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 "FormatInternal.h" 20 #include "FormatTokenLexer.h" 21 #include "NamespaceEndCommentsFixer.h" 22 #include "SortJavaScriptImports.h" 23 #include "TokenAnalyzer.h" 24 #include "TokenAnnotator.h" 25 #include "UnwrappedLineFormatter.h" 26 #include "UnwrappedLineParser.h" 27 #include "UsingDeclarationsSorter.h" 28 #include "WhitespaceManager.h" 29 #include "clang/Basic/Diagnostic.h" 30 #include "clang/Basic/DiagnosticOptions.h" 31 #include "clang/Basic/SourceManager.h" 32 #include "clang/Basic/VirtualFileSystem.h" 33 #include "clang/Lex/Lexer.h" 34 #include "llvm/ADT/STLExtras.h" 35 #include "llvm/Support/Allocator.h" 36 #include "llvm/Support/Debug.h" 37 #include "llvm/Support/Path.h" 38 #include "llvm/Support/Regex.h" 39 #include "llvm/Support/YAMLTraits.h" 40 #include <algorithm> 41 #include <memory> 42 #include <string> 43 44 #define DEBUG_TYPE "format-formatter" 45 46 using clang::format::FormatStyle; 47 48 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory) 49 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::RawStringFormat) 50 51 namespace llvm { 52 namespace yaml { 53 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> { 54 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) { 55 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp); 56 IO.enumCase(Value, "Java", FormatStyle::LK_Java); 57 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript); 58 IO.enumCase(Value, "ObjC", FormatStyle::LK_ObjC); 59 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto); 60 IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen); 61 IO.enumCase(Value, "TextProto", FormatStyle::LK_TextProto); 62 } 63 }; 64 65 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> { 66 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) { 67 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03); 68 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03); 69 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11); 70 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11); 71 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto); 72 } 73 }; 74 75 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> { 76 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) { 77 IO.enumCase(Value, "Never", FormatStyle::UT_Never); 78 IO.enumCase(Value, "false", FormatStyle::UT_Never); 79 IO.enumCase(Value, "Always", FormatStyle::UT_Always); 80 IO.enumCase(Value, "true", FormatStyle::UT_Always); 81 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation); 82 IO.enumCase(Value, "ForContinuationAndIndentation", 83 FormatStyle::UT_ForContinuationAndIndentation); 84 } 85 }; 86 87 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> { 88 static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) { 89 IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave); 90 IO.enumCase(Value, "Single", FormatStyle::JSQS_Single); 91 IO.enumCase(Value, "Double", FormatStyle::JSQS_Double); 92 } 93 }; 94 95 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> { 96 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) { 97 IO.enumCase(Value, "None", FormatStyle::SFS_None); 98 IO.enumCase(Value, "false", FormatStyle::SFS_None); 99 IO.enumCase(Value, "All", FormatStyle::SFS_All); 100 IO.enumCase(Value, "true", FormatStyle::SFS_All); 101 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline); 102 IO.enumCase(Value, "InlineOnly", FormatStyle::SFS_InlineOnly); 103 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty); 104 } 105 }; 106 107 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> { 108 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) { 109 IO.enumCase(Value, "All", FormatStyle::BOS_All); 110 IO.enumCase(Value, "true", FormatStyle::BOS_All); 111 IO.enumCase(Value, "None", FormatStyle::BOS_None); 112 IO.enumCase(Value, "false", FormatStyle::BOS_None); 113 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment); 114 } 115 }; 116 117 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> { 118 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) { 119 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach); 120 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux); 121 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla); 122 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup); 123 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman); 124 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU); 125 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit); 126 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom); 127 } 128 }; 129 130 template <> 131 struct ScalarEnumerationTraits<FormatStyle::BreakConstructorInitializersStyle> { 132 static void 133 enumeration(IO &IO, FormatStyle::BreakConstructorInitializersStyle &Value) { 134 IO.enumCase(Value, "BeforeColon", FormatStyle::BCIS_BeforeColon); 135 IO.enumCase(Value, "BeforeComma", FormatStyle::BCIS_BeforeComma); 136 IO.enumCase(Value, "AfterColon", FormatStyle::BCIS_AfterColon); 137 } 138 }; 139 140 template <> 141 struct ScalarEnumerationTraits<FormatStyle::PPDirectiveIndentStyle> { 142 static void enumeration(IO &IO, FormatStyle::PPDirectiveIndentStyle &Value) { 143 IO.enumCase(Value, "None", FormatStyle::PPDIS_None); 144 IO.enumCase(Value, "AfterHash", FormatStyle::PPDIS_AfterHash); 145 } 146 }; 147 148 template <> 149 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> { 150 static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) { 151 IO.enumCase(Value, "None", FormatStyle::RTBS_None); 152 IO.enumCase(Value, "All", FormatStyle::RTBS_All); 153 IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel); 154 IO.enumCase(Value, "TopLevelDefinitions", 155 FormatStyle::RTBS_TopLevelDefinitions); 156 IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions); 157 } 158 }; 159 160 template <> 161 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> { 162 static void 163 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) { 164 IO.enumCase(Value, "None", FormatStyle::DRTBS_None); 165 IO.enumCase(Value, "All", FormatStyle::DRTBS_All); 166 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel); 167 168 // For backward compatibility. 169 IO.enumCase(Value, "false", FormatStyle::DRTBS_None); 170 IO.enumCase(Value, "true", FormatStyle::DRTBS_All); 171 } 172 }; 173 174 template <> 175 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> { 176 static void enumeration(IO &IO, 177 FormatStyle::NamespaceIndentationKind &Value) { 178 IO.enumCase(Value, "None", FormatStyle::NI_None); 179 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner); 180 IO.enumCase(Value, "All", FormatStyle::NI_All); 181 } 182 }; 183 184 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> { 185 static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) { 186 IO.enumCase(Value, "Align", FormatStyle::BAS_Align); 187 IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign); 188 IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak); 189 190 // For backward compatibility. 191 IO.enumCase(Value, "true", FormatStyle::BAS_Align); 192 IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign); 193 } 194 }; 195 196 template <> 197 struct ScalarEnumerationTraits<FormatStyle::EscapedNewlineAlignmentStyle> { 198 static void enumeration(IO &IO, 199 FormatStyle::EscapedNewlineAlignmentStyle &Value) { 200 IO.enumCase(Value, "DontAlign", FormatStyle::ENAS_DontAlign); 201 IO.enumCase(Value, "Left", FormatStyle::ENAS_Left); 202 IO.enumCase(Value, "Right", FormatStyle::ENAS_Right); 203 204 // For backward compatibility. 205 IO.enumCase(Value, "true", FormatStyle::ENAS_Left); 206 IO.enumCase(Value, "false", FormatStyle::ENAS_Right); 207 } 208 }; 209 210 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> { 211 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) { 212 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle); 213 IO.enumCase(Value, "Left", FormatStyle::PAS_Left); 214 IO.enumCase(Value, "Right", FormatStyle::PAS_Right); 215 216 // For backward compatibility. 217 IO.enumCase(Value, "true", FormatStyle::PAS_Left); 218 IO.enumCase(Value, "false", FormatStyle::PAS_Right); 219 } 220 }; 221 222 template <> 223 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> { 224 static void enumeration(IO &IO, 225 FormatStyle::SpaceBeforeParensOptions &Value) { 226 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never); 227 IO.enumCase(Value, "ControlStatements", 228 FormatStyle::SBPO_ControlStatements); 229 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always); 230 231 // For backward compatibility. 232 IO.enumCase(Value, "false", FormatStyle::SBPO_Never); 233 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements); 234 } 235 }; 236 237 template <> struct MappingTraits<FormatStyle> { 238 static void mapping(IO &IO, FormatStyle &Style) { 239 // When reading, read the language first, we need it for getPredefinedStyle. 240 IO.mapOptional("Language", Style.Language); 241 242 if (IO.outputting()) { 243 StringRef StylesArray[] = {"LLVM", "Google", "Chromium", 244 "Mozilla", "WebKit", "GNU"}; 245 ArrayRef<StringRef> Styles(StylesArray); 246 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 247 StringRef StyleName(Styles[i]); 248 FormatStyle PredefinedStyle; 249 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) && 250 Style == PredefinedStyle) { 251 IO.mapOptional("# BasedOnStyle", StyleName); 252 break; 253 } 254 } 255 } else { 256 StringRef BasedOnStyle; 257 IO.mapOptional("BasedOnStyle", BasedOnStyle); 258 if (!BasedOnStyle.empty()) { 259 FormatStyle::LanguageKind OldLanguage = Style.Language; 260 FormatStyle::LanguageKind Language = 261 ((FormatStyle *)IO.getContext())->Language; 262 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) { 263 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 264 return; 265 } 266 Style.Language = OldLanguage; 267 } 268 } 269 270 // For backward compatibility. 271 if (!IO.outputting()) { 272 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlines); 273 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment); 274 IO.mapOptional("IndentFunctionDeclarationAfterType", 275 Style.IndentWrappedFunctionNames); 276 IO.mapOptional("PointerBindsToType", Style.PointerAlignment); 277 IO.mapOptional("SpaceAfterControlStatementKeyword", 278 Style.SpaceBeforeParens); 279 } 280 281 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 282 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket); 283 IO.mapOptional("AlignConsecutiveAssignments", 284 Style.AlignConsecutiveAssignments); 285 IO.mapOptional("AlignConsecutiveDeclarations", 286 Style.AlignConsecutiveDeclarations); 287 IO.mapOptional("AlignEscapedNewlines", Style.AlignEscapedNewlines); 288 IO.mapOptional("AlignOperands", Style.AlignOperands); 289 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 290 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 291 Style.AllowAllParametersOfDeclarationOnNextLine); 292 IO.mapOptional("AllowShortBlocksOnASingleLine", 293 Style.AllowShortBlocksOnASingleLine); 294 IO.mapOptional("AllowShortCaseLabelsOnASingleLine", 295 Style.AllowShortCaseLabelsOnASingleLine); 296 IO.mapOptional("AllowShortFunctionsOnASingleLine", 297 Style.AllowShortFunctionsOnASingleLine); 298 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 299 Style.AllowShortIfStatementsOnASingleLine); 300 IO.mapOptional("AllowShortLoopsOnASingleLine", 301 Style.AllowShortLoopsOnASingleLine); 302 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType", 303 Style.AlwaysBreakAfterDefinitionReturnType); 304 IO.mapOptional("AlwaysBreakAfterReturnType", 305 Style.AlwaysBreakAfterReturnType); 306 // If AlwaysBreakAfterDefinitionReturnType was specified but 307 // AlwaysBreakAfterReturnType was not, initialize the latter from the 308 // former for backwards compatibility. 309 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None && 310 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) { 311 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All) 312 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 313 else if (Style.AlwaysBreakAfterDefinitionReturnType == 314 FormatStyle::DRTBS_TopLevel) 315 Style.AlwaysBreakAfterReturnType = 316 FormatStyle::RTBS_TopLevelDefinitions; 317 } 318 319 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 320 Style.AlwaysBreakBeforeMultilineStrings); 321 IO.mapOptional("AlwaysBreakTemplateDeclarations", 322 Style.AlwaysBreakTemplateDeclarations); 323 IO.mapOptional("BinPackArguments", Style.BinPackArguments); 324 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 325 IO.mapOptional("BraceWrapping", Style.BraceWrapping); 326 IO.mapOptional("BreakBeforeBinaryOperators", 327 Style.BreakBeforeBinaryOperators); 328 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 329 IO.mapOptional("BreakBeforeInheritanceComma", 330 Style.BreakBeforeInheritanceComma); 331 IO.mapOptional("BreakBeforeTernaryOperators", 332 Style.BreakBeforeTernaryOperators); 333 334 bool BreakConstructorInitializersBeforeComma = false; 335 IO.mapOptional("BreakConstructorInitializersBeforeComma", 336 BreakConstructorInitializersBeforeComma); 337 IO.mapOptional("BreakConstructorInitializers", 338 Style.BreakConstructorInitializers); 339 // If BreakConstructorInitializersBeforeComma was specified but 340 // BreakConstructorInitializers was not, initialize the latter from the 341 // former for backwards compatibility. 342 if (BreakConstructorInitializersBeforeComma && 343 Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeColon) 344 Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma; 345 346 IO.mapOptional("BreakAfterJavaFieldAnnotations", 347 Style.BreakAfterJavaFieldAnnotations); 348 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals); 349 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 350 IO.mapOptional("CommentPragmas", Style.CommentPragmas); 351 IO.mapOptional("CompactNamespaces", Style.CompactNamespaces); 352 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 353 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 354 IO.mapOptional("ConstructorInitializerIndentWidth", 355 Style.ConstructorInitializerIndentWidth); 356 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 357 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 358 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment); 359 IO.mapOptional("DisableFormat", Style.DisableFormat); 360 IO.mapOptional("ExperimentalAutoDetectBinPacking", 361 Style.ExperimentalAutoDetectBinPacking); 362 IO.mapOptional("FixNamespaceComments", Style.FixNamespaceComments); 363 IO.mapOptional("ForEachMacros", Style.ForEachMacros); 364 IO.mapOptional("IncludeBlocks", Style.IncludeBlocks); 365 IO.mapOptional("IncludeCategories", Style.IncludeCategories); 366 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex); 367 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 368 IO.mapOptional("IndentPPDirectives", Style.IndentPPDirectives); 369 IO.mapOptional("IndentWidth", Style.IndentWidth); 370 IO.mapOptional("IndentWrappedFunctionNames", 371 Style.IndentWrappedFunctionNames); 372 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes); 373 IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports); 374 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks", 375 Style.KeepEmptyLinesAtTheStartOfBlocks); 376 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin); 377 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd); 378 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 379 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 380 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth); 381 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty); 382 IO.mapOptional("ObjCSpaceBeforeProtocolList", 383 Style.ObjCSpaceBeforeProtocolList); 384 IO.mapOptional("PenaltyBreakAssignment", Style.PenaltyBreakAssignment); 385 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 386 Style.PenaltyBreakBeforeFirstCallParameter); 387 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 388 IO.mapOptional("PenaltyBreakFirstLessLess", 389 Style.PenaltyBreakFirstLessLess); 390 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 391 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 392 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 393 Style.PenaltyReturnTypeOnItsOwnLine); 394 IO.mapOptional("PointerAlignment", Style.PointerAlignment); 395 IO.mapOptional("RawStringFormats", Style.RawStringFormats); 396 IO.mapOptional("ReflowComments", Style.ReflowComments); 397 IO.mapOptional("SortIncludes", Style.SortIncludes); 398 IO.mapOptional("SortUsingDeclarations", Style.SortUsingDeclarations); 399 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast); 400 IO.mapOptional("SpaceAfterTemplateKeyword", 401 Style.SpaceAfterTemplateKeyword); 402 IO.mapOptional("SpaceBeforeAssignmentOperators", 403 Style.SpaceBeforeAssignmentOperators); 404 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens); 405 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 406 IO.mapOptional("SpacesBeforeTrailingComments", 407 Style.SpacesBeforeTrailingComments); 408 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 409 IO.mapOptional("SpacesInContainerLiterals", 410 Style.SpacesInContainerLiterals); 411 IO.mapOptional("SpacesInCStyleCastParentheses", 412 Style.SpacesInCStyleCastParentheses); 413 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 414 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets); 415 IO.mapOptional("Standard", Style.Standard); 416 IO.mapOptional("TabWidth", Style.TabWidth); 417 IO.mapOptional("UseTab", Style.UseTab); 418 } 419 }; 420 421 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> { 422 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) { 423 IO.mapOptional("AfterClass", Wrapping.AfterClass); 424 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement); 425 IO.mapOptional("AfterEnum", Wrapping.AfterEnum); 426 IO.mapOptional("AfterFunction", Wrapping.AfterFunction); 427 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace); 428 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration); 429 IO.mapOptional("AfterStruct", Wrapping.AfterStruct); 430 IO.mapOptional("AfterUnion", Wrapping.AfterUnion); 431 IO.mapOptional("AfterExternBlock", Wrapping.AfterExternBlock); 432 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch); 433 IO.mapOptional("BeforeElse", Wrapping.BeforeElse); 434 IO.mapOptional("IndentBraces", Wrapping.IndentBraces); 435 IO.mapOptional("SplitEmptyFunction", Wrapping.SplitEmptyFunction); 436 IO.mapOptional("SplitEmptyRecord", Wrapping.SplitEmptyRecord); 437 IO.mapOptional("SplitEmptyNamespace", Wrapping.SplitEmptyNamespace); 438 } 439 }; 440 441 template <> struct MappingTraits<FormatStyle::IncludeCategory> { 442 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) { 443 IO.mapOptional("Regex", Category.Regex); 444 IO.mapOptional("Priority", Category.Priority); 445 } 446 }; 447 448 template <> struct ScalarEnumerationTraits<FormatStyle::IncludeBlocksStyle> { 449 static void enumeration(IO &IO, FormatStyle::IncludeBlocksStyle &Value) { 450 IO.enumCase(Value, "Preserve", FormatStyle::IBS_Preserve); 451 IO.enumCase(Value, "Merge", FormatStyle::IBS_Merge); 452 IO.enumCase(Value, "Regroup", FormatStyle::IBS_Regroup); 453 } 454 }; 455 456 template <> struct MappingTraits<FormatStyle::RawStringFormat> { 457 static void mapping(IO &IO, FormatStyle::RawStringFormat &Format) { 458 IO.mapOptional("Delimiter", Format.Delimiter); 459 IO.mapOptional("Language", Format.Language); 460 IO.mapOptional("BasedOnStyle", Format.BasedOnStyle); 461 } 462 }; 463 464 // Allows to read vector<FormatStyle> while keeping default values. 465 // IO.getContext() should contain a pointer to the FormatStyle structure, that 466 // will be used to get default values for missing keys. 467 // If the first element has no Language specified, it will be treated as the 468 // default one for the following elements. 469 template <> struct DocumentListTraits<std::vector<FormatStyle>> { 470 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) { 471 return Seq.size(); 472 } 473 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq, 474 size_t Index) { 475 if (Index >= Seq.size()) { 476 assert(Index == Seq.size()); 477 FormatStyle Template; 478 if (!Seq.empty() && Seq[0].Language == FormatStyle::LK_None) { 479 Template = Seq[0]; 480 } else { 481 Template = *((const FormatStyle *)IO.getContext()); 482 Template.Language = FormatStyle::LK_None; 483 } 484 Seq.resize(Index + 1, Template); 485 } 486 return Seq[Index]; 487 } 488 }; 489 } // namespace yaml 490 } // namespace llvm 491 492 namespace clang { 493 namespace format { 494 495 const std::error_category &getParseCategory() { 496 static ParseErrorCategory C; 497 return C; 498 } 499 std::error_code make_error_code(ParseError e) { 500 return std::error_code(static_cast<int>(e), getParseCategory()); 501 } 502 503 inline llvm::Error make_string_error(const llvm::Twine &Message) { 504 return llvm::make_error<llvm::StringError>(Message, 505 llvm::inconvertibleErrorCode()); 506 } 507 508 const char *ParseErrorCategory::name() const noexcept { 509 return "clang-format.parse_error"; 510 } 511 512 std::string ParseErrorCategory::message(int EV) const { 513 switch (static_cast<ParseError>(EV)) { 514 case ParseError::Success: 515 return "Success"; 516 case ParseError::Error: 517 return "Invalid argument"; 518 case ParseError::Unsuitable: 519 return "Unsuitable"; 520 } 521 llvm_unreachable("unexpected parse error"); 522 } 523 524 static FormatStyle expandPresets(const FormatStyle &Style) { 525 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom) 526 return Style; 527 FormatStyle Expanded = Style; 528 Expanded.BraceWrapping = {false, false, false, false, false, 529 false, false, false, false, false, 530 false, false, true, true, true}; 531 switch (Style.BreakBeforeBraces) { 532 case FormatStyle::BS_Linux: 533 Expanded.BraceWrapping.AfterClass = true; 534 Expanded.BraceWrapping.AfterFunction = true; 535 Expanded.BraceWrapping.AfterNamespace = true; 536 break; 537 case FormatStyle::BS_Mozilla: 538 Expanded.BraceWrapping.AfterClass = true; 539 Expanded.BraceWrapping.AfterEnum = true; 540 Expanded.BraceWrapping.AfterFunction = true; 541 Expanded.BraceWrapping.AfterStruct = true; 542 Expanded.BraceWrapping.AfterUnion = true; 543 Expanded.BraceWrapping.AfterExternBlock = true; 544 Expanded.BraceWrapping.SplitEmptyFunction = true; 545 Expanded.BraceWrapping.SplitEmptyRecord = false; 546 break; 547 case FormatStyle::BS_Stroustrup: 548 Expanded.BraceWrapping.AfterFunction = true; 549 Expanded.BraceWrapping.BeforeCatch = true; 550 Expanded.BraceWrapping.BeforeElse = true; 551 break; 552 case FormatStyle::BS_Allman: 553 Expanded.BraceWrapping.AfterClass = true; 554 Expanded.BraceWrapping.AfterControlStatement = true; 555 Expanded.BraceWrapping.AfterEnum = true; 556 Expanded.BraceWrapping.AfterFunction = true; 557 Expanded.BraceWrapping.AfterNamespace = true; 558 Expanded.BraceWrapping.AfterObjCDeclaration = true; 559 Expanded.BraceWrapping.AfterStruct = true; 560 Expanded.BraceWrapping.AfterExternBlock = true; 561 Expanded.BraceWrapping.BeforeCatch = true; 562 Expanded.BraceWrapping.BeforeElse = true; 563 break; 564 case FormatStyle::BS_GNU: 565 Expanded.BraceWrapping = {true, true, true, true, true, true, true, true, 566 true, true, true, true, true, true, true}; 567 break; 568 case FormatStyle::BS_WebKit: 569 Expanded.BraceWrapping.AfterFunction = true; 570 break; 571 default: 572 break; 573 } 574 return Expanded; 575 } 576 577 FormatStyle getLLVMStyle() { 578 FormatStyle LLVMStyle; 579 LLVMStyle.Language = FormatStyle::LK_Cpp; 580 LLVMStyle.AccessModifierOffset = -2; 581 LLVMStyle.AlignEscapedNewlines = FormatStyle::ENAS_Right; 582 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align; 583 LLVMStyle.AlignOperands = true; 584 LLVMStyle.AlignTrailingComments = true; 585 LLVMStyle.AlignConsecutiveAssignments = false; 586 LLVMStyle.AlignConsecutiveDeclarations = false; 587 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 588 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All; 589 LLVMStyle.AllowShortBlocksOnASingleLine = false; 590 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false; 591 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 592 LLVMStyle.AllowShortLoopsOnASingleLine = false; 593 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None; 594 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None; 595 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 596 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 597 LLVMStyle.BinPackArguments = true; 598 LLVMStyle.BinPackParameters = true; 599 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None; 600 LLVMStyle.BreakBeforeTernaryOperators = true; 601 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 602 LLVMStyle.BraceWrapping = {false, false, false, false, false, 603 false, false, false, false, false, 604 false, false, true, true, true}; 605 LLVMStyle.BreakAfterJavaFieldAnnotations = false; 606 LLVMStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeColon; 607 LLVMStyle.BreakBeforeInheritanceComma = false; 608 LLVMStyle.BreakStringLiterals = true; 609 LLVMStyle.ColumnLimit = 80; 610 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; 611 LLVMStyle.CompactNamespaces = false; 612 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 613 LLVMStyle.ConstructorInitializerIndentWidth = 4; 614 LLVMStyle.ContinuationIndentWidth = 4; 615 LLVMStyle.Cpp11BracedListStyle = true; 616 LLVMStyle.DerivePointerAlignment = false; 617 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 618 LLVMStyle.FixNamespaceComments = true; 619 LLVMStyle.ForEachMacros.push_back("foreach"); 620 LLVMStyle.ForEachMacros.push_back("Q_FOREACH"); 621 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH"); 622 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2}, 623 {"^(<|\"(gtest|gmock|isl|json)/)", 3}, 624 {".*", 1}}; 625 LLVMStyle.IncludeIsMainRegex = "(Test)?$"; 626 LLVMStyle.IncludeBlocks = FormatStyle::IBS_Preserve; 627 LLVMStyle.IndentCaseLabels = false; 628 LLVMStyle.IndentPPDirectives = FormatStyle::PPDIS_None; 629 LLVMStyle.IndentWrappedFunctionNames = false; 630 LLVMStyle.IndentWidth = 2; 631 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave; 632 LLVMStyle.JavaScriptWrapImports = true; 633 LLVMStyle.TabWidth = 8; 634 LLVMStyle.MaxEmptyLinesToKeep = 1; 635 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true; 636 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 637 LLVMStyle.ObjCBlockIndentWidth = 2; 638 LLVMStyle.ObjCSpaceAfterProperty = false; 639 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 640 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right; 641 LLVMStyle.SpacesBeforeTrailingComments = 1; 642 LLVMStyle.Standard = FormatStyle::LS_Cpp11; 643 LLVMStyle.UseTab = FormatStyle::UT_Never; 644 LLVMStyle.RawStringFormats = {{"pb", FormatStyle::LK_TextProto, "google"}}; 645 LLVMStyle.ReflowComments = true; 646 LLVMStyle.SpacesInParentheses = false; 647 LLVMStyle.SpacesInSquareBrackets = false; 648 LLVMStyle.SpaceInEmptyParentheses = false; 649 LLVMStyle.SpacesInContainerLiterals = true; 650 LLVMStyle.SpacesInCStyleCastParentheses = false; 651 LLVMStyle.SpaceAfterCStyleCast = false; 652 LLVMStyle.SpaceAfterTemplateKeyword = true; 653 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements; 654 LLVMStyle.SpaceBeforeAssignmentOperators = true; 655 LLVMStyle.SpacesInAngles = false; 656 657 LLVMStyle.PenaltyBreakAssignment = prec::Assignment; 658 LLVMStyle.PenaltyBreakComment = 300; 659 LLVMStyle.PenaltyBreakFirstLessLess = 120; 660 LLVMStyle.PenaltyBreakString = 1000; 661 LLVMStyle.PenaltyExcessCharacter = 1000000; 662 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 663 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 664 665 LLVMStyle.DisableFormat = false; 666 LLVMStyle.SortIncludes = true; 667 LLVMStyle.SortUsingDeclarations = true; 668 669 return LLVMStyle; 670 } 671 672 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) { 673 if (Language == FormatStyle::LK_TextProto) { 674 FormatStyle GoogleStyle = getGoogleStyle(FormatStyle::LK_Proto); 675 GoogleStyle.Language = FormatStyle::LK_TextProto; 676 return GoogleStyle; 677 } 678 679 FormatStyle GoogleStyle = getLLVMStyle(); 680 GoogleStyle.Language = Language; 681 682 GoogleStyle.AccessModifierOffset = -1; 683 GoogleStyle.AlignEscapedNewlines = FormatStyle::ENAS_Left; 684 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 685 GoogleStyle.AllowShortLoopsOnASingleLine = true; 686 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 687 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 688 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 689 GoogleStyle.DerivePointerAlignment = true; 690 GoogleStyle.IncludeCategories = { 691 {"^<ext/.*\\.h>", 2}, {"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}}; 692 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$"; 693 GoogleStyle.IndentCaseLabels = true; 694 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false; 695 GoogleStyle.ObjCSpaceAfterProperty = false; 696 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 697 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left; 698 GoogleStyle.SpacesBeforeTrailingComments = 2; 699 GoogleStyle.Standard = FormatStyle::LS_Auto; 700 701 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 702 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 703 704 if (Language == FormatStyle::LK_Java) { 705 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 706 GoogleStyle.AlignOperands = false; 707 GoogleStyle.AlignTrailingComments = false; 708 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 709 GoogleStyle.AllowShortIfStatementsOnASingleLine = false; 710 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 711 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment; 712 GoogleStyle.ColumnLimit = 100; 713 GoogleStyle.SpaceAfterCStyleCast = true; 714 GoogleStyle.SpacesBeforeTrailingComments = 1; 715 } else if (Language == FormatStyle::LK_JavaScript) { 716 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak; 717 GoogleStyle.AlignOperands = false; 718 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty; 719 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false; 720 GoogleStyle.BreakBeforeTernaryOperators = false; 721 // taze:, triple slash directives (`/// <...`), @tag followed by { for a lot 722 // of JSDoc tags, and @see, which is commonly followed by overlong URLs. 723 GoogleStyle.CommentPragmas = 724 "(taze:|^/[ \t]*<|(@[A-Za-z_0-9-]+[ \\t]*{)|@see)"; 725 GoogleStyle.MaxEmptyLinesToKeep = 3; 726 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All; 727 GoogleStyle.SpacesInContainerLiterals = false; 728 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single; 729 GoogleStyle.JavaScriptWrapImports = false; 730 } else if (Language == FormatStyle::LK_Proto) { 731 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None; 732 GoogleStyle.SpacesInContainerLiterals = false; 733 } else if (Language == FormatStyle::LK_ObjC) { 734 GoogleStyle.ColumnLimit = 100; 735 } 736 737 return GoogleStyle; 738 } 739 740 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) { 741 FormatStyle ChromiumStyle = getGoogleStyle(Language); 742 if (Language == FormatStyle::LK_Java) { 743 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true; 744 ChromiumStyle.BreakAfterJavaFieldAnnotations = true; 745 ChromiumStyle.ContinuationIndentWidth = 8; 746 ChromiumStyle.IndentWidth = 4; 747 } else if (Language == FormatStyle::LK_JavaScript) { 748 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 749 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 750 } else { 751 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 752 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 753 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 754 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 755 ChromiumStyle.BinPackParameters = false; 756 ChromiumStyle.DerivePointerAlignment = false; 757 if (Language == FormatStyle::LK_ObjC) 758 ChromiumStyle.ColumnLimit = 80; 759 } 760 return ChromiumStyle; 761 } 762 763 FormatStyle getMozillaStyle() { 764 FormatStyle MozillaStyle = getLLVMStyle(); 765 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 766 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline; 767 MozillaStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_TopLevel; 768 MozillaStyle.AlwaysBreakAfterDefinitionReturnType = 769 FormatStyle::DRTBS_TopLevel; 770 MozillaStyle.AlwaysBreakTemplateDeclarations = true; 771 MozillaStyle.BinPackParameters = false; 772 MozillaStyle.BinPackArguments = false; 773 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla; 774 MozillaStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma; 775 MozillaStyle.BreakBeforeInheritanceComma = true; 776 MozillaStyle.ConstructorInitializerIndentWidth = 2; 777 MozillaStyle.ContinuationIndentWidth = 2; 778 MozillaStyle.Cpp11BracedListStyle = false; 779 MozillaStyle.FixNamespaceComments = false; 780 MozillaStyle.IndentCaseLabels = true; 781 MozillaStyle.ObjCSpaceAfterProperty = true; 782 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 783 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 784 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left; 785 MozillaStyle.SpaceAfterTemplateKeyword = false; 786 return MozillaStyle; 787 } 788 789 FormatStyle getWebKitStyle() { 790 FormatStyle Style = getLLVMStyle(); 791 Style.AccessModifierOffset = -4; 792 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign; 793 Style.AlignOperands = false; 794 Style.AlignTrailingComments = false; 795 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 796 Style.BreakBeforeBraces = FormatStyle::BS_WebKit; 797 Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma; 798 Style.Cpp11BracedListStyle = false; 799 Style.ColumnLimit = 0; 800 Style.FixNamespaceComments = false; 801 Style.IndentWidth = 4; 802 Style.NamespaceIndentation = FormatStyle::NI_Inner; 803 Style.ObjCBlockIndentWidth = 4; 804 Style.ObjCSpaceAfterProperty = true; 805 Style.PointerAlignment = FormatStyle::PAS_Left; 806 return Style; 807 } 808 809 FormatStyle getGNUStyle() { 810 FormatStyle Style = getLLVMStyle(); 811 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All; 812 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions; 813 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All; 814 Style.BreakBeforeBraces = FormatStyle::BS_GNU; 815 Style.BreakBeforeTernaryOperators = true; 816 Style.Cpp11BracedListStyle = false; 817 Style.ColumnLimit = 79; 818 Style.FixNamespaceComments = false; 819 Style.SpaceBeforeParens = FormatStyle::SBPO_Always; 820 Style.Standard = FormatStyle::LS_Cpp03; 821 return Style; 822 } 823 824 FormatStyle getNoStyle() { 825 FormatStyle NoStyle = getLLVMStyle(); 826 NoStyle.DisableFormat = true; 827 NoStyle.SortIncludes = false; 828 NoStyle.SortUsingDeclarations = false; 829 return NoStyle; 830 } 831 832 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language, 833 FormatStyle *Style) { 834 if (Name.equals_lower("llvm")) { 835 *Style = getLLVMStyle(); 836 } else if (Name.equals_lower("chromium")) { 837 *Style = getChromiumStyle(Language); 838 } else if (Name.equals_lower("mozilla")) { 839 *Style = getMozillaStyle(); 840 } else if (Name.equals_lower("google")) { 841 *Style = getGoogleStyle(Language); 842 } else if (Name.equals_lower("webkit")) { 843 *Style = getWebKitStyle(); 844 } else if (Name.equals_lower("gnu")) { 845 *Style = getGNUStyle(); 846 } else if (Name.equals_lower("none")) { 847 *Style = getNoStyle(); 848 } else { 849 return false; 850 } 851 852 Style->Language = Language; 853 return true; 854 } 855 856 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 857 assert(Style); 858 FormatStyle::LanguageKind Language = Style->Language; 859 assert(Language != FormatStyle::LK_None); 860 if (Text.trim().empty()) 861 return make_error_code(ParseError::Error); 862 863 std::vector<FormatStyle> Styles; 864 llvm::yaml::Input Input(Text); 865 // DocumentListTraits<vector<FormatStyle>> uses the context to get default 866 // values for the fields, keys for which are missing from the configuration. 867 // Mapping also uses the context to get the language to find the correct 868 // base style. 869 Input.setContext(Style); 870 Input >> Styles; 871 if (Input.error()) 872 return Input.error(); 873 874 for (unsigned i = 0; i < Styles.size(); ++i) { 875 // Ensures that only the first configuration can skip the Language option. 876 if (Styles[i].Language == FormatStyle::LK_None && i != 0) 877 return make_error_code(ParseError::Error); 878 // Ensure that each language is configured at most once. 879 for (unsigned j = 0; j < i; ++j) { 880 if (Styles[i].Language == Styles[j].Language) { 881 DEBUG(llvm::dbgs() 882 << "Duplicate languages in the config file on positions " << j 883 << " and " << i << "\n"); 884 return make_error_code(ParseError::Error); 885 } 886 } 887 } 888 // Look for a suitable configuration starting from the end, so we can 889 // find the configuration for the specific language first, and the default 890 // configuration (which can only be at slot 0) after it. 891 for (int i = Styles.size() - 1; i >= 0; --i) { 892 if (Styles[i].Language == Language || 893 Styles[i].Language == FormatStyle::LK_None) { 894 *Style = Styles[i]; 895 Style->Language = Language; 896 return make_error_code(ParseError::Success); 897 } 898 } 899 return make_error_code(ParseError::Unsuitable); 900 } 901 902 std::string configurationAsText(const FormatStyle &Style) { 903 std::string Text; 904 llvm::raw_string_ostream Stream(Text); 905 llvm::yaml::Output Output(Stream); 906 // We use the same mapping method for input and output, so we need a non-const 907 // reference here. 908 FormatStyle NonConstStyle = expandPresets(Style); 909 Output << NonConstStyle; 910 return Stream.str(); 911 } 912 913 namespace { 914 915 class JavaScriptRequoter : public TokenAnalyzer { 916 public: 917 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style) 918 : TokenAnalyzer(Env, Style) {} 919 920 std::pair<tooling::Replacements, unsigned> 921 analyze(TokenAnnotator &Annotator, 922 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 923 FormatTokenLexer &Tokens) override { 924 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 925 AnnotatedLines.end()); 926 tooling::Replacements Result; 927 requoteJSStringLiteral(AnnotatedLines, Result); 928 return {Result, 0}; 929 } 930 931 private: 932 // Replaces double/single-quoted string literal as appropriate, re-escaping 933 // the contents in the process. 934 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines, 935 tooling::Replacements &Result) { 936 for (AnnotatedLine *Line : Lines) { 937 requoteJSStringLiteral(Line->Children, Result); 938 if (!Line->Affected) 939 continue; 940 for (FormatToken *FormatTok = Line->First; FormatTok; 941 FormatTok = FormatTok->Next) { 942 StringRef Input = FormatTok->TokenText; 943 if (FormatTok->Finalized || !FormatTok->isStringLiteral() || 944 // NB: testing for not starting with a double quote to avoid 945 // breaking `template strings`. 946 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single && 947 !Input.startswith("\"")) || 948 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double && 949 !Input.startswith("\'"))) 950 continue; 951 952 // Change start and end quote. 953 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single; 954 SourceLocation Start = FormatTok->Tok.getLocation(); 955 auto Replace = [&](SourceLocation Start, unsigned Length, 956 StringRef ReplacementText) { 957 auto Err = Result.add(tooling::Replacement( 958 Env.getSourceManager(), Start, Length, ReplacementText)); 959 // FIXME: handle error. For now, print error message and skip the 960 // replacement for release version. 961 if (Err) { 962 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 963 assert(false); 964 } 965 }; 966 Replace(Start, 1, IsSingle ? "'" : "\""); 967 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1, 968 IsSingle ? "'" : "\""); 969 970 // Escape internal quotes. 971 bool Escaped = false; 972 for (size_t i = 1; i < Input.size() - 1; i++) { 973 switch (Input[i]) { 974 case '\\': 975 if (!Escaped && i + 1 < Input.size() && 976 ((IsSingle && Input[i + 1] == '"') || 977 (!IsSingle && Input[i + 1] == '\''))) { 978 // Remove this \, it's escaping a " or ' that no longer needs 979 // escaping 980 Replace(Start.getLocWithOffset(i), 1, ""); 981 continue; 982 } 983 Escaped = !Escaped; 984 break; 985 case '\"': 986 case '\'': 987 if (!Escaped && IsSingle == (Input[i] == '\'')) { 988 // Escape the quote. 989 Replace(Start.getLocWithOffset(i), 0, "\\"); 990 } 991 Escaped = false; 992 break; 993 default: 994 Escaped = false; 995 break; 996 } 997 } 998 } 999 } 1000 } 1001 }; 1002 1003 class Formatter : public TokenAnalyzer { 1004 public: 1005 Formatter(const Environment &Env, const FormatStyle &Style, 1006 FormattingAttemptStatus *Status) 1007 : TokenAnalyzer(Env, Style), Status(Status) {} 1008 1009 std::pair<tooling::Replacements, unsigned> 1010 analyze(TokenAnnotator &Annotator, 1011 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1012 FormatTokenLexer &Tokens) override { 1013 tooling::Replacements Result; 1014 deriveLocalStyle(AnnotatedLines); 1015 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1016 AnnotatedLines.end()); 1017 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1018 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 1019 } 1020 Annotator.setCommentLineLevels(AnnotatedLines); 1021 1022 WhitespaceManager Whitespaces( 1023 Env.getSourceManager(), Style, 1024 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID()))); 1025 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), 1026 Env.getSourceManager(), Whitespaces, Encoding, 1027 BinPackInconclusiveFunctions); 1028 unsigned Penalty = 1029 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, 1030 Tokens.getKeywords(), Env.getSourceManager(), 1031 Status) 1032 .format(AnnotatedLines, /*DryRun=*/false, 1033 /*AdditionalIndent=*/0, 1034 /*FixBadIndentation=*/false, 1035 /*FirstStartColumn=*/Env.getFirstStartColumn(), 1036 /*NextStartColumn=*/Env.getNextStartColumn(), 1037 /*LastStartColumn=*/Env.getLastStartColumn()); 1038 for (const auto &R : Whitespaces.generateReplacements()) 1039 if (Result.add(R)) 1040 return std::make_pair(Result, 0); 1041 return std::make_pair(Result, Penalty); 1042 } 1043 1044 private: 1045 static bool inputUsesCRLF(StringRef Text) { 1046 return Text.count('\r') * 2 > Text.count('\n'); 1047 } 1048 1049 bool 1050 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) { 1051 for (const AnnotatedLine *Line : Lines) { 1052 if (hasCpp03IncompatibleFormat(Line->Children)) 1053 return true; 1054 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) { 1055 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1056 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener)) 1057 return true; 1058 if (Tok->is(TT_TemplateCloser) && 1059 Tok->Previous->is(TT_TemplateCloser)) 1060 return true; 1061 } 1062 } 1063 } 1064 return false; 1065 } 1066 1067 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) { 1068 int AlignmentDiff = 0; 1069 for (const AnnotatedLine *Line : Lines) { 1070 AlignmentDiff += countVariableAlignments(Line->Children); 1071 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) { 1072 if (!Tok->is(TT_PointerOrReference)) 1073 continue; 1074 bool SpaceBefore = 1075 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1076 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() != 1077 Tok->Next->WhitespaceRange.getEnd(); 1078 if (SpaceBefore && !SpaceAfter) 1079 ++AlignmentDiff; 1080 if (!SpaceBefore && SpaceAfter) 1081 --AlignmentDiff; 1082 } 1083 } 1084 return AlignmentDiff; 1085 } 1086 1087 void 1088 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1089 bool HasBinPackedFunction = false; 1090 bool HasOnePerLineFunction = false; 1091 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1092 if (!AnnotatedLines[i]->First->Next) 1093 continue; 1094 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1095 while (Tok->Next) { 1096 if (Tok->PackingKind == PPK_BinPacked) 1097 HasBinPackedFunction = true; 1098 if (Tok->PackingKind == PPK_OnePerLine) 1099 HasOnePerLineFunction = true; 1100 1101 Tok = Tok->Next; 1102 } 1103 } 1104 if (Style.DerivePointerAlignment) 1105 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0 1106 ? FormatStyle::PAS_Left 1107 : FormatStyle::PAS_Right; 1108 if (Style.Standard == FormatStyle::LS_Auto) 1109 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines) 1110 ? FormatStyle::LS_Cpp11 1111 : FormatStyle::LS_Cpp03; 1112 BinPackInconclusiveFunctions = 1113 HasBinPackedFunction || !HasOnePerLineFunction; 1114 } 1115 1116 bool BinPackInconclusiveFunctions; 1117 FormattingAttemptStatus *Status; 1118 }; 1119 1120 // This class clean up the erroneous/redundant code around the given ranges in 1121 // file. 1122 class Cleaner : public TokenAnalyzer { 1123 public: 1124 Cleaner(const Environment &Env, const FormatStyle &Style) 1125 : TokenAnalyzer(Env, Style), 1126 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {} 1127 1128 // FIXME: eliminate unused parameters. 1129 std::pair<tooling::Replacements, unsigned> 1130 analyze(TokenAnnotator &Annotator, 1131 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1132 FormatTokenLexer &Tokens) override { 1133 // FIXME: in the current implementation the granularity of affected range 1134 // is an annotated line. However, this is not sufficient. Furthermore, 1135 // redundant code introduced by replacements does not necessarily 1136 // intercept with ranges of replacements that result in the redundancy. 1137 // To determine if some redundant code is actually introduced by 1138 // replacements(e.g. deletions), we need to come up with a more 1139 // sophisticated way of computing affected ranges. 1140 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(), 1141 AnnotatedLines.end()); 1142 1143 checkEmptyNamespace(AnnotatedLines); 1144 1145 for (auto &Line : AnnotatedLines) { 1146 if (Line->Affected) { 1147 cleanupRight(Line->First, tok::comma, tok::comma); 1148 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma); 1149 cleanupRight(Line->First, tok::l_paren, tok::comma); 1150 cleanupLeft(Line->First, tok::comma, tok::r_paren); 1151 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace); 1152 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace); 1153 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal); 1154 } 1155 } 1156 1157 return {generateFixes(), 0}; 1158 } 1159 1160 private: 1161 bool containsOnlyComments(const AnnotatedLine &Line) { 1162 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) { 1163 if (Tok->isNot(tok::comment)) 1164 return false; 1165 } 1166 return true; 1167 } 1168 1169 // Iterate through all lines and remove any empty (nested) namespaces. 1170 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1171 std::set<unsigned> DeletedLines; 1172 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1173 auto &Line = *AnnotatedLines[i]; 1174 if (Line.startsWith(tok::kw_namespace) || 1175 Line.startsWith(tok::kw_inline, tok::kw_namespace)) { 1176 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines); 1177 } 1178 } 1179 1180 for (auto Line : DeletedLines) { 1181 FormatToken *Tok = AnnotatedLines[Line]->First; 1182 while (Tok) { 1183 deleteToken(Tok); 1184 Tok = Tok->Next; 1185 } 1186 } 1187 } 1188 1189 // The function checks if the namespace, which starts from \p CurrentLine, and 1190 // its nested namespaces are empty and delete them if they are empty. It also 1191 // sets \p NewLine to the last line checked. 1192 // Returns true if the current namespace is empty. 1193 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1194 unsigned CurrentLine, unsigned &NewLine, 1195 std::set<unsigned> &DeletedLines) { 1196 unsigned InitLine = CurrentLine, End = AnnotatedLines.size(); 1197 if (Style.BraceWrapping.AfterNamespace) { 1198 // If the left brace is in a new line, we should consume it first so that 1199 // it does not make the namespace non-empty. 1200 // FIXME: error handling if there is no left brace. 1201 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) { 1202 NewLine = CurrentLine; 1203 return false; 1204 } 1205 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) { 1206 return false; 1207 } 1208 while (++CurrentLine < End) { 1209 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace)) 1210 break; 1211 1212 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) || 1213 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline, 1214 tok::kw_namespace)) { 1215 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine, 1216 DeletedLines)) 1217 return false; 1218 CurrentLine = NewLine; 1219 continue; 1220 } 1221 1222 if (containsOnlyComments(*AnnotatedLines[CurrentLine])) 1223 continue; 1224 1225 // If there is anything other than comments or nested namespaces in the 1226 // current namespace, the namespace cannot be empty. 1227 NewLine = CurrentLine; 1228 return false; 1229 } 1230 1231 NewLine = CurrentLine; 1232 if (CurrentLine >= End) 1233 return false; 1234 1235 // Check if the empty namespace is actually affected by changed ranges. 1236 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange( 1237 AnnotatedLines[InitLine]->First->Tok.getLocation(), 1238 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc()))) 1239 return false; 1240 1241 for (unsigned i = InitLine; i <= CurrentLine; ++i) { 1242 DeletedLines.insert(i); 1243 } 1244 1245 return true; 1246 } 1247 1248 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one 1249 // of the token in the pair if the left token has \p LK token kind and the 1250 // right token has \p RK token kind. If \p DeleteLeft is true, the left token 1251 // is deleted on match; otherwise, the right token is deleted. 1252 template <typename LeftKind, typename RightKind> 1253 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK, 1254 bool DeleteLeft) { 1255 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * { 1256 for (auto *Res = Tok.Next; Res; Res = Res->Next) 1257 if (!Res->is(tok::comment) && 1258 DeletedTokens.find(Res) == DeletedTokens.end()) 1259 return Res; 1260 return nullptr; 1261 }; 1262 for (auto *Left = Start; Left;) { 1263 auto *Right = NextNotDeleted(*Left); 1264 if (!Right) 1265 break; 1266 if (Left->is(LK) && Right->is(RK)) { 1267 deleteToken(DeleteLeft ? Left : Right); 1268 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next) 1269 deleteToken(Tok); 1270 // If the right token is deleted, we should keep the left token 1271 // unchanged and pair it with the new right token. 1272 if (!DeleteLeft) 1273 continue; 1274 } 1275 Left = Right; 1276 } 1277 } 1278 1279 template <typename LeftKind, typename RightKind> 1280 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) { 1281 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true); 1282 } 1283 1284 template <typename LeftKind, typename RightKind> 1285 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) { 1286 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false); 1287 } 1288 1289 // Delete the given token. 1290 inline void deleteToken(FormatToken *Tok) { 1291 if (Tok) 1292 DeletedTokens.insert(Tok); 1293 } 1294 1295 tooling::Replacements generateFixes() { 1296 tooling::Replacements Fixes; 1297 std::vector<FormatToken *> Tokens; 1298 std::copy(DeletedTokens.begin(), DeletedTokens.end(), 1299 std::back_inserter(Tokens)); 1300 1301 // Merge multiple continuous token deletions into one big deletion so that 1302 // the number of replacements can be reduced. This makes computing affected 1303 // ranges more efficient when we run reformat on the changed code. 1304 unsigned Idx = 0; 1305 while (Idx < Tokens.size()) { 1306 unsigned St = Idx, End = Idx; 1307 while ((End + 1) < Tokens.size() && 1308 Tokens[End]->Next == Tokens[End + 1]) { 1309 End++; 1310 } 1311 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(), 1312 Tokens[End]->Tok.getEndLoc()); 1313 auto Err = 1314 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, "")); 1315 // FIXME: better error handling. for now just print error message and skip 1316 // for the release version. 1317 if (Err) { 1318 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1319 assert(false && "Fixes must not conflict!"); 1320 } 1321 Idx = End + 1; 1322 } 1323 1324 return Fixes; 1325 } 1326 1327 // Class for less-than inequality comparason for the set `RedundantTokens`. 1328 // We store tokens in the order they appear in the translation unit so that 1329 // we do not need to sort them in `generateFixes()`. 1330 struct FormatTokenLess { 1331 FormatTokenLess(const SourceManager &SM) : SM(SM) {} 1332 1333 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const { 1334 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(), 1335 RHS->Tok.getLocation()); 1336 } 1337 const SourceManager &SM; 1338 }; 1339 1340 // Tokens to be deleted. 1341 std::set<FormatToken *, FormatTokenLess> DeletedTokens; 1342 }; 1343 1344 struct IncludeDirective { 1345 StringRef Filename; 1346 StringRef Text; 1347 unsigned Offset; 1348 int Category; 1349 }; 1350 1351 } // end anonymous namespace 1352 1353 // Determines whether 'Ranges' intersects with ('Start', 'End'). 1354 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start, 1355 unsigned End) { 1356 for (auto Range : Ranges) { 1357 if (Range.getOffset() < End && 1358 Range.getOffset() + Range.getLength() > Start) 1359 return true; 1360 } 1361 return false; 1362 } 1363 1364 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor 1365 // before sorting/deduplicating. Index is the index of the include under the 1366 // cursor in the original set of includes. If this include has duplicates, it is 1367 // the index of the first of the duplicates as the others are going to be 1368 // removed. OffsetToEOL describes the cursor's position relative to the end of 1369 // its current line. 1370 // If `Cursor` is not on any #include, `Index` will be UINT_MAX. 1371 static std::pair<unsigned, unsigned> 1372 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes, 1373 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) { 1374 unsigned CursorIndex = UINT_MAX; 1375 unsigned OffsetToEOL = 0; 1376 for (int i = 0, e = Includes.size(); i != e; ++i) { 1377 unsigned Start = Includes[Indices[i]].Offset; 1378 unsigned End = Start + Includes[Indices[i]].Text.size(); 1379 if (!(Cursor >= Start && Cursor < End)) 1380 continue; 1381 CursorIndex = Indices[i]; 1382 OffsetToEOL = End - Cursor; 1383 // Put the cursor on the only remaining #include among the duplicate 1384 // #includes. 1385 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text) 1386 CursorIndex = i; 1387 break; 1388 } 1389 return std::make_pair(CursorIndex, OffsetToEOL); 1390 } 1391 1392 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically 1393 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict 1394 // source order. 1395 // #include directives with the same text will be deduplicated, and only the 1396 // first #include in the duplicate #includes remains. If the `Cursor` is 1397 // provided and put on a deleted #include, it will be moved to the remaining 1398 // #include in the duplicate #includes. 1399 static void sortCppIncludes(const FormatStyle &Style, 1400 const SmallVectorImpl<IncludeDirective> &Includes, 1401 ArrayRef<tooling::Range> Ranges, StringRef FileName, 1402 tooling::Replacements &Replaces, unsigned *Cursor) { 1403 unsigned IncludesBeginOffset = Includes.front().Offset; 1404 unsigned IncludesEndOffset = 1405 Includes.back().Offset + Includes.back().Text.size(); 1406 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset; 1407 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset)) 1408 return; 1409 SmallVector<unsigned, 16> Indices; 1410 for (unsigned i = 0, e = Includes.size(); i != e; ++i) 1411 Indices.push_back(i); 1412 std::stable_sort( 1413 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) { 1414 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) < 1415 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename); 1416 }); 1417 // The index of the include on which the cursor will be put after 1418 // sorting/deduplicating. 1419 unsigned CursorIndex; 1420 // The offset from cursor to the end of line. 1421 unsigned CursorToEOLOffset; 1422 if (Cursor) 1423 std::tie(CursorIndex, CursorToEOLOffset) = 1424 FindCursorIndex(Includes, Indices, *Cursor); 1425 1426 // Deduplicate #includes. 1427 Indices.erase(std::unique(Indices.begin(), Indices.end(), 1428 [&](unsigned LHSI, unsigned RHSI) { 1429 return Includes[LHSI].Text == Includes[RHSI].Text; 1430 }), 1431 Indices.end()); 1432 1433 int CurrentCategory = Includes.front().Category; 1434 1435 // If the #includes are out of order, we generate a single replacement fixing 1436 // the entire block. Otherwise, no replacement is generated. 1437 if (Indices.size() == Includes.size() && 1438 std::is_sorted(Indices.begin(), Indices.end()) && 1439 Style.IncludeBlocks == FormatStyle::IBS_Preserve) 1440 return; 1441 1442 std::string result; 1443 for (unsigned Index : Indices) { 1444 if (!result.empty()) { 1445 result += "\n"; 1446 if (Style.IncludeBlocks == FormatStyle::IBS_Regroup && 1447 CurrentCategory != Includes[Index].Category) 1448 result += "\n"; 1449 } 1450 result += Includes[Index].Text; 1451 if (Cursor && CursorIndex == Index) 1452 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset; 1453 CurrentCategory = Includes[Index].Category; 1454 } 1455 1456 auto Err = Replaces.add(tooling::Replacement( 1457 FileName, Includes.front().Offset, IncludesBlockSize, result)); 1458 // FIXME: better error handling. For now, just skip the replacement for the 1459 // release version. 1460 if (Err) { 1461 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1462 assert(false); 1463 } 1464 } 1465 1466 namespace { 1467 1468 // This class manages priorities of #include categories and calculates 1469 // priorities for headers. 1470 class IncludeCategoryManager { 1471 public: 1472 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName) 1473 : Style(Style), FileName(FileName) { 1474 FileStem = llvm::sys::path::stem(FileName); 1475 for (const auto &Category : Style.IncludeCategories) 1476 CategoryRegexs.emplace_back(Category.Regex, llvm::Regex::IgnoreCase); 1477 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") || 1478 FileName.endswith(".cpp") || FileName.endswith(".c++") || 1479 FileName.endswith(".cxx") || FileName.endswith(".m") || 1480 FileName.endswith(".mm"); 1481 } 1482 1483 // Returns the priority of the category which \p IncludeName belongs to. 1484 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns 1485 // 0. Otherwise, returns the priority of the matching category or INT_MAX. 1486 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) { 1487 int Ret = INT_MAX; 1488 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i) 1489 if (CategoryRegexs[i].match(IncludeName)) { 1490 Ret = Style.IncludeCategories[i].Priority; 1491 break; 1492 } 1493 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName)) 1494 Ret = 0; 1495 return Ret; 1496 } 1497 1498 private: 1499 bool isMainHeader(StringRef IncludeName) const { 1500 if (!IncludeName.startswith("\"")) 1501 return false; 1502 StringRef HeaderStem = 1503 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1)); 1504 if (FileStem.startswith(HeaderStem) || 1505 FileStem.startswith_lower(HeaderStem)) { 1506 llvm::Regex MainIncludeRegex( 1507 (HeaderStem + Style.IncludeIsMainRegex).str(), 1508 llvm::Regex::IgnoreCase); 1509 if (MainIncludeRegex.match(FileStem)) 1510 return true; 1511 } 1512 return false; 1513 } 1514 1515 const FormatStyle &Style; 1516 bool IsMainFile; 1517 StringRef FileName; 1518 StringRef FileStem; 1519 SmallVector<llvm::Regex, 4> CategoryRegexs; 1520 }; 1521 1522 const char IncludeRegexPattern[] = 1523 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))"; 1524 1525 } // anonymous namespace 1526 1527 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code, 1528 ArrayRef<tooling::Range> Ranges, 1529 StringRef FileName, 1530 tooling::Replacements &Replaces, 1531 unsigned *Cursor) { 1532 unsigned Prev = 0; 1533 unsigned SearchFrom = 0; 1534 llvm::Regex IncludeRegex(IncludeRegexPattern); 1535 SmallVector<StringRef, 4> Matches; 1536 SmallVector<IncludeDirective, 16> IncludesInBlock; 1537 1538 // In compiled files, consider the first #include to be the main #include of 1539 // the file if it is not a system #include. This ensures that the header 1540 // doesn't have hidden dependencies 1541 // (http://llvm.org/docs/CodingStandards.html#include-style). 1542 // 1543 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix 1544 // cases where the first #include is unlikely to be the main header. 1545 IncludeCategoryManager Categories(Style, FileName); 1546 bool FirstIncludeBlock = true; 1547 bool MainIncludeFound = false; 1548 bool FormattingOff = false; 1549 1550 for (;;) { 1551 auto Pos = Code.find('\n', SearchFrom); 1552 StringRef Line = 1553 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev); 1554 1555 StringRef Trimmed = Line.trim(); 1556 if (Trimmed == "// clang-format off") 1557 FormattingOff = true; 1558 else if (Trimmed == "// clang-format on") 1559 FormattingOff = false; 1560 1561 const bool EmptyLineSkipped = 1562 Trimmed.empty() && (Style.IncludeBlocks == FormatStyle::IBS_Merge || 1563 Style.IncludeBlocks == FormatStyle::IBS_Regroup); 1564 1565 if (!FormattingOff && !Line.endswith("\\")) { 1566 if (IncludeRegex.match(Line, &Matches)) { 1567 StringRef IncludeName = Matches[2]; 1568 int Category = Categories.getIncludePriority( 1569 IncludeName, 1570 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock); 1571 if (Category == 0) 1572 MainIncludeFound = true; 1573 IncludesInBlock.push_back({IncludeName, Line, Prev, Category}); 1574 } else if (!IncludesInBlock.empty() && !EmptyLineSkipped) { 1575 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, 1576 Cursor); 1577 IncludesInBlock.clear(); 1578 FirstIncludeBlock = false; 1579 } 1580 Prev = Pos + 1; 1581 } 1582 if (Pos == StringRef::npos || Pos + 1 == Code.size()) 1583 break; 1584 SearchFrom = Pos + 1; 1585 } 1586 if (!IncludesInBlock.empty()) 1587 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor); 1588 return Replaces; 1589 } 1590 1591 bool isMpegTS(StringRef Code) { 1592 // MPEG transport streams use the ".ts" file extension. clang-format should 1593 // not attempt to format those. MPEG TS' frame format starts with 0x47 every 1594 // 189 bytes - detect that and return. 1595 return Code.size() > 188 && Code[0] == 0x47 && Code[188] == 0x47; 1596 } 1597 1598 bool isLikelyXml(StringRef Code) { return Code.ltrim().startswith("<"); } 1599 1600 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code, 1601 ArrayRef<tooling::Range> Ranges, 1602 StringRef FileName, unsigned *Cursor) { 1603 tooling::Replacements Replaces; 1604 if (!Style.SortIncludes) 1605 return Replaces; 1606 if (isLikelyXml(Code)) 1607 return Replaces; 1608 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript && 1609 isMpegTS(Code)) 1610 return Replaces; 1611 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript) 1612 return sortJavaScriptImports(Style, Code, Ranges, FileName); 1613 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor); 1614 return Replaces; 1615 } 1616 1617 template <typename T> 1618 static llvm::Expected<tooling::Replacements> 1619 processReplacements(T ProcessFunc, StringRef Code, 1620 const tooling::Replacements &Replaces, 1621 const FormatStyle &Style) { 1622 if (Replaces.empty()) 1623 return tooling::Replacements(); 1624 1625 auto NewCode = applyAllReplacements(Code, Replaces); 1626 if (!NewCode) 1627 return NewCode.takeError(); 1628 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges(); 1629 StringRef FileName = Replaces.begin()->getFilePath(); 1630 1631 tooling::Replacements FormatReplaces = 1632 ProcessFunc(Style, *NewCode, ChangedRanges, FileName); 1633 1634 return Replaces.merge(FormatReplaces); 1635 } 1636 1637 llvm::Expected<tooling::Replacements> 1638 formatReplacements(StringRef Code, const tooling::Replacements &Replaces, 1639 const FormatStyle &Style) { 1640 // We need to use lambda function here since there are two versions of 1641 // `sortIncludes`. 1642 auto SortIncludes = [](const FormatStyle &Style, StringRef Code, 1643 std::vector<tooling::Range> Ranges, 1644 StringRef FileName) -> tooling::Replacements { 1645 return sortIncludes(Style, Code, Ranges, FileName); 1646 }; 1647 auto SortedReplaces = 1648 processReplacements(SortIncludes, Code, Replaces, Style); 1649 if (!SortedReplaces) 1650 return SortedReplaces.takeError(); 1651 1652 // We need to use lambda function here since there are two versions of 1653 // `reformat`. 1654 auto Reformat = [](const FormatStyle &Style, StringRef Code, 1655 std::vector<tooling::Range> Ranges, 1656 StringRef FileName) -> tooling::Replacements { 1657 return reformat(Style, Code, Ranges, FileName); 1658 }; 1659 return processReplacements(Reformat, Code, *SortedReplaces, Style); 1660 } 1661 1662 namespace { 1663 1664 inline bool isHeaderInsertion(const tooling::Replacement &Replace) { 1665 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 && 1666 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText()); 1667 } 1668 1669 inline bool isHeaderDeletion(const tooling::Replacement &Replace) { 1670 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1; 1671 } 1672 1673 // Returns the offset after skipping a sequence of tokens, matched by \p 1674 // GetOffsetAfterSequence, from the start of the code. 1675 // \p GetOffsetAfterSequence should be a function that matches a sequence of 1676 // tokens and returns an offset after the sequence. 1677 unsigned getOffsetAfterTokenSequence( 1678 StringRef FileName, StringRef Code, const FormatStyle &Style, 1679 llvm::function_ref<unsigned(const SourceManager &, Lexer &, Token &)> 1680 GetOffsetAfterSequence) { 1681 std::unique_ptr<Environment> Env = 1682 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{}); 1683 const SourceManager &SourceMgr = Env->getSourceManager(); 1684 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr, 1685 getFormattingLangOpts(Style)); 1686 Token Tok; 1687 // Get the first token. 1688 Lex.LexFromRawLexer(Tok); 1689 return GetOffsetAfterSequence(SourceMgr, Lex, Tok); 1690 } 1691 1692 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is, 1693 // \p Tok will be the token after this directive; otherwise, it can be any token 1694 // after the given \p Tok (including \p Tok). 1695 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) { 1696 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) && 1697 Tok.is(tok::raw_identifier) && 1698 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) && 1699 Tok.is(tok::raw_identifier); 1700 if (Matched) 1701 Lex.LexFromRawLexer(Tok); 1702 return Matched; 1703 } 1704 1705 void skipComments(Lexer &Lex, Token &Tok) { 1706 while (Tok.is(tok::comment)) 1707 if (Lex.LexFromRawLexer(Tok)) 1708 return; 1709 } 1710 1711 // Returns the offset after header guard directives and any comments 1712 // before/after header guards. If no header guard presents in the code, this 1713 // will returns the offset after skipping all comments from the start of the 1714 // code. 1715 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName, 1716 StringRef Code, 1717 const FormatStyle &Style) { 1718 return getOffsetAfterTokenSequence( 1719 FileName, Code, Style, 1720 [](const SourceManager &SM, Lexer &Lex, Token Tok) { 1721 skipComments(Lex, Tok); 1722 unsigned InitialOffset = SM.getFileOffset(Tok.getLocation()); 1723 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) { 1724 skipComments(Lex, Tok); 1725 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok)) 1726 return SM.getFileOffset(Tok.getLocation()); 1727 } 1728 return InitialOffset; 1729 }); 1730 } 1731 1732 // Check if a sequence of tokens is like 1733 // "#include ("header.h" | <header.h>)". 1734 // If it is, \p Tok will be the token after this directive; otherwise, it can be 1735 // any token after the given \p Tok (including \p Tok). 1736 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) { 1737 auto Matched = [&]() { 1738 Lex.LexFromRawLexer(Tok); 1739 return true; 1740 }; 1741 if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) && 1742 Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") { 1743 if (Lex.LexFromRawLexer(Tok)) 1744 return false; 1745 if (Tok.is(tok::string_literal)) 1746 return Matched(); 1747 if (Tok.is(tok::less)) { 1748 while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) { 1749 } 1750 if (Tok.is(tok::greater)) 1751 return Matched(); 1752 } 1753 } 1754 return false; 1755 } 1756 1757 // Returns the offset of the last #include directive after which a new 1758 // #include can be inserted. This ignores #include's after the #include block(s) 1759 // in the beginning of a file to avoid inserting headers into code sections 1760 // where new #include's should not be added by default. 1761 // These code sections include: 1762 // - raw string literals (containing #include). 1763 // - #if blocks. 1764 // - Special #include's among declarations (e.g. functions). 1765 // 1766 // If no #include after which a new #include can be inserted, this returns the 1767 // offset after skipping all comments from the start of the code. 1768 // Inserting after an #include is not allowed if it comes after code that is not 1769 // #include (e.g. pre-processing directive that is not #include, declarations). 1770 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code, 1771 const FormatStyle &Style) { 1772 return getOffsetAfterTokenSequence( 1773 FileName, Code, Style, 1774 [](const SourceManager &SM, Lexer &Lex, Token Tok) { 1775 skipComments(Lex, Tok); 1776 unsigned MaxOffset = SM.getFileOffset(Tok.getLocation()); 1777 while (checkAndConsumeInclusiveDirective(Lex, Tok)) 1778 MaxOffset = SM.getFileOffset(Tok.getLocation()); 1779 return MaxOffset; 1780 }); 1781 } 1782 1783 bool isDeletedHeader(llvm::StringRef HeaderName, 1784 const std::set<llvm::StringRef> &HeadersToDelete) { 1785 return HeadersToDelete.count(HeaderName) || 1786 HeadersToDelete.count(HeaderName.trim("\"<>")); 1787 } 1788 1789 // FIXME: insert empty lines between newly created blocks. 1790 tooling::Replacements 1791 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces, 1792 const FormatStyle &Style) { 1793 if (!Style.isCpp()) 1794 return Replaces; 1795 1796 tooling::Replacements HeaderInsertions; 1797 std::set<llvm::StringRef> HeadersToDelete; 1798 tooling::Replacements Result; 1799 for (const auto &R : Replaces) { 1800 if (isHeaderInsertion(R)) { 1801 // Replacements from \p Replaces must be conflict-free already, so we can 1802 // simply consume the error. 1803 llvm::consumeError(HeaderInsertions.add(R)); 1804 } else if (isHeaderDeletion(R)) { 1805 HeadersToDelete.insert(R.getReplacementText()); 1806 } else if (R.getOffset() == UINT_MAX) { 1807 llvm::errs() << "Insertions other than header #include insertion are " 1808 "not supported! " 1809 << R.getReplacementText() << "\n"; 1810 } else { 1811 llvm::consumeError(Result.add(R)); 1812 } 1813 } 1814 if (HeaderInsertions.empty() && HeadersToDelete.empty()) 1815 return Replaces; 1816 1817 llvm::Regex IncludeRegex(IncludeRegexPattern); 1818 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)"); 1819 SmallVector<StringRef, 4> Matches; 1820 1821 StringRef FileName = Replaces.begin()->getFilePath(); 1822 IncludeCategoryManager Categories(Style, FileName); 1823 1824 // Record the offset of the end of the last include in each category. 1825 std::map<int, int> CategoryEndOffsets; 1826 // All possible priorities. 1827 // Add 0 for main header and INT_MAX for headers that are not in any category. 1828 std::set<int> Priorities = {0, INT_MAX}; 1829 for (const auto &Category : Style.IncludeCategories) 1830 Priorities.insert(Category.Priority); 1831 int FirstIncludeOffset = -1; 1832 // All new headers should be inserted after this offset. 1833 unsigned MinInsertOffset = 1834 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style); 1835 StringRef TrimmedCode = Code.drop_front(MinInsertOffset); 1836 // Max insertion offset in the original code. 1837 unsigned MaxInsertOffset = 1838 MinInsertOffset + 1839 getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style); 1840 SmallVector<StringRef, 32> Lines; 1841 TrimmedCode.split(Lines, '\n'); 1842 unsigned Offset = MinInsertOffset; 1843 unsigned NextLineOffset; 1844 std::set<StringRef> ExistingIncludes; 1845 for (auto Line : Lines) { 1846 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1); 1847 if (IncludeRegex.match(Line, &Matches)) { 1848 // The header name with quotes or angle brackets. 1849 StringRef IncludeName = Matches[2]; 1850 ExistingIncludes.insert(IncludeName); 1851 // Only record the offset of current #include if we can insert after it. 1852 if (Offset <= MaxInsertOffset) { 1853 int Category = Categories.getIncludePriority( 1854 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0); 1855 CategoryEndOffsets[Category] = NextLineOffset; 1856 if (FirstIncludeOffset < 0) 1857 FirstIncludeOffset = Offset; 1858 } 1859 if (isDeletedHeader(IncludeName, HeadersToDelete)) { 1860 // If this is the last line without trailing newline, we need to make 1861 // sure we don't delete across the file boundary. 1862 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset); 1863 llvm::Error Err = 1864 Result.add(tooling::Replacement(FileName, Offset, Length, "")); 1865 if (Err) { 1866 // Ignore the deletion on conflict. 1867 llvm::errs() << "Failed to add header deletion replacement for " 1868 << IncludeName << ": " << llvm::toString(std::move(Err)) 1869 << "\n"; 1870 } 1871 } 1872 } 1873 Offset = NextLineOffset; 1874 } 1875 1876 // Populate CategoryEndOfssets: 1877 // - Ensure that CategoryEndOffset[Highest] is always populated. 1878 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that 1879 // is set, up to CategoryEndOffset[Highest]. 1880 auto Highest = Priorities.begin(); 1881 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) { 1882 if (FirstIncludeOffset >= 0) 1883 CategoryEndOffsets[*Highest] = FirstIncludeOffset; 1884 else 1885 CategoryEndOffsets[*Highest] = MinInsertOffset; 1886 } 1887 // By this point, CategoryEndOffset[Highest] is always set appropriately: 1888 // - to an appropriate location before/after existing #includes, or 1889 // - to right after the header guard, or 1890 // - to the beginning of the file. 1891 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I) 1892 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end()) 1893 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)]; 1894 1895 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n'; 1896 for (const auto &R : HeaderInsertions) { 1897 auto IncludeDirective = R.getReplacementText(); 1898 bool Matched = IncludeRegex.match(IncludeDirective, &Matches); 1899 assert(Matched && "Header insertion replacement must have replacement text " 1900 "'#include ...'"); 1901 (void)Matched; 1902 auto IncludeName = Matches[2]; 1903 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) { 1904 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName 1905 << "\n"); 1906 continue; 1907 } 1908 int Category = 1909 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true); 1910 Offset = CategoryEndOffsets[Category]; 1911 std::string NewInclude = !IncludeDirective.endswith("\n") 1912 ? (IncludeDirective + "\n").str() 1913 : IncludeDirective.str(); 1914 // When inserting headers at end of the code, also append '\n' to the code 1915 // if it does not end with '\n'. 1916 if (NeedNewLineAtEnd && Offset == Code.size()) { 1917 NewInclude = "\n" + NewInclude; 1918 NeedNewLineAtEnd = false; 1919 } 1920 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude); 1921 auto Err = Result.add(NewReplace); 1922 if (Err) { 1923 llvm::consumeError(std::move(Err)); 1924 unsigned NewOffset = Result.getShiftedCodePosition(Offset); 1925 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude); 1926 Result = Result.merge(tooling::Replacements(NewReplace)); 1927 } 1928 } 1929 return Result; 1930 } 1931 1932 } // anonymous namespace 1933 1934 llvm::Expected<tooling::Replacements> 1935 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces, 1936 const FormatStyle &Style) { 1937 // We need to use lambda function here since there are two versions of 1938 // `cleanup`. 1939 auto Cleanup = [](const FormatStyle &Style, StringRef Code, 1940 std::vector<tooling::Range> Ranges, 1941 StringRef FileName) -> tooling::Replacements { 1942 return cleanup(Style, Code, Ranges, FileName); 1943 }; 1944 // Make header insertion replacements insert new headers into correct blocks. 1945 tooling::Replacements NewReplaces = 1946 fixCppIncludeInsertions(Code, Replaces, Style); 1947 return processReplacements(Cleanup, Code, NewReplaces, Style); 1948 } 1949 1950 namespace internal { 1951 std::pair<tooling::Replacements, unsigned> 1952 reformat(const FormatStyle &Style, StringRef Code, 1953 ArrayRef<tooling::Range> Ranges, unsigned FirstStartColumn, 1954 unsigned NextStartColumn, unsigned LastStartColumn, StringRef FileName, 1955 FormattingAttemptStatus *Status) { 1956 FormatStyle Expanded = expandPresets(Style); 1957 if (Expanded.DisableFormat) 1958 return {tooling::Replacements(), 0}; 1959 if (isLikelyXml(Code)) 1960 return {tooling::Replacements(), 0}; 1961 if (Expanded.Language == FormatStyle::LK_JavaScript && isMpegTS(Code)) 1962 return {tooling::Replacements(), 0}; 1963 1964 typedef std::function<std::pair<tooling::Replacements, unsigned>( 1965 const Environment &)> 1966 AnalyzerPass; 1967 SmallVector<AnalyzerPass, 4> Passes; 1968 1969 if (Style.Language == FormatStyle::LK_Cpp) { 1970 if (Style.FixNamespaceComments) 1971 Passes.emplace_back([&](const Environment &Env) { 1972 return NamespaceEndCommentsFixer(Env, Expanded).process(); 1973 }); 1974 1975 if (Style.SortUsingDeclarations) 1976 Passes.emplace_back([&](const Environment &Env) { 1977 return UsingDeclarationsSorter(Env, Expanded).process(); 1978 }); 1979 } 1980 1981 if (Style.Language == FormatStyle::LK_JavaScript && 1982 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) 1983 Passes.emplace_back([&](const Environment &Env) { 1984 return JavaScriptRequoter(Env, Expanded).process(); 1985 }); 1986 1987 Passes.emplace_back([&](const Environment &Env) { 1988 return Formatter(Env, Expanded, Status).process(); 1989 }); 1990 1991 std::unique_ptr<Environment> Env = Environment::CreateVirtualEnvironment( 1992 Code, FileName, Ranges, FirstStartColumn, NextStartColumn, 1993 LastStartColumn); 1994 llvm::Optional<std::string> CurrentCode = None; 1995 tooling::Replacements Fixes; 1996 unsigned Penalty = 0; 1997 for (size_t I = 0, E = Passes.size(); I < E; ++I) { 1998 std::pair<tooling::Replacements, unsigned> PassFixes = Passes[I](*Env); 1999 auto NewCode = applyAllReplacements( 2000 CurrentCode ? StringRef(*CurrentCode) : Code, PassFixes.first); 2001 if (NewCode) { 2002 Fixes = Fixes.merge(PassFixes.first); 2003 Penalty += PassFixes.second; 2004 if (I + 1 < E) { 2005 CurrentCode = std::move(*NewCode); 2006 Env = Environment::CreateVirtualEnvironment( 2007 *CurrentCode, FileName, 2008 tooling::calculateRangesAfterReplacements(Fixes, Ranges), 2009 FirstStartColumn, NextStartColumn, LastStartColumn); 2010 } 2011 } 2012 } 2013 2014 return {Fixes, Penalty}; 2015 } 2016 } // namespace internal 2017 2018 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 2019 ArrayRef<tooling::Range> Ranges, 2020 StringRef FileName, 2021 FormattingAttemptStatus *Status) { 2022 return internal::reformat(Style, Code, Ranges, 2023 /*FirstStartColumn=*/0, 2024 /*NextStartColumn=*/0, 2025 /*LastStartColumn=*/0, FileName, Status) 2026 .first; 2027 } 2028 2029 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code, 2030 ArrayRef<tooling::Range> Ranges, 2031 StringRef FileName) { 2032 // cleanups only apply to C++ (they mostly concern ctor commas etc.) 2033 if (Style.Language != FormatStyle::LK_Cpp) 2034 return tooling::Replacements(); 2035 std::unique_ptr<Environment> Env = 2036 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 2037 Cleaner Clean(*Env, Style); 2038 return Clean.process().first; 2039 } 2040 2041 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 2042 ArrayRef<tooling::Range> Ranges, 2043 StringRef FileName, bool *IncompleteFormat) { 2044 FormattingAttemptStatus Status; 2045 auto Result = reformat(Style, Code, Ranges, FileName, &Status); 2046 if (!Status.FormatComplete) 2047 *IncompleteFormat = true; 2048 return Result; 2049 } 2050 2051 tooling::Replacements fixNamespaceEndComments(const FormatStyle &Style, 2052 StringRef Code, 2053 ArrayRef<tooling::Range> Ranges, 2054 StringRef FileName) { 2055 std::unique_ptr<Environment> Env = 2056 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 2057 NamespaceEndCommentsFixer Fix(*Env, Style); 2058 return Fix.process().first; 2059 } 2060 2061 tooling::Replacements sortUsingDeclarations(const FormatStyle &Style, 2062 StringRef Code, 2063 ArrayRef<tooling::Range> Ranges, 2064 StringRef FileName) { 2065 std::unique_ptr<Environment> Env = 2066 Environment::CreateVirtualEnvironment(Code, FileName, Ranges); 2067 UsingDeclarationsSorter Sorter(*Env, Style); 2068 return Sorter.process().first; 2069 } 2070 2071 LangOptions getFormattingLangOpts(const FormatStyle &Style) { 2072 LangOptions LangOpts; 2073 LangOpts.CPlusPlus = 1; 2074 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2075 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2076 LangOpts.CPlusPlus17 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2077 LangOpts.CPlusPlus2a = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 2078 LangOpts.LineComment = 1; 2079 bool AlternativeOperators = Style.isCpp(); 2080 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0; 2081 LangOpts.Bool = 1; 2082 LangOpts.ObjC1 = 1; 2083 LangOpts.ObjC2 = 1; 2084 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally. 2085 LangOpts.DeclSpecKeyword = 1; // To get __declspec. 2086 return LangOpts; 2087 } 2088 2089 const char *StyleOptionHelpDescription = 2090 "Coding style, currently supports:\n" 2091 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 2092 "Use -style=file to load style configuration from\n" 2093 ".clang-format file located in one of the parent\n" 2094 "directories of the source file (or current\n" 2095 "directory for stdin).\n" 2096 "Use -style=\"{key: value, ...}\" to set specific\n" 2097 "parameters, e.g.:\n" 2098 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 2099 2100 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) { 2101 if (FileName.endswith(".java")) 2102 return FormatStyle::LK_Java; 2103 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) 2104 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript. 2105 if (FileName.endswith(".m") || FileName.endswith(".mm")) 2106 return FormatStyle::LK_ObjC; 2107 if (FileName.endswith_lower(".proto") || 2108 FileName.endswith_lower(".protodevel")) 2109 return FormatStyle::LK_Proto; 2110 if (FileName.endswith_lower(".textpb") || 2111 FileName.endswith_lower(".pb.txt") || 2112 FileName.endswith_lower(".textproto") || 2113 FileName.endswith_lower(".asciipb")) 2114 return FormatStyle::LK_TextProto; 2115 if (FileName.endswith_lower(".td")) 2116 return FormatStyle::LK_TableGen; 2117 return FormatStyle::LK_Cpp; 2118 } 2119 2120 llvm::Expected<FormatStyle> getStyle(StringRef StyleName, StringRef FileName, 2121 StringRef FallbackStyleName, 2122 StringRef Code, vfs::FileSystem *FS) { 2123 if (!FS) { 2124 FS = vfs::getRealFileSystem().get(); 2125 } 2126 FormatStyle Style = getLLVMStyle(); 2127 Style.Language = getLanguageByFileName(FileName); 2128 2129 // This is a very crude detection of whether a header contains ObjC code that 2130 // should be improved over time and probably be done on tokens, not one the 2131 // bare content of the file. 2132 if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") && 2133 (Code.contains("\n- (") || Code.contains("\n+ (") || 2134 Code.contains("\n@end\n") || Code.contains("\n@end ") || 2135 Code.endswith("@end"))) 2136 Style.Language = FormatStyle::LK_ObjC; 2137 2138 FormatStyle FallbackStyle = getNoStyle(); 2139 if (!getPredefinedStyle(FallbackStyleName, Style.Language, &FallbackStyle)) 2140 return make_string_error("Invalid fallback style \"" + FallbackStyleName); 2141 2142 if (StyleName.startswith("{")) { 2143 // Parse YAML/JSON style from the command line. 2144 if (std::error_code ec = parseConfiguration(StyleName, &Style)) 2145 return make_string_error("Error parsing -style: " + ec.message()); 2146 return Style; 2147 } 2148 2149 if (!StyleName.equals_lower("file")) { 2150 if (!getPredefinedStyle(StyleName, Style.Language, &Style)) 2151 return make_string_error("Invalid value for -style"); 2152 return Style; 2153 } 2154 2155 // Look for .clang-format/_clang-format file in the file's parent directories. 2156 SmallString<128> UnsuitableConfigFiles; 2157 SmallString<128> Path(FileName); 2158 if (std::error_code EC = FS->makeAbsolute(Path)) 2159 return make_string_error(EC.message()); 2160 2161 for (StringRef Directory = Path; !Directory.empty(); 2162 Directory = llvm::sys::path::parent_path(Directory)) { 2163 2164 auto Status = FS->status(Directory); 2165 if (!Status || 2166 Status->getType() != llvm::sys::fs::file_type::directory_file) { 2167 continue; 2168 } 2169 2170 SmallString<128> ConfigFile(Directory); 2171 2172 llvm::sys::path::append(ConfigFile, ".clang-format"); 2173 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 2174 2175 Status = FS->status(ConfigFile.str()); 2176 bool FoundConfigFile = 2177 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file); 2178 if (!FoundConfigFile) { 2179 // Try _clang-format too, since dotfiles are not commonly used on Windows. 2180 ConfigFile = Directory; 2181 llvm::sys::path::append(ConfigFile, "_clang-format"); 2182 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 2183 Status = FS->status(ConfigFile.str()); 2184 FoundConfigFile = Status && (Status->getType() == 2185 llvm::sys::fs::file_type::regular_file); 2186 } 2187 2188 if (FoundConfigFile) { 2189 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text = 2190 FS->getBufferForFile(ConfigFile.str()); 2191 if (std::error_code EC = Text.getError()) 2192 return make_string_error(EC.message()); 2193 if (std::error_code ec = 2194 parseConfiguration(Text.get()->getBuffer(), &Style)) { 2195 if (ec == ParseError::Unsuitable) { 2196 if (!UnsuitableConfigFiles.empty()) 2197 UnsuitableConfigFiles.append(", "); 2198 UnsuitableConfigFiles.append(ConfigFile); 2199 continue; 2200 } 2201 return make_string_error("Error reading " + ConfigFile + ": " + 2202 ec.message()); 2203 } 2204 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 2205 return Style; 2206 } 2207 } 2208 if (!UnsuitableConfigFiles.empty()) 2209 return make_string_error("Configuration file(s) do(es) not support " + 2210 getLanguageName(Style.Language) + ": " + 2211 UnsuitableConfigFiles); 2212 return FallbackStyle; 2213 } 2214 2215 } // namespace format 2216 } // namespace clang 2217