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 LLVM_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_Inline; 614 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 615 GoogleStyle.BreakBeforeTernaryOperators = false; 616 GoogleStyle.CommentPragmas = "@(export|requirecss|return|see|visibility) "; 617 GoogleStyle.MaxEmptyLinesToKeep = 3; 618 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All; 619 GoogleStyle.SpacesInContainerLiterals = false; 620 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single; 621 GoogleStyle.JavaScriptWrapImports = false; 622 } else if (Language == FormatStyle::LK_Proto) { 623 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None; 624 GoogleStyle.SpacesInContainerLiterals = false; 625 } 626 627 return GoogleStyle; 628 } 629 630 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) { 631 FormatStyle ChromiumStyle = getGoogleStyle(Language); 632 if (Language == FormatStyle::LK_Java) { 633 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true; 634 ChromiumStyle.BreakAfterJavaFieldAnnotations = true; 635 ChromiumStyle.ContinuationIndentWidth = 8; 636 ChromiumStyle.IndentWidth = 4; 637 } else { 638 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 639 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 640 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 641 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 642 ChromiumStyle.BinPackParameters = false; 643 ChromiumStyle.DerivePointerAlignment = false; 644 } 645 ChromiumStyle.SortIncludes = false; 646 return ChromiumStyle; 647 } 648 649 FormatStyle getMozillaStyle() { 650 FormatStyle MozillaStyle = getLLVMStyle(); 651 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 652 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 653 MozillaStyle.AlwaysBreakAfterReturnType = 654 FormatStyle::RTBS_TopLevelDefinitions; 655 MozillaStyle.AlwaysBreakAfterDefinitionReturnType = 656 FormatStyle::DRTBS_TopLevel; 657 MozillaStyle.AlwaysBreakTemplateDeclarations = true; 658 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla; 659 MozillaStyle.BreakConstructorInitializersBeforeComma = true; 660 MozillaStyle.ConstructorInitializerIndentWidth = 2; 661 MozillaStyle.ContinuationIndentWidth = 2; 662 MozillaStyle.Cpp11BracedListStyle = false; 663 MozillaStyle.IndentCaseLabels = true; 664 MozillaStyle.ObjCSpaceAfterProperty = true; 665 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 666 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 667 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left; 668 MozillaStyle.SpaceAfterTemplateKeyword = false; 669 return MozillaStyle; 670 } 671 672 FormatStyle getWebKitStyle() { 673 FormatStyle Style = getLLVMStyle(); 674 Style.AccessModifierOffset = -4; 675 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 676 Style.AlignOperands = false; 677 Style.AlignTrailingComments = false; 678 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 679 Style.BreakBeforeBraces = FormatStyle::BS_WebKit; 680 Style.BreakConstructorInitializersBeforeComma = true; 681 Style.Cpp11BracedListStyle = false; 682 Style.ColumnLimit = 0; 683 Style.IndentWidth = 4; 684 Style.NamespaceIndentation = FormatStyle::NI_Inner; 685 Style.ObjCBlockIndentWidth = 4; 686 Style.ObjCSpaceAfterProperty = true; 687 Style.PointerAlignment = FormatStyle::PAS_Left; 688 return Style; 689 } 690 691 FormatStyle getGNUStyle() { 692 FormatStyle Style = getLLVMStyle(); 693 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All; 694 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 695 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 696 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 697 Style.BreakBeforeTernaryOperators = true; 698 Style.Cpp11BracedListStyle = false; 699 Style.ColumnLimit = 79; 700 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 701 Style.Standard = FormatStyle::LS_Cpp03; 702 return Style; 703 } 704 705 FormatStyle getNoStyle() { 706 FormatStyle NoStyle = getLLVMStyle(); 707 NoStyle.DisableFormat = true; 708 NoStyle.SortIncludes = false; 709 return NoStyle; 710 } 711 712 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 713 FormatStyle *Style) { 714 if (Name.equals_lower("llvm")) { 715 *Style = getLLVMStyle(); 716 } else if (Name.equals_lower("chromium")) { 717 *Style = getChromiumStyle(Language); 718 } else if (Name.equals_lower("mozilla")) { 719 *Style = getMozillaStyle(); 720 } else if (Name.equals_lower("google")) { 721 *Style = getGoogleStyle(Language); 722 } else if (Name.equals_lower("webkit")) { 723 *Style = getWebKitStyle(); 724 } else if (Name.equals_lower("gnu")) { 725 *Style = getGNUStyle(); 726 } else if (Name.equals_lower("none")) { 727 *Style = getNoStyle(); 728 } else { 729 return false; 730 } 731 732 Style->Language = Language; 733 return true; 734 } 735 736 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 737 assert(Style); 738 FormatStyle::LanguageKind Language = Style->Language; 739 assert(Language != FormatStyle::LK_None); 740 if (Text.trim().empty()) 741 return make_error_code(ParseError::Error); 742 743 std::vector<FormatStyle> Styles; 744 llvm::yaml::Input Input(Text); 745 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 746 // values for the fields, keys for which are missing from the configuration. 747 // Mapping also uses the context to get the language to find the correct 748 // base style. 749 Input.setContext(Style); 750 Input >> Styles; 751 if (Input.error()) 752 return Input.error(); 753 754 for (unsigned i = 0; i < Styles.size(); ++i) { 755 // Ensures that only the first configuration can skip the Language option. 756 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 757 return make_error_code(ParseError::Error); 758 // Ensure that each language is configured at most once. 759 for (unsigned j = 0; j < i; ++j) { 760 if (Styles[i].Language == Styles[j].Language) { 761 DEBUG(llvm::dbgs() 762 << "Duplicate languages in the config file on positions " << j 763 << " and " << i << "\n"); 764 return make_error_code(ParseError::Error); 765 } 766 } 767 } 768 // Look for a suitable configuration starting from the end, so we can 769 // find the configuration for the specific language first, and the default 770 // configuration (which can only be at slot 0) after it. 771 for (int i = Styles.size() - 1; i >= 0; --i) { 772 if (Styles[i].Language == Language || 773 Styles[i].Language == FormatStyle::LK_None) { 774 *Style = Styles[i]; 775 Style->Language = Language; 776 return make_error_code(ParseError::Success); 777 } 778 } 779 return make_error_code(ParseError::Unsuitable); 780 } 781 782 std::string configurationAsText(const FormatStyle &Style) { 783 std::string Text; 784 llvm::raw_string_ostream Stream(Text); 785 llvm::yaml::Output Output(Stream); 786 // We use the same mapping method for input and output, so we need a non-const 787 // reference here. 788 FormatStyle NonConstStyle = expandPresets(Style); 789 Output << NonConstStyle; 790 return Stream.str(); 791 } 792 793 namespace { 794 795 class Formatter : public TokenAnalyzer { 796 public: 797 Formatter(const Environment &Env, const FormatStyle &Style, 798 bool *IncompleteFormat) 799 : TokenAnalyzer(Env, Style), IncompleteFormat(IncompleteFormat) {} 800 801 tooling::Replacements 802 analyze(TokenAnnotator &Annotator, 803 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 804 FormatTokenLexer &Tokens, tooling::Replacements &Result) override { 805 deriveLocalStyle(AnnotatedLines); 806 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 807 AnnotatedLines.end()); 808 809 if (Style.Language == FormatStyle::LK_JavaScript && 810 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) 811 requoteJSStringLiteral(AnnotatedLines, Result); 812 813 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 814 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 815 } 816 817 Annotator.setCommentLineLevels(AnnotatedLines); 818 819 WhitespaceManager Whitespaces( 820 Env.getSourceManager(), Style, 821 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID()))); 822 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), 823 Env.getSourceManager(), Whitespaces, Encoding, 824 BinPackInconclusiveFunctions); 825 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(), 826 IncompleteFormat) 827 .format(AnnotatedLines); 828 return Whitespaces.generateReplacements(); 829 } 830 831 private: 832 // If the last token is a double/single-quoted string literal, generates a 833 // replacement with a single/double quoted string literal, re-escaping the 834 // contents in the process. 835 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines, 836 tooling::Replacements &Result) { 837 for (AnnotatedLine *Line : Lines) { 838 requoteJSStringLiteral(Line->Children, Result); 839 if (!Line->Affected) 840 continue; 841 for (FormatToken *FormatTok = Line->First; FormatTok; 842 FormatTok = FormatTok->Next) { 843 StringRef Input = FormatTok->TokenText; 844 if (FormatTok->Finalized || !FormatTok->isStringLiteral() || 845 // NB: testing for not starting with a double quote to avoid 846 // breaking 847 // `template strings`. 848 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single && 849 !Input.startswith("\"")) || 850 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double && 851 !Input.startswith("\'"))) 852 continue; 853 854 // Change start and end quote. 855 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single; 856 SourceLocation Start = FormatTok->Tok.getLocation(); 857 auto Replace = [&](SourceLocation Start, unsigned Length, 858 StringRef ReplacementText) { 859 auto Err = Result.add(tooling::Replacement( 860 Env.getSourceManager(), Start, Length, ReplacementText)); 861 // FIXME: handle error. For now, print error message and skip the 862 // replacement for release version. 863 if (Err) 864 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 865 assert(!Err); 866 }; 867 Replace(Start, 1, IsSingle ? "'" : "\""); 868 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1, 869 IsSingle ? "'" : "\""); 870 871 // Escape internal quotes. 872 size_t ColumnWidth = FormatTok->TokenText.size(); 873 bool Escaped = false; 874 for (size_t i = 1; i < Input.size() - 1; i++) { 875 switch (Input[i]) { 876 case '\\': 877 if (!Escaped && i + 1 < Input.size() && 878 ((IsSingle && Input[i + 1] == '"') || 879 (!IsSingle && Input[i + 1] == '\''))) { 880 // Remove this \, it's escaping a " or ' that no longer needs 881 // escaping 882 ColumnWidth--; 883 Replace(Start.getLocWithOffset(i), 1, ""); 884 continue; 885 } 886 Escaped = !Escaped; 887 break; 888 case '\"': 889 case '\'': 890 if (!Escaped && IsSingle == (Input[i] == '\'')) { 891 // Escape the quote. 892 Replace(Start.getLocWithOffset(i), 0, "\\"); 893 ColumnWidth++; 894 } 895 Escaped = false; 896 break; 897 default: 898 Escaped = false; 899 break; 900 } 901 } 902 903 // For formatting, count the number of non-escaped single quotes in them 904 // and adjust ColumnWidth to take the added escapes into account. 905 // FIXME(martinprobst): this might conflict with code breaking a long 906 // string literal (which clang-format doesn't do, yet). For that to 907 // work, this code would have to modify TokenText directly. 908 FormatTok->ColumnWidth = ColumnWidth; 909 } 910 } 911 } 912 913 static bool inputUsesCRLF(StringRef Text) { 914 return Text.count('\r') * 2 > Text.count('\n'); 915 } 916 917 bool 918 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) { 919 for (const AnnotatedLine *Line : Lines) { 920 if (hasCpp03IncompatibleFormat(Line->Children)) 921 return true; 922 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) { 923 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 924 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener)) 925 return true; 926 if (Tok->is(TT_TemplateCloser) && 927 Tok->Previous->is(TT_TemplateCloser)) 928 return true; 929 } 930 } 931 } 932 return false; 933 } 934 935 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) { 936 int AlignmentDiff = 0; 937 for (const AnnotatedLine *Line : Lines) { 938 AlignmentDiff += countVariableAlignments(Line->Children); 939 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) { 940 if (!Tok->is(TT_PointerOrReference)) 941 continue; 942 bool SpaceBefore = 943 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 944 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() != 945 Tok->Next->WhitespaceRange.getEnd(); 946 if (SpaceBefore && !SpaceAfter) 947 ++AlignmentDiff; 948 if (!SpaceBefore && SpaceAfter) 949 --AlignmentDiff; 950 } 951 } 952 return AlignmentDiff; 953 } 954 955 void 956 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 957 bool HasBinPackedFunction = false; 958 bool HasOnePerLineFunction = false; 959 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 960 if (!AnnotatedLines[i]->First->Next) 961 continue; 962 FormatToken *Tok = AnnotatedLines[i]->First->Next; 963 while (Tok->Next) { 964 if (Tok->PackingKind == PPK_BinPacked) 965 HasBinPackedFunction = true; 966 if (Tok->PackingKind == PPK_OnePerLine) 967 HasOnePerLineFunction = true; 968 969 Tok = Tok->Next; 970 } 971 } 972 if (Style.DerivePointerAlignment) 973 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0 974 ? FormatStyle::PAS_Left 975 : FormatStyle::PAS_Right; 976 if (Style.Standard == FormatStyle::LS_Auto) 977 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines) 978 ? FormatStyle::LS_Cpp11 979 : FormatStyle::LS_Cpp03; 980 BinPackInconclusiveFunctions = 981 HasBinPackedFunction || !HasOnePerLineFunction; 982 } 983 984 bool BinPackInconclusiveFunctions; 985 bool *IncompleteFormat; 986 }; 987 988 // This class clean up the erroneous/redundant code around the given ranges in 989 // file. 990 class Cleaner : public TokenAnalyzer { 991 public: 992 Cleaner(const Environment &Env, const FormatStyle &Style) 993 : TokenAnalyzer(Env, Style), 994 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {} 995 996 // FIXME: eliminate unused parameters. 997 tooling::Replacements 998 analyze(TokenAnnotator &Annotator, 999 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1000 FormatTokenLexer &Tokens, tooling::Replacements &Result) override { 1001 // FIXME: in the current implementation the granularity of affected range 1002 // is an annotated line. However, this is not sufficient. Furthermore, 1003 // redundant code introduced by replacements does not necessarily 1004 // intercept with ranges of replacements that result in the redundancy. 1005 // To determine if some redundant code is actually introduced by 1006 // replacements(e.g. deletions), we need to come up with a more 1007 // sophisticated way of computing affected ranges. 1008 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1009 AnnotatedLines.end()); 1010 1011 checkEmptyNamespace(AnnotatedLines); 1012 1013 for (auto &Line : AnnotatedLines) { 1014 if (Line->Affected) { 1015 cleanupRight(Line->First, tok::comma, tok::comma); 1016 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma); 1017 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace); 1018 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace); 1019 } 1020 } 1021 1022 return generateFixes(); 1023 } 1024 1025 private: 1026 bool containsOnlyComments(const AnnotatedLine &Line) { 1027 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) { 1028 if (Tok->isNot(tok::comment)) 1029 return false; 1030 } 1031 return true; 1032 } 1033 1034 // Iterate through all lines and remove any empty (nested) namespaces. 1035 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1036 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1037 auto &Line = *AnnotatedLines[i]; 1038 if (Line.startsWith(tok::kw_namespace) || 1039 Line.startsWith(tok::kw_inline, tok::kw_namespace)) { 1040 checkEmptyNamespace(AnnotatedLines, i, i); 1041 } 1042 } 1043 1044 for (auto Line : DeletedLines) { 1045 FormatToken *Tok = AnnotatedLines[Line]->First; 1046 while (Tok) { 1047 deleteToken(Tok); 1048 Tok = Tok->Next; 1049 } 1050 } 1051 } 1052 1053 // The function checks if the namespace, which starts from \p CurrentLine, and 1054 // its nested namespaces are empty and delete them if they are empty. It also 1055 // sets \p NewLine to the last line checked. 1056 // Returns true if the current namespace is empty. 1057 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1058 unsigned CurrentLine, unsigned &NewLine) { 1059 unsigned InitLine = CurrentLine, End = AnnotatedLines.size(); 1060 if (Style.BraceWrapping.AfterNamespace) { 1061 // If the left brace is in a new line, we should consume it first so that 1062 // it does not make the namespace non-empty. 1063 // FIXME: error handling if there is no left brace. 1064 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) { 1065 NewLine = CurrentLine; 1066 return false; 1067 } 1068 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) { 1069 return false; 1070 } 1071 while (++CurrentLine < End) { 1072 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace)) 1073 break; 1074 1075 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) || 1076 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline, 1077 tok::kw_namespace)) { 1078 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine)) 1079 return false; 1080 CurrentLine = NewLine; 1081 continue; 1082 } 1083 1084 if (containsOnlyComments(*AnnotatedLines[CurrentLine])) 1085 continue; 1086 1087 // If there is anything other than comments or nested namespaces in the 1088 // current namespace, the namespace cannot be empty. 1089 NewLine = CurrentLine; 1090 return false; 1091 } 1092 1093 NewLine = CurrentLine; 1094 if (CurrentLine >= End) 1095 return false; 1096 1097 // Check if the empty namespace is actually affected by changed ranges. 1098 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange( 1099 AnnotatedLines[InitLine]->First->Tok.getLocation(), 1100 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc()))) 1101 return false; 1102 1103 for (unsigned i = InitLine; i <= CurrentLine; ++i) { 1104 DeletedLines.insert(i); 1105 } 1106 1107 return true; 1108 } 1109 1110 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one 1111 // of the token in the pair if the left token has \p LK token kind and the 1112 // right token has \p RK token kind. If \p DeleteLeft is true, the left token 1113 // is deleted on match; otherwise, the right token is deleted. 1114 template <typename LeftKind, typename RightKind> 1115 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK, 1116 bool DeleteLeft) { 1117 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * { 1118 for (auto *Res = Tok.Next; Res; Res = Res->Next) 1119 if (!Res->is(tok::comment) && 1120 DeletedTokens.find(Res) == DeletedTokens.end()) 1121 return Res; 1122 return nullptr; 1123 }; 1124 for (auto *Left = Start; Left;) { 1125 auto *Right = NextNotDeleted(*Left); 1126 if (!Right) 1127 break; 1128 if (Left->is(LK) && Right->is(RK)) { 1129 deleteToken(DeleteLeft ? Left : Right); 1130 // If the right token is deleted, we should keep the left token 1131 // unchanged and pair it with the new right token. 1132 if (!DeleteLeft) 1133 continue; 1134 } 1135 Left = Right; 1136 } 1137 } 1138 1139 template <typename LeftKind, typename RightKind> 1140 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) { 1141 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true); 1142 } 1143 1144 template <typename LeftKind, typename RightKind> 1145 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) { 1146 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false); 1147 } 1148 1149 // Delete the given token. 1150 inline void deleteToken(FormatToken *Tok) { 1151 if (Tok) 1152 DeletedTokens.insert(Tok); 1153 } 1154 1155 tooling::Replacements generateFixes() { 1156 tooling::Replacements Fixes; 1157 std::vector<FormatToken *> Tokens; 1158 std::copy(DeletedTokens.begin(), DeletedTokens.end(), 1159 std::back_inserter(Tokens)); 1160 1161 // Merge multiple continuous token deletions into one big deletion so that 1162 // the number of replacements can be reduced. This makes computing affected 1163 // ranges more efficient when we run reformat on the changed code. 1164 unsigned Idx = 0; 1165 while (Idx < Tokens.size()) { 1166 unsigned St = Idx, End = Idx; 1167 while ((End + 1) < Tokens.size() && 1168 Tokens[End]->Next == Tokens[End + 1]) { 1169 End++; 1170 } 1171 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(), 1172 Tokens[End]->Tok.getEndLoc()); 1173 auto Err = 1174 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, "")); 1175 // FIXME: better error handling. for now just print error message and skip 1176 // for the release version. 1177 if (Err) 1178 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1179 assert(!Err && "Fixes must not conflict!"); 1180 Idx = End + 1; 1181 } 1182 1183 return Fixes; 1184 } 1185 1186 // Class for less-than inequality comparason for the set `RedundantTokens`. 1187 // We store tokens in the order they appear in the translation unit so that 1188 // we do not need to sort them in `generateFixes()`. 1189 struct FormatTokenLess { 1190 FormatTokenLess(const SourceManager &SM) : SM(SM) {} 1191 1192 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const { 1193 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(), 1194 RHS->Tok.getLocation()); 1195 } 1196 const SourceManager &SM; 1197 }; 1198 1199 // Tokens to be deleted. 1200 std::set<FormatToken *, FormatTokenLess> DeletedTokens; 1201 // The line numbers of lines to be deleted. 1202 std::set<unsigned> DeletedLines; 1203 }; 1204 1205 struct IncludeDirective { 1206 StringRef Filename; 1207 StringRef Text; 1208 unsigned Offset; 1209 int Category; 1210 }; 1211 1212 } // end anonymous namespace 1213 1214 // Determines whether 'Ranges' intersects with ('Start', 'End'). 1215 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start, 1216 unsigned End) { 1217 for (auto Range : Ranges) { 1218 if (Range.getOffset() < End && 1219 Range.getOffset() + Range.getLength() > Start) 1220 return true; 1221 } 1222 return false; 1223 } 1224 1225 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor 1226 // before sorting/deduplicating. Index is the index of the include under the 1227 // cursor in the original set of includes. If this include has duplicates, it is 1228 // the index of the first of the duplicates as the others are going to be 1229 // removed. OffsetToEOL describes the cursor's position relative to the end of 1230 // its current line. 1231 // If `Cursor` is not on any #include, `Index` will be UINT_MAX. 1232 static std::pair<unsigned, unsigned> 1233 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes, 1234 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) { 1235 unsigned CursorIndex = UINT_MAX; 1236 unsigned OffsetToEOL = 0; 1237 for (int i = 0, e = Includes.size(); i != e; ++i) { 1238 unsigned Start = Includes[Indices[i]].Offset; 1239 unsigned End = Start + Includes[Indices[i]].Text.size(); 1240 if (!(Cursor >= Start && Cursor < End)) 1241 continue; 1242 CursorIndex = Indices[i]; 1243 OffsetToEOL = End - Cursor; 1244 // Put the cursor on the only remaining #include among the duplicate 1245 // #includes. 1246 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text) 1247 CursorIndex = i; 1248 break; 1249 } 1250 return std::make_pair(CursorIndex, OffsetToEOL); 1251 } 1252 1253 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically 1254 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict 1255 // source order. 1256 // #include directives with the same text will be deduplicated, and only the 1257 // first #include in the duplicate #includes remains. If the `Cursor` is 1258 // provided and put on a deleted #include, it will be moved to the remaining 1259 // #include in the duplicate #includes. 1260 static void sortCppIncludes(const FormatStyle &Style, 1261 const SmallVectorImpl<IncludeDirective> &Includes, 1262 ArrayRef<tooling::Range> Ranges, StringRef FileName, 1263 tooling::Replacements &Replaces, unsigned *Cursor) { 1264 unsigned IncludesBeginOffset = Includes.front().Offset; 1265 unsigned IncludesEndOffset = 1266 Includes.back().Offset + Includes.back().Text.size(); 1267 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset; 1268 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset)) 1269 return; 1270 SmallVector<unsigned, 16> Indices; 1271 for (unsigned i = 0, e = Includes.size(); i != e; ++i) 1272 Indices.push_back(i); 1273 std::stable_sort( 1274 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) { 1275 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) < 1276 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename); 1277 }); 1278 // The index of the include on which the cursor will be put after 1279 // sorting/deduplicating. 1280 unsigned CursorIndex; 1281 // The offset from cursor to the end of line. 1282 unsigned CursorToEOLOffset; 1283 if (Cursor) 1284 std::tie(CursorIndex, CursorToEOLOffset) = 1285 FindCursorIndex(Includes, Indices, *Cursor); 1286 1287 // Deduplicate #includes. 1288 Indices.erase(std::unique(Indices.begin(), Indices.end(), 1289 [&](unsigned LHSI, unsigned RHSI) { 1290 return Includes[LHSI].Text == Includes[RHSI].Text; 1291 }), 1292 Indices.end()); 1293 1294 // If the #includes are out of order, we generate a single replacement fixing 1295 // the entire block. Otherwise, no replacement is generated. 1296 if (Indices.size() == Includes.size() && 1297 std::is_sorted(Indices.begin(), Indices.end())) 1298 return; 1299 1300 std::string result; 1301 for (unsigned Index : Indices) { 1302 if (!result.empty()) 1303 result += "\n"; 1304 result += Includes[Index].Text; 1305 if (Cursor && CursorIndex == Index) 1306 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset; 1307 } 1308 1309 auto Err = Replaces.add(tooling::Replacement( 1310 FileName, Includes.front().Offset, IncludesBlockSize, result)); 1311 // FIXME: better error handling. For now, just skip the replacement for the 1312 // release version. 1313 if (Err) 1314 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1315 assert(!Err); 1316 } 1317 1318 namespace { 1319 1320 // This class manages priorities of #include categories and calculates 1321 // priorities for headers. 1322 class IncludeCategoryManager { 1323 public: 1324 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName) 1325 : Style(Style), FileName(FileName) { 1326 FileStem = llvm::sys::path::stem(FileName); 1327 for (const auto &Category : Style.IncludeCategories) 1328 CategoryRegexs.emplace_back(Category.Regex); 1329 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") || 1330 FileName.endswith(".cpp") || FileName.endswith(".c++") || 1331 FileName.endswith(".cxx") || FileName.endswith(".m") || 1332 FileName.endswith(".mm"); 1333 } 1334 1335 // Returns the priority of the category which \p IncludeName belongs to. 1336 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns 1337 // 0. Otherwise, returns the priority of the matching category or INT_MAX. 1338 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) { 1339 int Ret = INT_MAX; 1340 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i) 1341 if (CategoryRegexs[i].match(IncludeName)) { 1342 Ret = Style.IncludeCategories[i].Priority; 1343 break; 1344 } 1345 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName)) 1346 Ret = 0; 1347 return Ret; 1348 } 1349 1350 private: 1351 bool isMainHeader(StringRef IncludeName) const { 1352 if (!IncludeName.startswith("\"")) 1353 return false; 1354 StringRef HeaderStem = 1355 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1)); 1356 if (FileStem.startswith(HeaderStem)) { 1357 llvm::Regex MainIncludeRegex( 1358 (HeaderStem + Style.IncludeIsMainRegex).str()); 1359 if (MainIncludeRegex.match(FileStem)) 1360 return true; 1361 } 1362 return false; 1363 } 1364 1365 const FormatStyle &Style; 1366 bool IsMainFile; 1367 StringRef FileName; 1368 StringRef FileStem; 1369 SmallVector<llvm::Regex, 4> CategoryRegexs; 1370 }; 1371 1372 const char IncludeRegexPattern[] = 1373 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))"; 1374 1375 } // anonymous namespace 1376 1377 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code, 1378 ArrayRef<tooling::Range> Ranges, 1379 StringRef FileName, 1380 tooling::Replacements &Replaces, 1381 unsigned *Cursor) { 1382 unsigned Prev = 0; 1383 unsigned SearchFrom = 0; 1384 llvm::Regex IncludeRegex(IncludeRegexPattern); 1385 SmallVector<StringRef, 4> Matches; 1386 SmallVector<IncludeDirective, 16> IncludesInBlock; 1387 1388 // In compiled files, consider the first #include to be the main #include of 1389 // the file if it is not a system #include. This ensures that the header 1390 // doesn't have hidden dependencies 1391 // (http://llvm.org/docs/CodingStandards.html#include-style). 1392 // 1393 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix 1394 // cases where the first #include is unlikely to be the main header. 1395 IncludeCategoryManager Categories(Style, FileName); 1396 bool FirstIncludeBlock = true; 1397 bool MainIncludeFound = false; 1398 bool FormattingOff = false; 1399 1400 for (;;) { 1401 auto Pos = Code.find('\n', SearchFrom); 1402 StringRef Line = 1403 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev); 1404 1405 StringRef Trimmed = Line.trim(); 1406 if (Trimmed == "// clang-format off") 1407 FormattingOff = true; 1408 else if (Trimmed == "// clang-format on") 1409 FormattingOff = false; 1410 1411 if (!FormattingOff && !Line.endswith("\\")) { 1412 if (IncludeRegex.match(Line, &Matches)) { 1413 StringRef IncludeName = Matches[2]; 1414 int Category = Categories.getIncludePriority( 1415 IncludeName, 1416 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock); 1417 if (Category == 0) 1418 MainIncludeFound = true; 1419 IncludesInBlock.push_back({IncludeName, Line, Prev, Category}); 1420 } else if (!IncludesInBlock.empty()) { 1421 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, 1422 Cursor); 1423 IncludesInBlock.clear(); 1424 FirstIncludeBlock = false; 1425 } 1426 Prev = Pos + 1; 1427 } 1428 if (Pos == StringRef::npos || Pos + 1 == Code.size()) 1429 break; 1430 SearchFrom = Pos + 1; 1431 } 1432 if (!IncludesInBlock.empty()) 1433 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor); 1434 return Replaces; 1435 } 1436 1437 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code, 1438 ArrayRef<tooling::Range> Ranges, 1439 StringRef FileName, unsigned *Cursor) { 1440 tooling::Replacements Replaces; 1441 if (!Style.SortIncludes) 1442 return Replaces; 1443 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript) 1444 return sortJavaScriptImports(Style, Code, Ranges, FileName); 1445 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor); 1446 return Replaces; 1447 } 1448 1449 template <typename T> 1450 static llvm::Expected<tooling::Replacements> 1451 processReplacements(T ProcessFunc, StringRef Code, 1452 const tooling::Replacements &Replaces, 1453 const FormatStyle &Style) { 1454 if (Replaces.empty()) 1455 return tooling::Replacements(); 1456 1457 auto NewCode = applyAllReplacements(Code, Replaces); 1458 if (!NewCode) 1459 return NewCode.takeError(); 1460 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges(); 1461 StringRef FileName = Replaces.begin()->getFilePath(); 1462 1463 tooling::Replacements FormatReplaces = 1464 ProcessFunc(Style, *NewCode, ChangedRanges, FileName); 1465 1466 return Replaces.merge(FormatReplaces); 1467 } 1468 1469 llvm::Expected<tooling::Replacements> 1470 formatReplacements(StringRef Code, const tooling::Replacements &Replaces, 1471 const FormatStyle &Style) { 1472 // We need to use lambda function here since there are two versions of 1473 // `sortIncludes`. 1474 auto SortIncludes = [](const FormatStyle &Style, StringRef Code, 1475 std::vector<tooling::Range> Ranges, 1476 StringRef FileName) -> tooling::Replacements { 1477 return sortIncludes(Style, Code, Ranges, FileName); 1478 }; 1479 auto SortedReplaces = 1480 processReplacements(SortIncludes, Code, Replaces, Style); 1481 if (!SortedReplaces) 1482 return SortedReplaces.takeError(); 1483 1484 // We need to use lambda function here since there are two versions of 1485 // `reformat`. 1486 auto Reformat = [](const FormatStyle &Style, StringRef Code, 1487 std::vector<tooling::Range> Ranges, 1488 StringRef FileName) -> tooling::Replacements { 1489 return reformat(Style, Code, Ranges, FileName); 1490 }; 1491 return processReplacements(Reformat, Code, *SortedReplaces, Style); 1492 } 1493 1494 namespace { 1495 1496 inline bool isHeaderInsertion(const tooling::Replacement &Replace) { 1497 return Replace.getOffset() == UINT_MAX && 1498 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText()); 1499 } 1500 1501 void skipComments(Lexer &Lex, Token &Tok) { 1502 while (Tok.is(tok::comment)) 1503 if (Lex.LexFromRawLexer(Tok)) 1504 return; 1505 } 1506 1507 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is, 1508 // \p Tok will be the token after this directive; otherwise, it can be any token 1509 // after the given \p Tok (including \p Tok). 1510 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) { 1511 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) && 1512 Tok.is(tok::raw_identifier) && 1513 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) && 1514 Tok.is(tok::raw_identifier); 1515 if (Matched) 1516 Lex.LexFromRawLexer(Tok); 1517 return Matched; 1518 } 1519 1520 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName, 1521 StringRef Code, 1522 const FormatStyle &Style) { 1523 std::unique_ptr<Environment> Env = 1524 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{}); 1525 const SourceManager &SourceMgr = Env->getSourceManager(); 1526 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr, 1527 getFormattingLangOpts(Style)); 1528 Token Tok; 1529 // Get the first token. 1530 Lex.LexFromRawLexer(Tok); 1531 skipComments(Lex, Tok); 1532 unsigned AfterComments = SourceMgr.getFileOffset(Tok.getLocation()); 1533 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) { 1534 skipComments(Lex, Tok); 1535 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok)) 1536 return SourceMgr.getFileOffset(Tok.getLocation()); 1537 } 1538 return AfterComments; 1539 } 1540 1541 // FIXME: we also need to insert a '\n' at the end of the code if we have an 1542 // insertion with offset Code.size(), and there is no '\n' at the end of the 1543 // code. 1544 // FIXME: do not insert headers into conditional #include blocks, e.g. #includes 1545 // surrounded by compile condition "#if...". 1546 // FIXME: insert empty lines between newly created blocks. 1547 tooling::Replacements 1548 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces, 1549 const FormatStyle &Style) { 1550 if (Style.Language != FormatStyle::LanguageKind::LK_Cpp) 1551 return Replaces; 1552 1553 tooling::Replacements HeaderInsertions; 1554 tooling::Replacements Result; 1555 for (const auto &R : Replaces) { 1556 if (isHeaderInsertion(R)) { 1557 // Replacements from \p Replaces must be conflict-free already, so we can 1558 // simply consume the error. 1559 llvm::consumeError(HeaderInsertions.add(R)); 1560 } else if (R.getOffset() == UINT_MAX) { 1561 llvm::errs() << "Insertions other than header #include insertion are " 1562 "not supported! " 1563 << R.getReplacementText() << "\n"; 1564 } else { 1565 llvm::consumeError(Result.add(R)); 1566 } 1567 } 1568 if (HeaderInsertions.empty()) 1569 return Replaces; 1570 1571 llvm::Regex IncludeRegex(IncludeRegexPattern); 1572 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)"); 1573 SmallVector<StringRef, 4> Matches; 1574 1575 StringRef FileName = Replaces.begin()->getFilePath(); 1576 IncludeCategoryManager Categories(Style, FileName); 1577 1578 // Record the offset of the end of the last include in each category. 1579 std::map<int, int> CategoryEndOffsets; 1580 // All possible priorities. 1581 // Add 0 for main header and INT_MAX for headers that are not in any category. 1582 std::set<int> Priorities = {0, INT_MAX}; 1583 for (const auto &Category : Style.IncludeCategories) 1584 Priorities.insert(Category.Priority); 1585 int FirstIncludeOffset = -1; 1586 // All new headers should be inserted after this offset. 1587 unsigned MinInsertOffset = 1588 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style); 1589 StringRef TrimmedCode = Code.drop_front(MinInsertOffset); 1590 SmallVector<StringRef, 32> Lines; 1591 TrimmedCode.split(Lines, '\n'); 1592 unsigned Offset = MinInsertOffset; 1593 unsigned NextLineOffset; 1594 std::set<StringRef> ExistingIncludes; 1595 for (auto Line : Lines) { 1596 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1); 1597 if (IncludeRegex.match(Line, &Matches)) { 1598 StringRef IncludeName = Matches[2]; 1599 ExistingIncludes.insert(IncludeName); 1600 int Category = Categories.getIncludePriority( 1601 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0); 1602 CategoryEndOffsets[Category] = NextLineOffset; 1603 if (FirstIncludeOffset < 0) 1604 FirstIncludeOffset = Offset; 1605 } 1606 Offset = NextLineOffset; 1607 } 1608 1609 // Populate CategoryEndOfssets: 1610 // - Ensure that CategoryEndOffset[Highest] is always populated. 1611 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that 1612 // is set, up to CategoryEndOffset[Highest]. 1613 auto Highest = Priorities.begin(); 1614 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) { 1615 if (FirstIncludeOffset >= 0) 1616 CategoryEndOffsets[*Highest] = FirstIncludeOffset; 1617 else 1618 CategoryEndOffsets[*Highest] = MinInsertOffset; 1619 } 1620 // By this point, CategoryEndOffset[Highest] is always set appropriately: 1621 // - to an appropriate location before/after existing #includes, or 1622 // - to right after the header guard, or 1623 // - to the beginning of the file. 1624 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I) 1625 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end()) 1626 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)]; 1627 1628 for (const auto &R : HeaderInsertions) { 1629 auto IncludeDirective = R.getReplacementText(); 1630 bool Matched = IncludeRegex.match(IncludeDirective, &Matches); 1631 assert(Matched && "Header insertion replacement must have replacement text " 1632 "'#include ...'"); 1633 (void)Matched; 1634 auto IncludeName = Matches[2]; 1635 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) { 1636 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName 1637 << "\n"); 1638 continue; 1639 } 1640 int Category = 1641 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true); 1642 Offset = CategoryEndOffsets[Category]; 1643 std::string NewInclude = !IncludeDirective.endswith("\n") 1644 ? (IncludeDirective + "\n").str() 1645 : IncludeDirective.str(); 1646 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude); 1647 auto Err = Result.add(NewReplace); 1648 if (Err) { 1649 llvm::consumeError(std::move(Err)); 1650 Result = Result.merge(tooling::Replacements(NewReplace)); 1651 } 1652 } 1653 return Result; 1654 } 1655 1656 } // anonymous namespace 1657 1658 llvm::Expected<tooling::Replacements> 1659 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces, 1660 const FormatStyle &Style) { 1661 // We need to use lambda function here since there are two versions of 1662 // `cleanup`. 1663 auto Cleanup = [](const FormatStyle &Style, StringRef Code, 1664 std::vector<tooling::Range> Ranges, 1665 StringRef FileName) -> tooling::Replacements { 1666 return cleanup(Style, Code, Ranges, FileName); 1667 }; 1668 // Make header insertion replacements insert new headers into correct blocks. 1669 tooling::Replacements NewReplaces = 1670 fixCppIncludeInsertions(Code, Replaces, Style); 1671 return processReplacements(Cleanup, Code, NewReplaces, Style); 1672 } 1673 1674 tooling::Replacements reformat(const FormatStyle &Style, SourceManager &SM, 1675 FileID ID, ArrayRef<CharSourceRange> Ranges, 1676 bool *IncompleteFormat) { 1677 FormatStyle Expanded = expandPresets(Style); 1678 if (Expanded.DisableFormat) 1679 return tooling::Replacements(); 1680 1681 Environment Env(SM, ID, Ranges); 1682 Formatter Format(Env, Expanded, IncompleteFormat); 1683 return Format.process(); 1684 } 1685 1686 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1687 ArrayRef<tooling::Range> Ranges, 1688 StringRef FileName, bool *IncompleteFormat) { 1689 FormatStyle Expanded = expandPresets(Style); 1690 if (Expanded.DisableFormat) 1691 return tooling::Replacements(); 1692 1693 std::unique_ptr<Environment> Env = 1694 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 1695 Formatter Format(*Env, Expanded, IncompleteFormat); 1696 return Format.process(); 1697 } 1698 1699 tooling::Replacements cleanup(const FormatStyle &Style, SourceManager &SM, 1700 FileID ID, ArrayRef<CharSourceRange> Ranges) { 1701 Environment Env(SM, ID, Ranges); 1702 Cleaner Clean(Env, Style); 1703 return Clean.process(); 1704 } 1705 1706 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code, 1707 ArrayRef<tooling::Range> Ranges, 1708 StringRef FileName) { 1709 std::unique_ptr<Environment> Env = 1710 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 1711 Cleaner Clean(*Env, Style); 1712 return Clean.process(); 1713 } 1714 1715 LangOptions getFormattingLangOpts(const FormatStyle &Style) { 1716 LangOptions LangOpts; 1717 LangOpts.CPlusPlus = 1; 1718 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1719 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1720 LangOpts.LineComment = 1; 1721 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp; 1722 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0; 1723 LangOpts.Bool = 1; 1724 LangOpts.ObjC1 = 1; 1725 LangOpts.ObjC2 = 1; 1726 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally. 1727 LangOpts.DeclSpecKeyword = 1; // To get __declspec. 1728 return LangOpts; 1729 } 1730 1731 const char *StyleOptionHelpDescription = 1732 "Coding style, currently supports:\n" 1733 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1734 "Use -style=file to load style configuration from\n" 1735 ".clang-format file located in one of the parent\n" 1736 "directories of the source file (or current\n" 1737 "directory for stdin).\n" 1738 "Use -style=\"{key: value, ...}\" to set specific\n" 1739 "parameters, e.g.:\n" 1740 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1741 1742 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 1743 if (FileName.endswith(".java")) 1744 return FormatStyle::LK_Java; 1745 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) 1746 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript. 1747 if (FileName.endswith_lower(".proto") || 1748 FileName.endswith_lower(".protodevel")) 1749 return FormatStyle::LK_Proto; 1750 if (FileName.endswith_lower(".td")) 1751 return FormatStyle::LK_TableGen; 1752 return FormatStyle::LK_Cpp; 1753 } 1754 1755 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 1756 StringRef FallbackStyle, vfs::FileSystem *FS) { 1757 if (!FS) { 1758 FS = vfs::getRealFileSystem().get(); 1759 } 1760 FormatStyle Style = getLLVMStyle(); 1761 Style.Language = getLanguageByFileName(FileName); 1762 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) { 1763 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 1764 << "\" using LLVM style\n"; 1765 return Style; 1766 } 1767 1768 if (StyleName.startswith("{")) { 1769 // Parse YAML/JSON style from the command line. 1770 if (std::error_code ec = parseConfiguration(StyleName, &Style)) { 1771 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1772 << FallbackStyle << " style\n"; 1773 } 1774 return Style; 1775 } 1776 1777 if (!StyleName.equals_lower("file")) { 1778 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 1779 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1780 << " style\n"; 1781 return Style; 1782 } 1783 1784 // Look for .clang-format/_clang-format file in the file's parent directories. 1785 SmallString<128> UnsuitableConfigFiles; 1786 SmallString<128> Path(FileName); 1787 llvm::sys::fs::make_absolute(Path); 1788 for (StringRef Directory = Path; !Directory.empty(); 1789 Directory = llvm::sys::path::parent_path(Directory)) { 1790 1791 auto Status = FS->status(Directory); 1792 if (!Status || 1793 Status->getType() != llvm::sys::fs::file_type::directory_file) { 1794 continue; 1795 } 1796 1797 SmallString<128> ConfigFile(Directory); 1798 1799 llvm::sys::path::append(ConfigFile, ".clang-format"); 1800 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1801 1802 Status = FS->status(ConfigFile.str()); 1803 bool IsFile = 1804 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file); 1805 if (!IsFile) { 1806 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1807 ConfigFile = Directory; 1808 llvm::sys::path::append(ConfigFile, "_clang-format"); 1809 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1810 Status = FS->status(ConfigFile.str()); 1811 IsFile = Status && 1812 (Status->getType() == llvm::sys::fs::file_type::regular_file); 1813 } 1814 1815 if (IsFile) { 1816 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text = 1817 FS->getBufferForFile(ConfigFile.str()); 1818 if (std::error_code EC = Text.getError()) { 1819 llvm::errs() << EC.message() << "\n"; 1820 break; 1821 } 1822 if (std::error_code ec = 1823 parseConfiguration(Text.get()->getBuffer(), &Style)) { 1824 if (ec == ParseError::Unsuitable) { 1825 if (!UnsuitableConfigFiles.empty()) 1826 UnsuitableConfigFiles.append(", "); 1827 UnsuitableConfigFiles.append(ConfigFile); 1828 continue; 1829 } 1830 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1831 << "\n"; 1832 break; 1833 } 1834 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1835 return Style; 1836 } 1837 } 1838 if (!UnsuitableConfigFiles.empty()) { 1839 llvm::errs() << "Configuration file(s) do(es) not support " 1840 << getLanguageName(Style.Language) << ": " 1841 << UnsuitableConfigFiles << "\n"; 1842 } 1843 return Style; 1844 } 1845 1846 } // namespace format 1847 } // namespace clang 1848