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::PointerAlignmentStyle> {
175 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
176 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
177 IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
178 IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
180 // For backward compatibility.
181 IO.enumCase(Value, "true", FormatStyle::PAS_Left);
182 IO.enumCase(Value, "false", FormatStyle::PAS_Right);
187 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
188 static void enumeration(IO &IO,
189 FormatStyle::SpaceBeforeParensOptions &Value) {
190 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
191 IO.enumCase(Value, "ControlStatements",
192 FormatStyle::SBPO_ControlStatements);
193 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
195 // For backward compatibility.
196 IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
197 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
201 template <> struct MappingTraits<FormatStyle> {
202 static void mapping(IO &IO, FormatStyle &Style) {
203 // When reading, read the language first, we need it for getPredefinedStyle.
204 IO.mapOptional("Language", Style.Language);
206 if (IO.outputting()) {
207 StringRef StylesArray[] = {"LLVM", "Google", "Chromium",
208 "Mozilla", "WebKit", "GNU"};
209 ArrayRef<StringRef> Styles(StylesArray);
210 for (size_t i = 0, e = Styles.size(); i < e; ++i) {
211 StringRef StyleName(Styles[i]);
212 FormatStyle PredefinedStyle;
213 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
214 Style == PredefinedStyle) {
215 IO.mapOptional("# BasedOnStyle", StyleName);
220 StringRef BasedOnStyle;
221 IO.mapOptional("BasedOnStyle", BasedOnStyle);
222 if (!BasedOnStyle.empty()) {
223 FormatStyle::LanguageKind OldLanguage = Style.Language;
224 FormatStyle::LanguageKind Language =
225 ((FormatStyle *)IO.getContext())->Language;
226 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
227 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
230 Style.Language = OldLanguage;
234 // For backward compatibility.
235 if (!IO.outputting()) {
236 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
237 IO.mapOptional("IndentFunctionDeclarationAfterType",
238 Style.IndentWrappedFunctionNames);
239 IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
240 IO.mapOptional("SpaceAfterControlStatementKeyword",
241 Style.SpaceBeforeParens);
244 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
245 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
246 IO.mapOptional("AlignConsecutiveAssignments",
247 Style.AlignConsecutiveAssignments);
248 IO.mapOptional("AlignConsecutiveDeclarations",
249 Style.AlignConsecutiveDeclarations);
250 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft);
251 IO.mapOptional("AlignOperands", Style.AlignOperands);
252 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
253 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
254 Style.AllowAllParametersOfDeclarationOnNextLine);
255 IO.mapOptional("AllowShortBlocksOnASingleLine",
256 Style.AllowShortBlocksOnASingleLine);
257 IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
258 Style.AllowShortCaseLabelsOnASingleLine);
259 IO.mapOptional("AllowShortFunctionsOnASingleLine",
260 Style.AllowShortFunctionsOnASingleLine);
261 IO.mapOptional("AllowShortIfStatementsOnASingleLine",
262 Style.AllowShortIfStatementsOnASingleLine);
263 IO.mapOptional("AllowShortLoopsOnASingleLine",
264 Style.AllowShortLoopsOnASingleLine);
265 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
266 Style.AlwaysBreakAfterDefinitionReturnType);
267 IO.mapOptional("AlwaysBreakAfterReturnType",
268 Style.AlwaysBreakAfterReturnType);
269 // If AlwaysBreakAfterDefinitionReturnType was specified but
270 // AlwaysBreakAfterReturnType was not, initialize the latter from the
271 // former for backwards compatibility.
272 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None &&
273 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) {
274 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All)
275 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
276 else if (Style.AlwaysBreakAfterDefinitionReturnType ==
277 FormatStyle::DRTBS_TopLevel)
278 Style.AlwaysBreakAfterReturnType =
279 FormatStyle::RTBS_TopLevelDefinitions;
282 IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
283 Style.AlwaysBreakBeforeMultilineStrings);
284 IO.mapOptional("AlwaysBreakTemplateDeclarations",
285 Style.AlwaysBreakTemplateDeclarations);
286 IO.mapOptional("BinPackArguments", Style.BinPackArguments);
287 IO.mapOptional("BinPackParameters", Style.BinPackParameters);
288 IO.mapOptional("BraceWrapping", Style.BraceWrapping);
289 IO.mapOptional("BreakBeforeBinaryOperators",
290 Style.BreakBeforeBinaryOperators);
291 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
292 IO.mapOptional("BreakBeforeTernaryOperators",
293 Style.BreakBeforeTernaryOperators);
294 IO.mapOptional("BreakConstructorInitializersBeforeComma",
295 Style.BreakConstructorInitializersBeforeComma);
296 IO.mapOptional("BreakAfterJavaFieldAnnotations",
297 Style.BreakAfterJavaFieldAnnotations);
298 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals);
299 IO.mapOptional("ColumnLimit", Style.ColumnLimit);
300 IO.mapOptional("CommentPragmas", Style.CommentPragmas);
301 IO.mapOptional("BreakBeforeInheritanceComma",
302 Style.BreakBeforeInheritanceComma);
303 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
304 Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
305 IO.mapOptional("ConstructorInitializerIndentWidth",
306 Style.ConstructorInitializerIndentWidth);
307 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
308 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
309 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
310 IO.mapOptional("DisableFormat", Style.DisableFormat);
311 IO.mapOptional("ExperimentalAutoDetectBinPacking",
312 Style.ExperimentalAutoDetectBinPacking);
313 IO.mapOptional("FixNamespaceComments", Style.FixNamespaceComments);
314 IO.mapOptional("ForEachMacros", Style.ForEachMacros);
315 IO.mapOptional("IncludeCategories", Style.IncludeCategories);
316 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex);
317 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
318 IO.mapOptional("IndentWidth", Style.IndentWidth);
319 IO.mapOptional("IndentWrappedFunctionNames",
320 Style.IndentWrappedFunctionNames);
321 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes);
322 IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports);
323 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
324 Style.KeepEmptyLinesAtTheStartOfBlocks);
325 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
326 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
327 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
328 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
329 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
330 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
331 IO.mapOptional("ObjCSpaceBeforeProtocolList",
332 Style.ObjCSpaceBeforeProtocolList);
333 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
334 Style.PenaltyBreakBeforeFirstCallParameter);
335 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
336 IO.mapOptional("PenaltyBreakFirstLessLess",
337 Style.PenaltyBreakFirstLessLess);
338 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
339 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
340 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
341 Style.PenaltyReturnTypeOnItsOwnLine);
342 IO.mapOptional("PointerAlignment", Style.PointerAlignment);
343 IO.mapOptional("ReflowComments", Style.ReflowComments);
344 IO.mapOptional("SortIncludes", Style.SortIncludes);
345 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
346 IO.mapOptional("SpaceAfterTemplateKeyword", Style.SpaceAfterTemplateKeyword);
347 IO.mapOptional("SpaceBeforeAssignmentOperators",
348 Style.SpaceBeforeAssignmentOperators);
349 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
350 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
351 IO.mapOptional("SpacesBeforeTrailingComments",
352 Style.SpacesBeforeTrailingComments);
353 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
354 IO.mapOptional("SpacesInContainerLiterals",
355 Style.SpacesInContainerLiterals);
356 IO.mapOptional("SpacesInCStyleCastParentheses",
357 Style.SpacesInCStyleCastParentheses);
358 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
359 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
360 IO.mapOptional("Standard", Style.Standard);
361 IO.mapOptional("TabWidth", Style.TabWidth);
362 IO.mapOptional("UseTab", Style.UseTab);
366 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
367 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
368 IO.mapOptional("AfterClass", Wrapping.AfterClass);
369 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
370 IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
371 IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
372 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
373 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
374 IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
375 IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
376 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
377 IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
378 IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
382 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
383 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
384 IO.mapOptional("Regex", Category.Regex);
385 IO.mapOptional("Priority", Category.Priority);
389 // Allows to read vector<FormatStyle> while keeping default values.
390 // IO.getContext() should contain a pointer to the FormatStyle structure, that
391 // will be used to get default values for missing keys.
392 // If the first element has no Language specified, it will be treated as the
393 // default one for the following elements.
394 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
395 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
398 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
400 if (Index >= Seq.size()) {
401 assert(Index == Seq.size());
402 FormatStyle Template;
403 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
406 Template = *((const FormatStyle *)IO.getContext());
407 Template.Language = FormatStyle::LK_None;
409 Seq.resize(Index + 1, Template);
420 const std::error_category &getParseCategory() {
421 static ParseErrorCategory C;
424 std::error_code make_error_code(ParseError e) {
425 return std::error_code(static_cast<int>(e), getParseCategory());
428 inline llvm::Error make_string_error(const llvm::Twine &Message) {
429 return llvm::make_error<llvm::StringError>(Message,
430 llvm::inconvertibleErrorCode());
433 const char *ParseErrorCategory::name() const noexcept {
434 return "clang-format.parse_error";
437 std::string ParseErrorCategory::message(int EV) const {
438 switch (static_cast<ParseError>(EV)) {
439 case ParseError::Success:
441 case ParseError::Error:
442 return "Invalid argument";
443 case ParseError::Unsuitable:
446 llvm_unreachable("unexpected parse error");
449 static FormatStyle expandPresets(const FormatStyle &Style) {
450 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
452 FormatStyle Expanded = Style;
453 Expanded.BraceWrapping = {false, false, false, false, false, false,
454 false, false, false, false, false};
455 switch (Style.BreakBeforeBraces) {
456 case FormatStyle::BS_Linux:
457 Expanded.BraceWrapping.AfterClass = true;
458 Expanded.BraceWrapping.AfterFunction = true;
459 Expanded.BraceWrapping.AfterNamespace = true;
461 case FormatStyle::BS_Mozilla:
462 Expanded.BraceWrapping.AfterClass = true;
463 Expanded.BraceWrapping.AfterEnum = true;
464 Expanded.BraceWrapping.AfterFunction = true;
465 Expanded.BraceWrapping.AfterStruct = true;
466 Expanded.BraceWrapping.AfterUnion = true;
468 case FormatStyle::BS_Stroustrup:
469 Expanded.BraceWrapping.AfterFunction = true;
470 Expanded.BraceWrapping.BeforeCatch = true;
471 Expanded.BraceWrapping.BeforeElse = true;
473 case FormatStyle::BS_Allman:
474 Expanded.BraceWrapping.AfterClass = true;
475 Expanded.BraceWrapping.AfterControlStatement = true;
476 Expanded.BraceWrapping.AfterEnum = true;
477 Expanded.BraceWrapping.AfterFunction = true;
478 Expanded.BraceWrapping.AfterNamespace = true;
479 Expanded.BraceWrapping.AfterObjCDeclaration = true;
480 Expanded.BraceWrapping.AfterStruct = true;
481 Expanded.BraceWrapping.BeforeCatch = true;
482 Expanded.BraceWrapping.BeforeElse = true;
484 case FormatStyle::BS_GNU:
485 Expanded.BraceWrapping = {true, true, true, true, true, true,
486 true, true, true, true, true};
488 case FormatStyle::BS_WebKit:
489 Expanded.BraceWrapping.AfterFunction = true;
497 FormatStyle getLLVMStyle() {
498 FormatStyle LLVMStyle;
499 LLVMStyle.Language = FormatStyle::LK_Cpp;
500 LLVMStyle.AccessModifierOffset = -2;
501 LLVMStyle.AlignEscapedNewlinesLeft = false;
502 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align;
503 LLVMStyle.AlignOperands = true;
504 LLVMStyle.AlignTrailingComments = true;
505 LLVMStyle.AlignConsecutiveAssignments = false;
506 LLVMStyle.AlignConsecutiveDeclarations = false;
507 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
508 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
509 LLVMStyle.AllowShortBlocksOnASingleLine = false;
510 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
511 LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
512 LLVMStyle.AllowShortLoopsOnASingleLine = false;
513 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
514 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
515 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
516 LLVMStyle.AlwaysBreakTemplateDeclarations = false;
517 LLVMStyle.BinPackParameters = true;
518 LLVMStyle.BinPackArguments = true;
519 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
520 LLVMStyle.BreakBeforeTernaryOperators = true;
521 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
522 LLVMStyle.BraceWrapping = {false, false, false, false, false, false,
523 false, false, false, false, false};
524 LLVMStyle.BreakAfterJavaFieldAnnotations = false;
525 LLVMStyle.BreakConstructorInitializersBeforeComma = false;
526 LLVMStyle.BreakBeforeInheritanceComma = false;
527 LLVMStyle.BreakStringLiterals = true;
528 LLVMStyle.ColumnLimit = 80;
529 LLVMStyle.CommentPragmas = "^ IWYU pragma:";
530 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
531 LLVMStyle.ConstructorInitializerIndentWidth = 4;
532 LLVMStyle.ContinuationIndentWidth = 4;
533 LLVMStyle.Cpp11BracedListStyle = true;
534 LLVMStyle.DerivePointerAlignment = false;
535 LLVMStyle.ExperimentalAutoDetectBinPacking = false;
536 LLVMStyle.FixNamespaceComments = true;
537 LLVMStyle.ForEachMacros.push_back("foreach");
538 LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
539 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
540 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
541 {"^(<|\"(gtest|isl|json)/)", 3},
543 LLVMStyle.IncludeIsMainRegex = "$";
544 LLVMStyle.IndentCaseLabels = false;
545 LLVMStyle.IndentWrappedFunctionNames = false;
546 LLVMStyle.IndentWidth = 2;
547 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
548 LLVMStyle.JavaScriptWrapImports = true;
549 LLVMStyle.TabWidth = 8;
550 LLVMStyle.MaxEmptyLinesToKeep = 1;
551 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
552 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
553 LLVMStyle.ObjCBlockIndentWidth = 2;
554 LLVMStyle.ObjCSpaceAfterProperty = false;
555 LLVMStyle.ObjCSpaceBeforeProtocolList = true;
556 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
557 LLVMStyle.SpacesBeforeTrailingComments = 1;
558 LLVMStyle.Standard = FormatStyle::LS_Cpp11;
559 LLVMStyle.UseTab = FormatStyle::UT_Never;
560 LLVMStyle.ReflowComments = true;
561 LLVMStyle.SpacesInParentheses = false;
562 LLVMStyle.SpacesInSquareBrackets = false;
563 LLVMStyle.SpaceInEmptyParentheses = false;
564 LLVMStyle.SpacesInContainerLiterals = true;
565 LLVMStyle.SpacesInCStyleCastParentheses = false;
566 LLVMStyle.SpaceAfterCStyleCast = false;
567 LLVMStyle.SpaceAfterTemplateKeyword = true;
568 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
569 LLVMStyle.SpaceBeforeAssignmentOperators = true;
570 LLVMStyle.SpacesInAngles = false;
572 LLVMStyle.PenaltyBreakComment = 300;
573 LLVMStyle.PenaltyBreakFirstLessLess = 120;
574 LLVMStyle.PenaltyBreakString = 1000;
575 LLVMStyle.PenaltyExcessCharacter = 1000000;
576 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
577 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
579 LLVMStyle.DisableFormat = false;
580 LLVMStyle.SortIncludes = true;
585 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
586 FormatStyle GoogleStyle = getLLVMStyle();
587 GoogleStyle.Language = Language;
589 GoogleStyle.AccessModifierOffset = -1;
590 GoogleStyle.AlignEscapedNewlinesLeft = true;
591 GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
592 GoogleStyle.AllowShortLoopsOnASingleLine = true;
593 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
594 GoogleStyle.AlwaysBreakTemplateDeclarations = true;
595 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
596 GoogleStyle.DerivePointerAlignment = true;
597 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
598 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$";
599 GoogleStyle.IndentCaseLabels = true;
600 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
601 GoogleStyle.ObjCSpaceAfterProperty = false;
602 GoogleStyle.ObjCSpaceBeforeProtocolList = false;
603 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
604 GoogleStyle.SpacesBeforeTrailingComments = 2;
605 GoogleStyle.Standard = FormatStyle::LS_Auto;
607 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
608 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
610 if (Language == FormatStyle::LK_Java) {
611 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
612 GoogleStyle.AlignOperands = false;
613 GoogleStyle.AlignTrailingComments = false;
614 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
615 GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
616 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
617 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
618 GoogleStyle.ColumnLimit = 100;
619 GoogleStyle.SpaceAfterCStyleCast = true;
620 GoogleStyle.SpacesBeforeTrailingComments = 1;
621 } else if (Language == FormatStyle::LK_JavaScript) {
622 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak;
623 GoogleStyle.AlignOperands = false;
624 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
625 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
626 GoogleStyle.BreakBeforeTernaryOperators = false;
627 // taze:, @tag followed by { for a lot of JSDoc tags, and @see, which is
628 // commonly followed by overlong URLs.
629 GoogleStyle.CommentPragmas = "(taze:|(@[A-Za-z_0-9-]+[ \\t]*{)|@see)";
630 GoogleStyle.MaxEmptyLinesToKeep = 3;
631 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All;
632 GoogleStyle.SpacesInContainerLiterals = false;
633 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single;
634 GoogleStyle.JavaScriptWrapImports = false;
635 } else if (Language == FormatStyle::LK_Proto) {
636 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
637 GoogleStyle.SpacesInContainerLiterals = false;
638 } else if (Language == FormatStyle::LK_ObjC) {
639 GoogleStyle.ColumnLimit = 100;
645 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
646 FormatStyle ChromiumStyle = getGoogleStyle(Language);
647 if (Language == FormatStyle::LK_Java) {
648 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
649 ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
650 ChromiumStyle.ContinuationIndentWidth = 8;
651 ChromiumStyle.IndentWidth = 4;
652 } else if (Language == FormatStyle::LK_JavaScript) {
653 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
654 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
656 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
657 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
658 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
659 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
660 ChromiumStyle.BinPackParameters = false;
661 ChromiumStyle.DerivePointerAlignment = false;
662 if (Language == FormatStyle::LK_ObjC)
663 ChromiumStyle.ColumnLimit = 80;
665 return ChromiumStyle;
668 FormatStyle getMozillaStyle() {
669 FormatStyle MozillaStyle = getLLVMStyle();
670 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
671 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
672 MozillaStyle.AlwaysBreakAfterReturnType =
673 FormatStyle::RTBS_TopLevel;
674 MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
675 FormatStyle::DRTBS_TopLevel;
676 MozillaStyle.AlwaysBreakTemplateDeclarations = true;
677 MozillaStyle.BinPackParameters = false;
678 MozillaStyle.BinPackArguments = false;
679 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
680 MozillaStyle.BreakConstructorInitializersBeforeComma = true;
681 MozillaStyle.BreakBeforeInheritanceComma = true;
682 MozillaStyle.ConstructorInitializerIndentWidth = 2;
683 MozillaStyle.ContinuationIndentWidth = 2;
684 MozillaStyle.Cpp11BracedListStyle = false;
685 MozillaStyle.FixNamespaceComments = false;
686 MozillaStyle.IndentCaseLabels = true;
687 MozillaStyle.ObjCSpaceAfterProperty = true;
688 MozillaStyle.ObjCSpaceBeforeProtocolList = false;
689 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
690 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
691 MozillaStyle.SpaceAfterTemplateKeyword = false;
695 FormatStyle getWebKitStyle() {
696 FormatStyle Style = getLLVMStyle();
697 Style.AccessModifierOffset = -4;
698 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
699 Style.AlignOperands = false;
700 Style.AlignTrailingComments = false;
701 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
702 Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
703 Style.BreakConstructorInitializersBeforeComma = true;
704 Style.Cpp11BracedListStyle = false;
705 Style.ColumnLimit = 0;
706 Style.FixNamespaceComments = false;
707 Style.IndentWidth = 4;
708 Style.NamespaceIndentation = FormatStyle::NI_Inner;
709 Style.ObjCBlockIndentWidth = 4;
710 Style.ObjCSpaceAfterProperty = true;
711 Style.PointerAlignment = FormatStyle::PAS_Left;
715 FormatStyle getGNUStyle() {
716 FormatStyle Style = getLLVMStyle();
717 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
718 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
719 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
720 Style.BreakBeforeBraces = FormatStyle::BS_GNU;
721 Style.BreakBeforeTernaryOperators = true;
722 Style.Cpp11BracedListStyle = false;
723 Style.ColumnLimit = 79;
724 Style.FixNamespaceComments = false;
725 Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
726 Style.Standard = FormatStyle::LS_Cpp03;
730 FormatStyle getNoStyle() {
731 FormatStyle NoStyle = getLLVMStyle();
732 NoStyle.DisableFormat = true;
733 NoStyle.SortIncludes = false;
737 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
738 FormatStyle *Style) {
739 if (Name.equals_lower("llvm")) {
740 *Style = getLLVMStyle();
741 } else if (Name.equals_lower("chromium")) {
742 *Style = getChromiumStyle(Language);
743 } else if (Name.equals_lower("mozilla")) {
744 *Style = getMozillaStyle();
745 } else if (Name.equals_lower("google")) {
746 *Style = getGoogleStyle(Language);
747 } else if (Name.equals_lower("webkit")) {
748 *Style = getWebKitStyle();
749 } else if (Name.equals_lower("gnu")) {
750 *Style = getGNUStyle();
751 } else if (Name.equals_lower("none")) {
752 *Style = getNoStyle();
757 Style->Language = Language;
761 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
763 FormatStyle::LanguageKind Language = Style->Language;
764 assert(Language != FormatStyle::LK_None);
765 if (Text.trim().empty())
766 return make_error_code(ParseError::Error);
768 std::vector<FormatStyle> Styles;
769 llvm::yaml::Input Input(Text);
770 // DocumentListTraits<vector<FormatStyle>> uses the context to get default
771 // values for the fields, keys for which are missing from the configuration.
772 // Mapping also uses the context to get the language to find the correct
774 Input.setContext(Style);
777 return Input.error();
779 for (unsigned i = 0; i < Styles.size(); ++i) {
780 // Ensures that only the first configuration can skip the Language option.
781 if (Styles[i].Language == FormatStyle::LK_None && i != 0)
782 return make_error_code(ParseError::Error);
783 // Ensure that each language is configured at most once.
784 for (unsigned j = 0; j < i; ++j) {
785 if (Styles[i].Language == Styles[j].Language) {
787 << "Duplicate languages in the config file on positions " << j
788 << " and " << i << "\n");
789 return make_error_code(ParseError::Error);
793 // Look for a suitable configuration starting from the end, so we can
794 // find the configuration for the specific language first, and the default
795 // configuration (which can only be at slot 0) after it.
796 for (int i = Styles.size() - 1; i >= 0; --i) {
797 if (Styles[i].Language == Language ||
798 Styles[i].Language == FormatStyle::LK_None) {
800 Style->Language = Language;
801 return make_error_code(ParseError::Success);
804 return make_error_code(ParseError::Unsuitable);
807 std::string configurationAsText(const FormatStyle &Style) {
809 llvm::raw_string_ostream Stream(Text);
810 llvm::yaml::Output Output(Stream);
811 // We use the same mapping method for input and output, so we need a non-const
813 FormatStyle NonConstStyle = expandPresets(Style);
814 Output << NonConstStyle;
820 class JavaScriptRequoter : public TokenAnalyzer {
822 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
823 : TokenAnalyzer(Env, Style) {}
825 tooling::Replacements
826 analyze(TokenAnnotator &Annotator,
827 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
828 FormatTokenLexer &Tokens) override {
829 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
830 AnnotatedLines.end());
831 tooling::Replacements Result;
832 requoteJSStringLiteral(AnnotatedLines, Result);
837 // Replaces double/single-quoted string literal as appropriate, re-escaping
838 // the contents in the process.
839 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
840 tooling::Replacements &Result) {
841 for (AnnotatedLine *Line : Lines) {
842 requoteJSStringLiteral(Line->Children, Result);
845 for (FormatToken *FormatTok = Line->First; FormatTok;
846 FormatTok = FormatTok->Next) {
847 StringRef Input = FormatTok->TokenText;
848 if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
849 // NB: testing for not starting with a double quote to avoid
850 // breaking `template strings`.
851 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
852 !Input.startswith("\"")) ||
853 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
854 !Input.startswith("\'")))
857 // Change start and end quote.
858 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
859 SourceLocation Start = FormatTok->Tok.getLocation();
860 auto Replace = [&](SourceLocation Start, unsigned Length,
861 StringRef ReplacementText) {
862 auto Err = Result.add(tooling::Replacement(
863 Env.getSourceManager(), Start, Length, ReplacementText));
864 // FIXME: handle error. For now, print error message and skip the
865 // replacement for release version.
867 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
871 Replace(Start, 1, IsSingle ? "'" : "\"");
872 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
873 IsSingle ? "'" : "\"");
875 // Escape internal quotes.
876 bool Escaped = false;
877 for (size_t i = 1; i < Input.size() - 1; i++) {
880 if (!Escaped && i + 1 < Input.size() &&
881 ((IsSingle && Input[i + 1] == '"') ||
882 (!IsSingle && Input[i + 1] == '\''))) {
883 // Remove this \, it's escaping a " or ' that no longer needs
885 Replace(Start.getLocWithOffset(i), 1, "");
892 if (!Escaped && IsSingle == (Input[i] == '\'')) {
894 Replace(Start.getLocWithOffset(i), 0, "\\");
908 class Formatter : public TokenAnalyzer {
910 Formatter(const Environment &Env, const FormatStyle &Style,
911 FormattingAttemptStatus *Status)
912 : TokenAnalyzer(Env, Style), Status(Status) {}
914 tooling::Replacements
915 analyze(TokenAnnotator &Annotator,
916 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
917 FormatTokenLexer &Tokens) override {
918 tooling::Replacements Result;
919 deriveLocalStyle(AnnotatedLines);
920 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
921 AnnotatedLines.end());
922 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
923 Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
925 Annotator.setCommentLineLevels(AnnotatedLines);
927 WhitespaceManager Whitespaces(
928 Env.getSourceManager(), Style,
929 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
930 ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
931 Env.getSourceManager(), Whitespaces, Encoding,
932 BinPackInconclusiveFunctions);
933 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
934 Env.getSourceManager(), Status)
935 .format(AnnotatedLines);
936 for (const auto &R : Whitespaces.generateReplacements())
944 static bool inputUsesCRLF(StringRef Text) {
945 return Text.count('\r') * 2 > Text.count('\n');
949 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
950 for (const AnnotatedLine *Line : Lines) {
951 if (hasCpp03IncompatibleFormat(Line->Children))
953 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
954 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
955 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
957 if (Tok->is(TT_TemplateCloser) &&
958 Tok->Previous->is(TT_TemplateCloser))
966 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
967 int AlignmentDiff = 0;
968 for (const AnnotatedLine *Line : Lines) {
969 AlignmentDiff += countVariableAlignments(Line->Children);
970 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
971 if (!Tok->is(TT_PointerOrReference))
974 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
975 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
976 Tok->Next->WhitespaceRange.getEnd();
977 if (SpaceBefore && !SpaceAfter)
979 if (!SpaceBefore && SpaceAfter)
983 return AlignmentDiff;
987 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
988 bool HasBinPackedFunction = false;
989 bool HasOnePerLineFunction = false;
990 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
991 if (!AnnotatedLines[i]->First->Next)
993 FormatToken *Tok = AnnotatedLines[i]->First->Next;
995 if (Tok->PackingKind == PPK_BinPacked)
996 HasBinPackedFunction = true;
997 if (Tok->PackingKind == PPK_OnePerLine)
998 HasOnePerLineFunction = true;
1003 if (Style.DerivePointerAlignment)
1004 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
1005 ? FormatStyle::PAS_Left
1006 : FormatStyle::PAS_Right;
1007 if (Style.Standard == FormatStyle::LS_Auto)
1008 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
1009 ? FormatStyle::LS_Cpp11
1010 : FormatStyle::LS_Cpp03;
1011 BinPackInconclusiveFunctions =
1012 HasBinPackedFunction || !HasOnePerLineFunction;
1015 bool BinPackInconclusiveFunctions;
1016 FormattingAttemptStatus *Status;
1019 // This class clean up the erroneous/redundant code around the given ranges in
1021 class Cleaner : public TokenAnalyzer {
1023 Cleaner(const Environment &Env, const FormatStyle &Style)
1024 : TokenAnalyzer(Env, Style),
1025 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1027 // FIXME: eliminate unused parameters.
1028 tooling::Replacements
1029 analyze(TokenAnnotator &Annotator,
1030 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1031 FormatTokenLexer &Tokens) override {
1032 // FIXME: in the current implementation the granularity of affected range
1033 // is an annotated line. However, this is not sufficient. Furthermore,
1034 // redundant code introduced by replacements does not necessarily
1035 // intercept with ranges of replacements that result in the redundancy.
1036 // To determine if some redundant code is actually introduced by
1037 // replacements(e.g. deletions), we need to come up with a more
1038 // sophisticated way of computing affected ranges.
1039 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1040 AnnotatedLines.end());
1042 checkEmptyNamespace(AnnotatedLines);
1044 for (auto &Line : AnnotatedLines) {
1045 if (Line->Affected) {
1046 cleanupRight(Line->First, tok::comma, tok::comma);
1047 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1048 cleanupRight(Line->First, tok::l_paren, tok::comma);
1049 cleanupLeft(Line->First, tok::comma, tok::r_paren);
1050 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1051 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1052 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1056 return generateFixes();
1060 bool containsOnlyComments(const AnnotatedLine &Line) {
1061 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1062 if (Tok->isNot(tok::comment))
1068 // Iterate through all lines and remove any empty (nested) namespaces.
1069 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1070 std::set<unsigned> DeletedLines;
1071 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1072 auto &Line = *AnnotatedLines[i];
1073 if (Line.startsWith(tok::kw_namespace) ||
1074 Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1075 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1079 for (auto Line : DeletedLines) {
1080 FormatToken *Tok = AnnotatedLines[Line]->First;
1088 // The function checks if the namespace, which starts from \p CurrentLine, and
1089 // its nested namespaces are empty and delete them if they are empty. It also
1090 // sets \p NewLine to the last line checked.
1091 // Returns true if the current namespace is empty.
1092 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1093 unsigned CurrentLine, unsigned &NewLine,
1094 std::set<unsigned> &DeletedLines) {
1095 unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1096 if (Style.BraceWrapping.AfterNamespace) {
1097 // If the left brace is in a new line, we should consume it first so that
1098 // it does not make the namespace non-empty.
1099 // FIXME: error handling if there is no left brace.
1100 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1101 NewLine = CurrentLine;
1104 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1107 while (++CurrentLine < End) {
1108 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1111 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1112 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1113 tok::kw_namespace)) {
1114 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1117 CurrentLine = NewLine;
1121 if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1124 // If there is anything other than comments or nested namespaces in the
1125 // current namespace, the namespace cannot be empty.
1126 NewLine = CurrentLine;
1130 NewLine = CurrentLine;
1131 if (CurrentLine >= End)
1134 // Check if the empty namespace is actually affected by changed ranges.
1135 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1136 AnnotatedLines[InitLine]->First->Tok.getLocation(),
1137 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1140 for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1141 DeletedLines.insert(i);
1147 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1148 // of the token in the pair if the left token has \p LK token kind and the
1149 // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1150 // is deleted on match; otherwise, the right token is deleted.
1151 template <typename LeftKind, typename RightKind>
1152 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1154 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1155 for (auto *Res = Tok.Next; Res; Res = Res->Next)
1156 if (!Res->is(tok::comment) &&
1157 DeletedTokens.find(Res) == DeletedTokens.end())
1161 for (auto *Left = Start; Left;) {
1162 auto *Right = NextNotDeleted(*Left);
1165 if (Left->is(LK) && Right->is(RK)) {
1166 deleteToken(DeleteLeft ? Left : Right);
1167 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1169 // If the right token is deleted, we should keep the left token
1170 // unchanged and pair it with the new right token.
1178 template <typename LeftKind, typename RightKind>
1179 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1180 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1183 template <typename LeftKind, typename RightKind>
1184 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1185 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1188 // Delete the given token.
1189 inline void deleteToken(FormatToken *Tok) {
1191 DeletedTokens.insert(Tok);
1194 tooling::Replacements generateFixes() {
1195 tooling::Replacements Fixes;
1196 std::vector<FormatToken *> Tokens;
1197 std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1198 std::back_inserter(Tokens));
1200 // Merge multiple continuous token deletions into one big deletion so that
1201 // the number of replacements can be reduced. This makes computing affected
1202 // ranges more efficient when we run reformat on the changed code.
1204 while (Idx < Tokens.size()) {
1205 unsigned St = Idx, End = Idx;
1206 while ((End + 1) < Tokens.size() &&
1207 Tokens[End]->Next == Tokens[End + 1]) {
1210 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1211 Tokens[End]->Tok.getEndLoc());
1213 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1214 // FIXME: better error handling. for now just print error message and skip
1215 // for the release version.
1217 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1218 assert(false && "Fixes must not conflict!");
1226 // Class for less-than inequality comparason for the set `RedundantTokens`.
1227 // We store tokens in the order they appear in the translation unit so that
1228 // we do not need to sort them in `generateFixes()`.
1229 struct FormatTokenLess {
1230 FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1232 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1233 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1234 RHS->Tok.getLocation());
1236 const SourceManager &SM;
1239 // Tokens to be deleted.
1240 std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1243 struct IncludeDirective {
1250 } // end anonymous namespace
1252 // Determines whether 'Ranges' intersects with ('Start', 'End').
1253 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1255 for (auto Range : Ranges) {
1256 if (Range.getOffset() < End &&
1257 Range.getOffset() + Range.getLength() > Start)
1263 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1264 // before sorting/deduplicating. Index is the index of the include under the
1265 // cursor in the original set of includes. If this include has duplicates, it is
1266 // the index of the first of the duplicates as the others are going to be
1267 // removed. OffsetToEOL describes the cursor's position relative to the end of
1268 // its current line.
1269 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1270 static std::pair<unsigned, unsigned>
1271 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1272 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1273 unsigned CursorIndex = UINT_MAX;
1274 unsigned OffsetToEOL = 0;
1275 for (int i = 0, e = Includes.size(); i != e; ++i) {
1276 unsigned Start = Includes[Indices[i]].Offset;
1277 unsigned End = Start + Includes[Indices[i]].Text.size();
1278 if (!(Cursor >= Start && Cursor < End))
1280 CursorIndex = Indices[i];
1281 OffsetToEOL = End - Cursor;
1282 // Put the cursor on the only remaining #include among the duplicate
1284 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1288 return std::make_pair(CursorIndex, OffsetToEOL);
1291 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1292 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1294 // #include directives with the same text will be deduplicated, and only the
1295 // first #include in the duplicate #includes remains. If the `Cursor` is
1296 // provided and put on a deleted #include, it will be moved to the remaining
1297 // #include in the duplicate #includes.
1298 static void sortCppIncludes(const FormatStyle &Style,
1299 const SmallVectorImpl<IncludeDirective> &Includes,
1300 ArrayRef<tooling::Range> Ranges, StringRef FileName,
1301 tooling::Replacements &Replaces, unsigned *Cursor) {
1302 unsigned IncludesBeginOffset = Includes.front().Offset;
1303 unsigned IncludesEndOffset =
1304 Includes.back().Offset + Includes.back().Text.size();
1305 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1306 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1308 SmallVector<unsigned, 16> Indices;
1309 for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1310 Indices.push_back(i);
1312 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1313 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1314 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1316 // The index of the include on which the cursor will be put after
1317 // sorting/deduplicating.
1318 unsigned CursorIndex;
1319 // The offset from cursor to the end of line.
1320 unsigned CursorToEOLOffset;
1322 std::tie(CursorIndex, CursorToEOLOffset) =
1323 FindCursorIndex(Includes, Indices, *Cursor);
1325 // Deduplicate #includes.
1326 Indices.erase(std::unique(Indices.begin(), Indices.end(),
1327 [&](unsigned LHSI, unsigned RHSI) {
1328 return Includes[LHSI].Text == Includes[RHSI].Text;
1332 // If the #includes are out of order, we generate a single replacement fixing
1333 // the entire block. Otherwise, no replacement is generated.
1334 if (Indices.size() == Includes.size() &&
1335 std::is_sorted(Indices.begin(), Indices.end()))
1339 for (unsigned Index : Indices) {
1340 if (!result.empty())
1342 result += Includes[Index].Text;
1343 if (Cursor && CursorIndex == Index)
1344 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1347 auto Err = Replaces.add(tooling::Replacement(
1348 FileName, Includes.front().Offset, IncludesBlockSize, result));
1349 // FIXME: better error handling. For now, just skip the replacement for the
1352 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1359 // This class manages priorities of #include categories and calculates
1360 // priorities for headers.
1361 class IncludeCategoryManager {
1363 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1364 : Style(Style), FileName(FileName) {
1365 FileStem = llvm::sys::path::stem(FileName);
1366 for (const auto &Category : Style.IncludeCategories)
1367 CategoryRegexs.emplace_back(Category.Regex);
1368 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1369 FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1370 FileName.endswith(".cxx") || FileName.endswith(".m") ||
1371 FileName.endswith(".mm");
1374 // Returns the priority of the category which \p IncludeName belongs to.
1375 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1376 // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1377 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1379 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1380 if (CategoryRegexs[i].match(IncludeName)) {
1381 Ret = Style.IncludeCategories[i].Priority;
1384 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1390 bool isMainHeader(StringRef IncludeName) const {
1391 if (!IncludeName.startswith("\""))
1393 StringRef HeaderStem =
1394 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1395 if (FileStem.startswith(HeaderStem)) {
1396 llvm::Regex MainIncludeRegex(
1397 (HeaderStem + Style.IncludeIsMainRegex).str());
1398 if (MainIncludeRegex.match(FileStem))
1404 const FormatStyle &Style;
1408 SmallVector<llvm::Regex, 4> CategoryRegexs;
1411 const char IncludeRegexPattern[] =
1412 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1414 } // anonymous namespace
1416 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1417 ArrayRef<tooling::Range> Ranges,
1419 tooling::Replacements &Replaces,
1422 unsigned SearchFrom = 0;
1423 llvm::Regex IncludeRegex(IncludeRegexPattern);
1424 SmallVector<StringRef, 4> Matches;
1425 SmallVector<IncludeDirective, 16> IncludesInBlock;
1427 // In compiled files, consider the first #include to be the main #include of
1428 // the file if it is not a system #include. This ensures that the header
1429 // doesn't have hidden dependencies
1430 // (http://llvm.org/docs/CodingStandards.html#include-style).
1432 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1433 // cases where the first #include is unlikely to be the main header.
1434 IncludeCategoryManager Categories(Style, FileName);
1435 bool FirstIncludeBlock = true;
1436 bool MainIncludeFound = false;
1437 bool FormattingOff = false;
1440 auto Pos = Code.find('\n', SearchFrom);
1442 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1444 StringRef Trimmed = Line.trim();
1445 if (Trimmed == "// clang-format off")
1446 FormattingOff = true;
1447 else if (Trimmed == "// clang-format on")
1448 FormattingOff = false;
1450 if (!FormattingOff && !Line.endswith("\\")) {
1451 if (IncludeRegex.match(Line, &Matches)) {
1452 StringRef IncludeName = Matches[2];
1453 int Category = Categories.getIncludePriority(
1455 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1457 MainIncludeFound = true;
1458 IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1459 } else if (!IncludesInBlock.empty()) {
1460 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1462 IncludesInBlock.clear();
1463 FirstIncludeBlock = false;
1467 if (Pos == StringRef::npos || Pos + 1 == Code.size())
1469 SearchFrom = Pos + 1;
1471 if (!IncludesInBlock.empty())
1472 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1476 bool isMpegTS(StringRef Code) {
1477 // MPEG transport streams use the ".ts" file extension. clang-format should
1478 // not attempt to format those. MPEG TS' frame format starts with 0x47 every
1479 // 189 bytes - detect that and return.
1480 return Code.size() > 188 && Code[0] == 0x47 && Code[188] == 0x47;
1483 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1484 ArrayRef<tooling::Range> Ranges,
1485 StringRef FileName, unsigned *Cursor) {
1486 tooling::Replacements Replaces;
1487 if (!Style.SortIncludes)
1489 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript &&
1492 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1493 return sortJavaScriptImports(Style, Code, Ranges, FileName);
1494 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1498 template <typename T>
1499 static llvm::Expected<tooling::Replacements>
1500 processReplacements(T ProcessFunc, StringRef Code,
1501 const tooling::Replacements &Replaces,
1502 const FormatStyle &Style) {
1503 if (Replaces.empty())
1504 return tooling::Replacements();
1506 auto NewCode = applyAllReplacements(Code, Replaces);
1508 return NewCode.takeError();
1509 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1510 StringRef FileName = Replaces.begin()->getFilePath();
1512 tooling::Replacements FormatReplaces =
1513 ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1515 return Replaces.merge(FormatReplaces);
1518 llvm::Expected<tooling::Replacements>
1519 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1520 const FormatStyle &Style) {
1521 // We need to use lambda function here since there are two versions of
1523 auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1524 std::vector<tooling::Range> Ranges,
1525 StringRef FileName) -> tooling::Replacements {
1526 return sortIncludes(Style, Code, Ranges, FileName);
1528 auto SortedReplaces =
1529 processReplacements(SortIncludes, Code, Replaces, Style);
1530 if (!SortedReplaces)
1531 return SortedReplaces.takeError();
1533 // We need to use lambda function here since there are two versions of
1535 auto Reformat = [](const FormatStyle &Style, StringRef Code,
1536 std::vector<tooling::Range> Ranges,
1537 StringRef FileName) -> tooling::Replacements {
1538 return reformat(Style, Code, Ranges, FileName);
1540 return processReplacements(Reformat, Code, *SortedReplaces, Style);
1545 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1546 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1547 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1550 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1551 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1554 // Returns the offset after skipping a sequence of tokens, matched by \p
1555 // GetOffsetAfterSequence, from the start of the code.
1556 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1557 // tokens and returns an offset after the sequence.
1558 unsigned getOffsetAfterTokenSequence(
1559 StringRef FileName, StringRef Code, const FormatStyle &Style,
1560 llvm::function_ref<unsigned(const SourceManager &, Lexer &, Token &)>
1561 GetOffsetAfterSequence) {
1562 std::unique_ptr<Environment> Env =
1563 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1564 const SourceManager &SourceMgr = Env->getSourceManager();
1565 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1566 getFormattingLangOpts(Style));
1568 // Get the first token.
1569 Lex.LexFromRawLexer(Tok);
1570 return GetOffsetAfterSequence(SourceMgr, Lex, Tok);
1573 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1574 // \p Tok will be the token after this directive; otherwise, it can be any token
1575 // after the given \p Tok (including \p Tok).
1576 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1577 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1578 Tok.is(tok::raw_identifier) &&
1579 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1580 Tok.is(tok::raw_identifier);
1582 Lex.LexFromRawLexer(Tok);
1586 void skipComments(Lexer &Lex, Token &Tok) {
1587 while (Tok.is(tok::comment))
1588 if (Lex.LexFromRawLexer(Tok))
1592 // Returns the offset after header guard directives and any comments
1593 // before/after header guards. If no header guard presents in the code, this
1594 // will returns the offset after skipping all comments from the start of the
1596 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1598 const FormatStyle &Style) {
1599 return getOffsetAfterTokenSequence(
1600 FileName, Code, Style,
1601 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1602 skipComments(Lex, Tok);
1603 unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1604 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1605 skipComments(Lex, Tok);
1606 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1607 return SM.getFileOffset(Tok.getLocation());
1609 return InitialOffset;
1613 // Check if a sequence of tokens is like
1614 // "#include ("header.h" | <header.h>)".
1615 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1616 // any token after the given \p Tok (including \p Tok).
1617 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1618 auto Matched = [&]() {
1619 Lex.LexFromRawLexer(Tok);
1622 if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1623 Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1624 if (Lex.LexFromRawLexer(Tok))
1626 if (Tok.is(tok::string_literal))
1628 if (Tok.is(tok::less)) {
1629 while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1631 if (Tok.is(tok::greater))
1638 // Returns the offset of the last #include directive after which a new
1639 // #include can be inserted. This ignores #include's after the #include block(s)
1640 // in the beginning of a file to avoid inserting headers into code sections
1641 // where new #include's should not be added by default.
1642 // These code sections include:
1643 // - raw string literals (containing #include).
1645 // - Special #include's among declarations (e.g. functions).
1647 // If no #include after which a new #include can be inserted, this returns the
1648 // offset after skipping all comments from the start of the code.
1649 // Inserting after an #include is not allowed if it comes after code that is not
1650 // #include (e.g. pre-processing directive that is not #include, declarations).
1651 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1652 const FormatStyle &Style) {
1653 return getOffsetAfterTokenSequence(
1654 FileName, Code, Style,
1655 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1656 skipComments(Lex, Tok);
1657 unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1658 while (checkAndConsumeInclusiveDirective(Lex, Tok))
1659 MaxOffset = SM.getFileOffset(Tok.getLocation());
1664 bool isDeletedHeader(llvm::StringRef HeaderName,
1665 const std::set<llvm::StringRef> &HeadersToDelete) {
1666 return HeadersToDelete.count(HeaderName) ||
1667 HeadersToDelete.count(HeaderName.trim("\"<>"));
1670 // FIXME: insert empty lines between newly created blocks.
1671 tooling::Replacements
1672 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1673 const FormatStyle &Style) {
1677 tooling::Replacements HeaderInsertions;
1678 std::set<llvm::StringRef> HeadersToDelete;
1679 tooling::Replacements Result;
1680 for (const auto &R : Replaces) {
1681 if (isHeaderInsertion(R)) {
1682 // Replacements from \p Replaces must be conflict-free already, so we can
1683 // simply consume the error.
1684 llvm::consumeError(HeaderInsertions.add(R));
1685 } else if (isHeaderDeletion(R)) {
1686 HeadersToDelete.insert(R.getReplacementText());
1687 } else if (R.getOffset() == UINT_MAX) {
1688 llvm::errs() << "Insertions other than header #include insertion are "
1690 << R.getReplacementText() << "\n";
1692 llvm::consumeError(Result.add(R));
1695 if (HeaderInsertions.empty() && HeadersToDelete.empty())
1698 llvm::Regex IncludeRegex(IncludeRegexPattern);
1699 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1700 SmallVector<StringRef, 4> Matches;
1702 StringRef FileName = Replaces.begin()->getFilePath();
1703 IncludeCategoryManager Categories(Style, FileName);
1705 // Record the offset of the end of the last include in each category.
1706 std::map<int, int> CategoryEndOffsets;
1707 // All possible priorities.
1708 // Add 0 for main header and INT_MAX for headers that are not in any category.
1709 std::set<int> Priorities = {0, INT_MAX};
1710 for (const auto &Category : Style.IncludeCategories)
1711 Priorities.insert(Category.Priority);
1712 int FirstIncludeOffset = -1;
1713 // All new headers should be inserted after this offset.
1714 unsigned MinInsertOffset =
1715 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1716 StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1717 // Max insertion offset in the original code.
1718 unsigned MaxInsertOffset =
1720 getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1721 SmallVector<StringRef, 32> Lines;
1722 TrimmedCode.split(Lines, '\n');
1723 unsigned Offset = MinInsertOffset;
1724 unsigned NextLineOffset;
1725 std::set<StringRef> ExistingIncludes;
1726 for (auto Line : Lines) {
1727 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1728 if (IncludeRegex.match(Line, &Matches)) {
1729 // The header name with quotes or angle brackets.
1730 StringRef IncludeName = Matches[2];
1731 ExistingIncludes.insert(IncludeName);
1732 // Only record the offset of current #include if we can insert after it.
1733 if (Offset <= MaxInsertOffset) {
1734 int Category = Categories.getIncludePriority(
1735 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1736 CategoryEndOffsets[Category] = NextLineOffset;
1737 if (FirstIncludeOffset < 0)
1738 FirstIncludeOffset = Offset;
1740 if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1741 // If this is the last line without trailing newline, we need to make
1742 // sure we don't delete across the file boundary.
1743 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1745 Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1747 // Ignore the deletion on conflict.
1748 llvm::errs() << "Failed to add header deletion replacement for "
1749 << IncludeName << ": " << llvm::toString(std::move(Err))
1754 Offset = NextLineOffset;
1757 // Populate CategoryEndOfssets:
1758 // - Ensure that CategoryEndOffset[Highest] is always populated.
1759 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1760 // is set, up to CategoryEndOffset[Highest].
1761 auto Highest = Priorities.begin();
1762 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1763 if (FirstIncludeOffset >= 0)
1764 CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1766 CategoryEndOffsets[*Highest] = MinInsertOffset;
1768 // By this point, CategoryEndOffset[Highest] is always set appropriately:
1769 // - to an appropriate location before/after existing #includes, or
1770 // - to right after the header guard, or
1771 // - to the beginning of the file.
1772 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1773 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1774 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1776 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1777 for (const auto &R : HeaderInsertions) {
1778 auto IncludeDirective = R.getReplacementText();
1779 bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1780 assert(Matched && "Header insertion replacement must have replacement text "
1783 auto IncludeName = Matches[2];
1784 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1785 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1790 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1791 Offset = CategoryEndOffsets[Category];
1792 std::string NewInclude = !IncludeDirective.endswith("\n")
1793 ? (IncludeDirective + "\n").str()
1794 : IncludeDirective.str();
1795 // When inserting headers at end of the code, also append '\n' to the code
1796 // if it does not end with '\n'.
1797 if (NeedNewLineAtEnd && Offset == Code.size()) {
1798 NewInclude = "\n" + NewInclude;
1799 NeedNewLineAtEnd = false;
1801 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1802 auto Err = Result.add(NewReplace);
1804 llvm::consumeError(std::move(Err));
1805 unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1806 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1807 Result = Result.merge(tooling::Replacements(NewReplace));
1813 } // anonymous namespace
1815 llvm::Expected<tooling::Replacements>
1816 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1817 const FormatStyle &Style) {
1818 // We need to use lambda function here since there are two versions of
1820 auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1821 std::vector<tooling::Range> Ranges,
1822 StringRef FileName) -> tooling::Replacements {
1823 return cleanup(Style, Code, Ranges, FileName);
1825 // Make header insertion replacements insert new headers into correct blocks.
1826 tooling::Replacements NewReplaces =
1827 fixCppIncludeInsertions(Code, Replaces, Style);
1828 return processReplacements(Cleanup, Code, NewReplaces, Style);
1831 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1832 ArrayRef<tooling::Range> Ranges,
1834 FormattingAttemptStatus *Status) {
1835 FormatStyle Expanded = expandPresets(Style);
1836 if (Expanded.DisableFormat)
1837 return tooling::Replacements();
1838 if (Expanded.Language == FormatStyle::LK_JavaScript && isMpegTS(Code))
1839 return tooling::Replacements();
1840 auto Env = Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1842 auto reformatAfterApplying = [&] (TokenAnalyzer& Fixer) {
1843 tooling::Replacements Fixes = Fixer.process();
1844 if (!Fixes.empty()) {
1845 auto NewCode = applyAllReplacements(Code, Fixes);
1847 auto NewEnv = Environment::CreateVirtualEnvironment(
1849 tooling::calculateRangesAfterReplacements(Fixes, Ranges));
1850 Formatter Format(*NewEnv, Expanded, Status);
1851 return Fixes.merge(Format.process());
1854 Formatter Format(*Env, Expanded, Status);
1855 return Format.process();
1858 if (Style.Language == FormatStyle::LK_Cpp &&
1859 Style.FixNamespaceComments) {
1860 NamespaceEndCommentsFixer CommentsFixer(*Env, Expanded);
1861 return reformatAfterApplying(CommentsFixer);
1864 if (Style.Language == FormatStyle::LK_JavaScript &&
1865 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) {
1866 JavaScriptRequoter Requoter(*Env, Expanded);
1867 return reformatAfterApplying(Requoter);
1870 Formatter Format(*Env, Expanded, Status);
1871 return Format.process();
1874 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
1875 ArrayRef<tooling::Range> Ranges,
1876 StringRef FileName) {
1877 std::unique_ptr<Environment> Env =
1878 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1879 Cleaner Clean(*Env, Style);
1880 return Clean.process();
1883 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1884 ArrayRef<tooling::Range> Ranges,
1885 StringRef FileName, bool *IncompleteFormat) {
1886 FormattingAttemptStatus Status;
1887 auto Result = reformat(Style, Code, Ranges, FileName, &Status);
1888 if (!Status.FormatComplete)
1889 *IncompleteFormat = true;
1893 tooling::Replacements fixNamespaceEndComments(const FormatStyle &Style,
1895 ArrayRef<tooling::Range> Ranges,
1896 StringRef FileName) {
1897 std::unique_ptr<Environment> Env =
1898 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1899 NamespaceEndCommentsFixer Fix(*Env, Style);
1900 return Fix.process();
1903 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1904 LangOptions LangOpts;
1905 LangOpts.CPlusPlus = 1;
1906 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1907 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1908 LangOpts.CPlusPlus1z = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1909 LangOpts.LineComment = 1;
1910 bool AlternativeOperators = Style.isCpp();
1911 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1915 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally.
1916 LangOpts.DeclSpecKeyword = 1; // To get __declspec.
1920 const char *StyleOptionHelpDescription =
1921 "Coding style, currently supports:\n"
1922 " LLVM, Google, Chromium, Mozilla, WebKit.\n"
1923 "Use -style=file to load style configuration from\n"
1924 ".clang-format file located in one of the parent\n"
1925 "directories of the source file (or current\n"
1926 "directory for stdin).\n"
1927 "Use -style=\"{key: value, ...}\" to set specific\n"
1928 "parameters, e.g.:\n"
1929 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1931 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1932 if (FileName.endswith(".java"))
1933 return FormatStyle::LK_Java;
1934 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
1935 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
1936 if (FileName.endswith(".m") || FileName.endswith(".mm"))
1937 return FormatStyle::LK_ObjC;
1938 if (FileName.endswith_lower(".proto") ||
1939 FileName.endswith_lower(".protodevel"))
1940 return FormatStyle::LK_Proto;
1941 if (FileName.endswith_lower(".td"))
1942 return FormatStyle::LK_TableGen;
1943 return FormatStyle::LK_Cpp;
1946 llvm::Expected<FormatStyle> getStyle(StringRef StyleName, StringRef FileName,
1947 StringRef FallbackStyleName,
1948 StringRef Code, vfs::FileSystem *FS) {
1950 FS = vfs::getRealFileSystem().get();
1952 FormatStyle Style = getLLVMStyle();
1953 Style.Language = getLanguageByFileName(FileName);
1955 // This is a very crude detection of whether a header contains ObjC code that
1956 // should be improved over time and probably be done on tokens, not one the
1957 // bare content of the file.
1958 if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
1959 (Code.contains("\n- (") || Code.contains("\n+ (")))
1960 Style.Language = FormatStyle::LK_ObjC;
1962 FormatStyle FallbackStyle = getNoStyle();
1963 if (!getPredefinedStyle(FallbackStyleName, Style.Language, &FallbackStyle))
1964 return make_string_error("Invalid fallback style \"" + FallbackStyleName);
1966 if (StyleName.startswith("{")) {
1967 // Parse YAML/JSON style from the command line.
1968 if (std::error_code ec = parseConfiguration(StyleName, &Style))
1969 return make_string_error("Error parsing -style: " + ec.message());
1973 if (!StyleName.equals_lower("file")) {
1974 if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1975 return make_string_error("Invalid value for -style");
1979 // Look for .clang-format/_clang-format file in the file's parent directories.
1980 SmallString<128> UnsuitableConfigFiles;
1981 SmallString<128> Path(FileName);
1982 if (std::error_code EC = FS->makeAbsolute(Path))
1983 return make_string_error(EC.message());
1985 for (StringRef Directory = Path; !Directory.empty();
1986 Directory = llvm::sys::path::parent_path(Directory)) {
1988 auto Status = FS->status(Directory);
1990 Status->getType() != llvm::sys::fs::file_type::directory_file) {
1994 SmallString<128> ConfigFile(Directory);
1996 llvm::sys::path::append(ConfigFile, ".clang-format");
1997 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1999 Status = FS->status(ConfigFile.str());
2000 bool FoundConfigFile =
2001 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
2002 if (!FoundConfigFile) {
2003 // Try _clang-format too, since dotfiles are not commonly used on Windows.
2004 ConfigFile = Directory;
2005 llvm::sys::path::append(ConfigFile, "_clang-format");
2006 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2007 Status = FS->status(ConfigFile.str());
2008 FoundConfigFile = Status && (Status->getType() ==
2009 llvm::sys::fs::file_type::regular_file);
2012 if (FoundConfigFile) {
2013 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
2014 FS->getBufferForFile(ConfigFile.str());
2015 if (std::error_code EC = Text.getError())
2016 return make_string_error(EC.message());
2017 if (std::error_code ec =
2018 parseConfiguration(Text.get()->getBuffer(), &Style)) {
2019 if (ec == ParseError::Unsuitable) {
2020 if (!UnsuitableConfigFiles.empty())
2021 UnsuitableConfigFiles.append(", ");
2022 UnsuitableConfigFiles.append(ConfigFile);
2025 return make_string_error("Error reading " + ConfigFile + ": " +
2028 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
2032 if (!UnsuitableConfigFiles.empty())
2033 return make_string_error("Configuration file(s) do(es) not support " +
2034 getLanguageName(Style.Language) + ": " +
2035 UnsuitableConfigFiles);
2036 return FallbackStyle;
2039 } // namespace format
2040 } // namespace clang