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 "AffectedRangeManager.h"
18 #include "ContinuationIndenter.h"
19 #include "FormatTokenLexer.h"
20 #include "NamespaceEndCommentsFixer.h"
21 #include "SortJavaScriptImports.h"
22 #include "TokenAnalyzer.h"
23 #include "TokenAnnotator.h"
24 #include "UnwrappedLineFormatter.h"
25 #include "UnwrappedLineParser.h"
26 #include "WhitespaceManager.h"
27 #include "clang/Basic/Diagnostic.h"
28 #include "clang/Basic/DiagnosticOptions.h"
29 #include "clang/Basic/SourceManager.h"
30 #include "clang/Basic/VirtualFileSystem.h"
31 #include "clang/Lex/Lexer.h"
32 #include "llvm/ADT/STLExtras.h"
33 #include "llvm/Support/Allocator.h"
34 #include "llvm/Support/Debug.h"
35 #include "llvm/Support/Path.h"
36 #include "llvm/Support/Regex.h"
37 #include "llvm/Support/YAMLTraits.h"
42 #define DEBUG_TYPE "format-formatter"
44 using clang::format::FormatStyle;
46 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string)
47 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory)
51 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
52 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
53 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
54 IO.enumCase(Value, "Java", FormatStyle::LK_Java);
55 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
56 IO.enumCase(Value, "ObjC", FormatStyle::LK_ObjC);
57 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
58 IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen);
62 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
63 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
64 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
65 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
66 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
67 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
68 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
72 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
73 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
74 IO.enumCase(Value, "Never", FormatStyle::UT_Never);
75 IO.enumCase(Value, "false", FormatStyle::UT_Never);
76 IO.enumCase(Value, "Always", FormatStyle::UT_Always);
77 IO.enumCase(Value, "true", FormatStyle::UT_Always);
78 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
79 IO.enumCase(Value, "ForContinuationAndIndentation",
80 FormatStyle::UT_ForContinuationAndIndentation);
84 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> {
85 static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) {
86 IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave);
87 IO.enumCase(Value, "Single", FormatStyle::JSQS_Single);
88 IO.enumCase(Value, "Double", FormatStyle::JSQS_Double);
92 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
93 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
94 IO.enumCase(Value, "None", FormatStyle::SFS_None);
95 IO.enumCase(Value, "false", FormatStyle::SFS_None);
96 IO.enumCase(Value, "All", FormatStyle::SFS_All);
97 IO.enumCase(Value, "true", FormatStyle::SFS_All);
98 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
99 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
103 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
104 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
105 IO.enumCase(Value, "All", FormatStyle::BOS_All);
106 IO.enumCase(Value, "true", FormatStyle::BOS_All);
107 IO.enumCase(Value, "None", FormatStyle::BOS_None);
108 IO.enumCase(Value, "false", FormatStyle::BOS_None);
109 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
113 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
114 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
115 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
116 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
117 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla);
118 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
119 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
120 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
121 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit);
122 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom);
127 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> {
128 static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) {
129 IO.enumCase(Value, "None", FormatStyle::RTBS_None);
130 IO.enumCase(Value, "All", FormatStyle::RTBS_All);
131 IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel);
132 IO.enumCase(Value, "TopLevelDefinitions",
133 FormatStyle::RTBS_TopLevelDefinitions);
134 IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions);
139 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> {
141 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) {
142 IO.enumCase(Value, "None", FormatStyle::DRTBS_None);
143 IO.enumCase(Value, "All", FormatStyle::DRTBS_All);
144 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel);
146 // For backward compatibility.
147 IO.enumCase(Value, "false", FormatStyle::DRTBS_None);
148 IO.enumCase(Value, "true", FormatStyle::DRTBS_All);
153 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
154 static void enumeration(IO &IO,
155 FormatStyle::NamespaceIndentationKind &Value) {
156 IO.enumCase(Value, "None", FormatStyle::NI_None);
157 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
158 IO.enumCase(Value, "All", FormatStyle::NI_All);
162 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> {
163 static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) {
164 IO.enumCase(Value, "Align", FormatStyle::BAS_Align);
165 IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign);
166 IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak);
168 // For backward compatibility.
169 IO.enumCase(Value, "true", FormatStyle::BAS_Align);
170 IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign);
174 template <> struct ScalarEnumerationTraits<FormatStyle::EscapedNewlineAlignmentStyle> {
175 static void enumeration(IO &IO, FormatStyle::EscapedNewlineAlignmentStyle &Value) {
176 IO.enumCase(Value, "DontAlign", FormatStyle::ENAS_DontAlign);
177 IO.enumCase(Value, "Left", FormatStyle::ENAS_Left);
178 IO.enumCase(Value, "Right", FormatStyle::ENAS_Right);
180 // For backward compatibility.
181 IO.enumCase(Value, "true", FormatStyle::ENAS_Left);
182 IO.enumCase(Value, "false", FormatStyle::ENAS_Right);
186 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
187 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
188 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
189 IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
190 IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
192 // For backward compatibility.
193 IO.enumCase(Value, "true", FormatStyle::PAS_Left);
194 IO.enumCase(Value, "false", FormatStyle::PAS_Right);
199 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
200 static void enumeration(IO &IO,
201 FormatStyle::SpaceBeforeParensOptions &Value) {
202 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
203 IO.enumCase(Value, "ControlStatements",
204 FormatStyle::SBPO_ControlStatements);
205 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
207 // For backward compatibility.
208 IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
209 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
213 template <> struct MappingTraits<FormatStyle> {
214 static void mapping(IO &IO, FormatStyle &Style) {
215 // When reading, read the language first, we need it for getPredefinedStyle.
216 IO.mapOptional("Language", Style.Language);
218 if (IO.outputting()) {
219 StringRef StylesArray[] = {"LLVM", "Google", "Chromium",
220 "Mozilla", "WebKit", "GNU"};
221 ArrayRef<StringRef> Styles(StylesArray);
222 for (size_t i = 0, e = Styles.size(); i < e; ++i) {
223 StringRef StyleName(Styles[i]);
224 FormatStyle PredefinedStyle;
225 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
226 Style == PredefinedStyle) {
227 IO.mapOptional("# BasedOnStyle", StyleName);
232 StringRef BasedOnStyle;
233 IO.mapOptional("BasedOnStyle", BasedOnStyle);
234 if (!BasedOnStyle.empty()) {
235 FormatStyle::LanguageKind OldLanguage = Style.Language;
236 FormatStyle::LanguageKind Language =
237 ((FormatStyle *)IO.getContext())->Language;
238 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
239 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
242 Style.Language = OldLanguage;
246 // For backward compatibility.
247 if (!IO.outputting()) {
248 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlines);
249 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
250 IO.mapOptional("IndentFunctionDeclarationAfterType",
251 Style.IndentWrappedFunctionNames);
252 IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
253 IO.mapOptional("SpaceAfterControlStatementKeyword",
254 Style.SpaceBeforeParens);
257 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
258 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
259 IO.mapOptional("AlignConsecutiveAssignments",
260 Style.AlignConsecutiveAssignments);
261 IO.mapOptional("AlignConsecutiveDeclarations",
262 Style.AlignConsecutiveDeclarations);
263 IO.mapOptional("AlignEscapedNewlines", Style.AlignEscapedNewlines);
264 IO.mapOptional("AlignOperands", Style.AlignOperands);
265 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
266 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
267 Style.AllowAllParametersOfDeclarationOnNextLine);
268 IO.mapOptional("AllowShortBlocksOnASingleLine",
269 Style.AllowShortBlocksOnASingleLine);
270 IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
271 Style.AllowShortCaseLabelsOnASingleLine);
272 IO.mapOptional("AllowShortFunctionsOnASingleLine",
273 Style.AllowShortFunctionsOnASingleLine);
274 IO.mapOptional("AllowShortIfStatementsOnASingleLine",
275 Style.AllowShortIfStatementsOnASingleLine);
276 IO.mapOptional("AllowShortLoopsOnASingleLine",
277 Style.AllowShortLoopsOnASingleLine);
278 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
279 Style.AlwaysBreakAfterDefinitionReturnType);
280 IO.mapOptional("AlwaysBreakAfterReturnType",
281 Style.AlwaysBreakAfterReturnType);
282 // If AlwaysBreakAfterDefinitionReturnType was specified but
283 // AlwaysBreakAfterReturnType was not, initialize the latter from the
284 // former for backwards compatibility.
285 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None &&
286 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) {
287 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All)
288 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
289 else if (Style.AlwaysBreakAfterDefinitionReturnType ==
290 FormatStyle::DRTBS_TopLevel)
291 Style.AlwaysBreakAfterReturnType =
292 FormatStyle::RTBS_TopLevelDefinitions;
295 IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
296 Style.AlwaysBreakBeforeMultilineStrings);
297 IO.mapOptional("AlwaysBreakTemplateDeclarations",
298 Style.AlwaysBreakTemplateDeclarations);
299 IO.mapOptional("BinPackArguments", Style.BinPackArguments);
300 IO.mapOptional("BinPackParameters", Style.BinPackParameters);
301 IO.mapOptional("BraceWrapping", Style.BraceWrapping);
302 IO.mapOptional("BreakBeforeBinaryOperators",
303 Style.BreakBeforeBinaryOperators);
304 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
305 IO.mapOptional("BreakBeforeTernaryOperators",
306 Style.BreakBeforeTernaryOperators);
307 IO.mapOptional("BreakConstructorInitializersBeforeComma",
308 Style.BreakConstructorInitializersBeforeComma);
309 IO.mapOptional("BreakAfterJavaFieldAnnotations",
310 Style.BreakAfterJavaFieldAnnotations);
311 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals);
312 IO.mapOptional("ColumnLimit", Style.ColumnLimit);
313 IO.mapOptional("CommentPragmas", Style.CommentPragmas);
314 IO.mapOptional("BreakBeforeInheritanceComma",
315 Style.BreakBeforeInheritanceComma);
316 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
317 Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
318 IO.mapOptional("ConstructorInitializerIndentWidth",
319 Style.ConstructorInitializerIndentWidth);
320 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
321 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
322 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
323 IO.mapOptional("DisableFormat", Style.DisableFormat);
324 IO.mapOptional("ExperimentalAutoDetectBinPacking",
325 Style.ExperimentalAutoDetectBinPacking);
326 IO.mapOptional("FixNamespaceComments", Style.FixNamespaceComments);
327 IO.mapOptional("ForEachMacros", Style.ForEachMacros);
328 IO.mapOptional("IncludeCategories", Style.IncludeCategories);
329 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex);
330 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
331 IO.mapOptional("IndentWidth", Style.IndentWidth);
332 IO.mapOptional("IndentWrappedFunctionNames",
333 Style.IndentWrappedFunctionNames);
334 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes);
335 IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports);
336 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
337 Style.KeepEmptyLinesAtTheStartOfBlocks);
338 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
339 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
340 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
341 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
342 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
343 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
344 IO.mapOptional("ObjCSpaceBeforeProtocolList",
345 Style.ObjCSpaceBeforeProtocolList);
346 IO.mapOptional("PenaltyBreakAssignment",
347 Style.PenaltyBreakAssignment);
348 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
349 Style.PenaltyBreakBeforeFirstCallParameter);
350 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
351 IO.mapOptional("PenaltyBreakFirstLessLess",
352 Style.PenaltyBreakFirstLessLess);
353 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
354 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
355 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
356 Style.PenaltyReturnTypeOnItsOwnLine);
357 IO.mapOptional("PointerAlignment", Style.PointerAlignment);
358 IO.mapOptional("ReflowComments", Style.ReflowComments);
359 IO.mapOptional("SortIncludes", Style.SortIncludes);
360 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
361 IO.mapOptional("SpaceAfterTemplateKeyword", Style.SpaceAfterTemplateKeyword);
362 IO.mapOptional("SpaceBeforeAssignmentOperators",
363 Style.SpaceBeforeAssignmentOperators);
364 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
365 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
366 IO.mapOptional("SpacesBeforeTrailingComments",
367 Style.SpacesBeforeTrailingComments);
368 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
369 IO.mapOptional("SpacesInContainerLiterals",
370 Style.SpacesInContainerLiterals);
371 IO.mapOptional("SpacesInCStyleCastParentheses",
372 Style.SpacesInCStyleCastParentheses);
373 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
374 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
375 IO.mapOptional("Standard", Style.Standard);
376 IO.mapOptional("TabWidth", Style.TabWidth);
377 IO.mapOptional("UseTab", Style.UseTab);
381 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
382 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
383 IO.mapOptional("AfterClass", Wrapping.AfterClass);
384 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
385 IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
386 IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
387 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
388 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
389 IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
390 IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
391 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
392 IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
393 IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
397 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
398 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
399 IO.mapOptional("Regex", Category.Regex);
400 IO.mapOptional("Priority", Category.Priority);
404 // Allows to read vector<FormatStyle> while keeping default values.
405 // IO.getContext() should contain a pointer to the FormatStyle structure, that
406 // will be used to get default values for missing keys.
407 // If the first element has no Language specified, it will be treated as the
408 // default one for the following elements.
409 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
410 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
413 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
415 if (Index >= Seq.size()) {
416 assert(Index == Seq.size());
417 FormatStyle Template;
418 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
421 Template = *((const FormatStyle *)IO.getContext());
422 Template.Language = FormatStyle::LK_None;
424 Seq.resize(Index + 1, Template);
435 const std::error_category &getParseCategory() {
436 static ParseErrorCategory C;
439 std::error_code make_error_code(ParseError e) {
440 return std::error_code(static_cast<int>(e), getParseCategory());
443 inline llvm::Error make_string_error(const llvm::Twine &Message) {
444 return llvm::make_error<llvm::StringError>(Message,
445 llvm::inconvertibleErrorCode());
448 const char *ParseErrorCategory::name() const noexcept {
449 return "clang-format.parse_error";
452 std::string ParseErrorCategory::message(int EV) const {
453 switch (static_cast<ParseError>(EV)) {
454 case ParseError::Success:
456 case ParseError::Error:
457 return "Invalid argument";
458 case ParseError::Unsuitable:
461 llvm_unreachable("unexpected parse error");
464 static FormatStyle expandPresets(const FormatStyle &Style) {
465 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
467 FormatStyle Expanded = Style;
468 Expanded.BraceWrapping = {false, false, false, false, false, false,
469 false, false, false, false, false};
470 switch (Style.BreakBeforeBraces) {
471 case FormatStyle::BS_Linux:
472 Expanded.BraceWrapping.AfterClass = true;
473 Expanded.BraceWrapping.AfterFunction = true;
474 Expanded.BraceWrapping.AfterNamespace = true;
476 case FormatStyle::BS_Mozilla:
477 Expanded.BraceWrapping.AfterClass = true;
478 Expanded.BraceWrapping.AfterEnum = true;
479 Expanded.BraceWrapping.AfterFunction = true;
480 Expanded.BraceWrapping.AfterStruct = true;
481 Expanded.BraceWrapping.AfterUnion = true;
483 case FormatStyle::BS_Stroustrup:
484 Expanded.BraceWrapping.AfterFunction = true;
485 Expanded.BraceWrapping.BeforeCatch = true;
486 Expanded.BraceWrapping.BeforeElse = true;
488 case FormatStyle::BS_Allman:
489 Expanded.BraceWrapping.AfterClass = true;
490 Expanded.BraceWrapping.AfterControlStatement = true;
491 Expanded.BraceWrapping.AfterEnum = true;
492 Expanded.BraceWrapping.AfterFunction = true;
493 Expanded.BraceWrapping.AfterNamespace = true;
494 Expanded.BraceWrapping.AfterObjCDeclaration = true;
495 Expanded.BraceWrapping.AfterStruct = true;
496 Expanded.BraceWrapping.BeforeCatch = true;
497 Expanded.BraceWrapping.BeforeElse = true;
499 case FormatStyle::BS_GNU:
500 Expanded.BraceWrapping = {true, true, true, true, true, true,
501 true, true, true, true, true};
503 case FormatStyle::BS_WebKit:
504 Expanded.BraceWrapping.AfterFunction = true;
512 FormatStyle getLLVMStyle() {
513 FormatStyle LLVMStyle;
514 LLVMStyle.Language = FormatStyle::LK_Cpp;
515 LLVMStyle.AccessModifierOffset = -2;
516 LLVMStyle.AlignEscapedNewlines = FormatStyle::ENAS_Right;
517 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align;
518 LLVMStyle.AlignOperands = true;
519 LLVMStyle.AlignTrailingComments = true;
520 LLVMStyle.AlignConsecutiveAssignments = false;
521 LLVMStyle.AlignConsecutiveDeclarations = false;
522 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
523 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
524 LLVMStyle.AllowShortBlocksOnASingleLine = false;
525 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
526 LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
527 LLVMStyle.AllowShortLoopsOnASingleLine = false;
528 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
529 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
530 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
531 LLVMStyle.AlwaysBreakTemplateDeclarations = false;
532 LLVMStyle.BinPackParameters = true;
533 LLVMStyle.BinPackArguments = true;
534 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
535 LLVMStyle.BreakBeforeTernaryOperators = true;
536 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
537 LLVMStyle.BraceWrapping = {false, false, false, false, false, false,
538 false, false, false, false, false};
539 LLVMStyle.BreakAfterJavaFieldAnnotations = false;
540 LLVMStyle.BreakConstructorInitializersBeforeComma = false;
541 LLVMStyle.BreakBeforeInheritanceComma = false;
542 LLVMStyle.BreakStringLiterals = true;
543 LLVMStyle.ColumnLimit = 80;
544 LLVMStyle.CommentPragmas = "^ IWYU pragma:";
545 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
546 LLVMStyle.ConstructorInitializerIndentWidth = 4;
547 LLVMStyle.ContinuationIndentWidth = 4;
548 LLVMStyle.Cpp11BracedListStyle = true;
549 LLVMStyle.DerivePointerAlignment = false;
550 LLVMStyle.ExperimentalAutoDetectBinPacking = false;
551 LLVMStyle.FixNamespaceComments = true;
552 LLVMStyle.ForEachMacros.push_back("foreach");
553 LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
554 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
555 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
556 {"^(<|\"(gtest|isl|json)/)", 3},
558 LLVMStyle.IncludeIsMainRegex = "$";
559 LLVMStyle.IndentCaseLabels = false;
560 LLVMStyle.IndentWrappedFunctionNames = false;
561 LLVMStyle.IndentWidth = 2;
562 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
563 LLVMStyle.JavaScriptWrapImports = true;
564 LLVMStyle.TabWidth = 8;
565 LLVMStyle.MaxEmptyLinesToKeep = 1;
566 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
567 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
568 LLVMStyle.ObjCBlockIndentWidth = 2;
569 LLVMStyle.ObjCSpaceAfterProperty = false;
570 LLVMStyle.ObjCSpaceBeforeProtocolList = true;
571 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
572 LLVMStyle.SpacesBeforeTrailingComments = 1;
573 LLVMStyle.Standard = FormatStyle::LS_Cpp11;
574 LLVMStyle.UseTab = FormatStyle::UT_Never;
575 LLVMStyle.ReflowComments = true;
576 LLVMStyle.SpacesInParentheses = false;
577 LLVMStyle.SpacesInSquareBrackets = false;
578 LLVMStyle.SpaceInEmptyParentheses = false;
579 LLVMStyle.SpacesInContainerLiterals = true;
580 LLVMStyle.SpacesInCStyleCastParentheses = false;
581 LLVMStyle.SpaceAfterCStyleCast = false;
582 LLVMStyle.SpaceAfterTemplateKeyword = true;
583 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
584 LLVMStyle.SpaceBeforeAssignmentOperators = true;
585 LLVMStyle.SpacesInAngles = false;
587 LLVMStyle.PenaltyBreakAssignment = prec::Assignment;
588 LLVMStyle.PenaltyBreakComment = 300;
589 LLVMStyle.PenaltyBreakFirstLessLess = 120;
590 LLVMStyle.PenaltyBreakString = 1000;
591 LLVMStyle.PenaltyExcessCharacter = 1000000;
592 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
593 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
595 LLVMStyle.DisableFormat = false;
596 LLVMStyle.SortIncludes = true;
601 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
602 FormatStyle GoogleStyle = getLLVMStyle();
603 GoogleStyle.Language = Language;
605 GoogleStyle.AccessModifierOffset = -1;
606 GoogleStyle.AlignEscapedNewlines = FormatStyle::ENAS_Left;
607 GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
608 GoogleStyle.AllowShortLoopsOnASingleLine = true;
609 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
610 GoogleStyle.AlwaysBreakTemplateDeclarations = true;
611 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
612 GoogleStyle.DerivePointerAlignment = true;
613 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
614 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$";
615 GoogleStyle.IndentCaseLabels = true;
616 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
617 GoogleStyle.ObjCSpaceAfterProperty = false;
618 GoogleStyle.ObjCSpaceBeforeProtocolList = false;
619 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
620 GoogleStyle.SpacesBeforeTrailingComments = 2;
621 GoogleStyle.Standard = FormatStyle::LS_Auto;
623 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
624 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
626 if (Language == FormatStyle::LK_Java) {
627 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
628 GoogleStyle.AlignOperands = false;
629 GoogleStyle.AlignTrailingComments = false;
630 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
631 GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
632 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
633 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
634 GoogleStyle.ColumnLimit = 100;
635 GoogleStyle.SpaceAfterCStyleCast = true;
636 GoogleStyle.SpacesBeforeTrailingComments = 1;
637 } else if (Language == FormatStyle::LK_JavaScript) {
638 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak;
639 GoogleStyle.AlignOperands = false;
640 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
641 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
642 GoogleStyle.BreakBeforeTernaryOperators = false;
643 // taze:, triple slash directives (`/// <...`), @tag followed by { for a lot
644 // of JSDoc tags, and @see, which is commonly followed by overlong URLs.
645 GoogleStyle.CommentPragmas =
646 "(taze:|^/[ \t]*<|(@[A-Za-z_0-9-]+[ \\t]*{)|@see)";
647 GoogleStyle.MaxEmptyLinesToKeep = 3;
648 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All;
649 GoogleStyle.SpacesInContainerLiterals = false;
650 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single;
651 GoogleStyle.JavaScriptWrapImports = false;
652 } else if (Language == FormatStyle::LK_Proto) {
653 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
654 GoogleStyle.SpacesInContainerLiterals = false;
655 } else if (Language == FormatStyle::LK_ObjC) {
656 GoogleStyle.ColumnLimit = 100;
662 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
663 FormatStyle ChromiumStyle = getGoogleStyle(Language);
664 if (Language == FormatStyle::LK_Java) {
665 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
666 ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
667 ChromiumStyle.ContinuationIndentWidth = 8;
668 ChromiumStyle.IndentWidth = 4;
669 } else if (Language == FormatStyle::LK_JavaScript) {
670 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
671 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
673 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
674 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
675 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
676 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
677 ChromiumStyle.BinPackParameters = false;
678 ChromiumStyle.DerivePointerAlignment = false;
679 if (Language == FormatStyle::LK_ObjC)
680 ChromiumStyle.ColumnLimit = 80;
682 return ChromiumStyle;
685 FormatStyle getMozillaStyle() {
686 FormatStyle MozillaStyle = getLLVMStyle();
687 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
688 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
689 MozillaStyle.AlwaysBreakAfterReturnType =
690 FormatStyle::RTBS_TopLevel;
691 MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
692 FormatStyle::DRTBS_TopLevel;
693 MozillaStyle.AlwaysBreakTemplateDeclarations = true;
694 MozillaStyle.BinPackParameters = false;
695 MozillaStyle.BinPackArguments = false;
696 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
697 MozillaStyle.BreakConstructorInitializersBeforeComma = true;
698 MozillaStyle.BreakBeforeInheritanceComma = true;
699 MozillaStyle.ConstructorInitializerIndentWidth = 2;
700 MozillaStyle.ContinuationIndentWidth = 2;
701 MozillaStyle.Cpp11BracedListStyle = false;
702 MozillaStyle.FixNamespaceComments = false;
703 MozillaStyle.IndentCaseLabels = true;
704 MozillaStyle.ObjCSpaceAfterProperty = true;
705 MozillaStyle.ObjCSpaceBeforeProtocolList = false;
706 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
707 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
708 MozillaStyle.SpaceAfterTemplateKeyword = false;
712 FormatStyle getWebKitStyle() {
713 FormatStyle Style = getLLVMStyle();
714 Style.AccessModifierOffset = -4;
715 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
716 Style.AlignOperands = false;
717 Style.AlignTrailingComments = false;
718 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
719 Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
720 Style.BreakConstructorInitializersBeforeComma = true;
721 Style.Cpp11BracedListStyle = false;
722 Style.ColumnLimit = 0;
723 Style.FixNamespaceComments = false;
724 Style.IndentWidth = 4;
725 Style.NamespaceIndentation = FormatStyle::NI_Inner;
726 Style.ObjCBlockIndentWidth = 4;
727 Style.ObjCSpaceAfterProperty = true;
728 Style.PointerAlignment = FormatStyle::PAS_Left;
732 FormatStyle getGNUStyle() {
733 FormatStyle Style = getLLVMStyle();
734 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
735 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
736 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
737 Style.BreakBeforeBraces = FormatStyle::BS_GNU;
738 Style.BreakBeforeTernaryOperators = true;
739 Style.Cpp11BracedListStyle = false;
740 Style.ColumnLimit = 79;
741 Style.FixNamespaceComments = false;
742 Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
743 Style.Standard = FormatStyle::LS_Cpp03;
747 FormatStyle getNoStyle() {
748 FormatStyle NoStyle = getLLVMStyle();
749 NoStyle.DisableFormat = true;
750 NoStyle.SortIncludes = false;
754 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
755 FormatStyle *Style) {
756 if (Name.equals_lower("llvm")) {
757 *Style = getLLVMStyle();
758 } else if (Name.equals_lower("chromium")) {
759 *Style = getChromiumStyle(Language);
760 } else if (Name.equals_lower("mozilla")) {
761 *Style = getMozillaStyle();
762 } else if (Name.equals_lower("google")) {
763 *Style = getGoogleStyle(Language);
764 } else if (Name.equals_lower("webkit")) {
765 *Style = getWebKitStyle();
766 } else if (Name.equals_lower("gnu")) {
767 *Style = getGNUStyle();
768 } else if (Name.equals_lower("none")) {
769 *Style = getNoStyle();
774 Style->Language = Language;
778 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
780 FormatStyle::LanguageKind Language = Style->Language;
781 assert(Language != FormatStyle::LK_None);
782 if (Text.trim().empty())
783 return make_error_code(ParseError::Error);
785 std::vector<FormatStyle> Styles;
786 llvm::yaml::Input Input(Text);
787 // DocumentListTraits<vector<FormatStyle>> uses the context to get default
788 // values for the fields, keys for which are missing from the configuration.
789 // Mapping also uses the context to get the language to find the correct
791 Input.setContext(Style);
794 return Input.error();
796 for (unsigned i = 0; i < Styles.size(); ++i) {
797 // Ensures that only the first configuration can skip the Language option.
798 if (Styles[i].Language == FormatStyle::LK_None && i != 0)
799 return make_error_code(ParseError::Error);
800 // Ensure that each language is configured at most once.
801 for (unsigned j = 0; j < i; ++j) {
802 if (Styles[i].Language == Styles[j].Language) {
804 << "Duplicate languages in the config file on positions " << j
805 << " and " << i << "\n");
806 return make_error_code(ParseError::Error);
810 // Look for a suitable configuration starting from the end, so we can
811 // find the configuration for the specific language first, and the default
812 // configuration (which can only be at slot 0) after it.
813 for (int i = Styles.size() - 1; i >= 0; --i) {
814 if (Styles[i].Language == Language ||
815 Styles[i].Language == FormatStyle::LK_None) {
817 Style->Language = Language;
818 return make_error_code(ParseError::Success);
821 return make_error_code(ParseError::Unsuitable);
824 std::string configurationAsText(const FormatStyle &Style) {
826 llvm::raw_string_ostream Stream(Text);
827 llvm::yaml::Output Output(Stream);
828 // We use the same mapping method for input and output, so we need a non-const
830 FormatStyle NonConstStyle = expandPresets(Style);
831 Output << NonConstStyle;
837 class JavaScriptRequoter : public TokenAnalyzer {
839 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
840 : TokenAnalyzer(Env, Style) {}
842 tooling::Replacements
843 analyze(TokenAnnotator &Annotator,
844 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
845 FormatTokenLexer &Tokens) override {
846 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
847 AnnotatedLines.end());
848 tooling::Replacements Result;
849 requoteJSStringLiteral(AnnotatedLines, Result);
854 // Replaces double/single-quoted string literal as appropriate, re-escaping
855 // the contents in the process.
856 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
857 tooling::Replacements &Result) {
858 for (AnnotatedLine *Line : Lines) {
859 requoteJSStringLiteral(Line->Children, Result);
862 for (FormatToken *FormatTok = Line->First; FormatTok;
863 FormatTok = FormatTok->Next) {
864 StringRef Input = FormatTok->TokenText;
865 if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
866 // NB: testing for not starting with a double quote to avoid
867 // breaking `template strings`.
868 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
869 !Input.startswith("\"")) ||
870 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
871 !Input.startswith("\'")))
874 // Change start and end quote.
875 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
876 SourceLocation Start = FormatTok->Tok.getLocation();
877 auto Replace = [&](SourceLocation Start, unsigned Length,
878 StringRef ReplacementText) {
879 auto Err = Result.add(tooling::Replacement(
880 Env.getSourceManager(), Start, Length, ReplacementText));
881 // FIXME: handle error. For now, print error message and skip the
882 // replacement for release version.
884 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
888 Replace(Start, 1, IsSingle ? "'" : "\"");
889 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
890 IsSingle ? "'" : "\"");
892 // Escape internal quotes.
893 bool Escaped = false;
894 for (size_t i = 1; i < Input.size() - 1; i++) {
897 if (!Escaped && i + 1 < Input.size() &&
898 ((IsSingle && Input[i + 1] == '"') ||
899 (!IsSingle && Input[i + 1] == '\''))) {
900 // Remove this \, it's escaping a " or ' that no longer needs
902 Replace(Start.getLocWithOffset(i), 1, "");
909 if (!Escaped && IsSingle == (Input[i] == '\'')) {
911 Replace(Start.getLocWithOffset(i), 0, "\\");
925 class Formatter : public TokenAnalyzer {
927 Formatter(const Environment &Env, const FormatStyle &Style,
928 FormattingAttemptStatus *Status)
929 : TokenAnalyzer(Env, Style), Status(Status) {}
931 tooling::Replacements
932 analyze(TokenAnnotator &Annotator,
933 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
934 FormatTokenLexer &Tokens) override {
935 tooling::Replacements Result;
936 deriveLocalStyle(AnnotatedLines);
937 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
938 AnnotatedLines.end());
939 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
940 Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
942 Annotator.setCommentLineLevels(AnnotatedLines);
944 WhitespaceManager Whitespaces(
945 Env.getSourceManager(), Style,
946 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
947 ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
948 Env.getSourceManager(), Whitespaces, Encoding,
949 BinPackInconclusiveFunctions);
950 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
951 Env.getSourceManager(), Status)
952 .format(AnnotatedLines);
953 for (const auto &R : Whitespaces.generateReplacements())
961 static bool inputUsesCRLF(StringRef Text) {
962 return Text.count('\r') * 2 > Text.count('\n');
966 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
967 for (const AnnotatedLine *Line : Lines) {
968 if (hasCpp03IncompatibleFormat(Line->Children))
970 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
971 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
972 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
974 if (Tok->is(TT_TemplateCloser) &&
975 Tok->Previous->is(TT_TemplateCloser))
983 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
984 int AlignmentDiff = 0;
985 for (const AnnotatedLine *Line : Lines) {
986 AlignmentDiff += countVariableAlignments(Line->Children);
987 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
988 if (!Tok->is(TT_PointerOrReference))
991 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
992 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
993 Tok->Next->WhitespaceRange.getEnd();
994 if (SpaceBefore && !SpaceAfter)
996 if (!SpaceBefore && SpaceAfter)
1000 return AlignmentDiff;
1004 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1005 bool HasBinPackedFunction = false;
1006 bool HasOnePerLineFunction = false;
1007 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1008 if (!AnnotatedLines[i]->First->Next)
1010 FormatToken *Tok = AnnotatedLines[i]->First->Next;
1012 if (Tok->PackingKind == PPK_BinPacked)
1013 HasBinPackedFunction = true;
1014 if (Tok->PackingKind == PPK_OnePerLine)
1015 HasOnePerLineFunction = true;
1020 if (Style.DerivePointerAlignment)
1021 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
1022 ? FormatStyle::PAS_Left
1023 : FormatStyle::PAS_Right;
1024 if (Style.Standard == FormatStyle::LS_Auto)
1025 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
1026 ? FormatStyle::LS_Cpp11
1027 : FormatStyle::LS_Cpp03;
1028 BinPackInconclusiveFunctions =
1029 HasBinPackedFunction || !HasOnePerLineFunction;
1032 bool BinPackInconclusiveFunctions;
1033 FormattingAttemptStatus *Status;
1036 // This class clean up the erroneous/redundant code around the given ranges in
1038 class Cleaner : public TokenAnalyzer {
1040 Cleaner(const Environment &Env, const FormatStyle &Style)
1041 : TokenAnalyzer(Env, Style),
1042 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1044 // FIXME: eliminate unused parameters.
1045 tooling::Replacements
1046 analyze(TokenAnnotator &Annotator,
1047 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1048 FormatTokenLexer &Tokens) override {
1049 // FIXME: in the current implementation the granularity of affected range
1050 // is an annotated line. However, this is not sufficient. Furthermore,
1051 // redundant code introduced by replacements does not necessarily
1052 // intercept with ranges of replacements that result in the redundancy.
1053 // To determine if some redundant code is actually introduced by
1054 // replacements(e.g. deletions), we need to come up with a more
1055 // sophisticated way of computing affected ranges.
1056 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1057 AnnotatedLines.end());
1059 checkEmptyNamespace(AnnotatedLines);
1061 for (auto &Line : AnnotatedLines) {
1062 if (Line->Affected) {
1063 cleanupRight(Line->First, tok::comma, tok::comma);
1064 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1065 cleanupRight(Line->First, tok::l_paren, tok::comma);
1066 cleanupLeft(Line->First, tok::comma, tok::r_paren);
1067 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1068 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1069 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1073 return generateFixes();
1077 bool containsOnlyComments(const AnnotatedLine &Line) {
1078 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1079 if (Tok->isNot(tok::comment))
1085 // Iterate through all lines and remove any empty (nested) namespaces.
1086 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1087 std::set<unsigned> DeletedLines;
1088 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1089 auto &Line = *AnnotatedLines[i];
1090 if (Line.startsWith(tok::kw_namespace) ||
1091 Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1092 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1096 for (auto Line : DeletedLines) {
1097 FormatToken *Tok = AnnotatedLines[Line]->First;
1105 // The function checks if the namespace, which starts from \p CurrentLine, and
1106 // its nested namespaces are empty and delete them if they are empty. It also
1107 // sets \p NewLine to the last line checked.
1108 // Returns true if the current namespace is empty.
1109 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1110 unsigned CurrentLine, unsigned &NewLine,
1111 std::set<unsigned> &DeletedLines) {
1112 unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1113 if (Style.BraceWrapping.AfterNamespace) {
1114 // If the left brace is in a new line, we should consume it first so that
1115 // it does not make the namespace non-empty.
1116 // FIXME: error handling if there is no left brace.
1117 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1118 NewLine = CurrentLine;
1121 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1124 while (++CurrentLine < End) {
1125 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1128 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1129 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1130 tok::kw_namespace)) {
1131 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1134 CurrentLine = NewLine;
1138 if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1141 // If there is anything other than comments or nested namespaces in the
1142 // current namespace, the namespace cannot be empty.
1143 NewLine = CurrentLine;
1147 NewLine = CurrentLine;
1148 if (CurrentLine >= End)
1151 // Check if the empty namespace is actually affected by changed ranges.
1152 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1153 AnnotatedLines[InitLine]->First->Tok.getLocation(),
1154 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1157 for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1158 DeletedLines.insert(i);
1164 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1165 // of the token in the pair if the left token has \p LK token kind and the
1166 // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1167 // is deleted on match; otherwise, the right token is deleted.
1168 template <typename LeftKind, typename RightKind>
1169 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1171 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1172 for (auto *Res = Tok.Next; Res; Res = Res->Next)
1173 if (!Res->is(tok::comment) &&
1174 DeletedTokens.find(Res) == DeletedTokens.end())
1178 for (auto *Left = Start; Left;) {
1179 auto *Right = NextNotDeleted(*Left);
1182 if (Left->is(LK) && Right->is(RK)) {
1183 deleteToken(DeleteLeft ? Left : Right);
1184 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1186 // If the right token is deleted, we should keep the left token
1187 // unchanged and pair it with the new right token.
1195 template <typename LeftKind, typename RightKind>
1196 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1197 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1200 template <typename LeftKind, typename RightKind>
1201 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1202 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1205 // Delete the given token.
1206 inline void deleteToken(FormatToken *Tok) {
1208 DeletedTokens.insert(Tok);
1211 tooling::Replacements generateFixes() {
1212 tooling::Replacements Fixes;
1213 std::vector<FormatToken *> Tokens;
1214 std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1215 std::back_inserter(Tokens));
1217 // Merge multiple continuous token deletions into one big deletion so that
1218 // the number of replacements can be reduced. This makes computing affected
1219 // ranges more efficient when we run reformat on the changed code.
1221 while (Idx < Tokens.size()) {
1222 unsigned St = Idx, End = Idx;
1223 while ((End + 1) < Tokens.size() &&
1224 Tokens[End]->Next == Tokens[End + 1]) {
1227 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1228 Tokens[End]->Tok.getEndLoc());
1230 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1231 // FIXME: better error handling. for now just print error message and skip
1232 // for the release version.
1234 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1235 assert(false && "Fixes must not conflict!");
1243 // Class for less-than inequality comparason for the set `RedundantTokens`.
1244 // We store tokens in the order they appear in the translation unit so that
1245 // we do not need to sort them in `generateFixes()`.
1246 struct FormatTokenLess {
1247 FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1249 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1250 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1251 RHS->Tok.getLocation());
1253 const SourceManager &SM;
1256 // Tokens to be deleted.
1257 std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1260 struct IncludeDirective {
1267 } // end anonymous namespace
1269 // Determines whether 'Ranges' intersects with ('Start', 'End').
1270 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1272 for (auto Range : Ranges) {
1273 if (Range.getOffset() < End &&
1274 Range.getOffset() + Range.getLength() > Start)
1280 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1281 // before sorting/deduplicating. Index is the index of the include under the
1282 // cursor in the original set of includes. If this include has duplicates, it is
1283 // the index of the first of the duplicates as the others are going to be
1284 // removed. OffsetToEOL describes the cursor's position relative to the end of
1285 // its current line.
1286 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1287 static std::pair<unsigned, unsigned>
1288 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1289 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1290 unsigned CursorIndex = UINT_MAX;
1291 unsigned OffsetToEOL = 0;
1292 for (int i = 0, e = Includes.size(); i != e; ++i) {
1293 unsigned Start = Includes[Indices[i]].Offset;
1294 unsigned End = Start + Includes[Indices[i]].Text.size();
1295 if (!(Cursor >= Start && Cursor < End))
1297 CursorIndex = Indices[i];
1298 OffsetToEOL = End - Cursor;
1299 // Put the cursor on the only remaining #include among the duplicate
1301 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1305 return std::make_pair(CursorIndex, OffsetToEOL);
1308 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1309 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1311 // #include directives with the same text will be deduplicated, and only the
1312 // first #include in the duplicate #includes remains. If the `Cursor` is
1313 // provided and put on a deleted #include, it will be moved to the remaining
1314 // #include in the duplicate #includes.
1315 static void sortCppIncludes(const FormatStyle &Style,
1316 const SmallVectorImpl<IncludeDirective> &Includes,
1317 ArrayRef<tooling::Range> Ranges, StringRef FileName,
1318 tooling::Replacements &Replaces, unsigned *Cursor) {
1319 unsigned IncludesBeginOffset = Includes.front().Offset;
1320 unsigned IncludesEndOffset =
1321 Includes.back().Offset + Includes.back().Text.size();
1322 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1323 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1325 SmallVector<unsigned, 16> Indices;
1326 for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1327 Indices.push_back(i);
1329 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1330 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1331 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1333 // The index of the include on which the cursor will be put after
1334 // sorting/deduplicating.
1335 unsigned CursorIndex;
1336 // The offset from cursor to the end of line.
1337 unsigned CursorToEOLOffset;
1339 std::tie(CursorIndex, CursorToEOLOffset) =
1340 FindCursorIndex(Includes, Indices, *Cursor);
1342 // Deduplicate #includes.
1343 Indices.erase(std::unique(Indices.begin(), Indices.end(),
1344 [&](unsigned LHSI, unsigned RHSI) {
1345 return Includes[LHSI].Text == Includes[RHSI].Text;
1349 // If the #includes are out of order, we generate a single replacement fixing
1350 // the entire block. Otherwise, no replacement is generated.
1351 if (Indices.size() == Includes.size() &&
1352 std::is_sorted(Indices.begin(), Indices.end()))
1356 for (unsigned Index : Indices) {
1357 if (!result.empty())
1359 result += Includes[Index].Text;
1360 if (Cursor && CursorIndex == Index)
1361 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1364 auto Err = Replaces.add(tooling::Replacement(
1365 FileName, Includes.front().Offset, IncludesBlockSize, result));
1366 // FIXME: better error handling. For now, just skip the replacement for the
1369 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1376 // This class manages priorities of #include categories and calculates
1377 // priorities for headers.
1378 class IncludeCategoryManager {
1380 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1381 : Style(Style), FileName(FileName) {
1382 FileStem = llvm::sys::path::stem(FileName);
1383 for (const auto &Category : Style.IncludeCategories)
1384 CategoryRegexs.emplace_back(Category.Regex);
1385 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1386 FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1387 FileName.endswith(".cxx") || FileName.endswith(".m") ||
1388 FileName.endswith(".mm");
1391 // Returns the priority of the category which \p IncludeName belongs to.
1392 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1393 // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1394 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1396 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1397 if (CategoryRegexs[i].match(IncludeName)) {
1398 Ret = Style.IncludeCategories[i].Priority;
1401 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1407 bool isMainHeader(StringRef IncludeName) const {
1408 if (!IncludeName.startswith("\""))
1410 StringRef HeaderStem =
1411 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1412 if (FileStem.startswith(HeaderStem)) {
1413 llvm::Regex MainIncludeRegex(
1414 (HeaderStem + Style.IncludeIsMainRegex).str());
1415 if (MainIncludeRegex.match(FileStem))
1421 const FormatStyle &Style;
1425 SmallVector<llvm::Regex, 4> CategoryRegexs;
1428 const char IncludeRegexPattern[] =
1429 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1431 } // anonymous namespace
1433 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1434 ArrayRef<tooling::Range> Ranges,
1436 tooling::Replacements &Replaces,
1439 unsigned SearchFrom = 0;
1440 llvm::Regex IncludeRegex(IncludeRegexPattern);
1441 SmallVector<StringRef, 4> Matches;
1442 SmallVector<IncludeDirective, 16> IncludesInBlock;
1444 // In compiled files, consider the first #include to be the main #include of
1445 // the file if it is not a system #include. This ensures that the header
1446 // doesn't have hidden dependencies
1447 // (http://llvm.org/docs/CodingStandards.html#include-style).
1449 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1450 // cases where the first #include is unlikely to be the main header.
1451 IncludeCategoryManager Categories(Style, FileName);
1452 bool FirstIncludeBlock = true;
1453 bool MainIncludeFound = false;
1454 bool FormattingOff = false;
1457 auto Pos = Code.find('\n', SearchFrom);
1459 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1461 StringRef Trimmed = Line.trim();
1462 if (Trimmed == "// clang-format off")
1463 FormattingOff = true;
1464 else if (Trimmed == "// clang-format on")
1465 FormattingOff = false;
1467 if (!FormattingOff && !Line.endswith("\\")) {
1468 if (IncludeRegex.match(Line, &Matches)) {
1469 StringRef IncludeName = Matches[2];
1470 int Category = Categories.getIncludePriority(
1472 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1474 MainIncludeFound = true;
1475 IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1476 } else if (!IncludesInBlock.empty()) {
1477 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1479 IncludesInBlock.clear();
1480 FirstIncludeBlock = false;
1484 if (Pos == StringRef::npos || Pos + 1 == Code.size())
1486 SearchFrom = Pos + 1;
1488 if (!IncludesInBlock.empty())
1489 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1493 bool isMpegTS(StringRef Code) {
1494 // MPEG transport streams use the ".ts" file extension. clang-format should
1495 // not attempt to format those. MPEG TS' frame format starts with 0x47 every
1496 // 189 bytes - detect that and return.
1497 return Code.size() > 188 && Code[0] == 0x47 && Code[188] == 0x47;
1500 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1501 ArrayRef<tooling::Range> Ranges,
1502 StringRef FileName, unsigned *Cursor) {
1503 tooling::Replacements Replaces;
1504 if (!Style.SortIncludes)
1506 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript &&
1509 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1510 return sortJavaScriptImports(Style, Code, Ranges, FileName);
1511 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1515 template <typename T>
1516 static llvm::Expected<tooling::Replacements>
1517 processReplacements(T ProcessFunc, StringRef Code,
1518 const tooling::Replacements &Replaces,
1519 const FormatStyle &Style) {
1520 if (Replaces.empty())
1521 return tooling::Replacements();
1523 auto NewCode = applyAllReplacements(Code, Replaces);
1525 return NewCode.takeError();
1526 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1527 StringRef FileName = Replaces.begin()->getFilePath();
1529 tooling::Replacements FormatReplaces =
1530 ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1532 return Replaces.merge(FormatReplaces);
1535 llvm::Expected<tooling::Replacements>
1536 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1537 const FormatStyle &Style) {
1538 // We need to use lambda function here since there are two versions of
1540 auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1541 std::vector<tooling::Range> Ranges,
1542 StringRef FileName) -> tooling::Replacements {
1543 return sortIncludes(Style, Code, Ranges, FileName);
1545 auto SortedReplaces =
1546 processReplacements(SortIncludes, Code, Replaces, Style);
1547 if (!SortedReplaces)
1548 return SortedReplaces.takeError();
1550 // We need to use lambda function here since there are two versions of
1552 auto Reformat = [](const FormatStyle &Style, StringRef Code,
1553 std::vector<tooling::Range> Ranges,
1554 StringRef FileName) -> tooling::Replacements {
1555 return reformat(Style, Code, Ranges, FileName);
1557 return processReplacements(Reformat, Code, *SortedReplaces, Style);
1562 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1563 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1564 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1567 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1568 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1571 // Returns the offset after skipping a sequence of tokens, matched by \p
1572 // GetOffsetAfterSequence, from the start of the code.
1573 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1574 // tokens and returns an offset after the sequence.
1575 unsigned getOffsetAfterTokenSequence(
1576 StringRef FileName, StringRef Code, const FormatStyle &Style,
1577 llvm::function_ref<unsigned(const SourceManager &, Lexer &, Token &)>
1578 GetOffsetAfterSequence) {
1579 std::unique_ptr<Environment> Env =
1580 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1581 const SourceManager &SourceMgr = Env->getSourceManager();
1582 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1583 getFormattingLangOpts(Style));
1585 // Get the first token.
1586 Lex.LexFromRawLexer(Tok);
1587 return GetOffsetAfterSequence(SourceMgr, Lex, Tok);
1590 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1591 // \p Tok will be the token after this directive; otherwise, it can be any token
1592 // after the given \p Tok (including \p Tok).
1593 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1594 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1595 Tok.is(tok::raw_identifier) &&
1596 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1597 Tok.is(tok::raw_identifier);
1599 Lex.LexFromRawLexer(Tok);
1603 void skipComments(Lexer &Lex, Token &Tok) {
1604 while (Tok.is(tok::comment))
1605 if (Lex.LexFromRawLexer(Tok))
1609 // Returns the offset after header guard directives and any comments
1610 // before/after header guards. If no header guard presents in the code, this
1611 // will returns the offset after skipping all comments from the start of the
1613 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1615 const FormatStyle &Style) {
1616 return getOffsetAfterTokenSequence(
1617 FileName, Code, Style,
1618 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1619 skipComments(Lex, Tok);
1620 unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1621 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1622 skipComments(Lex, Tok);
1623 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1624 return SM.getFileOffset(Tok.getLocation());
1626 return InitialOffset;
1630 // Check if a sequence of tokens is like
1631 // "#include ("header.h" | <header.h>)".
1632 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1633 // any token after the given \p Tok (including \p Tok).
1634 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1635 auto Matched = [&]() {
1636 Lex.LexFromRawLexer(Tok);
1639 if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1640 Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1641 if (Lex.LexFromRawLexer(Tok))
1643 if (Tok.is(tok::string_literal))
1645 if (Tok.is(tok::less)) {
1646 while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1648 if (Tok.is(tok::greater))
1655 // Returns the offset of the last #include directive after which a new
1656 // #include can be inserted. This ignores #include's after the #include block(s)
1657 // in the beginning of a file to avoid inserting headers into code sections
1658 // where new #include's should not be added by default.
1659 // These code sections include:
1660 // - raw string literals (containing #include).
1662 // - Special #include's among declarations (e.g. functions).
1664 // If no #include after which a new #include can be inserted, this returns the
1665 // offset after skipping all comments from the start of the code.
1666 // Inserting after an #include is not allowed if it comes after code that is not
1667 // #include (e.g. pre-processing directive that is not #include, declarations).
1668 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1669 const FormatStyle &Style) {
1670 return getOffsetAfterTokenSequence(
1671 FileName, Code, Style,
1672 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1673 skipComments(Lex, Tok);
1674 unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1675 while (checkAndConsumeInclusiveDirective(Lex, Tok))
1676 MaxOffset = SM.getFileOffset(Tok.getLocation());
1681 bool isDeletedHeader(llvm::StringRef HeaderName,
1682 const std::set<llvm::StringRef> &HeadersToDelete) {
1683 return HeadersToDelete.count(HeaderName) ||
1684 HeadersToDelete.count(HeaderName.trim("\"<>"));
1687 // FIXME: insert empty lines between newly created blocks.
1688 tooling::Replacements
1689 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1690 const FormatStyle &Style) {
1694 tooling::Replacements HeaderInsertions;
1695 std::set<llvm::StringRef> HeadersToDelete;
1696 tooling::Replacements Result;
1697 for (const auto &R : Replaces) {
1698 if (isHeaderInsertion(R)) {
1699 // Replacements from \p Replaces must be conflict-free already, so we can
1700 // simply consume the error.
1701 llvm::consumeError(HeaderInsertions.add(R));
1702 } else if (isHeaderDeletion(R)) {
1703 HeadersToDelete.insert(R.getReplacementText());
1704 } else if (R.getOffset() == UINT_MAX) {
1705 llvm::errs() << "Insertions other than header #include insertion are "
1707 << R.getReplacementText() << "\n";
1709 llvm::consumeError(Result.add(R));
1712 if (HeaderInsertions.empty() && HeadersToDelete.empty())
1715 llvm::Regex IncludeRegex(IncludeRegexPattern);
1716 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1717 SmallVector<StringRef, 4> Matches;
1719 StringRef FileName = Replaces.begin()->getFilePath();
1720 IncludeCategoryManager Categories(Style, FileName);
1722 // Record the offset of the end of the last include in each category.
1723 std::map<int, int> CategoryEndOffsets;
1724 // All possible priorities.
1725 // Add 0 for main header and INT_MAX for headers that are not in any category.
1726 std::set<int> Priorities = {0, INT_MAX};
1727 for (const auto &Category : Style.IncludeCategories)
1728 Priorities.insert(Category.Priority);
1729 int FirstIncludeOffset = -1;
1730 // All new headers should be inserted after this offset.
1731 unsigned MinInsertOffset =
1732 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1733 StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1734 // Max insertion offset in the original code.
1735 unsigned MaxInsertOffset =
1737 getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1738 SmallVector<StringRef, 32> Lines;
1739 TrimmedCode.split(Lines, '\n');
1740 unsigned Offset = MinInsertOffset;
1741 unsigned NextLineOffset;
1742 std::set<StringRef> ExistingIncludes;
1743 for (auto Line : Lines) {
1744 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1745 if (IncludeRegex.match(Line, &Matches)) {
1746 // The header name with quotes or angle brackets.
1747 StringRef IncludeName = Matches[2];
1748 ExistingIncludes.insert(IncludeName);
1749 // Only record the offset of current #include if we can insert after it.
1750 if (Offset <= MaxInsertOffset) {
1751 int Category = Categories.getIncludePriority(
1752 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1753 CategoryEndOffsets[Category] = NextLineOffset;
1754 if (FirstIncludeOffset < 0)
1755 FirstIncludeOffset = Offset;
1757 if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1758 // If this is the last line without trailing newline, we need to make
1759 // sure we don't delete across the file boundary.
1760 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1762 Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1764 // Ignore the deletion on conflict.
1765 llvm::errs() << "Failed to add header deletion replacement for "
1766 << IncludeName << ": " << llvm::toString(std::move(Err))
1771 Offset = NextLineOffset;
1774 // Populate CategoryEndOfssets:
1775 // - Ensure that CategoryEndOffset[Highest] is always populated.
1776 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1777 // is set, up to CategoryEndOffset[Highest].
1778 auto Highest = Priorities.begin();
1779 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1780 if (FirstIncludeOffset >= 0)
1781 CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1783 CategoryEndOffsets[*Highest] = MinInsertOffset;
1785 // By this point, CategoryEndOffset[Highest] is always set appropriately:
1786 // - to an appropriate location before/after existing #includes, or
1787 // - to right after the header guard, or
1788 // - to the beginning of the file.
1789 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1790 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1791 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1793 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1794 for (const auto &R : HeaderInsertions) {
1795 auto IncludeDirective = R.getReplacementText();
1796 bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1797 assert(Matched && "Header insertion replacement must have replacement text "
1800 auto IncludeName = Matches[2];
1801 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1802 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1807 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1808 Offset = CategoryEndOffsets[Category];
1809 std::string NewInclude = !IncludeDirective.endswith("\n")
1810 ? (IncludeDirective + "\n").str()
1811 : IncludeDirective.str();
1812 // When inserting headers at end of the code, also append '\n' to the code
1813 // if it does not end with '\n'.
1814 if (NeedNewLineAtEnd && Offset == Code.size()) {
1815 NewInclude = "\n" + NewInclude;
1816 NeedNewLineAtEnd = false;
1818 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1819 auto Err = Result.add(NewReplace);
1821 llvm::consumeError(std::move(Err));
1822 unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1823 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1824 Result = Result.merge(tooling::Replacements(NewReplace));
1830 } // anonymous namespace
1832 llvm::Expected<tooling::Replacements>
1833 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1834 const FormatStyle &Style) {
1835 // We need to use lambda function here since there are two versions of
1837 auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1838 std::vector<tooling::Range> Ranges,
1839 StringRef FileName) -> tooling::Replacements {
1840 return cleanup(Style, Code, Ranges, FileName);
1842 // Make header insertion replacements insert new headers into correct blocks.
1843 tooling::Replacements NewReplaces =
1844 fixCppIncludeInsertions(Code, Replaces, Style);
1845 return processReplacements(Cleanup, Code, NewReplaces, Style);
1848 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1849 ArrayRef<tooling::Range> Ranges,
1851 FormattingAttemptStatus *Status) {
1852 FormatStyle Expanded = expandPresets(Style);
1853 if (Expanded.DisableFormat)
1854 return tooling::Replacements();
1855 if (Expanded.Language == FormatStyle::LK_JavaScript && isMpegTS(Code))
1856 return tooling::Replacements();
1857 auto Env = Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1859 auto reformatAfterApplying = [&] (TokenAnalyzer& Fixer) {
1860 tooling::Replacements Fixes = Fixer.process();
1861 if (!Fixes.empty()) {
1862 auto NewCode = applyAllReplacements(Code, Fixes);
1864 auto NewEnv = Environment::CreateVirtualEnvironment(
1866 tooling::calculateRangesAfterReplacements(Fixes, Ranges));
1867 Formatter Format(*NewEnv, Expanded, Status);
1868 return Fixes.merge(Format.process());
1871 Formatter Format(*Env, Expanded, Status);
1872 return Format.process();
1875 if (Style.Language == FormatStyle::LK_Cpp &&
1876 Style.FixNamespaceComments) {
1877 NamespaceEndCommentsFixer CommentsFixer(*Env, Expanded);
1878 return reformatAfterApplying(CommentsFixer);
1881 if (Style.Language == FormatStyle::LK_JavaScript &&
1882 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) {
1883 JavaScriptRequoter Requoter(*Env, Expanded);
1884 return reformatAfterApplying(Requoter);
1887 Formatter Format(*Env, Expanded, Status);
1888 return Format.process();
1891 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
1892 ArrayRef<tooling::Range> Ranges,
1893 StringRef FileName) {
1894 std::unique_ptr<Environment> Env =
1895 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1896 Cleaner Clean(*Env, Style);
1897 return Clean.process();
1900 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1901 ArrayRef<tooling::Range> Ranges,
1902 StringRef FileName, bool *IncompleteFormat) {
1903 FormattingAttemptStatus Status;
1904 auto Result = reformat(Style, Code, Ranges, FileName, &Status);
1905 if (!Status.FormatComplete)
1906 *IncompleteFormat = true;
1910 tooling::Replacements fixNamespaceEndComments(const FormatStyle &Style,
1912 ArrayRef<tooling::Range> Ranges,
1913 StringRef FileName) {
1914 std::unique_ptr<Environment> Env =
1915 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1916 NamespaceEndCommentsFixer Fix(*Env, Style);
1917 return Fix.process();
1920 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1921 LangOptions LangOpts;
1922 LangOpts.CPlusPlus = 1;
1923 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1924 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1925 LangOpts.CPlusPlus1z = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1926 LangOpts.LineComment = 1;
1927 bool AlternativeOperators = Style.isCpp();
1928 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1932 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally.
1933 LangOpts.DeclSpecKeyword = 1; // To get __declspec.
1937 const char *StyleOptionHelpDescription =
1938 "Coding style, currently supports:\n"
1939 " LLVM, Google, Chromium, Mozilla, WebKit.\n"
1940 "Use -style=file to load style configuration from\n"
1941 ".clang-format file located in one of the parent\n"
1942 "directories of the source file (or current\n"
1943 "directory for stdin).\n"
1944 "Use -style=\"{key: value, ...}\" to set specific\n"
1945 "parameters, e.g.:\n"
1946 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1948 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1949 if (FileName.endswith(".java"))
1950 return FormatStyle::LK_Java;
1951 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
1952 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
1953 if (FileName.endswith(".m") || FileName.endswith(".mm"))
1954 return FormatStyle::LK_ObjC;
1955 if (FileName.endswith_lower(".proto") ||
1956 FileName.endswith_lower(".protodevel"))
1957 return FormatStyle::LK_Proto;
1958 if (FileName.endswith_lower(".td"))
1959 return FormatStyle::LK_TableGen;
1960 return FormatStyle::LK_Cpp;
1963 llvm::Expected<FormatStyle> getStyle(StringRef StyleName, StringRef FileName,
1964 StringRef FallbackStyleName,
1965 StringRef Code, vfs::FileSystem *FS) {
1967 FS = vfs::getRealFileSystem().get();
1969 FormatStyle Style = getLLVMStyle();
1970 Style.Language = getLanguageByFileName(FileName);
1972 // This is a very crude detection of whether a header contains ObjC code that
1973 // should be improved over time and probably be done on tokens, not one the
1974 // bare content of the file.
1975 if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
1976 (Code.contains("\n- (") || Code.contains("\n+ (")))
1977 Style.Language = FormatStyle::LK_ObjC;
1979 FormatStyle FallbackStyle = getNoStyle();
1980 if (!getPredefinedStyle(FallbackStyleName, Style.Language, &FallbackStyle))
1981 return make_string_error("Invalid fallback style \"" + FallbackStyleName);
1983 if (StyleName.startswith("{")) {
1984 // Parse YAML/JSON style from the command line.
1985 if (std::error_code ec = parseConfiguration(StyleName, &Style))
1986 return make_string_error("Error parsing -style: " + ec.message());
1990 if (!StyleName.equals_lower("file")) {
1991 if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1992 return make_string_error("Invalid value for -style");
1996 // Look for .clang-format/_clang-format file in the file's parent directories.
1997 SmallString<128> UnsuitableConfigFiles;
1998 SmallString<128> Path(FileName);
1999 if (std::error_code EC = FS->makeAbsolute(Path))
2000 return make_string_error(EC.message());
2002 for (StringRef Directory = Path; !Directory.empty();
2003 Directory = llvm::sys::path::parent_path(Directory)) {
2005 auto Status = FS->status(Directory);
2007 Status->getType() != llvm::sys::fs::file_type::directory_file) {
2011 SmallString<128> ConfigFile(Directory);
2013 llvm::sys::path::append(ConfigFile, ".clang-format");
2014 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2016 Status = FS->status(ConfigFile.str());
2017 bool FoundConfigFile =
2018 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
2019 if (!FoundConfigFile) {
2020 // Try _clang-format too, since dotfiles are not commonly used on Windows.
2021 ConfigFile = Directory;
2022 llvm::sys::path::append(ConfigFile, "_clang-format");
2023 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2024 Status = FS->status(ConfigFile.str());
2025 FoundConfigFile = Status && (Status->getType() ==
2026 llvm::sys::fs::file_type::regular_file);
2029 if (FoundConfigFile) {
2030 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
2031 FS->getBufferForFile(ConfigFile.str());
2032 if (std::error_code EC = Text.getError())
2033 return make_string_error(EC.message());
2034 if (std::error_code ec =
2035 parseConfiguration(Text.get()->getBuffer(), &Style)) {
2036 if (ec == ParseError::Unsuitable) {
2037 if (!UnsuitableConfigFiles.empty())
2038 UnsuitableConfigFiles.append(", ");
2039 UnsuitableConfigFiles.append(ConfigFile);
2042 return make_string_error("Error reading " + ConfigFile + ": " +
2045 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
2049 if (!UnsuitableConfigFiles.empty())
2050 return make_string_error("Configuration file(s) do(es) not support " +
2051 getLanguageName(Style.Language) + ": " +
2052 UnsuitableConfigFiles);
2053 return FallbackStyle;
2056 } // namespace format
2057 } // namespace clang