1 //===--- Format.cpp - Format C++ code -------------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
11 /// \brief This file implements functions declared in Format.h. This will be
12 /// split into separate files as we go.
14 //===----------------------------------------------------------------------===//
16 #include "clang/Format/Format.h"
17 #include "ContinuationIndenter.h"
18 #include "TokenAnnotator.h"
19 #include "UnwrappedLineFormatter.h"
20 #include "UnwrappedLineParser.h"
21 #include "WhitespaceManager.h"
22 #include "clang/Basic/Diagnostic.h"
23 #include "clang/Basic/DiagnosticOptions.h"
24 #include "clang/Basic/SourceManager.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/Path.h"
30 #include "llvm/Support/Regex.h"
31 #include "llvm/Support/YAMLTraits.h"
35 #define DEBUG_TYPE "format-formatter"
37 using clang::format::FormatStyle;
39 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string)
40 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory)
44 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
45 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
46 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
47 IO.enumCase(Value, "Java", FormatStyle::LK_Java);
48 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
49 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
53 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
54 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
55 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
56 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
57 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
58 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
59 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
63 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
64 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
65 IO.enumCase(Value, "Never", FormatStyle::UT_Never);
66 IO.enumCase(Value, "false", FormatStyle::UT_Never);
67 IO.enumCase(Value, "Always", FormatStyle::UT_Always);
68 IO.enumCase(Value, "true", FormatStyle::UT_Always);
69 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
73 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
74 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
75 IO.enumCase(Value, "None", FormatStyle::SFS_None);
76 IO.enumCase(Value, "false", FormatStyle::SFS_None);
77 IO.enumCase(Value, "All", FormatStyle::SFS_All);
78 IO.enumCase(Value, "true", FormatStyle::SFS_All);
79 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
80 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
84 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
85 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
86 IO.enumCase(Value, "All", FormatStyle::BOS_All);
87 IO.enumCase(Value, "true", FormatStyle::BOS_All);
88 IO.enumCase(Value, "None", FormatStyle::BOS_None);
89 IO.enumCase(Value, "false", FormatStyle::BOS_None);
90 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
94 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
95 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
96 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
97 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
98 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla);
99 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
100 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
101 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
102 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit);
103 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom);
108 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> {
110 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) {
111 IO.enumCase(Value, "None", FormatStyle::DRTBS_None);
112 IO.enumCase(Value, "All", FormatStyle::DRTBS_All);
113 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel);
115 // For backward compatibility.
116 IO.enumCase(Value, "false", FormatStyle::DRTBS_None);
117 IO.enumCase(Value, "true", FormatStyle::DRTBS_All);
122 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
123 static void enumeration(IO &IO,
124 FormatStyle::NamespaceIndentationKind &Value) {
125 IO.enumCase(Value, "None", FormatStyle::NI_None);
126 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
127 IO.enumCase(Value, "All", FormatStyle::NI_All);
131 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
132 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
133 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
134 IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
135 IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
137 // For backward compatibility.
138 IO.enumCase(Value, "true", FormatStyle::PAS_Left);
139 IO.enumCase(Value, "false", FormatStyle::PAS_Right);
144 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
145 static void enumeration(IO &IO,
146 FormatStyle::SpaceBeforeParensOptions &Value) {
147 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
148 IO.enumCase(Value, "ControlStatements",
149 FormatStyle::SBPO_ControlStatements);
150 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
152 // For backward compatibility.
153 IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
154 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
158 template <> struct MappingTraits<FormatStyle> {
159 static void mapping(IO &IO, FormatStyle &Style) {
160 // When reading, read the language first, we need it for getPredefinedStyle.
161 IO.mapOptional("Language", Style.Language);
163 if (IO.outputting()) {
164 StringRef StylesArray[] = {"LLVM", "Google", "Chromium",
165 "Mozilla", "WebKit", "GNU"};
166 ArrayRef<StringRef> Styles(StylesArray);
167 for (size_t i = 0, e = Styles.size(); i < e; ++i) {
168 StringRef StyleName(Styles[i]);
169 FormatStyle PredefinedStyle;
170 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
171 Style == PredefinedStyle) {
172 IO.mapOptional("# BasedOnStyle", StyleName);
177 StringRef BasedOnStyle;
178 IO.mapOptional("BasedOnStyle", BasedOnStyle);
179 if (!BasedOnStyle.empty()) {
180 FormatStyle::LanguageKind OldLanguage = Style.Language;
181 FormatStyle::LanguageKind Language =
182 ((FormatStyle *)IO.getContext())->Language;
183 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
184 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
187 Style.Language = OldLanguage;
191 // For backward compatibility.
192 if (!IO.outputting()) {
193 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
194 IO.mapOptional("IndentFunctionDeclarationAfterType",
195 Style.IndentWrappedFunctionNames);
196 IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
197 IO.mapOptional("SpaceAfterControlStatementKeyword",
198 Style.SpaceBeforeParens);
201 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
202 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
203 IO.mapOptional("AlignConsecutiveAssignments",
204 Style.AlignConsecutiveAssignments);
205 IO.mapOptional("AlignConsecutiveDeclarations",
206 Style.AlignConsecutiveDeclarations);
207 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft);
208 IO.mapOptional("AlignOperands", Style.AlignOperands);
209 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
210 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
211 Style.AllowAllParametersOfDeclarationOnNextLine);
212 IO.mapOptional("AllowShortBlocksOnASingleLine",
213 Style.AllowShortBlocksOnASingleLine);
214 IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
215 Style.AllowShortCaseLabelsOnASingleLine);
216 IO.mapOptional("AllowShortFunctionsOnASingleLine",
217 Style.AllowShortFunctionsOnASingleLine);
218 IO.mapOptional("AllowShortIfStatementsOnASingleLine",
219 Style.AllowShortIfStatementsOnASingleLine);
220 IO.mapOptional("AllowShortLoopsOnASingleLine",
221 Style.AllowShortLoopsOnASingleLine);
222 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
223 Style.AlwaysBreakAfterDefinitionReturnType);
224 IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
225 Style.AlwaysBreakBeforeMultilineStrings);
226 IO.mapOptional("AlwaysBreakTemplateDeclarations",
227 Style.AlwaysBreakTemplateDeclarations);
228 IO.mapOptional("BinPackArguments", Style.BinPackArguments);
229 IO.mapOptional("BinPackParameters", Style.BinPackParameters);
230 IO.mapOptional("BraceWrapping", Style.BraceWrapping);
231 IO.mapOptional("BreakBeforeBinaryOperators",
232 Style.BreakBeforeBinaryOperators);
233 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
234 IO.mapOptional("BreakBeforeTernaryOperators",
235 Style.BreakBeforeTernaryOperators);
236 IO.mapOptional("BreakConstructorInitializersBeforeComma",
237 Style.BreakConstructorInitializersBeforeComma);
238 IO.mapOptional("ColumnLimit", Style.ColumnLimit);
239 IO.mapOptional("CommentPragmas", Style.CommentPragmas);
240 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
241 Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
242 IO.mapOptional("ConstructorInitializerIndentWidth",
243 Style.ConstructorInitializerIndentWidth);
244 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
245 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
246 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
247 IO.mapOptional("DisableFormat", Style.DisableFormat);
248 IO.mapOptional("ExperimentalAutoDetectBinPacking",
249 Style.ExperimentalAutoDetectBinPacking);
250 IO.mapOptional("ForEachMacros", Style.ForEachMacros);
251 IO.mapOptional("IncludeCategories", Style.IncludeCategories);
252 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
253 IO.mapOptional("IndentWidth", Style.IndentWidth);
254 IO.mapOptional("IndentWrappedFunctionNames",
255 Style.IndentWrappedFunctionNames);
256 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
257 Style.KeepEmptyLinesAtTheStartOfBlocks);
258 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
259 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
260 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
261 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
262 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
263 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
264 IO.mapOptional("ObjCSpaceBeforeProtocolList",
265 Style.ObjCSpaceBeforeProtocolList);
266 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
267 Style.PenaltyBreakBeforeFirstCallParameter);
268 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
269 IO.mapOptional("PenaltyBreakFirstLessLess",
270 Style.PenaltyBreakFirstLessLess);
271 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
272 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
273 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
274 Style.PenaltyReturnTypeOnItsOwnLine);
275 IO.mapOptional("PointerAlignment", Style.PointerAlignment);
276 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
277 IO.mapOptional("SpaceBeforeAssignmentOperators",
278 Style.SpaceBeforeAssignmentOperators);
279 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
280 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
281 IO.mapOptional("SpacesBeforeTrailingComments",
282 Style.SpacesBeforeTrailingComments);
283 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
284 IO.mapOptional("SpacesInContainerLiterals",
285 Style.SpacesInContainerLiterals);
286 IO.mapOptional("SpacesInCStyleCastParentheses",
287 Style.SpacesInCStyleCastParentheses);
288 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
289 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
290 IO.mapOptional("Standard", Style.Standard);
291 IO.mapOptional("TabWidth", Style.TabWidth);
292 IO.mapOptional("UseTab", Style.UseTab);
296 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
297 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
298 IO.mapOptional("AfterClass", Wrapping.AfterClass);
299 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
300 IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
301 IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
302 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
303 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
304 IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
305 IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
306 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
307 IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
308 IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
312 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
313 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
314 IO.mapOptional("Regex", Category.Regex);
315 IO.mapOptional("Priority", Category.Priority);
319 // Allows to read vector<FormatStyle> while keeping default values.
320 // IO.getContext() should contain a pointer to the FormatStyle structure, that
321 // will be used to get default values for missing keys.
322 // If the first element has no Language specified, it will be treated as the
323 // default one for the following elements.
324 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
325 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
328 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
330 if (Index >= Seq.size()) {
331 assert(Index == Seq.size());
332 FormatStyle Template;
333 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
336 Template = *((const FormatStyle *)IO.getContext());
337 Template.Language = FormatStyle::LK_None;
339 Seq.resize(Index + 1, Template);
350 const std::error_category &getParseCategory() {
351 static ParseErrorCategory C;
354 std::error_code make_error_code(ParseError e) {
355 return std::error_code(static_cast<int>(e), getParseCategory());
358 const char *ParseErrorCategory::name() const LLVM_NOEXCEPT {
359 return "clang-format.parse_error";
362 std::string ParseErrorCategory::message(int EV) const {
363 switch (static_cast<ParseError>(EV)) {
364 case ParseError::Success:
366 case ParseError::Error:
367 return "Invalid argument";
368 case ParseError::Unsuitable:
371 llvm_unreachable("unexpected parse error");
374 static FormatStyle expandPresets(const FormatStyle &Style) {
375 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
377 FormatStyle Expanded = Style;
378 Expanded.BraceWrapping = {false, false, false, false, false, false,
379 false, false, false, false, false};
380 switch (Style.BreakBeforeBraces) {
381 case FormatStyle::BS_Linux:
382 Expanded.BraceWrapping.AfterClass = true;
383 Expanded.BraceWrapping.AfterFunction = true;
384 Expanded.BraceWrapping.AfterNamespace = true;
385 Expanded.BraceWrapping.BeforeElse = true;
387 case FormatStyle::BS_Mozilla:
388 Expanded.BraceWrapping.AfterClass = true;
389 Expanded.BraceWrapping.AfterEnum = true;
390 Expanded.BraceWrapping.AfterFunction = true;
391 Expanded.BraceWrapping.AfterStruct = true;
392 Expanded.BraceWrapping.AfterUnion = true;
394 case FormatStyle::BS_Stroustrup:
395 Expanded.BraceWrapping.AfterFunction = true;
396 Expanded.BraceWrapping.BeforeCatch = true;
397 Expanded.BraceWrapping.BeforeElse = true;
399 case FormatStyle::BS_Allman:
400 Expanded.BraceWrapping.AfterClass = true;
401 Expanded.BraceWrapping.AfterControlStatement = true;
402 Expanded.BraceWrapping.AfterEnum = true;
403 Expanded.BraceWrapping.AfterFunction = true;
404 Expanded.BraceWrapping.AfterNamespace = true;
405 Expanded.BraceWrapping.AfterObjCDeclaration = true;
406 Expanded.BraceWrapping.AfterStruct = true;
407 Expanded.BraceWrapping.BeforeCatch = true;
408 Expanded.BraceWrapping.BeforeElse = true;
410 case FormatStyle::BS_GNU:
411 Expanded.BraceWrapping = {true, true, true, true, true, true,
412 true, true, true, true, true};
414 case FormatStyle::BS_WebKit:
415 Expanded.BraceWrapping.AfterFunction = true;
423 FormatStyle getLLVMStyle() {
424 FormatStyle LLVMStyle;
425 LLVMStyle.Language = FormatStyle::LK_Cpp;
426 LLVMStyle.AccessModifierOffset = -2;
427 LLVMStyle.AlignEscapedNewlinesLeft = false;
428 LLVMStyle.AlignAfterOpenBracket = true;
429 LLVMStyle.AlignOperands = true;
430 LLVMStyle.AlignTrailingComments = true;
431 LLVMStyle.AlignConsecutiveAssignments = false;
432 LLVMStyle.AlignConsecutiveDeclarations = false;
433 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
434 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
435 LLVMStyle.AllowShortBlocksOnASingleLine = false;
436 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
437 LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
438 LLVMStyle.AllowShortLoopsOnASingleLine = false;
439 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
440 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
441 LLVMStyle.AlwaysBreakTemplateDeclarations = false;
442 LLVMStyle.BinPackParameters = true;
443 LLVMStyle.BinPackArguments = true;
444 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
445 LLVMStyle.BreakBeforeTernaryOperators = true;
446 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
447 LLVMStyle.BraceWrapping = {false, false, false, false, false, false,
448 false, false, false, false, false};
449 LLVMStyle.BreakConstructorInitializersBeforeComma = false;
450 LLVMStyle.BreakAfterJavaFieldAnnotations = false;
451 LLVMStyle.ColumnLimit = 80;
452 LLVMStyle.CommentPragmas = "^ IWYU pragma:";
453 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
454 LLVMStyle.ConstructorInitializerIndentWidth = 4;
455 LLVMStyle.ContinuationIndentWidth = 4;
456 LLVMStyle.Cpp11BracedListStyle = true;
457 LLVMStyle.DerivePointerAlignment = false;
458 LLVMStyle.ExperimentalAutoDetectBinPacking = false;
459 LLVMStyle.ForEachMacros.push_back("foreach");
460 LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
461 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
462 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
463 {"^(<|\"(gtest|isl|json)/)", 3},
465 LLVMStyle.IndentCaseLabels = false;
466 LLVMStyle.IndentWrappedFunctionNames = false;
467 LLVMStyle.IndentWidth = 2;
468 LLVMStyle.TabWidth = 8;
469 LLVMStyle.MaxEmptyLinesToKeep = 1;
470 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
471 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
472 LLVMStyle.ObjCBlockIndentWidth = 2;
473 LLVMStyle.ObjCSpaceAfterProperty = false;
474 LLVMStyle.ObjCSpaceBeforeProtocolList = true;
475 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
476 LLVMStyle.SpacesBeforeTrailingComments = 1;
477 LLVMStyle.Standard = FormatStyle::LS_Cpp11;
478 LLVMStyle.UseTab = FormatStyle::UT_Never;
479 LLVMStyle.SpacesInParentheses = false;
480 LLVMStyle.SpacesInSquareBrackets = false;
481 LLVMStyle.SpaceInEmptyParentheses = false;
482 LLVMStyle.SpacesInContainerLiterals = true;
483 LLVMStyle.SpacesInCStyleCastParentheses = false;
484 LLVMStyle.SpaceAfterCStyleCast = false;
485 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
486 LLVMStyle.SpaceBeforeAssignmentOperators = true;
487 LLVMStyle.SpacesInAngles = false;
489 LLVMStyle.PenaltyBreakComment = 300;
490 LLVMStyle.PenaltyBreakFirstLessLess = 120;
491 LLVMStyle.PenaltyBreakString = 1000;
492 LLVMStyle.PenaltyExcessCharacter = 1000000;
493 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
494 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
496 LLVMStyle.DisableFormat = false;
501 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
502 FormatStyle GoogleStyle = getLLVMStyle();
503 GoogleStyle.Language = Language;
505 GoogleStyle.AccessModifierOffset = -1;
506 GoogleStyle.AlignEscapedNewlinesLeft = true;
507 GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
508 GoogleStyle.AllowShortLoopsOnASingleLine = true;
509 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
510 GoogleStyle.AlwaysBreakTemplateDeclarations = true;
511 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
512 GoogleStyle.DerivePointerAlignment = true;
513 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
514 GoogleStyle.IndentCaseLabels = true;
515 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
516 GoogleStyle.ObjCSpaceAfterProperty = false;
517 GoogleStyle.ObjCSpaceBeforeProtocolList = false;
518 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
519 GoogleStyle.SpacesBeforeTrailingComments = 2;
520 GoogleStyle.Standard = FormatStyle::LS_Auto;
522 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
523 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
525 if (Language == FormatStyle::LK_Java) {
526 GoogleStyle.AlignAfterOpenBracket = false;
527 GoogleStyle.AlignOperands = false;
528 GoogleStyle.AlignTrailingComments = false;
529 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
530 GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
531 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
532 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
533 GoogleStyle.ColumnLimit = 100;
534 GoogleStyle.SpaceAfterCStyleCast = true;
535 GoogleStyle.SpacesBeforeTrailingComments = 1;
536 } else if (Language == FormatStyle::LK_JavaScript) {
537 GoogleStyle.BreakBeforeTernaryOperators = false;
538 GoogleStyle.MaxEmptyLinesToKeep = 3;
539 GoogleStyle.SpacesInContainerLiterals = false;
540 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
541 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
542 } else if (Language == FormatStyle::LK_Proto) {
543 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
544 GoogleStyle.SpacesInContainerLiterals = false;
550 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
551 FormatStyle ChromiumStyle = getGoogleStyle(Language);
552 if (Language == FormatStyle::LK_Java) {
553 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
554 ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
555 ChromiumStyle.ContinuationIndentWidth = 8;
556 ChromiumStyle.IndentWidth = 4;
558 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
559 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
560 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
561 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
562 ChromiumStyle.BinPackParameters = false;
563 ChromiumStyle.DerivePointerAlignment = false;
565 return ChromiumStyle;
568 FormatStyle getMozillaStyle() {
569 FormatStyle MozillaStyle = getLLVMStyle();
570 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
571 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
572 MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
573 FormatStyle::DRTBS_TopLevel;
574 MozillaStyle.AlwaysBreakTemplateDeclarations = true;
575 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
576 MozillaStyle.BreakConstructorInitializersBeforeComma = true;
577 MozillaStyle.ConstructorInitializerIndentWidth = 2;
578 MozillaStyle.ContinuationIndentWidth = 2;
579 MozillaStyle.Cpp11BracedListStyle = false;
580 MozillaStyle.IndentCaseLabels = true;
581 MozillaStyle.ObjCSpaceAfterProperty = true;
582 MozillaStyle.ObjCSpaceBeforeProtocolList = false;
583 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
584 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
588 FormatStyle getWebKitStyle() {
589 FormatStyle Style = getLLVMStyle();
590 Style.AccessModifierOffset = -4;
591 Style.AlignAfterOpenBracket = false;
592 Style.AlignOperands = false;
593 Style.AlignTrailingComments = false;
594 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
595 Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
596 Style.BreakConstructorInitializersBeforeComma = true;
597 Style.Cpp11BracedListStyle = false;
598 Style.ColumnLimit = 0;
599 Style.IndentWidth = 4;
600 Style.NamespaceIndentation = FormatStyle::NI_Inner;
601 Style.ObjCBlockIndentWidth = 4;
602 Style.ObjCSpaceAfterProperty = true;
603 Style.PointerAlignment = FormatStyle::PAS_Left;
604 Style.Standard = FormatStyle::LS_Cpp03;
608 FormatStyle getGNUStyle() {
609 FormatStyle Style = getLLVMStyle();
610 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
611 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
612 Style.BreakBeforeBraces = FormatStyle::BS_GNU;
613 Style.BreakBeforeTernaryOperators = true;
614 Style.Cpp11BracedListStyle = false;
615 Style.ColumnLimit = 79;
616 Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
617 Style.Standard = FormatStyle::LS_Cpp03;
621 FormatStyle getNoStyle() {
622 FormatStyle NoStyle = getLLVMStyle();
623 NoStyle.DisableFormat = true;
627 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
628 FormatStyle *Style) {
629 if (Name.equals_lower("llvm")) {
630 *Style = getLLVMStyle();
631 } else if (Name.equals_lower("chromium")) {
632 *Style = getChromiumStyle(Language);
633 } else if (Name.equals_lower("mozilla")) {
634 *Style = getMozillaStyle();
635 } else if (Name.equals_lower("google")) {
636 *Style = getGoogleStyle(Language);
637 } else if (Name.equals_lower("webkit")) {
638 *Style = getWebKitStyle();
639 } else if (Name.equals_lower("gnu")) {
640 *Style = getGNUStyle();
641 } else if (Name.equals_lower("none")) {
642 *Style = getNoStyle();
647 Style->Language = Language;
651 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
653 FormatStyle::LanguageKind Language = Style->Language;
654 assert(Language != FormatStyle::LK_None);
655 if (Text.trim().empty())
656 return make_error_code(ParseError::Error);
658 std::vector<FormatStyle> Styles;
659 llvm::yaml::Input Input(Text);
660 // DocumentListTraits<vector<FormatStyle>> uses the context to get default
661 // values for the fields, keys for which are missing from the configuration.
662 // Mapping also uses the context to get the language to find the correct
664 Input.setContext(Style);
667 return Input.error();
669 for (unsigned i = 0; i < Styles.size(); ++i) {
670 // Ensures that only the first configuration can skip the Language option.
671 if (Styles[i].Language == FormatStyle::LK_None && i != 0)
672 return make_error_code(ParseError::Error);
673 // Ensure that each language is configured at most once.
674 for (unsigned j = 0; j < i; ++j) {
675 if (Styles[i].Language == Styles[j].Language) {
677 << "Duplicate languages in the config file on positions " << j
678 << " and " << i << "\n");
679 return make_error_code(ParseError::Error);
683 // Look for a suitable configuration starting from the end, so we can
684 // find the configuration for the specific language first, and the default
685 // configuration (which can only be at slot 0) after it.
686 for (int i = Styles.size() - 1; i >= 0; --i) {
687 if (Styles[i].Language == Language ||
688 Styles[i].Language == FormatStyle::LK_None) {
690 Style->Language = Language;
691 return make_error_code(ParseError::Success);
694 return make_error_code(ParseError::Unsuitable);
697 std::string configurationAsText(const FormatStyle &Style) {
699 llvm::raw_string_ostream Stream(Text);
700 llvm::yaml::Output Output(Stream);
701 // We use the same mapping method for input and output, so we need a non-const
703 FormatStyle NonConstStyle = expandPresets(Style);
704 Output << NonConstStyle;
710 class FormatTokenLexer {
712 FormatTokenLexer(SourceManager &SourceMgr, FileID ID, FormatStyle &Style,
713 encoding::Encoding Encoding)
714 : FormatTok(nullptr), IsFirstToken(true), GreaterStashed(false),
715 LessStashed(false), Column(0), TrailingWhitespace(0),
716 SourceMgr(SourceMgr), ID(ID), Style(Style),
717 IdentTable(getFormattingLangOpts(Style)), Keywords(IdentTable),
718 Encoding(Encoding), FirstInLineIndex(0), FormattingDisabled(false),
719 MacroBlockBeginRegex(Style.MacroBlockBegin),
720 MacroBlockEndRegex(Style.MacroBlockEnd) {
721 Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr,
722 getFormattingLangOpts(Style)));
723 Lex->SetKeepWhitespaceMode(true);
725 for (const std::string &ForEachMacro : Style.ForEachMacros)
726 ForEachMacros.push_back(&IdentTable.get(ForEachMacro));
727 std::sort(ForEachMacros.begin(), ForEachMacros.end());
730 ArrayRef<FormatToken *> lex() {
731 assert(Tokens.empty());
732 assert(FirstInLineIndex == 0);
734 Tokens.push_back(getNextToken());
735 if (Style.Language == FormatStyle::LK_JavaScript)
736 tryParseJSRegexLiteral();
737 tryMergePreviousTokens();
738 if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline)
739 FirstInLineIndex = Tokens.size() - 1;
740 } while (Tokens.back()->Tok.isNot(tok::eof));
744 const AdditionalKeywords &getKeywords() { return Keywords; }
747 void tryMergePreviousTokens() {
748 if (tryMerge_TMacro())
750 if (tryMergeConflictMarkers())
752 if (tryMergeLessLess())
755 if (Style.Language == FormatStyle::LK_JavaScript) {
756 if (tryMergeTemplateString())
759 static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal};
760 static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal,
762 static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater,
764 static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater};
765 // FIXME: Investigate what token type gives the correct operator priority.
766 if (tryMergeTokens(JSIdentity, TT_BinaryOperator))
768 if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator))
770 if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator))
772 if (tryMergeTokens(JSRightArrow, TT_JsFatArrow))
777 bool tryMergeLessLess() {
778 // Merge X,less,less,Y into X,lessless,Y unless X or Y is less.
779 if (Tokens.size() < 3)
782 bool FourthTokenIsLess = false;
783 if (Tokens.size() > 3)
784 FourthTokenIsLess = (Tokens.end() - 4)[0]->is(tok::less);
786 auto First = Tokens.end() - 3;
787 if (First[2]->is(tok::less) || First[1]->isNot(tok::less) ||
788 First[0]->isNot(tok::less) || FourthTokenIsLess)
791 // Only merge if there currently is no whitespace between the two "<".
792 if (First[1]->WhitespaceRange.getBegin() !=
793 First[1]->WhitespaceRange.getEnd())
796 First[0]->Tok.setKind(tok::lessless);
797 First[0]->TokenText = "<<";
798 First[0]->ColumnWidth += 1;
799 Tokens.erase(Tokens.end() - 2);
803 bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds, TokenType NewType) {
804 if (Tokens.size() < Kinds.size())
807 SmallVectorImpl<FormatToken *>::const_iterator First =
808 Tokens.end() - Kinds.size();
809 if (!First[0]->is(Kinds[0]))
811 unsigned AddLength = 0;
812 for (unsigned i = 1; i < Kinds.size(); ++i) {
813 if (!First[i]->is(Kinds[i]) ||
814 First[i]->WhitespaceRange.getBegin() !=
815 First[i]->WhitespaceRange.getEnd())
817 AddLength += First[i]->TokenText.size();
819 Tokens.resize(Tokens.size() - Kinds.size() + 1);
820 First[0]->TokenText = StringRef(First[0]->TokenText.data(),
821 First[0]->TokenText.size() + AddLength);
822 First[0]->ColumnWidth += AddLength;
823 First[0]->Type = NewType;
827 // Returns \c true if \p Tok can only be followed by an operand in JavaScript.
828 bool precedesOperand(FormatToken *Tok) {
829 // NB: This is not entirely correct, as an r_paren can introduce an operand
830 // location in e.g. `if (foo) /bar/.exec(...);`. That is a rare enough
831 // corner case to not matter in practice, though.
832 return Tok->isOneOf(tok::period, tok::l_paren, tok::comma, tok::l_brace,
833 tok::r_brace, tok::l_square, tok::semi, tok::exclaim,
834 tok::colon, tok::question, tok::tilde) ||
835 Tok->isOneOf(tok::kw_return, tok::kw_do, tok::kw_case, tok::kw_throw,
836 tok::kw_else, tok::kw_new, tok::kw_delete, tok::kw_void,
837 tok::kw_typeof, Keywords.kw_instanceof,
839 Tok->isBinaryOperator();
842 bool canPrecedeRegexLiteral(FormatToken *Prev) {
846 // Regex literals can only follow after prefix unary operators, not after
847 // postfix unary operators. If the '++' is followed by a non-operand
848 // introducing token, the slash here is the operand and not the start of a
850 if (Prev->isOneOf(tok::plusplus, tok::minusminus))
851 return (Tokens.size() < 3 || precedesOperand(Tokens[Tokens.size() - 3]));
853 // The previous token must introduce an operand location where regex
854 // literals can occur.
855 if (!precedesOperand(Prev))
861 // Tries to parse a JavaScript Regex literal starting at the current token,
862 // if that begins with a slash and is in a location where JavaScript allows
863 // regex literals. Changes the current token to a regex literal and updates
864 // its text if successful.
865 void tryParseJSRegexLiteral() {
866 FormatToken *RegexToken = Tokens.back();
867 if (!RegexToken->isOneOf(tok::slash, tok::slashequal))
870 FormatToken *Prev = nullptr;
871 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) {
872 // NB: Because previous pointers are not initialized yet, this cannot use
873 // Token.getPreviousNonComment.
874 if ((*I)->isNot(tok::comment)) {
880 if (!canPrecedeRegexLiteral(Prev))
883 // 'Manually' lex ahead in the current file buffer.
884 const char *Offset = Lex->getBufferLocation();
885 const char *RegexBegin = Offset - RegexToken->TokenText.size();
886 StringRef Buffer = Lex->getBuffer();
887 bool InCharacterClass = false;
888 bool HaveClosingSlash = false;
889 for (; !HaveClosingSlash && Offset != Buffer.end(); ++Offset) {
890 // Regular expressions are terminated with a '/', which can only be
891 // escaped using '\' or a character class between '[' and ']'.
892 // See http://www.ecma-international.org/ecma-262/5.1/#sec-7.8.5.
895 // Skip the escaped character.
899 InCharacterClass = true;
902 InCharacterClass = false;
905 if (!InCharacterClass)
906 HaveClosingSlash = true;
911 RegexToken->Type = TT_RegexLiteral;
912 // Treat regex literals like other string_literals.
913 RegexToken->Tok.setKind(tok::string_literal);
914 RegexToken->TokenText = StringRef(RegexBegin, Offset - RegexBegin);
915 RegexToken->ColumnWidth = RegexToken->TokenText.size();
917 resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset)));
920 bool tryMergeTemplateString() {
921 if (Tokens.size() < 2)
924 FormatToken *EndBacktick = Tokens.back();
925 // Backticks get lexed as tok::unknown tokens. If a template string contains
926 // a comment start, it gets lexed as a tok::comment, or tok::unknown if
928 if (!EndBacktick->isOneOf(tok::comment, tok::string_literal,
929 tok::char_constant, tok::unknown))
931 size_t CommentBacktickPos = EndBacktick->TokenText.find('`');
932 // Unknown token that's not actually a backtick, or a comment that doesn't
933 // contain a backtick.
934 if (CommentBacktickPos == StringRef::npos)
937 unsigned TokenCount = 0;
938 bool IsMultiline = false;
939 unsigned EndColumnInFirstLine =
940 EndBacktick->OriginalColumn + EndBacktick->ColumnWidth;
941 for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; I++) {
943 if (I[0]->IsMultiline)
946 // If there was a preceding template string, this must be the start of a
947 // template string, not the end.
948 if (I[0]->is(TT_TemplateString))
951 if (I[0]->isNot(tok::unknown) || I[0]->TokenText != "`") {
952 // Keep track of the rhs offset of the last token to wrap across lines -
953 // its the rhs offset of the first line of the template string, used to
954 // determine its width.
955 if (I[0]->IsMultiline)
956 EndColumnInFirstLine = I[0]->OriginalColumn + I[0]->ColumnWidth;
957 // If the token has newlines, the token before it (if it exists) is the
958 // rhs end of the previous line.
959 if (I[0]->NewlinesBefore > 0 && (I + 1 != E)) {
960 EndColumnInFirstLine = I[1]->OriginalColumn + I[1]->ColumnWidth;
966 Tokens.resize(Tokens.size() - TokenCount);
967 Tokens.back()->Type = TT_TemplateString;
968 const char *EndOffset =
969 EndBacktick->TokenText.data() + 1 + CommentBacktickPos;
970 if (CommentBacktickPos != 0) {
971 // If the backtick was not the first character (e.g. in a comment),
972 // re-lex after the backtick position.
973 SourceLocation Loc = EndBacktick->Tok.getLocation();
974 resetLexer(SourceMgr.getFileOffset(Loc) + CommentBacktickPos + 1);
976 Tokens.back()->TokenText =
977 StringRef(Tokens.back()->TokenText.data(),
978 EndOffset - Tokens.back()->TokenText.data());
980 unsigned EndOriginalColumn = EndBacktick->OriginalColumn;
981 if (EndOriginalColumn == 0) {
982 SourceLocation Loc = EndBacktick->Tok.getLocation();
983 EndOriginalColumn = SourceMgr.getSpellingColumnNumber(Loc);
985 // If the ` is further down within the token (e.g. in a comment).
986 EndOriginalColumn += CommentBacktickPos;
989 // ColumnWidth is from backtick to last token in line.
990 // LastLineColumnWidth is 0 to backtick.
993 Tokens.back()->ColumnWidth =
994 EndColumnInFirstLine - Tokens.back()->OriginalColumn;
995 // +1 for the ` itself.
996 Tokens.back()->LastLineColumnWidth = EndOriginalColumn + 1;
997 Tokens.back()->IsMultiline = true;
999 // Token simply spans from start to end, +1 for the ` itself.
1000 Tokens.back()->ColumnWidth =
1001 EndOriginalColumn - Tokens.back()->OriginalColumn + 1;
1008 bool tryMerge_TMacro() {
1009 if (Tokens.size() < 4)
1011 FormatToken *Last = Tokens.back();
1012 if (!Last->is(tok::r_paren))
1015 FormatToken *String = Tokens[Tokens.size() - 2];
1016 if (!String->is(tok::string_literal) || String->IsMultiline)
1019 if (!Tokens[Tokens.size() - 3]->is(tok::l_paren))
1022 FormatToken *Macro = Tokens[Tokens.size() - 4];
1023 if (Macro->TokenText != "_T")
1026 const char *Start = Macro->TokenText.data();
1027 const char *End = Last->TokenText.data() + Last->TokenText.size();
1028 String->TokenText = StringRef(Start, End - Start);
1029 String->IsFirst = Macro->IsFirst;
1030 String->LastNewlineOffset = Macro->LastNewlineOffset;
1031 String->WhitespaceRange = Macro->WhitespaceRange;
1032 String->OriginalColumn = Macro->OriginalColumn;
1033 String->ColumnWidth = encoding::columnWidthWithTabs(
1034 String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding);
1035 String->NewlinesBefore = Macro->NewlinesBefore;
1036 String->HasUnescapedNewline = Macro->HasUnescapedNewline;
1041 Tokens.back() = String;
1045 bool tryMergeConflictMarkers() {
1046 if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof))
1049 // Conflict lines look like:
1050 // <marker> <text from the vcs>
1052 // >>>>>>> /file/in/file/system at revision 1234
1054 // We merge all tokens in a line that starts with a conflict marker
1055 // into a single token with a special token type that the unwrapped line
1056 // parser will use to correctly rebuild the underlying code.
1059 // Get the position of the first token in the line.
1060 unsigned FirstInLineOffset;
1061 std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc(
1062 Tokens[FirstInLineIndex]->getStartOfNonWhitespace());
1063 StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer();
1064 // Calculate the offset of the start of the current line.
1065 auto LineOffset = Buffer.rfind('\n', FirstInLineOffset);
1066 if (LineOffset == StringRef::npos) {
1072 auto FirstSpace = Buffer.find_first_of(" \n", LineOffset);
1073 StringRef LineStart;
1074 if (FirstSpace == StringRef::npos) {
1075 LineStart = Buffer.substr(LineOffset);
1077 LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset);
1080 TokenType Type = TT_Unknown;
1081 if (LineStart == "<<<<<<<" || LineStart == ">>>>") {
1082 Type = TT_ConflictStart;
1083 } else if (LineStart == "|||||||" || LineStart == "=======" ||
1084 LineStart == "====") {
1085 Type = TT_ConflictAlternative;
1086 } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") {
1087 Type = TT_ConflictEnd;
1090 if (Type != TT_Unknown) {
1091 FormatToken *Next = Tokens.back();
1093 Tokens.resize(FirstInLineIndex + 1);
1094 // We do not need to build a complete token here, as we will skip it
1095 // during parsing anyway (as we must not touch whitespace around conflict
1097 Tokens.back()->Type = Type;
1098 Tokens.back()->Tok.setKind(tok::kw___unknown_anytype);
1100 Tokens.push_back(Next);
1107 FormatToken *getStashedToken() {
1108 // Create a synthesized second '>' or '<' token.
1109 Token Tok = FormatTok->Tok;
1110 StringRef TokenText = FormatTok->TokenText;
1112 unsigned OriginalColumn = FormatTok->OriginalColumn;
1113 FormatTok = new (Allocator.Allocate()) FormatToken;
1114 FormatTok->Tok = Tok;
1115 SourceLocation TokLocation =
1116 FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1);
1117 FormatTok->Tok.setLocation(TokLocation);
1118 FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation);
1119 FormatTok->TokenText = TokenText;
1120 FormatTok->ColumnWidth = 1;
1121 FormatTok->OriginalColumn = OriginalColumn + 1;
1126 FormatToken *getNextToken() {
1127 if (GreaterStashed) {
1128 GreaterStashed = false;
1129 return getStashedToken();
1132 LessStashed = false;
1133 return getStashedToken();
1136 FormatTok = new (Allocator.Allocate()) FormatToken;
1137 readRawToken(*FormatTok);
1138 SourceLocation WhitespaceStart =
1139 FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace);
1140 FormatTok->IsFirst = IsFirstToken;
1141 IsFirstToken = false;
1143 // Consume and record whitespace until we find a significant token.
1144 unsigned WhitespaceLength = TrailingWhitespace;
1145 while (FormatTok->Tok.is(tok::unknown)) {
1146 StringRef Text = FormatTok->TokenText;
1147 auto EscapesNewline = [&](int pos) {
1148 // A '\r' here is just part of '\r\n'. Skip it.
1149 if (pos >= 0 && Text[pos] == '\r')
1151 // See whether there is an odd number of '\' before this.
1153 for (; pos >= 0; --pos, ++count)
1154 if (Text[pos] != '\\')
1158 // FIXME: This miscounts tok:unknown tokens that are not just
1159 // whitespace, e.g. a '`' character.
1160 for (int i = 0, e = Text.size(); i != e; ++i) {
1163 ++FormatTok->NewlinesBefore;
1164 FormatTok->HasUnescapedNewline = !EscapesNewline(i - 1);
1165 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
1169 FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
1180 Column += Style.TabWidth - Column % Style.TabWidth;
1183 if (i + 1 == e || (Text[i + 1] != '\r' && Text[i + 1] != '\n'))
1184 FormatTok->Type = TT_ImplicitStringLiteral;
1187 FormatTok->Type = TT_ImplicitStringLiteral;
1192 if (FormatTok->is(TT_ImplicitStringLiteral))
1194 WhitespaceLength += FormatTok->Tok.getLength();
1196 readRawToken(*FormatTok);
1199 // In case the token starts with escaped newlines, we want to
1200 // take them into account as whitespace - this pattern is quite frequent
1201 // in macro definitions.
1202 // FIXME: Add a more explicit test.
1203 while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' &&
1204 FormatTok->TokenText[1] == '\n') {
1205 ++FormatTok->NewlinesBefore;
1206 WhitespaceLength += 2;
1207 FormatTok->LastNewlineOffset = 2;
1209 FormatTok->TokenText = FormatTok->TokenText.substr(2);
1212 FormatTok->WhitespaceRange = SourceRange(
1213 WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength));
1215 FormatTok->OriginalColumn = Column;
1217 TrailingWhitespace = 0;
1218 if (FormatTok->Tok.is(tok::comment)) {
1219 // FIXME: Add the trimmed whitespace to Column.
1220 StringRef UntrimmedText = FormatTok->TokenText;
1221 FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f");
1222 TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size();
1223 } else if (FormatTok->Tok.is(tok::raw_identifier)) {
1224 IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText);
1225 FormatTok->Tok.setIdentifierInfo(&Info);
1226 FormatTok->Tok.setKind(Info.getTokenID());
1227 if (Style.Language == FormatStyle::LK_Java &&
1228 FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete)) {
1229 FormatTok->Tok.setKind(tok::identifier);
1230 FormatTok->Tok.setIdentifierInfo(nullptr);
1232 } else if (FormatTok->Tok.is(tok::greatergreater)) {
1233 FormatTok->Tok.setKind(tok::greater);
1234 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
1235 GreaterStashed = true;
1236 } else if (FormatTok->Tok.is(tok::lessless)) {
1237 FormatTok->Tok.setKind(tok::less);
1238 FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
1242 // Now FormatTok is the next non-whitespace token.
1244 StringRef Text = FormatTok->TokenText;
1245 size_t FirstNewlinePos = Text.find('\n');
1246 if (FirstNewlinePos == StringRef::npos) {
1247 // FIXME: ColumnWidth actually depends on the start column, we need to
1248 // take this into account when the token is moved.
1249 FormatTok->ColumnWidth =
1250 encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding);
1251 Column += FormatTok->ColumnWidth;
1253 FormatTok->IsMultiline = true;
1254 // FIXME: ColumnWidth actually depends on the start column, we need to
1255 // take this into account when the token is moved.
1256 FormatTok->ColumnWidth = encoding::columnWidthWithTabs(
1257 Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding);
1259 // The last line of the token always starts in column 0.
1260 // Thus, the length can be precomputed even in the presence of tabs.
1261 FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs(
1262 Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth,
1264 Column = FormatTok->LastLineColumnWidth;
1267 if (Style.Language == FormatStyle::LK_Cpp) {
1268 if (!(Tokens.size() > 0 && Tokens.back()->Tok.getIdentifierInfo() &&
1269 Tokens.back()->Tok.getIdentifierInfo()->getPPKeywordID() ==
1271 std::find(ForEachMacros.begin(), ForEachMacros.end(),
1272 FormatTok->Tok.getIdentifierInfo()) != ForEachMacros.end()) {
1273 FormatTok->Type = TT_ForEachMacro;
1274 } else if (FormatTok->is(tok::identifier)) {
1275 if (MacroBlockBeginRegex.match(Text)) {
1276 FormatTok->Type = TT_MacroBlockBegin;
1277 } else if (MacroBlockEndRegex.match(Text)) {
1278 FormatTok->Type = TT_MacroBlockEnd;
1286 FormatToken *FormatTok;
1288 bool GreaterStashed, LessStashed;
1290 unsigned TrailingWhitespace;
1291 std::unique_ptr<Lexer> Lex;
1292 SourceManager &SourceMgr;
1295 IdentifierTable IdentTable;
1296 AdditionalKeywords Keywords;
1297 encoding::Encoding Encoding;
1298 llvm::SpecificBumpPtrAllocator<FormatToken> Allocator;
1299 // Index (in 'Tokens') of the last token that starts a new line.
1300 unsigned FirstInLineIndex;
1301 SmallVector<FormatToken *, 16> Tokens;
1302 SmallVector<IdentifierInfo *, 8> ForEachMacros;
1304 bool FormattingDisabled;
1306 llvm::Regex MacroBlockBeginRegex;
1307 llvm::Regex MacroBlockEndRegex;
1309 void readRawToken(FormatToken &Tok) {
1310 Lex->LexFromRawLexer(Tok.Tok);
1311 Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()),
1312 Tok.Tok.getLength());
1313 // For formatting, treat unterminated string literals like normal string
1315 if (Tok.is(tok::unknown)) {
1316 if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') {
1317 Tok.Tok.setKind(tok::string_literal);
1318 Tok.IsUnterminatedLiteral = true;
1319 } else if (Style.Language == FormatStyle::LK_JavaScript &&
1320 Tok.TokenText == "''") {
1321 Tok.Tok.setKind(tok::char_constant);
1325 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" ||
1326 Tok.TokenText == "/* clang-format on */")) {
1327 FormattingDisabled = false;
1330 Tok.Finalized = FormattingDisabled;
1332 if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" ||
1333 Tok.TokenText == "/* clang-format off */")) {
1334 FormattingDisabled = true;
1338 void resetLexer(unsigned Offset) {
1339 StringRef Buffer = SourceMgr.getBufferData(ID);
1340 Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID),
1341 getFormattingLangOpts(Style), Buffer.begin(),
1342 Buffer.begin() + Offset, Buffer.end()));
1343 Lex->SetKeepWhitespaceMode(true);
1344 TrailingWhitespace = 0;
1348 static StringRef getLanguageName(FormatStyle::LanguageKind Language) {
1350 case FormatStyle::LK_Cpp:
1352 case FormatStyle::LK_Java:
1354 case FormatStyle::LK_JavaScript:
1355 return "JavaScript";
1356 case FormatStyle::LK_Proto:
1363 class Formatter : public UnwrappedLineConsumer {
1365 Formatter(const FormatStyle &Style, SourceManager &SourceMgr, FileID ID,
1366 ArrayRef<CharSourceRange> Ranges)
1367 : Style(Style), ID(ID), SourceMgr(SourceMgr),
1368 Whitespaces(SourceMgr, Style,
1369 inputUsesCRLF(SourceMgr.getBufferData(ID))),
1370 Ranges(Ranges.begin(), Ranges.end()), UnwrappedLines(1),
1371 Encoding(encoding::detectEncoding(SourceMgr.getBufferData(ID))) {
1372 DEBUG(llvm::dbgs() << "File encoding: "
1373 << (Encoding == encoding::Encoding_UTF8 ? "UTF8"
1376 DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language)
1380 tooling::Replacements format(bool *IncompleteFormat) {
1381 tooling::Replacements Result;
1382 FormatTokenLexer Tokens(SourceMgr, ID, Style, Encoding);
1384 UnwrappedLineParser Parser(Style, Tokens.getKeywords(), Tokens.lex(),
1387 assert(UnwrappedLines.rbegin()->empty());
1388 for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE;
1390 DEBUG(llvm::dbgs() << "Run " << Run << "...\n");
1391 SmallVector<AnnotatedLine *, 16> AnnotatedLines;
1392 for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) {
1393 AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i]));
1395 tooling::Replacements RunResult =
1396 format(AnnotatedLines, Tokens, IncompleteFormat);
1398 llvm::dbgs() << "Replacements for run " << Run << ":\n";
1399 for (tooling::Replacements::iterator I = RunResult.begin(),
1400 E = RunResult.end();
1402 llvm::dbgs() << I->toString() << "\n";
1405 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1406 delete AnnotatedLines[i];
1408 Result.insert(RunResult.begin(), RunResult.end());
1409 Whitespaces.reset();
1414 tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1415 FormatTokenLexer &Tokens,
1416 bool *IncompleteFormat) {
1417 TokenAnnotator Annotator(Style, Tokens.getKeywords());
1418 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1419 Annotator.annotate(*AnnotatedLines[i]);
1421 deriveLocalStyle(AnnotatedLines);
1422 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1423 Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
1425 computeAffectedLines(AnnotatedLines.begin(), AnnotatedLines.end());
1427 Annotator.setCommentLineLevels(AnnotatedLines);
1428 ContinuationIndenter Indenter(Style, Tokens.getKeywords(), SourceMgr,
1429 Whitespaces, Encoding,
1430 BinPackInconclusiveFunctions);
1431 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
1433 .format(AnnotatedLines);
1434 return Whitespaces.generateReplacements();
1438 // Determines which lines are affected by the SourceRanges given as input.
1439 // Returns \c true if at least one line between I and E or one of their
1440 // children is affected.
1441 bool computeAffectedLines(SmallVectorImpl<AnnotatedLine *>::iterator I,
1442 SmallVectorImpl<AnnotatedLine *>::iterator E) {
1443 bool SomeLineAffected = false;
1444 const AnnotatedLine *PreviousLine = nullptr;
1446 AnnotatedLine *Line = *I;
1447 Line->LeadingEmptyLinesAffected = affectsLeadingEmptyLines(*Line->First);
1449 // If a line is part of a preprocessor directive, it needs to be formatted
1450 // if any token within the directive is affected.
1451 if (Line->InPPDirective) {
1452 FormatToken *Last = Line->Last;
1453 SmallVectorImpl<AnnotatedLine *>::iterator PPEnd = I + 1;
1454 while (PPEnd != E && !(*PPEnd)->First->HasUnescapedNewline) {
1455 Last = (*PPEnd)->Last;
1459 if (affectsTokenRange(*Line->First, *Last,
1460 /*IncludeLeadingNewlines=*/false)) {
1461 SomeLineAffected = true;
1462 markAllAsAffected(I, PPEnd);
1468 if (nonPPLineAffected(Line, PreviousLine))
1469 SomeLineAffected = true;
1471 PreviousLine = Line;
1474 return SomeLineAffected;
1477 // Determines whether 'Line' is affected by the SourceRanges given as input.
1478 // Returns \c true if line or one if its children is affected.
1479 bool nonPPLineAffected(AnnotatedLine *Line,
1480 const AnnotatedLine *PreviousLine) {
1481 bool SomeLineAffected = false;
1482 Line->ChildrenAffected =
1483 computeAffectedLines(Line->Children.begin(), Line->Children.end());
1484 if (Line->ChildrenAffected)
1485 SomeLineAffected = true;
1487 // Stores whether one of the line's tokens is directly affected.
1488 bool SomeTokenAffected = false;
1489 // Stores whether we need to look at the leading newlines of the next token
1490 // in order to determine whether it was affected.
1491 bool IncludeLeadingNewlines = false;
1493 // Stores whether the first child line of any of this line's tokens is
1495 bool SomeFirstChildAffected = false;
1497 for (FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) {
1498 // Determine whether 'Tok' was affected.
1499 if (affectsTokenRange(*Tok, *Tok, IncludeLeadingNewlines))
1500 SomeTokenAffected = true;
1502 // Determine whether the first child of 'Tok' was affected.
1503 if (!Tok->Children.empty() && Tok->Children.front()->Affected)
1504 SomeFirstChildAffected = true;
1506 IncludeLeadingNewlines = Tok->Children.empty();
1509 // Was this line moved, i.e. has it previously been on the same line as an
1511 bool LineMoved = PreviousLine && PreviousLine->Affected &&
1512 Line->First->NewlinesBefore == 0;
1514 bool IsContinuedComment =
1515 Line->First->is(tok::comment) && Line->First->Next == nullptr &&
1516 Line->First->NewlinesBefore < 2 && PreviousLine &&
1517 PreviousLine->Affected && PreviousLine->Last->is(tok::comment);
1519 if (SomeTokenAffected || SomeFirstChildAffected || LineMoved ||
1520 IsContinuedComment) {
1521 Line->Affected = true;
1522 SomeLineAffected = true;
1524 return SomeLineAffected;
1527 // Marks all lines between I and E as well as all their children as affected.
1528 void markAllAsAffected(SmallVectorImpl<AnnotatedLine *>::iterator I,
1529 SmallVectorImpl<AnnotatedLine *>::iterator E) {
1531 (*I)->Affected = true;
1532 markAllAsAffected((*I)->Children.begin(), (*I)->Children.end());
1537 // Returns true if the range from 'First' to 'Last' intersects with one of the
1539 bool affectsTokenRange(const FormatToken &First, const FormatToken &Last,
1540 bool IncludeLeadingNewlines) {
1541 SourceLocation Start = First.WhitespaceRange.getBegin();
1542 if (!IncludeLeadingNewlines)
1543 Start = Start.getLocWithOffset(First.LastNewlineOffset);
1544 SourceLocation End = Last.getStartOfNonWhitespace();
1545 End = End.getLocWithOffset(Last.TokenText.size());
1546 CharSourceRange Range = CharSourceRange::getCharRange(Start, End);
1547 return affectsCharSourceRange(Range);
1550 // Returns true if one of the input ranges intersect the leading empty lines
1552 bool affectsLeadingEmptyLines(const FormatToken &Tok) {
1553 CharSourceRange EmptyLineRange = CharSourceRange::getCharRange(
1554 Tok.WhitespaceRange.getBegin(),
1555 Tok.WhitespaceRange.getBegin().getLocWithOffset(Tok.LastNewlineOffset));
1556 return affectsCharSourceRange(EmptyLineRange);
1559 // Returns true if 'Range' intersects with one of the input ranges.
1560 bool affectsCharSourceRange(const CharSourceRange &Range) {
1561 for (SmallVectorImpl<CharSourceRange>::const_iterator I = Ranges.begin(),
1564 if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), I->getBegin()) &&
1565 !SourceMgr.isBeforeInTranslationUnit(I->getEnd(), Range.getBegin()))
1571 static bool inputUsesCRLF(StringRef Text) {
1572 return Text.count('\r') * 2 > Text.count('\n');
1576 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1577 for (const AnnotatedLine* Line : Lines) {
1578 if (hasCpp03IncompatibleFormat(Line->Children))
1580 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
1581 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
1582 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
1584 if (Tok->is(TT_TemplateCloser) &&
1585 Tok->Previous->is(TT_TemplateCloser))
1593 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1594 int AlignmentDiff = 0;
1595 for (const AnnotatedLine* Line : Lines) {
1596 AlignmentDiff += countVariableAlignments(Line->Children);
1597 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
1598 if (!Tok->is(TT_PointerOrReference))
1601 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
1602 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
1603 Tok->Next->WhitespaceRange.getEnd();
1604 if (SpaceBefore && !SpaceAfter)
1606 if (!SpaceBefore && SpaceAfter)
1610 return AlignmentDiff;
1614 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1615 bool HasBinPackedFunction = false;
1616 bool HasOnePerLineFunction = false;
1617 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1618 if (!AnnotatedLines[i]->First->Next)
1620 FormatToken *Tok = AnnotatedLines[i]->First->Next;
1622 if (Tok->PackingKind == PPK_BinPacked)
1623 HasBinPackedFunction = true;
1624 if (Tok->PackingKind == PPK_OnePerLine)
1625 HasOnePerLineFunction = true;
1630 if (Style.DerivePointerAlignment)
1631 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
1632 ? FormatStyle::PAS_Left
1633 : FormatStyle::PAS_Right;
1634 if (Style.Standard == FormatStyle::LS_Auto)
1635 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
1636 ? FormatStyle::LS_Cpp11
1637 : FormatStyle::LS_Cpp03;
1638 BinPackInconclusiveFunctions =
1639 HasBinPackedFunction || !HasOnePerLineFunction;
1642 void consumeUnwrappedLine(const UnwrappedLine &TheLine) override {
1643 assert(!UnwrappedLines.empty());
1644 UnwrappedLines.back().push_back(TheLine);
1647 void finishRun() override {
1648 UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>());
1653 SourceManager &SourceMgr;
1654 WhitespaceManager Whitespaces;
1655 SmallVector<CharSourceRange, 8> Ranges;
1656 SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines;
1658 encoding::Encoding Encoding;
1659 bool BinPackInconclusiveFunctions;
1662 struct IncludeDirective {
1669 } // end anonymous namespace
1671 // Determines whether 'Ranges' intersects with ('Start', 'End').
1672 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1674 for (auto Range : Ranges) {
1675 if (Range.getOffset() < End &&
1676 Range.getOffset() + Range.getLength() > Start)
1682 // Sorts a block of includes given by 'Includes' alphabetically adding the
1683 // necessary replacement to 'Replaces'. 'Includes' must be in strict source
1685 static void sortIncludes(const FormatStyle &Style,
1686 const SmallVectorImpl<IncludeDirective> &Includes,
1687 ArrayRef<tooling::Range> Ranges, StringRef FileName,
1688 tooling::Replacements &Replaces) {
1689 if (!affectsRange(Ranges, Includes.front().Offset,
1690 Includes.back().Offset + Includes.back().Text.size()))
1692 SmallVector<unsigned, 16> Indices;
1693 for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1694 Indices.push_back(i);
1695 std::sort(Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1696 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1697 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1700 // If the #includes are out of order, we generate a single replacement fixing
1701 // the entire block. Otherwise, no replacement is generated.
1702 bool OutOfOrder = false;
1703 for (unsigned i = 1, e = Indices.size(); i != e; ++i) {
1704 if (Indices[i] != i) {
1712 std::string result = Includes[Indices[0]].Text;
1713 for (unsigned i = 1, e = Indices.size(); i != e; ++i) {
1715 result += Includes[Indices[i]].Text;
1718 // Sorting #includes shouldn't change their total number of characters.
1719 // This would otherwise mess up 'Ranges'.
1720 assert(result.size() ==
1721 Includes.back().Offset + Includes.back().Text.size() -
1722 Includes.front().Offset);
1724 Replaces.insert(tooling::Replacement(FileName, Includes.front().Offset,
1725 result.size(), result));
1728 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1729 ArrayRef<tooling::Range> Ranges,
1730 StringRef FileName) {
1731 tooling::Replacements Replaces;
1733 unsigned SearchFrom = 0;
1734 llvm::Regex IncludeRegex(
1735 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))");
1736 SmallVector<StringRef, 4> Matches;
1737 SmallVector<IncludeDirective, 16> IncludesInBlock;
1739 // In compiled files, consider the first #include to be the main #include of
1740 // the file if it is not a system #include. This ensures that the header
1741 // doesn't have hidden dependencies
1742 // (http://llvm.org/docs/CodingStandards.html#include-style).
1744 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1745 // cases where the first #include is unlikely to be the main header.
1746 bool LookForMainHeader = FileName.endswith(".c") ||
1747 FileName.endswith(".cc") ||
1748 FileName.endswith(".cpp")||
1749 FileName.endswith(".c++")||
1750 FileName.endswith(".cxx") ||
1751 FileName.endswith(".m")||
1752 FileName.endswith(".mm");
1754 // Create pre-compiled regular expressions for the #include categories.
1755 SmallVector<llvm::Regex, 4> CategoryRegexs;
1756 for (const auto &Category : Style.IncludeCategories)
1757 CategoryRegexs.emplace_back(Category.Regex);
1760 auto Pos = Code.find('\n', SearchFrom);
1762 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1763 if (!Line.endswith("\\")) {
1764 if (IncludeRegex.match(Line, &Matches)) {
1765 StringRef IncludeName = Matches[2];
1767 if (LookForMainHeader && !IncludeName.startswith("<")) {
1770 Category = UINT_MAX;
1771 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i) {
1772 if (CategoryRegexs[i].match(IncludeName)) {
1773 Category = Style.IncludeCategories[i].Priority;
1778 LookForMainHeader = false;
1779 IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1780 } else if (!IncludesInBlock.empty()) {
1781 sortIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces);
1782 IncludesInBlock.clear();
1786 if (Pos == StringRef::npos || Pos + 1 == Code.size())
1788 SearchFrom = Pos + 1;
1790 if (!IncludesInBlock.empty())
1791 sortIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces);
1795 tooling::Replacements reformat(const FormatStyle &Style,
1796 SourceManager &SourceMgr, FileID ID,
1797 ArrayRef<CharSourceRange> Ranges,
1798 bool *IncompleteFormat) {
1799 FormatStyle Expanded = expandPresets(Style);
1800 if (Expanded.DisableFormat)
1801 return tooling::Replacements();
1802 Formatter formatter(Expanded, SourceMgr, ID, Ranges);
1803 return formatter.format(IncompleteFormat);
1806 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1807 ArrayRef<tooling::Range> Ranges,
1808 StringRef FileName, bool *IncompleteFormat) {
1809 if (Style.DisableFormat)
1810 return tooling::Replacements();
1812 IntrusiveRefCntPtr<vfs::InMemoryFileSystem> InMemoryFileSystem(
1813 new vfs::InMemoryFileSystem);
1814 FileManager Files(FileSystemOptions(), InMemoryFileSystem);
1815 DiagnosticsEngine Diagnostics(
1816 IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs),
1817 new DiagnosticOptions);
1818 SourceManager SourceMgr(Diagnostics, Files);
1819 InMemoryFileSystem->addFile(FileName, 0,
1820 llvm::MemoryBuffer::getMemBuffer(Code, FileName));
1821 FileID ID = SourceMgr.createFileID(Files.getFile(FileName), SourceLocation(),
1822 clang::SrcMgr::C_User);
1823 SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID);
1824 std::vector<CharSourceRange> CharRanges;
1825 for (const tooling::Range &Range : Ranges) {
1826 SourceLocation Start = StartOfFile.getLocWithOffset(Range.getOffset());
1827 SourceLocation End = Start.getLocWithOffset(Range.getLength());
1828 CharRanges.push_back(CharSourceRange::getCharRange(Start, End));
1830 return reformat(Style, SourceMgr, ID, CharRanges, IncompleteFormat);
1833 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1834 LangOptions LangOpts;
1835 LangOpts.CPlusPlus = 1;
1836 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1837 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1838 LangOpts.LineComment = 1;
1839 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp;
1840 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1844 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally.
1845 LangOpts.DeclSpecKeyword = 1; // To get __declspec.
1849 const char *StyleOptionHelpDescription =
1850 "Coding style, currently supports:\n"
1851 " LLVM, Google, Chromium, Mozilla, WebKit.\n"
1852 "Use -style=file to load style configuration from\n"
1853 ".clang-format file located in one of the parent\n"
1854 "directories of the source file (or current\n"
1855 "directory for stdin).\n"
1856 "Use -style=\"{key: value, ...}\" to set specific\n"
1857 "parameters, e.g.:\n"
1858 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1860 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1861 if (FileName.endswith(".java")) {
1862 return FormatStyle::LK_Java;
1863 } else if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts")) {
1864 // JavaScript or TypeScript.
1865 return FormatStyle::LK_JavaScript;
1866 } else if (FileName.endswith_lower(".proto") ||
1867 FileName.endswith_lower(".protodevel")) {
1868 return FormatStyle::LK_Proto;
1870 return FormatStyle::LK_Cpp;
1873 FormatStyle getStyle(StringRef StyleName, StringRef FileName,
1874 StringRef FallbackStyle) {
1875 FormatStyle Style = getLLVMStyle();
1876 Style.Language = getLanguageByFileName(FileName);
1877 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) {
1878 llvm::errs() << "Invalid fallback style \"" << FallbackStyle
1879 << "\" using LLVM style\n";
1883 if (StyleName.startswith("{")) {
1884 // Parse YAML/JSON style from the command line.
1885 if (std::error_code ec = parseConfiguration(StyleName, &Style)) {
1886 llvm::errs() << "Error parsing -style: " << ec.message() << ", using "
1887 << FallbackStyle << " style\n";
1892 if (!StyleName.equals_lower("file")) {
1893 if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1894 llvm::errs() << "Invalid value for -style, using " << FallbackStyle
1899 // Look for .clang-format/_clang-format file in the file's parent directories.
1900 SmallString<128> UnsuitableConfigFiles;
1901 SmallString<128> Path(FileName);
1902 llvm::sys::fs::make_absolute(Path);
1903 for (StringRef Directory = Path; !Directory.empty();
1904 Directory = llvm::sys::path::parent_path(Directory)) {
1905 if (!llvm::sys::fs::is_directory(Directory))
1907 SmallString<128> ConfigFile(Directory);
1909 llvm::sys::path::append(ConfigFile, ".clang-format");
1910 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1911 bool IsFile = false;
1912 // Ignore errors from is_regular_file: we only need to know if we can read
1914 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile);
1917 // Try _clang-format too, since dotfiles are not commonly used on Windows.
1918 ConfigFile = Directory;
1919 llvm::sys::path::append(ConfigFile, "_clang-format");
1920 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1921 llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile);
1925 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
1926 llvm::MemoryBuffer::getFile(ConfigFile.c_str());
1927 if (std::error_code EC = Text.getError()) {
1928 llvm::errs() << EC.message() << "\n";
1931 if (std::error_code ec =
1932 parseConfiguration(Text.get()->getBuffer(), &Style)) {
1933 if (ec == ParseError::Unsuitable) {
1934 if (!UnsuitableConfigFiles.empty())
1935 UnsuitableConfigFiles.append(", ");
1936 UnsuitableConfigFiles.append(ConfigFile);
1939 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message()
1943 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
1947 if (!UnsuitableConfigFiles.empty()) {
1948 llvm::errs() << "Configuration file(s) do(es) not support "
1949 << getLanguageName(Style.Language) << ": "
1950 << UnsuitableConfigFiles << "\n";
1955 } // namespace format
1956 } // namespace clang