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 #define DEBUG_TYPE "format-formatter" 17 18 #include "ContinuationIndenter.h" 19 #include "TokenAnnotator.h" 20 #include "UnwrappedLineParser.h" 21 #include "WhitespaceManager.h" 22 #include "clang/Basic/Diagnostic.h" 23 #include "clang/Basic/SourceManager.h" 24 #include "clang/Format/Format.h" 25 #include "clang/Lex/Lexer.h" 26 #include "llvm/ADT/STLExtras.h" 27 #include "llvm/Support/Allocator.h" 28 #include "llvm/Support/Debug.h" 29 #include "llvm/Support/YAMLTraits.h" 30 #include "llvm/Support/Path.h" 31 #include <queue> 32 #include <string> 33 34 namespace llvm { 35 namespace yaml { 36 template <> 37 struct ScalarEnumerationTraits<clang::format::FormatStyle::LanguageKind> { 38 static void enumeration(IO &IO, 39 clang::format::FormatStyle::LanguageKind &Value) { 40 IO.enumCase(Value, "Cpp", clang::format::FormatStyle::LK_Cpp); 41 IO.enumCase(Value, "JavaScript", clang::format::FormatStyle::LK_JavaScript); 42 } 43 }; 44 45 template <> 46 struct ScalarEnumerationTraits<clang::format::FormatStyle::LanguageStandard> { 47 static void enumeration(IO &IO, 48 clang::format::FormatStyle::LanguageStandard &Value) { 49 IO.enumCase(Value, "Cpp03", clang::format::FormatStyle::LS_Cpp03); 50 IO.enumCase(Value, "C++03", clang::format::FormatStyle::LS_Cpp03); 51 IO.enumCase(Value, "Cpp11", clang::format::FormatStyle::LS_Cpp11); 52 IO.enumCase(Value, "C++11", clang::format::FormatStyle::LS_Cpp11); 53 IO.enumCase(Value, "Auto", clang::format::FormatStyle::LS_Auto); 54 } 55 }; 56 57 template <> 58 struct ScalarEnumerationTraits<clang::format::FormatStyle::UseTabStyle> { 59 static void enumeration(IO &IO, 60 clang::format::FormatStyle::UseTabStyle &Value) { 61 IO.enumCase(Value, "Never", clang::format::FormatStyle::UT_Never); 62 IO.enumCase(Value, "false", clang::format::FormatStyle::UT_Never); 63 IO.enumCase(Value, "Always", clang::format::FormatStyle::UT_Always); 64 IO.enumCase(Value, "true", clang::format::FormatStyle::UT_Always); 65 IO.enumCase(Value, "ForIndentation", 66 clang::format::FormatStyle::UT_ForIndentation); 67 } 68 }; 69 70 template <> 71 struct ScalarEnumerationTraits<clang::format::FormatStyle::BraceBreakingStyle> { 72 static void 73 enumeration(IO &IO, clang::format::FormatStyle::BraceBreakingStyle &Value) { 74 IO.enumCase(Value, "Attach", clang::format::FormatStyle::BS_Attach); 75 IO.enumCase(Value, "Linux", clang::format::FormatStyle::BS_Linux); 76 IO.enumCase(Value, "Stroustrup", clang::format::FormatStyle::BS_Stroustrup); 77 IO.enumCase(Value, "Allman", clang::format::FormatStyle::BS_Allman); 78 } 79 }; 80 81 template <> 82 struct ScalarEnumerationTraits< 83 clang::format::FormatStyle::NamespaceIndentationKind> { 84 static void 85 enumeration(IO &IO, 86 clang::format::FormatStyle::NamespaceIndentationKind &Value) { 87 IO.enumCase(Value, "None", clang::format::FormatStyle::NI_None); 88 IO.enumCase(Value, "Inner", clang::format::FormatStyle::NI_Inner); 89 IO.enumCase(Value, "All", clang::format::FormatStyle::NI_All); 90 } 91 }; 92 93 template <> struct MappingTraits<clang::format::FormatStyle> { 94 static void mapping(llvm::yaml::IO &IO, clang::format::FormatStyle &Style) { 95 if (IO.outputting()) { 96 StringRef StylesArray[] = { "LLVM", "Google", "Chromium", 97 "Mozilla", "WebKit" }; 98 ArrayRef<StringRef> Styles(StylesArray); 99 for (size_t i = 0, e = Styles.size(); i < e; ++i) { 100 StringRef StyleName(Styles[i]); 101 clang::format::FormatStyle PredefinedStyle; 102 if (clang::format::getPredefinedStyle(StyleName, &PredefinedStyle) && 103 Style == PredefinedStyle) { 104 IO.mapOptional("# BasedOnStyle", StyleName); 105 break; 106 } 107 } 108 } else { 109 StringRef BasedOnStyle; 110 IO.mapOptional("BasedOnStyle", BasedOnStyle); 111 if (!BasedOnStyle.empty()) { 112 clang::format::FormatStyle::LanguageKind Language = Style.Language; 113 if (!clang::format::getPredefinedStyle(BasedOnStyle, &Style)) { 114 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle)); 115 return; 116 } 117 Style.Language = Language; 118 } 119 } 120 121 IO.mapOptional("Language", Style.Language); 122 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset); 123 IO.mapOptional("ConstructorInitializerIndentWidth", 124 Style.ConstructorInitializerIndentWidth); 125 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft); 126 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments); 127 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine", 128 Style.AllowAllParametersOfDeclarationOnNextLine); 129 IO.mapOptional("AllowShortIfStatementsOnASingleLine", 130 Style.AllowShortIfStatementsOnASingleLine); 131 IO.mapOptional("AllowShortLoopsOnASingleLine", 132 Style.AllowShortLoopsOnASingleLine); 133 IO.mapOptional("AllowShortFunctionsOnASingleLine", 134 Style.AllowShortFunctionsOnASingleLine); 135 IO.mapOptional("AlwaysBreakTemplateDeclarations", 136 Style.AlwaysBreakTemplateDeclarations); 137 IO.mapOptional("AlwaysBreakBeforeMultilineStrings", 138 Style.AlwaysBreakBeforeMultilineStrings); 139 IO.mapOptional("BreakBeforeBinaryOperators", 140 Style.BreakBeforeBinaryOperators); 141 IO.mapOptional("BreakBeforeTernaryOperators", 142 Style.BreakBeforeTernaryOperators); 143 IO.mapOptional("BreakConstructorInitializersBeforeComma", 144 Style.BreakConstructorInitializersBeforeComma); 145 IO.mapOptional("BinPackParameters", Style.BinPackParameters); 146 IO.mapOptional("ColumnLimit", Style.ColumnLimit); 147 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine", 148 Style.ConstructorInitializerAllOnOneLineOrOnePerLine); 149 IO.mapOptional("DerivePointerBinding", Style.DerivePointerBinding); 150 IO.mapOptional("ExperimentalAutoDetectBinPacking", 151 Style.ExperimentalAutoDetectBinPacking); 152 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels); 153 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep); 154 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation); 155 IO.mapOptional("ObjCSpaceBeforeProtocolList", 156 Style.ObjCSpaceBeforeProtocolList); 157 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter", 158 Style.PenaltyBreakBeforeFirstCallParameter); 159 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment); 160 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString); 161 IO.mapOptional("PenaltyBreakFirstLessLess", 162 Style.PenaltyBreakFirstLessLess); 163 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter); 164 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine", 165 Style.PenaltyReturnTypeOnItsOwnLine); 166 IO.mapOptional("PointerBindsToType", Style.PointerBindsToType); 167 IO.mapOptional("SpacesBeforeTrailingComments", 168 Style.SpacesBeforeTrailingComments); 169 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle); 170 IO.mapOptional("Standard", Style.Standard); 171 IO.mapOptional("IndentWidth", Style.IndentWidth); 172 IO.mapOptional("TabWidth", Style.TabWidth); 173 IO.mapOptional("UseTab", Style.UseTab); 174 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces); 175 IO.mapOptional("IndentFunctionDeclarationAfterType", 176 Style.IndentFunctionDeclarationAfterType); 177 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses); 178 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles); 179 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses); 180 IO.mapOptional("SpacesInCStyleCastParentheses", 181 Style.SpacesInCStyleCastParentheses); 182 IO.mapOptional("SpaceAfterControlStatementKeyword", 183 Style.SpaceAfterControlStatementKeyword); 184 IO.mapOptional("SpaceBeforeAssignmentOperators", 185 Style.SpaceBeforeAssignmentOperators); 186 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth); 187 } 188 }; 189 190 // Allows to read vector<FormatStyle> while keeping default values. 191 // Elements will be written or read starting from the 1st element. 192 // When writing, the 0th element is ignored. 193 // When reading, keys that are not present in the serialized form will be 194 // copied from the 0th element of the vector. If the first element had no 195 // Language specified, it will be treated as the default one for the following 196 // elements. 197 template <> 198 struct DocumentListTraits<std::vector<clang::format::FormatStyle> > { 199 static size_t size(IO &io, std::vector<clang::format::FormatStyle> &Seq) { 200 return Seq.size() - 1; 201 } 202 static clang::format::FormatStyle & 203 element(IO &io, std::vector<clang::format::FormatStyle> &Seq, size_t Index) { 204 if (Index + 2 > Seq.size()) { 205 assert(Index + 2 == Seq.size() + 1); 206 clang::format::FormatStyle Template; 207 if (Seq.size() > 1 && 208 Seq[1].Language == clang::format::FormatStyle::LK_None) { 209 Template = Seq[1]; 210 } else { 211 Template = Seq[0]; 212 Template.Language = clang::format::FormatStyle::LK_None; 213 } 214 Seq.resize(Index + 2, Template); 215 } 216 return Seq[Index + 1]; 217 } 218 }; 219 } 220 } 221 222 namespace clang { 223 namespace format { 224 225 void setDefaultPenalties(FormatStyle &Style) { 226 Style.PenaltyBreakComment = 60; 227 Style.PenaltyBreakFirstLessLess = 120; 228 Style.PenaltyBreakString = 1000; 229 Style.PenaltyExcessCharacter = 1000000; 230 } 231 232 FormatStyle getLLVMStyle() { 233 FormatStyle LLVMStyle; 234 LLVMStyle.Language = FormatStyle::LK_Cpp; 235 LLVMStyle.AccessModifierOffset = -2; 236 LLVMStyle.AlignEscapedNewlinesLeft = false; 237 LLVMStyle.AlignTrailingComments = true; 238 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true; 239 LLVMStyle.AllowShortFunctionsOnASingleLine = true; 240 LLVMStyle.AllowShortIfStatementsOnASingleLine = false; 241 LLVMStyle.AllowShortLoopsOnASingleLine = false; 242 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false; 243 LLVMStyle.AlwaysBreakTemplateDeclarations = false; 244 LLVMStyle.BinPackParameters = true; 245 LLVMStyle.BreakBeforeBinaryOperators = false; 246 LLVMStyle.BreakBeforeTernaryOperators = true; 247 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 248 LLVMStyle.BreakConstructorInitializersBeforeComma = false; 249 LLVMStyle.ColumnLimit = 80; 250 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false; 251 LLVMStyle.ConstructorInitializerIndentWidth = 4; 252 LLVMStyle.Cpp11BracedListStyle = false; 253 LLVMStyle.DerivePointerBinding = false; 254 LLVMStyle.ExperimentalAutoDetectBinPacking = false; 255 LLVMStyle.IndentCaseLabels = false; 256 LLVMStyle.IndentFunctionDeclarationAfterType = false; 257 LLVMStyle.IndentWidth = 2; 258 LLVMStyle.TabWidth = 8; 259 LLVMStyle.MaxEmptyLinesToKeep = 1; 260 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None; 261 LLVMStyle.ObjCSpaceBeforeProtocolList = true; 262 LLVMStyle.PointerBindsToType = false; 263 LLVMStyle.SpacesBeforeTrailingComments = 1; 264 LLVMStyle.Standard = FormatStyle::LS_Cpp03; 265 LLVMStyle.UseTab = FormatStyle::UT_Never; 266 LLVMStyle.SpacesInParentheses = false; 267 LLVMStyle.SpaceInEmptyParentheses = false; 268 LLVMStyle.SpacesInCStyleCastParentheses = false; 269 LLVMStyle.SpaceAfterControlStatementKeyword = true; 270 LLVMStyle.SpaceBeforeAssignmentOperators = true; 271 LLVMStyle.ContinuationIndentWidth = 4; 272 LLVMStyle.SpacesInAngles = false; 273 274 setDefaultPenalties(LLVMStyle); 275 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60; 276 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19; 277 278 return LLVMStyle; 279 } 280 281 FormatStyle getGoogleStyle() { 282 FormatStyle GoogleStyle; 283 GoogleStyle.Language = FormatStyle::LK_Cpp; 284 GoogleStyle.AccessModifierOffset = -1; 285 GoogleStyle.AlignEscapedNewlinesLeft = true; 286 GoogleStyle.AlignTrailingComments = true; 287 GoogleStyle.AllowAllParametersOfDeclarationOnNextLine = true; 288 GoogleStyle.AllowShortFunctionsOnASingleLine = true; 289 GoogleStyle.AllowShortIfStatementsOnASingleLine = true; 290 GoogleStyle.AllowShortLoopsOnASingleLine = true; 291 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true; 292 GoogleStyle.AlwaysBreakTemplateDeclarations = true; 293 GoogleStyle.BinPackParameters = true; 294 GoogleStyle.BreakBeforeBinaryOperators = false; 295 GoogleStyle.BreakBeforeTernaryOperators = true; 296 GoogleStyle.BreakBeforeBraces = FormatStyle::BS_Attach; 297 GoogleStyle.BreakConstructorInitializersBeforeComma = false; 298 GoogleStyle.ColumnLimit = 80; 299 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 300 GoogleStyle.ConstructorInitializerIndentWidth = 4; 301 GoogleStyle.Cpp11BracedListStyle = true; 302 GoogleStyle.DerivePointerBinding = true; 303 GoogleStyle.ExperimentalAutoDetectBinPacking = false; 304 GoogleStyle.IndentCaseLabels = true; 305 GoogleStyle.IndentFunctionDeclarationAfterType = true; 306 GoogleStyle.IndentWidth = 2; 307 GoogleStyle.TabWidth = 8; 308 GoogleStyle.MaxEmptyLinesToKeep = 1; 309 GoogleStyle.NamespaceIndentation = FormatStyle::NI_None; 310 GoogleStyle.ObjCSpaceBeforeProtocolList = false; 311 GoogleStyle.PointerBindsToType = true; 312 GoogleStyle.SpacesBeforeTrailingComments = 2; 313 GoogleStyle.Standard = FormatStyle::LS_Auto; 314 GoogleStyle.UseTab = FormatStyle::UT_Never; 315 GoogleStyle.SpacesInParentheses = false; 316 GoogleStyle.SpaceInEmptyParentheses = false; 317 GoogleStyle.SpacesInCStyleCastParentheses = false; 318 GoogleStyle.SpaceAfterControlStatementKeyword = true; 319 GoogleStyle.SpaceBeforeAssignmentOperators = true; 320 GoogleStyle.ContinuationIndentWidth = 4; 321 GoogleStyle.SpacesInAngles = false; 322 323 setDefaultPenalties(GoogleStyle); 324 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200; 325 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1; 326 327 return GoogleStyle; 328 } 329 330 FormatStyle getChromiumStyle() { 331 FormatStyle ChromiumStyle = getGoogleStyle(); 332 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false; 333 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false; 334 ChromiumStyle.AllowShortLoopsOnASingleLine = false; 335 ChromiumStyle.BinPackParameters = false; 336 ChromiumStyle.DerivePointerBinding = false; 337 ChromiumStyle.Standard = FormatStyle::LS_Cpp03; 338 return ChromiumStyle; 339 } 340 341 FormatStyle getMozillaStyle() { 342 FormatStyle MozillaStyle = getLLVMStyle(); 343 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false; 344 MozillaStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true; 345 MozillaStyle.DerivePointerBinding = true; 346 MozillaStyle.IndentCaseLabels = true; 347 MozillaStyle.ObjCSpaceBeforeProtocolList = false; 348 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200; 349 MozillaStyle.PointerBindsToType = true; 350 return MozillaStyle; 351 } 352 353 FormatStyle getWebKitStyle() { 354 FormatStyle Style = getLLVMStyle(); 355 Style.AccessModifierOffset = -4; 356 Style.AlignTrailingComments = false; 357 Style.BreakBeforeBinaryOperators = true; 358 Style.BreakBeforeBraces = FormatStyle::BS_Stroustrup; 359 Style.BreakConstructorInitializersBeforeComma = true; 360 Style.ColumnLimit = 0; 361 Style.IndentWidth = 4; 362 Style.NamespaceIndentation = FormatStyle::NI_Inner; 363 Style.PointerBindsToType = true; 364 return Style; 365 } 366 367 bool getPredefinedStyle(StringRef Name, FormatStyle *Style) { 368 if (Name.equals_lower("llvm")) 369 *Style = getLLVMStyle(); 370 else if (Name.equals_lower("chromium")) 371 *Style = getChromiumStyle(); 372 else if (Name.equals_lower("mozilla")) 373 *Style = getMozillaStyle(); 374 else if (Name.equals_lower("google")) 375 *Style = getGoogleStyle(); 376 else if (Name.equals_lower("webkit")) 377 *Style = getWebKitStyle(); 378 else 379 return false; 380 381 return true; 382 } 383 384 llvm::error_code parseConfiguration(StringRef Text, FormatStyle *Style) { 385 assert(Style); 386 assert(Style->Language != FormatStyle::LK_None); 387 if (Text.trim().empty()) 388 return llvm::make_error_code(llvm::errc::invalid_argument); 389 390 std::vector<FormatStyle> Styles; 391 // DocumentListTraits<vector<FormatStyle>> uses 0th element as the default one 392 // for the fields, keys for which are missing from the configuration. 393 Styles.push_back(*Style); 394 llvm::yaml::Input Input(Text); 395 Input >> Styles; 396 if (Input.error()) 397 return Input.error(); 398 399 for (unsigned i = 1; i < Styles.size(); ++i) { 400 // Ensures that only the first configuration can skip the Language option. 401 if (Styles[i].Language == FormatStyle::LK_None && i != 1) 402 return llvm::make_error_code(llvm::errc::invalid_argument); 403 // Ensure that each language is configured at most once. 404 for (unsigned j = 1; j < i; ++j) { 405 if (Styles[i].Language == Styles[j].Language) 406 return llvm::make_error_code(llvm::errc::invalid_argument); 407 } 408 } 409 // Look for a suitable configuration starting from the end, so we can 410 // find the configuration for the specific language first, and the default 411 // configuration (which can only be at slot 1) after it. 412 for (unsigned i = Styles.size() - 1; i > 0; --i) { 413 if (Styles[i].Language == Styles[0].Language || 414 Styles[i].Language == FormatStyle::LK_None) { 415 *Style = Styles[i]; 416 Style->Language = Styles[0].Language; 417 return llvm::make_error_code(llvm::errc::success); 418 } 419 } 420 return llvm::make_error_code(llvm::errc::not_supported); 421 } 422 423 std::string configurationAsText(const FormatStyle &Style) { 424 std::string Text; 425 llvm::raw_string_ostream Stream(Text); 426 llvm::yaml::Output Output(Stream); 427 // We use the same mapping method for input and output, so we need a non-const 428 // reference here. 429 FormatStyle NonConstStyle = Style; 430 Output << NonConstStyle; 431 return Stream.str(); 432 } 433 434 namespace { 435 436 class NoColumnLimitFormatter { 437 public: 438 NoColumnLimitFormatter(ContinuationIndenter *Indenter) : Indenter(Indenter) {} 439 440 /// \brief Formats the line starting at \p State, simply keeping all of the 441 /// input's line breaking decisions. 442 void format(unsigned FirstIndent, const AnnotatedLine *Line, 443 bool LineIsMerged) { 444 LineState State = 445 Indenter->getInitialState(FirstIndent, Line, /*DryRun=*/false); 446 while (State.NextToken != NULL) { 447 bool Newline = 448 (!LineIsMerged && Indenter->mustBreak(State)) || 449 (Indenter->canBreak(State) && State.NextToken->NewlinesBefore > 0); 450 Indenter->addTokenToState(State, Newline, /*DryRun=*/false); 451 } 452 } 453 454 private: 455 ContinuationIndenter *Indenter; 456 }; 457 458 class LineJoiner { 459 public: 460 LineJoiner(const FormatStyle &Style) : Style(Style) {} 461 462 /// \brief Calculates how many lines can be merged into 1 starting at \p I. 463 unsigned 464 tryFitMultipleLinesInOne(unsigned Indent, 465 SmallVectorImpl<AnnotatedLine *>::const_iterator I, 466 SmallVectorImpl<AnnotatedLine *>::const_iterator E) { 467 // We can never merge stuff if there are trailing line comments. 468 AnnotatedLine *TheLine = *I; 469 if (TheLine->Last->Type == TT_LineComment) 470 return 0; 471 472 if (Style.ColumnLimit > 0 && Indent > Style.ColumnLimit) 473 return 0; 474 475 unsigned Limit = 476 Style.ColumnLimit == 0 ? UINT_MAX : Style.ColumnLimit - Indent; 477 // If we already exceed the column limit, we set 'Limit' to 0. The different 478 // tryMerge..() functions can then decide whether to still do merging. 479 Limit = TheLine->Last->TotalLength > Limit 480 ? 0 481 : Limit - TheLine->Last->TotalLength; 482 483 if (I + 1 == E || I[1]->Type == LT_Invalid) 484 return 0; 485 486 if (TheLine->Last->Type == TT_FunctionLBrace) { 487 return Style.AllowShortFunctionsOnASingleLine 488 ? tryMergeSimpleBlock(I, E, Limit) 489 : 0; 490 } 491 if (TheLine->Last->is(tok::l_brace)) { 492 return Style.BreakBeforeBraces == clang::format::FormatStyle::BS_Attach 493 ? tryMergeSimpleBlock(I, E, Limit) 494 : 0; 495 } 496 if (I[1]->First->Type == TT_FunctionLBrace && 497 Style.BreakBeforeBraces != FormatStyle::BS_Attach) { 498 // Reduce the column limit by the number of spaces we need to insert 499 // around braces. 500 Limit = Limit > 3 ? Limit - 3 : 0; 501 unsigned MergedLines = 0; 502 if (Style.AllowShortFunctionsOnASingleLine) { 503 MergedLines = tryMergeSimpleBlock(I + 1, E, Limit); 504 // If we managed to merge the block, count the function header, which is 505 // on a separate line. 506 if (MergedLines > 0) 507 ++MergedLines; 508 } 509 return MergedLines; 510 } 511 if (TheLine->First->is(tok::kw_if)) { 512 return Style.AllowShortIfStatementsOnASingleLine 513 ? tryMergeSimpleControlStatement(I, E, Limit) 514 : 0; 515 } 516 if (TheLine->First->isOneOf(tok::kw_for, tok::kw_while)) { 517 return Style.AllowShortLoopsOnASingleLine 518 ? tryMergeSimpleControlStatement(I, E, Limit) 519 : 0; 520 } 521 if (TheLine->InPPDirective && 522 (TheLine->First->HasUnescapedNewline || TheLine->First->IsFirst)) { 523 return tryMergeSimplePPDirective(I, E, Limit); 524 } 525 return 0; 526 } 527 528 private: 529 unsigned 530 tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 531 SmallVectorImpl<AnnotatedLine *>::const_iterator E, 532 unsigned Limit) { 533 if (Limit == 0) 534 return 0; 535 if (!I[1]->InPPDirective || I[1]->First->HasUnescapedNewline) 536 return 0; 537 if (I + 2 != E && I[2]->InPPDirective && !I[2]->First->HasUnescapedNewline) 538 return 0; 539 if (1 + I[1]->Last->TotalLength > Limit) 540 return 0; 541 return 1; 542 } 543 544 unsigned tryMergeSimpleControlStatement( 545 SmallVectorImpl<AnnotatedLine *>::const_iterator I, 546 SmallVectorImpl<AnnotatedLine *>::const_iterator E, unsigned Limit) { 547 if (Limit == 0) 548 return 0; 549 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman && 550 I[1]->First->is(tok::l_brace)) 551 return 0; 552 if (I[1]->InPPDirective != (*I)->InPPDirective || 553 (I[1]->InPPDirective && I[1]->First->HasUnescapedNewline)) 554 return 0; 555 AnnotatedLine &Line = **I; 556 if (Line.Last->isNot(tok::r_paren)) 557 return 0; 558 if (1 + I[1]->Last->TotalLength > Limit) 559 return 0; 560 if (I[1]->First->isOneOf(tok::semi, tok::kw_if, tok::kw_for, 561 tok::kw_while) || 562 I[1]->First->Type == TT_LineComment) 563 return 0; 564 // Only inline simple if's (no nested if or else). 565 if (I + 2 != E && Line.First->is(tok::kw_if) && 566 I[2]->First->is(tok::kw_else)) 567 return 0; 568 return 1; 569 } 570 571 unsigned 572 tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 573 SmallVectorImpl<AnnotatedLine *>::const_iterator E, 574 unsigned Limit) { 575 // First, check that the current line allows merging. This is the case if 576 // we're not in a control flow statement and the last token is an opening 577 // brace. 578 AnnotatedLine &Line = **I; 579 if (Line.First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_do, tok::r_brace, 580 tok::kw_else, tok::kw_try, tok::kw_catch, 581 tok::kw_for, 582 // This gets rid of all ObjC @ keywords and methods. 583 tok::at, tok::minus, tok::plus)) 584 return 0; 585 586 FormatToken *Tok = I[1]->First; 587 if (Tok->is(tok::r_brace) && !Tok->MustBreakBefore && 588 (Tok->getNextNonComment() == NULL || 589 Tok->getNextNonComment()->is(tok::semi))) { 590 // We merge empty blocks even if the line exceeds the column limit. 591 Tok->SpacesRequiredBefore = 0; 592 Tok->CanBreakBefore = true; 593 return 1; 594 } else if (Limit != 0 && Line.First->isNot(tok::kw_namespace)) { 595 // Check that we still have three lines and they fit into the limit. 596 if (I + 2 == E || I[2]->Type == LT_Invalid) 597 return 0; 598 599 if (!nextTwoLinesFitInto(I, Limit)) 600 return 0; 601 602 // Second, check that the next line does not contain any braces - if it 603 // does, readability declines when putting it into a single line. 604 if (I[1]->Last->Type == TT_LineComment || Tok->MustBreakBefore) 605 return 0; 606 do { 607 if (Tok->isOneOf(tok::l_brace, tok::r_brace)) 608 return 0; 609 Tok = Tok->Next; 610 } while (Tok != NULL); 611 612 // Last, check that the third line contains a single closing brace. 613 Tok = I[2]->First; 614 if (Tok->getNextNonComment() != NULL || Tok->isNot(tok::r_brace) || 615 Tok->MustBreakBefore) 616 return 0; 617 618 return 2; 619 } 620 return 0; 621 } 622 623 bool nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine *>::const_iterator I, 624 unsigned Limit) { 625 return 1 + I[1]->Last->TotalLength + 1 + I[2]->Last->TotalLength <= Limit; 626 } 627 628 const FormatStyle &Style; 629 }; 630 631 class UnwrappedLineFormatter { 632 public: 633 UnwrappedLineFormatter(ContinuationIndenter *Indenter, 634 WhitespaceManager *Whitespaces, 635 const FormatStyle &Style) 636 : Indenter(Indenter), Whitespaces(Whitespaces), Style(Style), 637 Joiner(Style) {} 638 639 unsigned format(const SmallVectorImpl<AnnotatedLine *> &Lines, bool DryRun, 640 int AdditionalIndent = 0, bool FixBadIndentation = false) { 641 assert(!Lines.empty()); 642 unsigned Penalty = 0; 643 std::vector<int> IndentForLevel; 644 for (unsigned i = 0, e = Lines[0]->Level; i != e; ++i) 645 IndentForLevel.push_back(Style.IndentWidth * i + AdditionalIndent); 646 const AnnotatedLine *PreviousLine = NULL; 647 for (SmallVectorImpl<AnnotatedLine *>::const_iterator I = Lines.begin(), 648 E = Lines.end(); 649 I != E; ++I) { 650 const AnnotatedLine &TheLine = **I; 651 const FormatToken *FirstTok = TheLine.First; 652 int Offset = getIndentOffset(*FirstTok); 653 654 // Determine indent and try to merge multiple unwrapped lines. 655 while (IndentForLevel.size() <= TheLine.Level) 656 IndentForLevel.push_back(-1); 657 IndentForLevel.resize(TheLine.Level + 1); 658 unsigned Indent = getIndent(IndentForLevel, TheLine.Level); 659 if (static_cast<int>(Indent) + Offset >= 0) 660 Indent += Offset; 661 unsigned MergedLines = Joiner.tryFitMultipleLinesInOne(Indent, I, E); 662 if (MergedLines > 0 && Style.ColumnLimit == 0) { 663 // Disallow line merging if there is a break at the start of one of the 664 // input lines. 665 for (unsigned i = 0; i < MergedLines; ++i) { 666 if (I[i + 1]->First->NewlinesBefore > 0) 667 MergedLines = 0; 668 } 669 } 670 if (!DryRun) { 671 for (unsigned i = 0; i < MergedLines; ++i) { 672 join(*I[i], *I[i + 1]); 673 } 674 } 675 I += MergedLines; 676 677 unsigned LevelIndent = getIndent(IndentForLevel, TheLine.Level); 678 bool FixIndentation = 679 FixBadIndentation && (LevelIndent != FirstTok->OriginalColumn); 680 if (TheLine.First->is(tok::eof)) { 681 if (PreviousLine && PreviousLine->Affected && !DryRun) { 682 // Remove the file's trailing whitespace. 683 unsigned Newlines = std::min(FirstTok->NewlinesBefore, 1u); 684 Whitespaces->replaceWhitespace(*TheLine.First, Newlines, 685 /*IndentLevel=*/0, /*Spaces=*/0, 686 /*TargetColumn=*/0); 687 } 688 } else if (TheLine.Type != LT_Invalid && 689 (TheLine.Affected || FixIndentation)) { 690 if (FirstTok->WhitespaceRange.isValid()) { 691 if (!DryRun) 692 formatFirstToken(*TheLine.First, PreviousLine, TheLine.Level, 693 Indent, TheLine.InPPDirective); 694 } else { 695 Indent = LevelIndent = FirstTok->OriginalColumn; 696 } 697 698 // If everything fits on a single line, just put it there. 699 unsigned ColumnLimit = Style.ColumnLimit; 700 if (I + 1 != E) { 701 AnnotatedLine *NextLine = I[1]; 702 if (NextLine->InPPDirective && !NextLine->First->HasUnescapedNewline) 703 ColumnLimit = getColumnLimit(TheLine.InPPDirective); 704 } 705 706 if (TheLine.Last->TotalLength + Indent <= ColumnLimit) { 707 LineState State = Indenter->getInitialState(Indent, &TheLine, DryRun); 708 while (State.NextToken != NULL) 709 Indenter->addTokenToState(State, /*Newline=*/false, DryRun); 710 } else if (Style.ColumnLimit == 0) { 711 // FIXME: Implement nested blocks for ColumnLimit = 0. 712 NoColumnLimitFormatter Formatter(Indenter); 713 if (!DryRun) 714 Formatter.format(Indent, &TheLine, 715 /*LineIsMerged=*/MergedLines > 0); 716 } else { 717 Penalty += format(TheLine, Indent, DryRun); 718 } 719 720 IndentForLevel[TheLine.Level] = LevelIndent; 721 } else if (TheLine.ChildrenAffected) { 722 format(TheLine.Children, DryRun); 723 } else { 724 // Format the first token if necessary, and notify the WhitespaceManager 725 // about the unchanged whitespace. 726 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 727 if (Tok == TheLine.First && 728 (Tok->NewlinesBefore > 0 || Tok->IsFirst)) { 729 unsigned LevelIndent = Tok->OriginalColumn; 730 if (!DryRun) { 731 // Remove trailing whitespace of the previous line. 732 if ((PreviousLine && PreviousLine->Affected) || 733 TheLine.LeadingEmptyLinesAffected) { 734 formatFirstToken(*Tok, PreviousLine, TheLine.Level, LevelIndent, 735 TheLine.InPPDirective); 736 } else { 737 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective); 738 } 739 } 740 741 if (static_cast<int>(LevelIndent) - Offset >= 0) 742 LevelIndent -= Offset; 743 if (Tok->isNot(tok::comment)) 744 IndentForLevel[TheLine.Level] = LevelIndent; 745 } else if (!DryRun) { 746 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective); 747 } 748 } 749 } 750 if (!DryRun) { 751 for (FormatToken *Tok = TheLine.First; Tok != NULL; Tok = Tok->Next) { 752 Tok->Finalized = true; 753 } 754 } 755 PreviousLine = *I; 756 } 757 return Penalty; 758 } 759 760 private: 761 /// \brief Formats an \c AnnotatedLine and returns the penalty. 762 /// 763 /// If \p DryRun is \c false, directly applies the changes. 764 unsigned format(const AnnotatedLine &Line, unsigned FirstIndent, 765 bool DryRun) { 766 LineState State = Indenter->getInitialState(FirstIndent, &Line, DryRun); 767 768 // If the ObjC method declaration does not fit on a line, we should format 769 // it with one arg per line. 770 if (State.Line->Type == LT_ObjCMethodDecl) 771 State.Stack.back().BreakBeforeParameter = true; 772 773 // Find best solution in solution space. 774 return analyzeSolutionSpace(State, DryRun); 775 } 776 777 /// \brief An edge in the solution space from \c Previous->State to \c State, 778 /// inserting a newline dependent on the \c NewLine. 779 struct StateNode { 780 StateNode(const LineState &State, bool NewLine, StateNode *Previous) 781 : State(State), NewLine(NewLine), Previous(Previous) {} 782 LineState State; 783 bool NewLine; 784 StateNode *Previous; 785 }; 786 787 /// \brief A pair of <penalty, count> that is used to prioritize the BFS on. 788 /// 789 /// In case of equal penalties, we want to prefer states that were inserted 790 /// first. During state generation we make sure that we insert states first 791 /// that break the line as late as possible. 792 typedef std::pair<unsigned, unsigned> OrderedPenalty; 793 794 /// \brief An item in the prioritized BFS search queue. The \c StateNode's 795 /// \c State has the given \c OrderedPenalty. 796 typedef std::pair<OrderedPenalty, StateNode *> QueueItem; 797 798 /// \brief The BFS queue type. 799 typedef std::priority_queue<QueueItem, std::vector<QueueItem>, 800 std::greater<QueueItem> > QueueType; 801 802 /// \brief Get the offset of the line relatively to the level. 803 /// 804 /// For example, 'public:' labels in classes are offset by 1 or 2 805 /// characters to the left from their level. 806 int getIndentOffset(const FormatToken &RootToken) { 807 if (RootToken.isAccessSpecifier(false) || RootToken.isObjCAccessSpecifier()) 808 return Style.AccessModifierOffset; 809 return 0; 810 } 811 812 /// \brief Add a new line and the required indent before the first Token 813 /// of the \c UnwrappedLine if there was no structural parsing error. 814 void formatFirstToken(FormatToken &RootToken, 815 const AnnotatedLine *PreviousLine, unsigned IndentLevel, 816 unsigned Indent, bool InPPDirective) { 817 unsigned Newlines = 818 std::min(RootToken.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1); 819 // Remove empty lines before "}" where applicable. 820 if (RootToken.is(tok::r_brace) && 821 (!RootToken.Next || 822 (RootToken.Next->is(tok::semi) && !RootToken.Next->Next))) 823 Newlines = std::min(Newlines, 1u); 824 if (Newlines == 0 && !RootToken.IsFirst) 825 Newlines = 1; 826 827 // Insert extra new line before access specifiers. 828 if (PreviousLine && PreviousLine->Last->isOneOf(tok::semi, tok::r_brace) && 829 RootToken.isAccessSpecifier() && RootToken.NewlinesBefore == 1) 830 ++Newlines; 831 832 // Remove empty lines after access specifiers. 833 if (PreviousLine && PreviousLine->First->isAccessSpecifier()) 834 Newlines = std::min(1u, Newlines); 835 836 Whitespaces->replaceWhitespace(RootToken, Newlines, IndentLevel, Indent, 837 Indent, InPPDirective && 838 !RootToken.HasUnescapedNewline); 839 } 840 841 /// \brief Get the indent of \p Level from \p IndentForLevel. 842 /// 843 /// \p IndentForLevel must contain the indent for the level \c l 844 /// at \p IndentForLevel[l], or a value < 0 if the indent for 845 /// that level is unknown. 846 unsigned getIndent(const std::vector<int> IndentForLevel, unsigned Level) { 847 if (IndentForLevel[Level] != -1) 848 return IndentForLevel[Level]; 849 if (Level == 0) 850 return 0; 851 return getIndent(IndentForLevel, Level - 1) + Style.IndentWidth; 852 } 853 854 void join(AnnotatedLine &A, const AnnotatedLine &B) { 855 assert(!A.Last->Next); 856 assert(!B.First->Previous); 857 if (B.Affected) 858 A.Affected = true; 859 A.Last->Next = B.First; 860 B.First->Previous = A.Last; 861 B.First->CanBreakBefore = true; 862 unsigned LengthA = A.Last->TotalLength + B.First->SpacesRequiredBefore; 863 for (FormatToken *Tok = B.First; Tok; Tok = Tok->Next) { 864 Tok->TotalLength += LengthA; 865 A.Last = Tok; 866 } 867 } 868 869 unsigned getColumnLimit(bool InPPDirective) const { 870 // In preprocessor directives reserve two chars for trailing " \" 871 return Style.ColumnLimit - (InPPDirective ? 2 : 0); 872 } 873 874 /// \brief Analyze the entire solution space starting from \p InitialState. 875 /// 876 /// This implements a variant of Dijkstra's algorithm on the graph that spans 877 /// the solution space (\c LineStates are the nodes). The algorithm tries to 878 /// find the shortest path (the one with lowest penalty) from \p InitialState 879 /// to a state where all tokens are placed. Returns the penalty. 880 /// 881 /// If \p DryRun is \c false, directly applies the changes. 882 unsigned analyzeSolutionSpace(LineState &InitialState, bool DryRun = false) { 883 std::set<LineState> Seen; 884 885 // Increasing count of \c StateNode items we have created. This is used to 886 // create a deterministic order independent of the container. 887 unsigned Count = 0; 888 QueueType Queue; 889 890 // Insert start element into queue. 891 StateNode *Node = 892 new (Allocator.Allocate()) StateNode(InitialState, false, NULL); 893 Queue.push(QueueItem(OrderedPenalty(0, Count), Node)); 894 ++Count; 895 896 unsigned Penalty = 0; 897 898 // While not empty, take first element and follow edges. 899 while (!Queue.empty()) { 900 Penalty = Queue.top().first.first; 901 StateNode *Node = Queue.top().second; 902 if (Node->State.NextToken == NULL) { 903 DEBUG(llvm::dbgs() << "\n---\nPenalty for line: " << Penalty << "\n"); 904 break; 905 } 906 Queue.pop(); 907 908 // Cut off the analysis of certain solutions if the analysis gets too 909 // complex. See description of IgnoreStackForComparison. 910 if (Count > 10000) 911 Node->State.IgnoreStackForComparison = true; 912 913 if (!Seen.insert(Node->State).second) 914 // State already examined with lower penalty. 915 continue; 916 917 FormatDecision LastFormat = Node->State.NextToken->Decision; 918 if (LastFormat == FD_Unformatted || LastFormat == FD_Continue) 919 addNextStateToQueue(Penalty, Node, /*NewLine=*/false, &Count, &Queue); 920 if (LastFormat == FD_Unformatted || LastFormat == FD_Break) 921 addNextStateToQueue(Penalty, Node, /*NewLine=*/true, &Count, &Queue); 922 } 923 924 if (Queue.empty()) { 925 // We were unable to find a solution, do nothing. 926 // FIXME: Add diagnostic? 927 DEBUG(llvm::dbgs() << "Could not find a solution.\n"); 928 return 0; 929 } 930 931 // Reconstruct the solution. 932 if (!DryRun) 933 reconstructPath(InitialState, Queue.top().second); 934 935 DEBUG(llvm::dbgs() << "Total number of analyzed states: " << Count << "\n"); 936 DEBUG(llvm::dbgs() << "---\n"); 937 938 return Penalty; 939 } 940 941 void reconstructPath(LineState &State, StateNode *Current) { 942 std::deque<StateNode *> Path; 943 // We do not need a break before the initial token. 944 while (Current->Previous) { 945 Path.push_front(Current); 946 Current = Current->Previous; 947 } 948 for (std::deque<StateNode *>::iterator I = Path.begin(), E = Path.end(); 949 I != E; ++I) { 950 unsigned Penalty = 0; 951 formatChildren(State, (*I)->NewLine, /*DryRun=*/false, Penalty); 952 Penalty += Indenter->addTokenToState(State, (*I)->NewLine, false); 953 954 DEBUG({ 955 if ((*I)->NewLine) { 956 llvm::dbgs() << "Penalty for placing " 957 << (*I)->Previous->State.NextToken->Tok.getName() << ": " 958 << Penalty << "\n"; 959 } 960 }); 961 } 962 } 963 964 /// \brief Add the following state to the analysis queue \c Queue. 965 /// 966 /// Assume the current state is \p PreviousNode and has been reached with a 967 /// penalty of \p Penalty. Insert a line break if \p NewLine is \c true. 968 void addNextStateToQueue(unsigned Penalty, StateNode *PreviousNode, 969 bool NewLine, unsigned *Count, QueueType *Queue) { 970 if (NewLine && !Indenter->canBreak(PreviousNode->State)) 971 return; 972 if (!NewLine && Indenter->mustBreak(PreviousNode->State)) 973 return; 974 975 StateNode *Node = new (Allocator.Allocate()) 976 StateNode(PreviousNode->State, NewLine, PreviousNode); 977 if (!formatChildren(Node->State, NewLine, /*DryRun=*/true, Penalty)) 978 return; 979 980 Penalty += Indenter->addTokenToState(Node->State, NewLine, true); 981 982 Queue->push(QueueItem(OrderedPenalty(Penalty, *Count), Node)); 983 ++(*Count); 984 } 985 986 /// \brief If the \p State's next token is an r_brace closing a nested block, 987 /// format the nested block before it. 988 /// 989 /// Returns \c true if all children could be placed successfully and adapts 990 /// \p Penalty as well as \p State. If \p DryRun is false, also directly 991 /// creates changes using \c Whitespaces. 992 /// 993 /// The crucial idea here is that children always get formatted upon 994 /// encountering the closing brace right after the nested block. Now, if we 995 /// are currently trying to keep the "}" on the same line (i.e. \p NewLine is 996 /// \c false), the entire block has to be kept on the same line (which is only 997 /// possible if it fits on the line, only contains a single statement, etc. 998 /// 999 /// If \p NewLine is true, we format the nested block on separate lines, i.e. 1000 /// break after the "{", format all lines with correct indentation and the put 1001 /// the closing "}" on yet another new line. 1002 /// 1003 /// This enables us to keep the simple structure of the 1004 /// \c UnwrappedLineFormatter, where we only have two options for each token: 1005 /// break or don't break. 1006 bool formatChildren(LineState &State, bool NewLine, bool DryRun, 1007 unsigned &Penalty) { 1008 FormatToken &Previous = *State.NextToken->Previous; 1009 const FormatToken *LBrace = State.NextToken->getPreviousNonComment(); 1010 if (!LBrace || LBrace->isNot(tok::l_brace) || 1011 LBrace->BlockKind != BK_Block || Previous.Children.size() == 0) 1012 // The previous token does not open a block. Nothing to do. We don't 1013 // assert so that we can simply call this function for all tokens. 1014 return true; 1015 1016 if (NewLine) { 1017 int AdditionalIndent = State.Stack.back().Indent - 1018 Previous.Children[0]->Level * Style.IndentWidth; 1019 Penalty += format(Previous.Children, DryRun, AdditionalIndent, 1020 /*FixBadIndentation=*/true); 1021 return true; 1022 } 1023 1024 // Cannot merge multiple statements into a single line. 1025 if (Previous.Children.size() > 1) 1026 return false; 1027 1028 // We can't put the closing "}" on a line with a trailing comment. 1029 if (Previous.Children[0]->Last->isTrailingComment()) 1030 return false; 1031 1032 if (!DryRun) { 1033 Whitespaces->replaceWhitespace( 1034 *Previous.Children[0]->First, 1035 /*Newlines=*/0, /*IndentLevel=*/0, /*Spaces=*/1, 1036 /*StartOfTokenColumn=*/State.Column, State.Line->InPPDirective); 1037 } 1038 Penalty += format(*Previous.Children[0], State.Column + 1, DryRun); 1039 1040 State.Column += 1 + Previous.Children[0]->Last->TotalLength; 1041 return true; 1042 } 1043 1044 ContinuationIndenter *Indenter; 1045 WhitespaceManager *Whitespaces; 1046 FormatStyle Style; 1047 LineJoiner Joiner; 1048 1049 llvm::SpecificBumpPtrAllocator<StateNode> Allocator; 1050 }; 1051 1052 class FormatTokenLexer { 1053 public: 1054 FormatTokenLexer(Lexer &Lex, SourceManager &SourceMgr, FormatStyle &Style, 1055 encoding::Encoding Encoding) 1056 : FormatTok(NULL), IsFirstToken(true), GreaterStashed(false), Column(0), 1057 TrailingWhitespace(0), Lex(Lex), SourceMgr(SourceMgr), Style(Style), 1058 IdentTable(getFormattingLangOpts()), Encoding(Encoding) { 1059 Lex.SetKeepWhitespaceMode(true); 1060 } 1061 1062 ArrayRef<FormatToken *> lex() { 1063 assert(Tokens.empty()); 1064 do { 1065 Tokens.push_back(getNextToken()); 1066 tryMergePreviousTokens(); 1067 } while (Tokens.back()->Tok.isNot(tok::eof)); 1068 return Tokens; 1069 } 1070 1071 IdentifierTable &getIdentTable() { return IdentTable; } 1072 1073 private: 1074 void tryMergePreviousTokens() { 1075 if (tryMerge_TMacro()) 1076 return; 1077 1078 if (Style.Language == FormatStyle::LK_JavaScript) { 1079 static tok::TokenKind JSIdentity[] = { tok::equalequal, tok::equal }; 1080 static tok::TokenKind JSNotIdentity[] = { tok::exclaimequal, tok::equal }; 1081 static tok::TokenKind JSShiftEqual[] = { tok::greater, tok::greater, 1082 tok::greaterequal }; 1083 // FIXME: We probably need to change token type to mimic operator with the 1084 // correct priority. 1085 if (tryMergeTokens(JSIdentity)) 1086 return; 1087 if (tryMergeTokens(JSNotIdentity)) 1088 return; 1089 if (tryMergeTokens(JSShiftEqual)) 1090 return; 1091 } 1092 } 1093 1094 bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds) { 1095 if (Tokens.size() < Kinds.size()) 1096 return false; 1097 1098 SmallVectorImpl<FormatToken *>::const_iterator First = 1099 Tokens.end() - Kinds.size(); 1100 if (!First[0]->is(Kinds[0])) 1101 return false; 1102 unsigned AddLength = 0; 1103 for (unsigned i = 1; i < Kinds.size(); ++i) { 1104 if (!First[i]->is(Kinds[i]) || First[i]->WhitespaceRange.getBegin() != 1105 First[i]->WhitespaceRange.getEnd()) 1106 return false; 1107 AddLength += First[i]->TokenText.size(); 1108 } 1109 Tokens.resize(Tokens.size() - Kinds.size() + 1); 1110 First[0]->TokenText = StringRef(First[0]->TokenText.data(), 1111 First[0]->TokenText.size() + AddLength); 1112 First[0]->ColumnWidth += AddLength; 1113 return true; 1114 } 1115 1116 bool tryMerge_TMacro() { 1117 if (Tokens.size() < 4) 1118 return false; 1119 FormatToken *Last = Tokens.back(); 1120 if (!Last->is(tok::r_paren)) 1121 return false; 1122 1123 FormatToken *String = Tokens[Tokens.size() - 2]; 1124 if (!String->is(tok::string_literal) || String->IsMultiline) 1125 return false; 1126 1127 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) 1128 return false; 1129 1130 FormatToken *Macro = Tokens[Tokens.size() - 4]; 1131 if (Macro->TokenText != "_T") 1132 return false; 1133 1134 const char *Start = Macro->TokenText.data(); 1135 const char *End = Last->TokenText.data() + Last->TokenText.size(); 1136 String->TokenText = StringRef(Start, End - Start); 1137 String->IsFirst = Macro->IsFirst; 1138 String->LastNewlineOffset = Macro->LastNewlineOffset; 1139 String->WhitespaceRange = Macro->WhitespaceRange; 1140 String->OriginalColumn = Macro->OriginalColumn; 1141 String->ColumnWidth = encoding::columnWidthWithTabs( 1142 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); 1143 1144 Tokens.pop_back(); 1145 Tokens.pop_back(); 1146 Tokens.pop_back(); 1147 Tokens.back() = String; 1148 return true; 1149 } 1150 1151 FormatToken *getNextToken() { 1152 if (GreaterStashed) { 1153 // Create a synthesized second '>' token. 1154 // FIXME: Increment Column and set OriginalColumn. 1155 Token Greater = FormatTok->Tok; 1156 FormatTok = new (Allocator.Allocate()) FormatToken; 1157 FormatTok->Tok = Greater; 1158 SourceLocation GreaterLocation = 1159 FormatTok->Tok.getLocation().getLocWithOffset(1); 1160 FormatTok->WhitespaceRange = 1161 SourceRange(GreaterLocation, GreaterLocation); 1162 FormatTok->TokenText = ">"; 1163 FormatTok->ColumnWidth = 1; 1164 GreaterStashed = false; 1165 return FormatTok; 1166 } 1167 1168 FormatTok = new (Allocator.Allocate()) FormatToken; 1169 readRawToken(*FormatTok); 1170 SourceLocation WhitespaceStart = 1171 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); 1172 FormatTok->IsFirst = IsFirstToken; 1173 IsFirstToken = false; 1174 1175 // Consume and record whitespace until we find a significant token. 1176 unsigned WhitespaceLength = TrailingWhitespace; 1177 while (FormatTok->Tok.is(tok::unknown)) { 1178 for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) { 1179 switch (FormatTok->TokenText[i]) { 1180 case '\n': 1181 ++FormatTok->NewlinesBefore; 1182 // FIXME: This is technically incorrect, as it could also 1183 // be a literal backslash at the end of the line. 1184 if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' && 1185 (FormatTok->TokenText[i - 1] != '\r' || i == 1 || 1186 FormatTok->TokenText[i - 2] != '\\'))) 1187 FormatTok->HasUnescapedNewline = true; 1188 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; 1189 Column = 0; 1190 break; 1191 case '\r': 1192 case '\f': 1193 case '\v': 1194 Column = 0; 1195 break; 1196 case ' ': 1197 ++Column; 1198 break; 1199 case '\t': 1200 Column += Style.TabWidth - Column % Style.TabWidth; 1201 break; 1202 case '\\': 1203 ++Column; 1204 if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' && 1205 FormatTok->TokenText[i + 1] != '\n')) 1206 FormatTok->Type = TT_ImplicitStringLiteral; 1207 break; 1208 default: 1209 FormatTok->Type = TT_ImplicitStringLiteral; 1210 ++Column; 1211 break; 1212 } 1213 } 1214 1215 if (FormatTok->Type == TT_ImplicitStringLiteral) 1216 break; 1217 WhitespaceLength += FormatTok->Tok.getLength(); 1218 1219 readRawToken(*FormatTok); 1220 } 1221 1222 // In case the token starts with escaped newlines, we want to 1223 // take them into account as whitespace - this pattern is quite frequent 1224 // in macro definitions. 1225 // FIXME: Add a more explicit test. 1226 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' && 1227 FormatTok->TokenText[1] == '\n') { 1228 // FIXME: ++FormatTok->NewlinesBefore is missing... 1229 WhitespaceLength += 2; 1230 Column = 0; 1231 FormatTok->TokenText = FormatTok->TokenText.substr(2); 1232 } 1233 1234 FormatTok->WhitespaceRange = SourceRange( 1235 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); 1236 1237 FormatTok->OriginalColumn = Column; 1238 1239 TrailingWhitespace = 0; 1240 if (FormatTok->Tok.is(tok::comment)) { 1241 // FIXME: Add the trimmed whitespace to Column. 1242 StringRef UntrimmedText = FormatTok->TokenText; 1243 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); 1244 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); 1245 } else if (FormatTok->Tok.is(tok::raw_identifier)) { 1246 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); 1247 FormatTok->Tok.setIdentifierInfo(&Info); 1248 FormatTok->Tok.setKind(Info.getTokenID()); 1249 } else if (FormatTok->Tok.is(tok::greatergreater)) { 1250 FormatTok->Tok.setKind(tok::greater); 1251 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); 1252 GreaterStashed = true; 1253 } 1254 1255 // Now FormatTok is the next non-whitespace token. 1256 1257 StringRef Text = FormatTok->TokenText; 1258 size_t FirstNewlinePos = Text.find('\n'); 1259 if (FirstNewlinePos == StringRef::npos) { 1260 // FIXME: ColumnWidth actually depends on the start column, we need to 1261 // take this into account when the token is moved. 1262 FormatTok->ColumnWidth = 1263 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); 1264 Column += FormatTok->ColumnWidth; 1265 } else { 1266 FormatTok->IsMultiline = true; 1267 // FIXME: ColumnWidth actually depends on the start column, we need to 1268 // take this into account when the token is moved. 1269 FormatTok->ColumnWidth = encoding::columnWidthWithTabs( 1270 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); 1271 1272 // The last line of the token always starts in column 0. 1273 // Thus, the length can be precomputed even in the presence of tabs. 1274 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( 1275 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, 1276 Encoding); 1277 Column = FormatTok->LastLineColumnWidth; 1278 } 1279 1280 return FormatTok; 1281 } 1282 1283 FormatToken *FormatTok; 1284 bool IsFirstToken; 1285 bool GreaterStashed; 1286 unsigned Column; 1287 unsigned TrailingWhitespace; 1288 Lexer &Lex; 1289 SourceManager &SourceMgr; 1290 FormatStyle &Style; 1291 IdentifierTable IdentTable; 1292 encoding::Encoding Encoding; 1293 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator; 1294 SmallVector<FormatToken *, 16> Tokens; 1295 1296 void readRawToken(FormatToken &Tok) { 1297 Lex.LexFromRawLexer(Tok.Tok); 1298 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), 1299 Tok.Tok.getLength()); 1300 // For formatting, treat unterminated string literals like normal string 1301 // literals. 1302 if (Tok.is(tok::unknown) && !Tok.TokenText.empty() && 1303 Tok.TokenText[0] == '"') { 1304 Tok.Tok.setKind(tok::string_literal); 1305 Tok.IsUnterminatedLiteral = true; 1306 } 1307 } 1308 }; 1309 1310 static StringRef getLanguageName(FormatStyle::LanguageKind Language) { 1311 switch (Language) { 1312 case FormatStyle::LK_Cpp: 1313 return "C++"; 1314 case FormatStyle::LK_JavaScript: 1315 return "JavaScript"; 1316 default: 1317 return "Unknown"; 1318 } 1319 } 1320 1321 class Formatter : public UnwrappedLineConsumer { 1322 public: 1323 Formatter(const FormatStyle &Style, Lexer &Lex, SourceManager &SourceMgr, 1324 const std::vector<CharSourceRange> &Ranges) 1325 : Style(Style), Lex(Lex), SourceMgr(SourceMgr), 1326 Whitespaces(SourceMgr, Style, inputUsesCRLF(Lex.getBuffer())), 1327 Ranges(Ranges.begin(), Ranges.end()), UnwrappedLines(1), 1328 Encoding(encoding::detectEncoding(Lex.getBuffer())) { 1329 DEBUG(llvm::dbgs() << "File encoding: " 1330 << (Encoding == encoding::Encoding_UTF8 ? "UTF8" 1331 : "unknown") 1332 << "\n"); 1333 DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language) 1334 << "\n"); 1335 } 1336 1337 tooling::Replacements format() { 1338 tooling::Replacements Result; 1339 FormatTokenLexer Tokens(Lex, SourceMgr, Style, Encoding); 1340 1341 UnwrappedLineParser Parser(Style, Tokens.lex(), *this); 1342 bool StructuralError = Parser.parse(); 1343 assert(UnwrappedLines.rbegin()->empty()); 1344 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE; 1345 ++Run) { 1346 DEBUG(llvm::dbgs() << "Run " << Run << "...\n"); 1347 SmallVector<AnnotatedLine *, 16> AnnotatedLines; 1348 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) { 1349 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i])); 1350 } 1351 tooling::Replacements RunResult = 1352 format(AnnotatedLines, StructuralError, Tokens); 1353 DEBUG({ 1354 llvm::dbgs() << "Replacements for run " << Run << ":\n"; 1355 for (tooling::Replacements::iterator I = RunResult.begin(), 1356 E = RunResult.end(); 1357 I != E; ++I) { 1358 llvm::dbgs() << I->toString() << "\n"; 1359 } 1360 }); 1361 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1362 delete AnnotatedLines[i]; 1363 } 1364 Result.insert(RunResult.begin(), RunResult.end()); 1365 Whitespaces.reset(); 1366 } 1367 return Result; 1368 } 1369 1370 tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines, 1371 bool StructuralError, FormatTokenLexer &Tokens) { 1372 TokenAnnotator Annotator(Style, Tokens.getIdentTable().get("in")); 1373 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1374 Annotator.annotate(*AnnotatedLines[i]); 1375 } 1376 deriveLocalStyle(AnnotatedLines); 1377 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1378 Annotator.calculateFormattingInformation(*AnnotatedLines[i]); 1379 } 1380 computeAffectedLines(AnnotatedLines.begin(), AnnotatedLines.end()); 1381 1382 Annotator.setCommentLineLevels(AnnotatedLines); 1383 ContinuationIndenter Indenter(Style, SourceMgr, Whitespaces, Encoding, 1384 BinPackInconclusiveFunctions); 1385 UnwrappedLineFormatter Formatter(&Indenter, &Whitespaces, Style); 1386 Formatter.format(AnnotatedLines, /*DryRun=*/false); 1387 return Whitespaces.generateReplacements(); 1388 } 1389 1390 private: 1391 // Determines which lines are affected by the SourceRanges given as input. 1392 // Returns \c true if at least one line between I and E or one of their 1393 // children is affected. 1394 bool computeAffectedLines(SmallVectorImpl<AnnotatedLine *>::iterator I, 1395 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1396 bool SomeLineAffected = false; 1397 const AnnotatedLine *PreviousLine = NULL; 1398 while (I != E) { 1399 AnnotatedLine *Line = *I; 1400 Line->LeadingEmptyLinesAffected = affectsLeadingEmptyLines(*Line->First); 1401 1402 // If a line is part of a preprocessor directive, it needs to be formatted 1403 // if any token within the directive is affected. 1404 if (Line->InPPDirective) { 1405 FormatToken *Last = Line->Last; 1406 SmallVectorImpl<AnnotatedLine *>::iterator PPEnd = I + 1; 1407 while (PPEnd != E && !(*PPEnd)->First->HasUnescapedNewline) { 1408 Last = (*PPEnd)->Last; 1409 ++PPEnd; 1410 } 1411 1412 if (affectsTokenRange(*Line->First, *Last, 1413 /*IncludeLeadingNewlines=*/false)) { 1414 SomeLineAffected = true; 1415 markAllAsAffected(I, PPEnd); 1416 } 1417 I = PPEnd; 1418 continue; 1419 } 1420 1421 if (nonPPLineAffected(Line, PreviousLine)) 1422 SomeLineAffected = true; 1423 1424 PreviousLine = Line; 1425 ++I; 1426 } 1427 return SomeLineAffected; 1428 } 1429 1430 // Determines whether 'Line' is affected by the SourceRanges given as input. 1431 // Returns \c true if line or one if its children is affected. 1432 bool nonPPLineAffected(AnnotatedLine *Line, 1433 const AnnotatedLine *PreviousLine) { 1434 bool SomeLineAffected = false; 1435 Line->ChildrenAffected = 1436 computeAffectedLines(Line->Children.begin(), Line->Children.end()); 1437 if (Line->ChildrenAffected) 1438 SomeLineAffected = true; 1439 1440 // Stores whether one of the line's tokens is directly affected. 1441 bool SomeTokenAffected = false; 1442 // Stores whether we need to look at the leading newlines of the next token 1443 // in order to determine whether it was affected. 1444 bool IncludeLeadingNewlines = false; 1445 1446 // Stores whether the first child line of any of this line's tokens is 1447 // affected. 1448 bool SomeFirstChildAffected = false; 1449 1450 for (FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) { 1451 // Determine whether 'Tok' was affected. 1452 if (affectsTokenRange(*Tok, *Tok, IncludeLeadingNewlines)) 1453 SomeTokenAffected = true; 1454 1455 // Determine whether the first child of 'Tok' was affected. 1456 if (!Tok->Children.empty() && Tok->Children.front()->Affected) 1457 SomeFirstChildAffected = true; 1458 1459 IncludeLeadingNewlines = Tok->Children.empty(); 1460 } 1461 1462 // Was this line moved, i.e. has it previously been on the same line as an 1463 // affected line? 1464 bool LineMoved = PreviousLine && PreviousLine->Affected && 1465 Line->First->NewlinesBefore == 0; 1466 1467 bool IsContinuedComment = Line->First->is(tok::comment) && 1468 Line->First->Next == NULL && 1469 Line->First->NewlinesBefore < 2 && PreviousLine && 1470 PreviousLine->Affected && 1471 PreviousLine->Last->is(tok::comment); 1472 1473 if (SomeTokenAffected || SomeFirstChildAffected || LineMoved || 1474 IsContinuedComment) { 1475 Line->Affected = true; 1476 SomeLineAffected = true; 1477 } 1478 return SomeLineAffected; 1479 } 1480 1481 // Marks all lines between I and E as well as all their children as affected. 1482 void markAllAsAffected(SmallVectorImpl<AnnotatedLine *>::iterator I, 1483 SmallVectorImpl<AnnotatedLine *>::iterator E) { 1484 while (I != E) { 1485 (*I)->Affected = true; 1486 markAllAsAffected((*I)->Children.begin(), (*I)->Children.end()); 1487 ++I; 1488 } 1489 } 1490 1491 // Returns true if the range from 'First' to 'Last' intersects with one of the 1492 // input ranges. 1493 bool affectsTokenRange(const FormatToken &First, const FormatToken &Last, 1494 bool IncludeLeadingNewlines) { 1495 SourceLocation Start = First.WhitespaceRange.getBegin(); 1496 if (!IncludeLeadingNewlines) 1497 Start = Start.getLocWithOffset(First.LastNewlineOffset); 1498 SourceLocation End = Last.getStartOfNonWhitespace(); 1499 if (Last.TokenText.size() > 0) 1500 End = End.getLocWithOffset(Last.TokenText.size() - 1); 1501 CharSourceRange Range = CharSourceRange::getCharRange(Start, End); 1502 return affectsCharSourceRange(Range); 1503 } 1504 1505 // Returns true if one of the input ranges intersect the leading empty lines 1506 // before 'Tok'. 1507 bool affectsLeadingEmptyLines(const FormatToken &Tok) { 1508 CharSourceRange EmptyLineRange = CharSourceRange::getCharRange( 1509 Tok.WhitespaceRange.getBegin(), 1510 Tok.WhitespaceRange.getBegin().getLocWithOffset(Tok.LastNewlineOffset)); 1511 return affectsCharSourceRange(EmptyLineRange); 1512 } 1513 1514 // Returns true if 'Range' intersects with one of the input ranges. 1515 bool affectsCharSourceRange(const CharSourceRange &Range) { 1516 for (SmallVectorImpl<CharSourceRange>::const_iterator I = Ranges.begin(), 1517 E = Ranges.end(); 1518 I != E; ++I) { 1519 if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), I->getBegin()) && 1520 !SourceMgr.isBeforeInTranslationUnit(I->getEnd(), Range.getBegin())) 1521 return true; 1522 } 1523 return false; 1524 } 1525 1526 static bool inputUsesCRLF(StringRef Text) { 1527 return Text.count('\r') * 2 > Text.count('\n'); 1528 } 1529 1530 void 1531 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) { 1532 unsigned CountBoundToVariable = 0; 1533 unsigned CountBoundToType = 0; 1534 bool HasCpp03IncompatibleFormat = false; 1535 bool HasBinPackedFunction = false; 1536 bool HasOnePerLineFunction = false; 1537 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) { 1538 if (!AnnotatedLines[i]->First->Next) 1539 continue; 1540 FormatToken *Tok = AnnotatedLines[i]->First->Next; 1541 while (Tok->Next) { 1542 if (Tok->Type == TT_PointerOrReference) { 1543 bool SpacesBefore = 1544 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd(); 1545 bool SpacesAfter = Tok->Next->WhitespaceRange.getBegin() != 1546 Tok->Next->WhitespaceRange.getEnd(); 1547 if (SpacesBefore && !SpacesAfter) 1548 ++CountBoundToVariable; 1549 else if (!SpacesBefore && SpacesAfter) 1550 ++CountBoundToType; 1551 } 1552 1553 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) { 1554 if (Tok->is(tok::coloncolon) && 1555 Tok->Previous->Type == TT_TemplateOpener) 1556 HasCpp03IncompatibleFormat = true; 1557 if (Tok->Type == TT_TemplateCloser && 1558 Tok->Previous->Type == TT_TemplateCloser) 1559 HasCpp03IncompatibleFormat = true; 1560 } 1561 1562 if (Tok->PackingKind == PPK_BinPacked) 1563 HasBinPackedFunction = true; 1564 if (Tok->PackingKind == PPK_OnePerLine) 1565 HasOnePerLineFunction = true; 1566 1567 Tok = Tok->Next; 1568 } 1569 } 1570 if (Style.DerivePointerBinding) { 1571 if (CountBoundToType > CountBoundToVariable) 1572 Style.PointerBindsToType = true; 1573 else if (CountBoundToType < CountBoundToVariable) 1574 Style.PointerBindsToType = false; 1575 } 1576 if (Style.Standard == FormatStyle::LS_Auto) { 1577 Style.Standard = HasCpp03IncompatibleFormat ? FormatStyle::LS_Cpp11 1578 : FormatStyle::LS_Cpp03; 1579 } 1580 BinPackInconclusiveFunctions = 1581 HasBinPackedFunction || !HasOnePerLineFunction; 1582 } 1583 1584 virtual void consumeUnwrappedLine(const UnwrappedLine &TheLine) { 1585 assert(!UnwrappedLines.empty()); 1586 UnwrappedLines.back().push_back(TheLine); 1587 } 1588 1589 virtual void finishRun() { 1590 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>()); 1591 } 1592 1593 FormatStyle Style; 1594 Lexer &Lex; 1595 SourceManager &SourceMgr; 1596 WhitespaceManager Whitespaces; 1597 SmallVector<CharSourceRange, 8> Ranges; 1598 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines; 1599 1600 encoding::Encoding Encoding; 1601 bool BinPackInconclusiveFunctions; 1602 }; 1603 1604 } // end anonymous namespace 1605 1606 tooling::Replacements reformat(const FormatStyle &Style, Lexer &Lex, 1607 SourceManager &SourceMgr, 1608 std::vector<CharSourceRange> Ranges) { 1609 Formatter formatter(Style, Lex, SourceMgr, Ranges); 1610 return formatter.format(); 1611 } 1612 1613 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code, 1614 std::vector<tooling::Range> Ranges, 1615 StringRef FileName) { 1616 FileManager Files((FileSystemOptions())); 1617 DiagnosticsEngine Diagnostics( 1618 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), 1619 new DiagnosticOptions); 1620 SourceManager SourceMgr(Diagnostics, Files); 1621 llvm::MemoryBuffer *Buf = llvm::MemoryBuffer::getMemBuffer(Code, FileName); 1622 const clang::FileEntry *Entry = 1623 Files.getVirtualFile(FileName, Buf->getBufferSize(), 0); 1624 SourceMgr.overrideFileContents(Entry, Buf); 1625 FileID ID = 1626 SourceMgr.createFileID(Entry, SourceLocation(), clang::SrcMgr::C_User); 1627 Lexer Lex(ID, SourceMgr.getBuffer(ID), SourceMgr, 1628 getFormattingLangOpts(Style.Standard)); 1629 SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID); 1630 std::vector<CharSourceRange> CharRanges; 1631 for (unsigned i = 0, e = Ranges.size(); i != e; ++i) { 1632 SourceLocation Start = StartOfFile.getLocWithOffset(Ranges[i].getOffset()); 1633 SourceLocation End = Start.getLocWithOffset(Ranges[i].getLength()); 1634 CharRanges.push_back(CharSourceRange::getCharRange(Start, End)); 1635 } 1636 return reformat(Style, Lex, SourceMgr, CharRanges); 1637 } 1638 1639 LangOptions getFormattingLangOpts(FormatStyle::LanguageStandard Standard) { 1640 LangOptions LangOpts; 1641 LangOpts.CPlusPlus = 1; 1642 LangOpts.CPlusPlus11 = Standard == FormatStyle::LS_Cpp03 ? 0 : 1; 1643 LangOpts.LineComment = 1; 1644 LangOpts.Bool = 1; 1645 LangOpts.ObjC1 = 1; 1646 LangOpts.ObjC2 = 1; 1647 return LangOpts; 1648 } 1649 1650 const char *StyleOptionHelpDescription = 1651 "Coding style, currently supports:\n" 1652 " LLVM, Google, Chromium, Mozilla, WebKit.\n" 1653 "Use -style=file to load style configuration from\n" 1654 ".clang-format file located in one of the parent\n" 1655 "directories of the source file (or current\n" 1656 "directory for stdin).\n" 1657 "Use -style=\"{key: value, ...}\" to set specific\n" 1658 "parameters, e.g.:\n" 1659 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\""; 1660 1661 static void fillLanguageByFileName(StringRef FileName, FormatStyle *Style) { 1662 if (FileName.endswith_lower(".c") || FileName.endswith_lower(".h") || 1663 FileName.endswith_lower(".cpp") || FileName.endswith_lower(".hpp") || 1664 FileName.endswith_lower(".cc") || FileName.endswith_lower(".hh") || 1665 FileName.endswith_lower(".cxx") || FileName.endswith_lower(".hxx") || 1666 FileName.endswith_lower(".m") || FileName.endswith_lower(".mm")) { 1667 Style->Language = FormatStyle::LK_Cpp; 1668 } 1669 if (FileName.endswith_lower(".js")) { 1670 Style->Language = FormatStyle::LK_JavaScript; 1671 } 1672 } 1673 1674 FormatStyle getStyle(StringRef StyleName, StringRef FileName, 1675 StringRef FallbackStyle) { 1676 FormatStyle Style; 1677 if (!getPredefinedStyle(FallbackStyle, &Style)) { 1678 llvm::errs() << "Invalid fallback style \"" << FallbackStyle 1679 << "\" using LLVM style\n"; 1680 return getLLVMStyle(); 1681 } 1682 fillLanguageByFileName(FileName, &Style); 1683 1684 if (StyleName.startswith("{")) { 1685 // Parse YAML/JSON style from the command line. 1686 if (llvm::error_code ec = parseConfiguration(StyleName, &Style)) { 1687 llvm::errs() << "Error parsing -style: " << ec.message() << ", using " 1688 << FallbackStyle << " style\n"; 1689 } 1690 return Style; 1691 } 1692 1693 if (!StyleName.equals_lower("file")) { 1694 if (!getPredefinedStyle(StyleName, &Style)) 1695 llvm::errs() << "Invalid value for -style, using " << FallbackStyle 1696 << " style\n"; 1697 fillLanguageByFileName(FileName, &Style); 1698 return Style; 1699 } 1700 1701 SmallString<128> UnsuitableConfigFiles; 1702 SmallString<128> Path(FileName); 1703 llvm::sys::fs::make_absolute(Path); 1704 for (StringRef Directory = Path; !Directory.empty(); 1705 Directory = llvm::sys::path::parent_path(Directory)) { 1706 if (!llvm::sys::fs::is_directory(Directory)) 1707 continue; 1708 SmallString<128> ConfigFile(Directory); 1709 1710 llvm::sys::path::append(ConfigFile, ".clang-format"); 1711 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1712 bool IsFile = false; 1713 // Ignore errors from is_regular_file: we only need to know if we can read 1714 // the file or not. 1715 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1716 1717 if (!IsFile) { 1718 // Try _clang-format too, since dotfiles are not commonly used on Windows. 1719 ConfigFile = Directory; 1720 llvm::sys::path::append(ConfigFile, "_clang-format"); 1721 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n"); 1722 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile); 1723 } 1724 1725 if (IsFile) { 1726 OwningPtr<llvm::MemoryBuffer> Text; 1727 if (llvm::error_code ec = 1728 llvm::MemoryBuffer::getFile(ConfigFile.c_str(), Text)) { 1729 llvm::errs() << ec.message() << "\n"; 1730 break; 1731 } 1732 if (llvm::error_code ec = parseConfiguration(Text->getBuffer(), &Style)) { 1733 if (ec == llvm::errc::not_supported) { 1734 if (!UnsuitableConfigFiles.empty()) 1735 UnsuitableConfigFiles.append(", "); 1736 UnsuitableConfigFiles.append(ConfigFile); 1737 continue; 1738 } 1739 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message() 1740 << "\n"; 1741 break; 1742 } 1743 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n"); 1744 return Style; 1745 } 1746 } 1747 llvm::errs() << "Can't find usable .clang-format, using " << FallbackStyle 1748 << " style\n"; 1749 if (!UnsuitableConfigFiles.empty()) { 1750 llvm::errs() << "Configuration file(s) do(es) not support " 1751 << getLanguageName(Style.Language) << ": " 1752 << UnsuitableConfigFiles << "\n"; 1753 } 1754 return Style; 1755 } 1756 1757 } // namespace format 1758 } // namespace clang 1759