/after existing #includes, or 1661 // - to right after the header guard, or 1662 // - to the beginning of the file. 1663 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I) 1664 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end()) 1665 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)]; 1666 1667 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n'; 1668 for (const auto &R : HeaderInsertions) { 1669 auto IncludeDirective = R.getReplacementText(); 1670 bool Matched = IncludeRegex.match(IncludeDirective, &Matches); 1671 assert(Matched && "Header insertion replacement must have replacement text " 1672 "'#include ...'"); 1673 (void)Matched; 1674 auto IncludeName = Matches[2]; 1675 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) { 1676 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName 1677 << "\n"); 1678 continue; 1679 } 1680 int Category = 1681 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true); 1682 Offset = CategoryEndOffsets[Category]; 1683 std::string NewInclude = !IncludeDirective.endswith("\n") 1684 ? (IncludeDirective + "\n").str() 1685 : IncludeDirective.str(); 1686 // When inserting headers at end of the code, also append '\n' to the code 1687 // if it does not end with '\n'. 1688 if (NeedNewLineAtEnd && Offset == Code.size()) { 1689 NewInclude = "\n" + NewInclude; 1690 NeedNewLineAtEnd = false; 1691 } 1692 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude); 1693 auto Err = Result.add(NewReplace); 1694 if (Err) { 1695 llvm::consumeError(std::move(Err)); 1696 unsigned NewOffset = Result.getShiftedCodePosition(Offset); 1697 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude); 1698 Result = Result.merge(tooling::Replacements(NewReplace)); 1699 } 1700 } 1701 return Result; 1702 } 1703 1704 } // anonymous namespace 1705 1706 llvm::Expected<tooling::Replacements> 1707 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces, 1708 const FormatStyle &Style) { 1709 // We need to use lambda function here since there are two versions of 1710 // `cleanup`. 1711 auto Cleanup = [](const FormatStyle &Style, StringRef Code, 1712 std::vector<tooling::Range> Ranges, 1713 StringRef FileName) -> tooling::Replacements { 1714 return cleanup(Style, Code, Ranges, FileName); 1715 }; 1716 // Make header insertion replacements insert new headers into correct blocks. 1717 tooling::Replacements NewReplaces = 1718 fixCppIncludeInsertions(Code, Replaces, Style); 1719 return processReplacements(Cleanup, Code, NewReplaces, Style); 1720 } 1721 1722 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1723 ArrayRef<tooling::Range> Ranges, 1724 StringRef FileName, bool *IncompleteFormat) { 1725 FormatStyle Expanded = expandPresets(Style); 1726 if (Expanded.DisableFormat) 1727 return tooling::Replacements(); 1728 1729 auto Env = Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 1730 1731 if (Style.Language == FormatStyle::LK_JavaScript && 1732 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) { 1733 JavaScriptRequoter Requoter(*Env, Expanded); 1734 tooling::Replacements Requotes = Requoter.process(); 1735 if (!Requotes.empty()) { 1736 auto NewCode = applyAllReplacements(Code, Requotes); 1737 if (NewCode) { 1738 auto NewEnv = Environment::CreateVirtualEnvironment( 1739 *NewCode, FileName, 1740 tooling::calculateRangesAfterReplacements(Requotes, Ranges)); 1741 Formatter Format(*NewEnv, Expanded, IncompleteFormat); 1742 return Requotes.merge(Format.process()); 1743 } 1744 } 1745 } 1746 1747 Formatter Format(*Env, Expanded, IncompleteFormat); 1748 return Format.process(); 1749 } 1750 1751 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code, 1752 ArrayRef<tooling::Range> Ranges, 1753 StringRef FileName) { 1754 std::unique_ptr<Environment> Env = 1755 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 1756 Cleaner Clean(*Env, Style); 1757 return Clean.process(); 1758 } 1759 1760 LangOptions getFormattingLangOpts(const FormatStyle &Style) { 1761 LangOptions LangOpts; 1762 LangOpts.CPlusPlus = 1; 1763 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1764 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1765 LangOpts.LineComment = 1; 1766 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp; 1767 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0; 1768 LangOpts.Bool = 1; 1769 LangOpts.ObjC1 = 1; 1770 LangOpts.ObjC2 = 1; 1771 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally. 1772 LangOpts.DeclSpecKeyword = 1; // To get __declspec. 1773 return LangOpts; 1774 } 1775 1776 const char *StyleOptionHelpDescription = 1777 "Coding style, currently supports:\n" 1778 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1779 "Use -style=file to load style configuration from\n" 1780 ".clang-format file located in one of the parent\n" 1781 "directories of the source file (or current\n" 1782 "directory for stdin).\n" 1783 "Use -style=\"{key: value, ...}\" to set specific\n" 1784 "parameters, e.g.:\n" 1785 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1786 1787 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 1788 if (FileName.endswith(".java")) 1789 return FormatStyle::LK_Java; 1790 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) 1791 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript. 1792 if (FileName.endswith_lower(".proto") || 1793 FileName.endswith_lower(".protodevel")) 1794 return FormatStyle::LK_Proto; 1795 if (FileName.endswith_lower(".td")) 1796 return FormatStyle::LK_TableGen; 1797 return FormatStyle::LK_Cpp; 1798 } 1799 1800 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 1801 StringRef FallbackStyle, vfs::FileSystem *FS) { 1802 if (!FS) { 1803 FS = vfs::getRealFileSystem().get(); 1804 } 1805 FormatStyle Style = getLLVMStyle(); 1806 Style.Language = getLanguageByFileName(FileName); 1807 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) { 1808 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 1809 << "\" using LLVM style\n"; 1810 return Style; 1811 } 1812 1813 if (StyleName.startswith("{")) { 1814 // Parse YAML/JSON style from the command line. 1815 if (std::error_code ec = parseConfiguration(StyleName, &Style)) { 1816 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1817 << FallbackStyle << " style\n"; 1818 } 1819 return Style; 1820 } 1821 1822 if (!StyleName.equals_lower("file")) { 1823 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 1824 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1825 << " style\n"; 1826 return Style; 1827 } 1828 1829 // Look for .clang-format/_clang-format file in the file's parent directories. 1830 SmallString<128> UnsuitableConfigFiles; 1831 SmallString<128> Path(FileName); 1832 llvm::sys::fs::make_absolute(Path); 1833 for (StringRef Directory = Path; !Directory.empty(); 1834 Directory = llvm::sys::path::parent_path(Directory)) { 1835 1836 auto Status = FS->status(Directory); 1837 if (!Status || 1838 Status->getType() != llvm::sys::fs::file_type::directory_file) { 1839 continue; 1840 } 1841 1842 SmallString<128> ConfigFile(Directory); 1843 1844 llvm::sys::path::append(ConfigFile, ".clang-format"); 1845 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1846 1847 Status = FS->status(ConfigFile.str()); 1848 bool IsFile = 1849 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file); 1850 if (!IsFile) { 1851 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1852 ConfigFile = Directory; 1853 llvm::sys::path::append(ConfigFile, "_clang-format"); 1854 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1855 Status = FS->status(ConfigFile.str()); 1856 IsFile = Status && 1857 (Status->getType() == llvm::sys::fs::file_type::regular_file); 1858 } 1859 1860 if (IsFile) { 1861 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text = 1862 FS->getBufferForFile(ConfigFile.str()); 1863 if (std::error_code EC = Text.getError()) { 1864 llvm::errs() << EC.message() << "\n"; 1865 break; 1866 } 1867 if (std::error_code ec = 1868 parseConfiguration(Text.get()->getBuffer(), &Style)) { 1869 if (ec == ParseError::Unsuitable) { 1870 if (!UnsuitableConfigFiles.empty()) 1871 UnsuitableConfigFiles.append(", "); 1872 UnsuitableConfigFiles.append(ConfigFile); 1873 continue; 1874 } 1875 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1876 << "\n"; 1877 break; 1878 } 1879 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1880 return Style; 1881 } 1882 } 1883 if (!UnsuitableConfigFiles.empty()) { 1884 llvm::errs() << "Configuration file(s) do(es) not support " 1885 << getLanguageName(Style.Language) << ": " 1886 << UnsuitableConfigFiles << "\n"; 1887 } 1888 return Style; 1889 } 1890 1891 } // namespace format 1892 } // namespace clang 1893