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 "SortJavaScriptImports.h"
21 #include "TokenAnalyzer.h"
22 #include "TokenAnnotator.h"
23 #include "UnwrappedLineFormatter.h"
24 #include "UnwrappedLineParser.h"
25 #include "WhitespaceManager.h"
26 #include "clang/Basic/Diagnostic.h"
27 #include "clang/Basic/DiagnosticOptions.h"
28 #include "clang/Basic/SourceManager.h"
29 #include "clang/Basic/VirtualFileSystem.h"
30 #include "clang/Lex/Lexer.h"
31 #include "llvm/ADT/STLExtras.h"
32 #include "llvm/Support/Allocator.h"
33 #include "llvm/Support/Debug.h"
34 #include "llvm/Support/Path.h"
35 #include "llvm/Support/Regex.h"
36 #include "llvm/Support/YAMLTraits.h"
41 #define DEBUG_TYPE "format-formatter"
43 using clang::format::FormatStyle;
45 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string)
46 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory)
50 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
51 static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
52 IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
53 IO.enumCase(Value, "Java", FormatStyle::LK_Java);
54 IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
55 IO.enumCase(Value, "ObjC", FormatStyle::LK_ObjC);
56 IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
57 IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen);
61 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
62 static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
63 IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
64 IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
65 IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
66 IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
67 IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
71 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
72 static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
73 IO.enumCase(Value, "Never", FormatStyle::UT_Never);
74 IO.enumCase(Value, "false", FormatStyle::UT_Never);
75 IO.enumCase(Value, "Always", FormatStyle::UT_Always);
76 IO.enumCase(Value, "true", FormatStyle::UT_Always);
77 IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
78 IO.enumCase(Value, "ForContinuationAndIndentation",
79 FormatStyle::UT_ForContinuationAndIndentation);
83 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> {
84 static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) {
85 IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave);
86 IO.enumCase(Value, "Single", FormatStyle::JSQS_Single);
87 IO.enumCase(Value, "Double", FormatStyle::JSQS_Double);
91 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
92 static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
93 IO.enumCase(Value, "None", FormatStyle::SFS_None);
94 IO.enumCase(Value, "false", FormatStyle::SFS_None);
95 IO.enumCase(Value, "All", FormatStyle::SFS_All);
96 IO.enumCase(Value, "true", FormatStyle::SFS_All);
97 IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
98 IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
102 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
103 static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
104 IO.enumCase(Value, "All", FormatStyle::BOS_All);
105 IO.enumCase(Value, "true", FormatStyle::BOS_All);
106 IO.enumCase(Value, "None", FormatStyle::BOS_None);
107 IO.enumCase(Value, "false", FormatStyle::BOS_None);
108 IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
112 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
113 static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
114 IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
115 IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
116 IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla);
117 IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
118 IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
119 IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
120 IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit);
121 IO.enumCase(Value, "Custom", FormatStyle::BS_Custom);
126 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> {
127 static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) {
128 IO.enumCase(Value, "None", FormatStyle::RTBS_None);
129 IO.enumCase(Value, "All", FormatStyle::RTBS_All);
130 IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel);
131 IO.enumCase(Value, "TopLevelDefinitions",
132 FormatStyle::RTBS_TopLevelDefinitions);
133 IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions);
138 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> {
140 enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) {
141 IO.enumCase(Value, "None", FormatStyle::DRTBS_None);
142 IO.enumCase(Value, "All", FormatStyle::DRTBS_All);
143 IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel);
145 // For backward compatibility.
146 IO.enumCase(Value, "false", FormatStyle::DRTBS_None);
147 IO.enumCase(Value, "true", FormatStyle::DRTBS_All);
152 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
153 static void enumeration(IO &IO,
154 FormatStyle::NamespaceIndentationKind &Value) {
155 IO.enumCase(Value, "None", FormatStyle::NI_None);
156 IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
157 IO.enumCase(Value, "All", FormatStyle::NI_All);
161 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> {
162 static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) {
163 IO.enumCase(Value, "Align", FormatStyle::BAS_Align);
164 IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign);
165 IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak);
167 // For backward compatibility.
168 IO.enumCase(Value, "true", FormatStyle::BAS_Align);
169 IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign);
173 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
174 static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
175 IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
176 IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
177 IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
179 // For backward compatibility.
180 IO.enumCase(Value, "true", FormatStyle::PAS_Left);
181 IO.enumCase(Value, "false", FormatStyle::PAS_Right);
186 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
187 static void enumeration(IO &IO,
188 FormatStyle::SpaceBeforeParensOptions &Value) {
189 IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
190 IO.enumCase(Value, "ControlStatements",
191 FormatStyle::SBPO_ControlStatements);
192 IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
194 // For backward compatibility.
195 IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
196 IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
200 template <> struct MappingTraits<FormatStyle> {
201 static void mapping(IO &IO, FormatStyle &Style) {
202 // When reading, read the language first, we need it for getPredefinedStyle.
203 IO.mapOptional("Language", Style.Language);
205 if (IO.outputting()) {
206 StringRef StylesArray[] = {"LLVM", "Google", "Chromium",
207 "Mozilla", "WebKit", "GNU"};
208 ArrayRef<StringRef> Styles(StylesArray);
209 for (size_t i = 0, e = Styles.size(); i < e; ++i) {
210 StringRef StyleName(Styles[i]);
211 FormatStyle PredefinedStyle;
212 if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
213 Style == PredefinedStyle) {
214 IO.mapOptional("# BasedOnStyle", StyleName);
219 StringRef BasedOnStyle;
220 IO.mapOptional("BasedOnStyle", BasedOnStyle);
221 if (!BasedOnStyle.empty()) {
222 FormatStyle::LanguageKind OldLanguage = Style.Language;
223 FormatStyle::LanguageKind Language =
224 ((FormatStyle *)IO.getContext())->Language;
225 if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
226 IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
229 Style.Language = OldLanguage;
233 // For backward compatibility.
234 if (!IO.outputting()) {
235 IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
236 IO.mapOptional("IndentFunctionDeclarationAfterType",
237 Style.IndentWrappedFunctionNames);
238 IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
239 IO.mapOptional("SpaceAfterControlStatementKeyword",
240 Style.SpaceBeforeParens);
243 IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
244 IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
245 IO.mapOptional("AlignConsecutiveAssignments",
246 Style.AlignConsecutiveAssignments);
247 IO.mapOptional("AlignConsecutiveDeclarations",
248 Style.AlignConsecutiveDeclarations);
249 IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft);
250 IO.mapOptional("AlignOperands", Style.AlignOperands);
251 IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
252 IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
253 Style.AllowAllParametersOfDeclarationOnNextLine);
254 IO.mapOptional("AllowShortBlocksOnASingleLine",
255 Style.AllowShortBlocksOnASingleLine);
256 IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
257 Style.AllowShortCaseLabelsOnASingleLine);
258 IO.mapOptional("AllowShortFunctionsOnASingleLine",
259 Style.AllowShortFunctionsOnASingleLine);
260 IO.mapOptional("AllowShortIfStatementsOnASingleLine",
261 Style.AllowShortIfStatementsOnASingleLine);
262 IO.mapOptional("AllowShortLoopsOnASingleLine",
263 Style.AllowShortLoopsOnASingleLine);
264 IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
265 Style.AlwaysBreakAfterDefinitionReturnType);
266 IO.mapOptional("AlwaysBreakAfterReturnType",
267 Style.AlwaysBreakAfterReturnType);
268 // If AlwaysBreakAfterDefinitionReturnType was specified but
269 // AlwaysBreakAfterReturnType was not, initialize the latter from the
270 // former for backwards compatibility.
271 if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None &&
272 Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) {
273 if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All)
274 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
275 else if (Style.AlwaysBreakAfterDefinitionReturnType ==
276 FormatStyle::DRTBS_TopLevel)
277 Style.AlwaysBreakAfterReturnType =
278 FormatStyle::RTBS_TopLevelDefinitions;
281 IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
282 Style.AlwaysBreakBeforeMultilineStrings);
283 IO.mapOptional("AlwaysBreakTemplateDeclarations",
284 Style.AlwaysBreakTemplateDeclarations);
285 IO.mapOptional("BinPackArguments", Style.BinPackArguments);
286 IO.mapOptional("BinPackParameters", Style.BinPackParameters);
287 IO.mapOptional("BraceWrapping", Style.BraceWrapping);
288 IO.mapOptional("BreakBeforeBinaryOperators",
289 Style.BreakBeforeBinaryOperators);
290 IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
291 IO.mapOptional("BreakBeforeTernaryOperators",
292 Style.BreakBeforeTernaryOperators);
293 IO.mapOptional("BreakConstructorInitializersBeforeComma",
294 Style.BreakConstructorInitializersBeforeComma);
295 IO.mapOptional("BreakAfterJavaFieldAnnotations",
296 Style.BreakAfterJavaFieldAnnotations);
297 IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals);
298 IO.mapOptional("ColumnLimit", Style.ColumnLimit);
299 IO.mapOptional("CommentPragmas", Style.CommentPragmas);
300 IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
301 Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
302 IO.mapOptional("ConstructorInitializerIndentWidth",
303 Style.ConstructorInitializerIndentWidth);
304 IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
305 IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
306 IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
307 IO.mapOptional("DisableFormat", Style.DisableFormat);
308 IO.mapOptional("ExperimentalAutoDetectBinPacking",
309 Style.ExperimentalAutoDetectBinPacking);
310 IO.mapOptional("ForEachMacros", Style.ForEachMacros);
311 IO.mapOptional("IncludeCategories", Style.IncludeCategories);
312 IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex);
313 IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
314 IO.mapOptional("IndentWidth", Style.IndentWidth);
315 IO.mapOptional("IndentWrappedFunctionNames",
316 Style.IndentWrappedFunctionNames);
317 IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes);
318 IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports);
319 IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
320 Style.KeepEmptyLinesAtTheStartOfBlocks);
321 IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
322 IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
323 IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
324 IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
325 IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
326 IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
327 IO.mapOptional("ObjCSpaceBeforeProtocolList",
328 Style.ObjCSpaceBeforeProtocolList);
329 IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
330 Style.PenaltyBreakBeforeFirstCallParameter);
331 IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
332 IO.mapOptional("PenaltyBreakFirstLessLess",
333 Style.PenaltyBreakFirstLessLess);
334 IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
335 IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
336 IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
337 Style.PenaltyReturnTypeOnItsOwnLine);
338 IO.mapOptional("PointerAlignment", Style.PointerAlignment);
339 IO.mapOptional("ReflowComments", Style.ReflowComments);
340 IO.mapOptional("SortIncludes", Style.SortIncludes);
341 IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
342 IO.mapOptional("SpaceAfterTemplateKeyword", Style.SpaceAfterTemplateKeyword);
343 IO.mapOptional("SpaceBeforeAssignmentOperators",
344 Style.SpaceBeforeAssignmentOperators);
345 IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
346 IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
347 IO.mapOptional("SpacesBeforeTrailingComments",
348 Style.SpacesBeforeTrailingComments);
349 IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
350 IO.mapOptional("SpacesInContainerLiterals",
351 Style.SpacesInContainerLiterals);
352 IO.mapOptional("SpacesInCStyleCastParentheses",
353 Style.SpacesInCStyleCastParentheses);
354 IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
355 IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
356 IO.mapOptional("Standard", Style.Standard);
357 IO.mapOptional("TabWidth", Style.TabWidth);
358 IO.mapOptional("UseTab", Style.UseTab);
362 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
363 static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
364 IO.mapOptional("AfterClass", Wrapping.AfterClass);
365 IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
366 IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
367 IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
368 IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
369 IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
370 IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
371 IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
372 IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
373 IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
374 IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
378 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
379 static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
380 IO.mapOptional("Regex", Category.Regex);
381 IO.mapOptional("Priority", Category.Priority);
385 // Allows to read vector<FormatStyle> while keeping default values.
386 // IO.getContext() should contain a pointer to the FormatStyle structure, that
387 // will be used to get default values for missing keys.
388 // If the first element has no Language specified, it will be treated as the
389 // default one for the following elements.
390 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
391 static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
394 static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
396 if (Index >= Seq.size()) {
397 assert(Index == Seq.size());
398 FormatStyle Template;
399 if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
402 Template = *((const FormatStyle *)IO.getContext());
403 Template.Language = FormatStyle::LK_None;
405 Seq.resize(Index + 1, Template);
416 const std::error_category &getParseCategory() {
417 static ParseErrorCategory C;
420 std::error_code make_error_code(ParseError e) {
421 return std::error_code(static_cast<int>(e), getParseCategory());
424 const char *ParseErrorCategory::name() const noexcept {
425 return "clang-format.parse_error";
428 std::string ParseErrorCategory::message(int EV) const {
429 switch (static_cast<ParseError>(EV)) {
430 case ParseError::Success:
432 case ParseError::Error:
433 return "Invalid argument";
434 case ParseError::Unsuitable:
437 llvm_unreachable("unexpected parse error");
440 static FormatStyle expandPresets(const FormatStyle &Style) {
441 if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
443 FormatStyle Expanded = Style;
444 Expanded.BraceWrapping = {false, false, false, false, false, false,
445 false, false, false, false, false};
446 switch (Style.BreakBeforeBraces) {
447 case FormatStyle::BS_Linux:
448 Expanded.BraceWrapping.AfterClass = true;
449 Expanded.BraceWrapping.AfterFunction = true;
450 Expanded.BraceWrapping.AfterNamespace = true;
452 case FormatStyle::BS_Mozilla:
453 Expanded.BraceWrapping.AfterClass = true;
454 Expanded.BraceWrapping.AfterEnum = true;
455 Expanded.BraceWrapping.AfterFunction = true;
456 Expanded.BraceWrapping.AfterStruct = true;
457 Expanded.BraceWrapping.AfterUnion = true;
459 case FormatStyle::BS_Stroustrup:
460 Expanded.BraceWrapping.AfterFunction = true;
461 Expanded.BraceWrapping.BeforeCatch = true;
462 Expanded.BraceWrapping.BeforeElse = true;
464 case FormatStyle::BS_Allman:
465 Expanded.BraceWrapping.AfterClass = true;
466 Expanded.BraceWrapping.AfterControlStatement = true;
467 Expanded.BraceWrapping.AfterEnum = true;
468 Expanded.BraceWrapping.AfterFunction = true;
469 Expanded.BraceWrapping.AfterNamespace = true;
470 Expanded.BraceWrapping.AfterObjCDeclaration = true;
471 Expanded.BraceWrapping.AfterStruct = true;
472 Expanded.BraceWrapping.BeforeCatch = true;
473 Expanded.BraceWrapping.BeforeElse = true;
475 case FormatStyle::BS_GNU:
476 Expanded.BraceWrapping = {true, true, true, true, true, true,
477 true, true, true, true, true};
479 case FormatStyle::BS_WebKit:
480 Expanded.BraceWrapping.AfterFunction = true;
488 FormatStyle getLLVMStyle() {
489 FormatStyle LLVMStyle;
490 LLVMStyle.Language = FormatStyle::LK_Cpp;
491 LLVMStyle.AccessModifierOffset = -2;
492 LLVMStyle.AlignEscapedNewlinesLeft = false;
493 LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align;
494 LLVMStyle.AlignOperands = true;
495 LLVMStyle.AlignTrailingComments = true;
496 LLVMStyle.AlignConsecutiveAssignments = false;
497 LLVMStyle.AlignConsecutiveDeclarations = false;
498 LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
499 LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
500 LLVMStyle.AllowShortBlocksOnASingleLine = false;
501 LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
502 LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
503 LLVMStyle.AllowShortLoopsOnASingleLine = false;
504 LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
505 LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
506 LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
507 LLVMStyle.AlwaysBreakTemplateDeclarations = false;
508 LLVMStyle.BinPackParameters = true;
509 LLVMStyle.BinPackArguments = true;
510 LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
511 LLVMStyle.BreakBeforeTernaryOperators = true;
512 LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
513 LLVMStyle.BraceWrapping = {false, false, false, false, false, false,
514 false, false, false, false, false};
515 LLVMStyle.BreakAfterJavaFieldAnnotations = false;
516 LLVMStyle.BreakConstructorInitializersBeforeComma = false;
517 LLVMStyle.BreakStringLiterals = true;
518 LLVMStyle.ColumnLimit = 80;
519 LLVMStyle.CommentPragmas = "^ IWYU pragma:";
520 LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
521 LLVMStyle.ConstructorInitializerIndentWidth = 4;
522 LLVMStyle.ContinuationIndentWidth = 4;
523 LLVMStyle.Cpp11BracedListStyle = true;
524 LLVMStyle.DerivePointerAlignment = false;
525 LLVMStyle.ExperimentalAutoDetectBinPacking = false;
526 LLVMStyle.ForEachMacros.push_back("foreach");
527 LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
528 LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
529 LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
530 {"^(<|\"(gtest|isl|json)/)", 3},
532 LLVMStyle.IncludeIsMainRegex = "$";
533 LLVMStyle.IndentCaseLabels = false;
534 LLVMStyle.IndentWrappedFunctionNames = false;
535 LLVMStyle.IndentWidth = 2;
536 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
537 LLVMStyle.JavaScriptWrapImports = true;
538 LLVMStyle.TabWidth = 8;
539 LLVMStyle.MaxEmptyLinesToKeep = 1;
540 LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
541 LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
542 LLVMStyle.ObjCBlockIndentWidth = 2;
543 LLVMStyle.ObjCSpaceAfterProperty = false;
544 LLVMStyle.ObjCSpaceBeforeProtocolList = true;
545 LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
546 LLVMStyle.SpacesBeforeTrailingComments = 1;
547 LLVMStyle.Standard = FormatStyle::LS_Cpp11;
548 LLVMStyle.UseTab = FormatStyle::UT_Never;
549 LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
550 LLVMStyle.ReflowComments = true;
551 LLVMStyle.SpacesInParentheses = false;
552 LLVMStyle.SpacesInSquareBrackets = false;
553 LLVMStyle.SpaceInEmptyParentheses = false;
554 LLVMStyle.SpacesInContainerLiterals = true;
555 LLVMStyle.SpacesInCStyleCastParentheses = false;
556 LLVMStyle.SpaceAfterCStyleCast = false;
557 LLVMStyle.SpaceAfterTemplateKeyword = true;
558 LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
559 LLVMStyle.SpaceBeforeAssignmentOperators = true;
560 LLVMStyle.SpacesInAngles = false;
562 LLVMStyle.PenaltyBreakComment = 300;
563 LLVMStyle.PenaltyBreakFirstLessLess = 120;
564 LLVMStyle.PenaltyBreakString = 1000;
565 LLVMStyle.PenaltyExcessCharacter = 1000000;
566 LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
567 LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
569 LLVMStyle.DisableFormat = false;
570 LLVMStyle.SortIncludes = true;
575 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
576 FormatStyle GoogleStyle = getLLVMStyle();
577 GoogleStyle.Language = Language;
579 GoogleStyle.AccessModifierOffset = -1;
580 GoogleStyle.AlignEscapedNewlinesLeft = true;
581 GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
582 GoogleStyle.AllowShortLoopsOnASingleLine = true;
583 GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
584 GoogleStyle.AlwaysBreakTemplateDeclarations = true;
585 GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
586 GoogleStyle.DerivePointerAlignment = true;
587 GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
588 GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$";
589 GoogleStyle.IndentCaseLabels = true;
590 GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
591 GoogleStyle.ObjCSpaceAfterProperty = false;
592 GoogleStyle.ObjCSpaceBeforeProtocolList = false;
593 GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
594 GoogleStyle.SpacesBeforeTrailingComments = 2;
595 GoogleStyle.Standard = FormatStyle::LS_Auto;
597 GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
598 GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
600 if (Language == FormatStyle::LK_Java) {
601 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
602 GoogleStyle.AlignOperands = false;
603 GoogleStyle.AlignTrailingComments = false;
604 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
605 GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
606 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
607 GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
608 GoogleStyle.ColumnLimit = 100;
609 GoogleStyle.SpaceAfterCStyleCast = true;
610 GoogleStyle.SpacesBeforeTrailingComments = 1;
611 } else if (Language == FormatStyle::LK_JavaScript) {
612 GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak;
613 GoogleStyle.AlignOperands = false;
614 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
615 GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
616 GoogleStyle.BreakBeforeTernaryOperators = false;
617 GoogleStyle.CommentPragmas =
618 "(taze:|@(export|requirecss|return|returns|see|visibility)) ";
619 GoogleStyle.MaxEmptyLinesToKeep = 3;
620 GoogleStyle.NamespaceIndentation = FormatStyle::NI_All;
621 GoogleStyle.SpacesInContainerLiterals = false;
622 GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single;
623 GoogleStyle.JavaScriptWrapImports = false;
624 } else if (Language == FormatStyle::LK_Proto) {
625 GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
626 GoogleStyle.SpacesInContainerLiterals = false;
627 } else if (Language == FormatStyle::LK_ObjC) {
628 GoogleStyle.ColumnLimit = 100;
634 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
635 FormatStyle ChromiumStyle = getGoogleStyle(Language);
636 if (Language == FormatStyle::LK_Java) {
637 ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
638 ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
639 ChromiumStyle.ContinuationIndentWidth = 8;
640 ChromiumStyle.IndentWidth = 4;
641 } else if (Language == FormatStyle::LK_JavaScript) {
642 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
643 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
645 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
646 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
647 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
648 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
649 ChromiumStyle.BinPackParameters = false;
650 ChromiumStyle.DerivePointerAlignment = false;
652 ChromiumStyle.SortIncludes = false;
653 return ChromiumStyle;
656 FormatStyle getMozillaStyle() {
657 FormatStyle MozillaStyle = getLLVMStyle();
658 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
659 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
660 MozillaStyle.AlwaysBreakAfterReturnType =
661 FormatStyle::RTBS_TopLevel;
662 MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
663 FormatStyle::DRTBS_TopLevel;
664 MozillaStyle.AlwaysBreakTemplateDeclarations = true;
665 MozillaStyle.BinPackParameters = false;
666 MozillaStyle.BinPackArguments = false;
667 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
668 MozillaStyle.BreakConstructorInitializersBeforeComma = true;
669 MozillaStyle.ConstructorInitializerIndentWidth = 2;
670 MozillaStyle.ContinuationIndentWidth = 2;
671 MozillaStyle.Cpp11BracedListStyle = false;
672 MozillaStyle.IndentCaseLabels = true;
673 MozillaStyle.ObjCSpaceAfterProperty = true;
674 MozillaStyle.ObjCSpaceBeforeProtocolList = false;
675 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
676 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
677 MozillaStyle.SpaceAfterTemplateKeyword = false;
681 FormatStyle getWebKitStyle() {
682 FormatStyle Style = getLLVMStyle();
683 Style.AccessModifierOffset = -4;
684 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
685 Style.AlignOperands = false;
686 Style.AlignTrailingComments = false;
687 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
688 Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
689 Style.BreakConstructorInitializersBeforeComma = true;
690 Style.Cpp11BracedListStyle = false;
691 Style.ColumnLimit = 0;
692 Style.IndentWidth = 4;
693 Style.NamespaceIndentation = FormatStyle::NI_Inner;
694 Style.ObjCBlockIndentWidth = 4;
695 Style.ObjCSpaceAfterProperty = true;
696 Style.PointerAlignment = FormatStyle::PAS_Left;
700 FormatStyle getGNUStyle() {
701 FormatStyle Style = getLLVMStyle();
702 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
703 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
704 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
705 Style.BreakBeforeBraces = FormatStyle::BS_GNU;
706 Style.BreakBeforeTernaryOperators = true;
707 Style.Cpp11BracedListStyle = false;
708 Style.ColumnLimit = 79;
709 Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
710 Style.Standard = FormatStyle::LS_Cpp03;
714 FormatStyle getNoStyle() {
715 FormatStyle NoStyle = getLLVMStyle();
716 NoStyle.DisableFormat = true;
717 NoStyle.SortIncludes = false;
721 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
722 FormatStyle *Style) {
723 if (Name.equals_lower("llvm")) {
724 *Style = getLLVMStyle();
725 } else if (Name.equals_lower("chromium")) {
726 *Style = getChromiumStyle(Language);
727 } else if (Name.equals_lower("mozilla")) {
728 *Style = getMozillaStyle();
729 } else if (Name.equals_lower("google")) {
730 *Style = getGoogleStyle(Language);
731 } else if (Name.equals_lower("webkit")) {
732 *Style = getWebKitStyle();
733 } else if (Name.equals_lower("gnu")) {
734 *Style = getGNUStyle();
735 } else if (Name.equals_lower("none")) {
736 *Style = getNoStyle();
741 Style->Language = Language;
745 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
747 FormatStyle::LanguageKind Language = Style->Language;
748 assert(Language != FormatStyle::LK_None);
749 if (Text.trim().empty())
750 return make_error_code(ParseError::Error);
752 std::vector<FormatStyle> Styles;
753 llvm::yaml::Input Input(Text);
754 // DocumentListTraits<vector<FormatStyle>> uses the context to get default
755 // values for the fields, keys for which are missing from the configuration.
756 // Mapping also uses the context to get the language to find the correct
758 Input.setContext(Style);
761 return Input.error();
763 for (unsigned i = 0; i < Styles.size(); ++i) {
764 // Ensures that only the first configuration can skip the Language option.
765 if (Styles[i].Language == FormatStyle::LK_None && i != 0)
766 return make_error_code(ParseError::Error);
767 // Ensure that each language is configured at most once.
768 for (unsigned j = 0; j < i; ++j) {
769 if (Styles[i].Language == Styles[j].Language) {
771 << "Duplicate languages in the config file on positions " << j
772 << " and " << i << "\n");
773 return make_error_code(ParseError::Error);
777 // Look for a suitable configuration starting from the end, so we can
778 // find the configuration for the specific language first, and the default
779 // configuration (which can only be at slot 0) after it.
780 for (int i = Styles.size() - 1; i >= 0; --i) {
781 if (Styles[i].Language == Language ||
782 Styles[i].Language == FormatStyle::LK_None) {
784 Style->Language = Language;
785 return make_error_code(ParseError::Success);
788 return make_error_code(ParseError::Unsuitable);
791 std::string configurationAsText(const FormatStyle &Style) {
793 llvm::raw_string_ostream Stream(Text);
794 llvm::yaml::Output Output(Stream);
795 // We use the same mapping method for input and output, so we need a non-const
797 FormatStyle NonConstStyle = expandPresets(Style);
798 Output << NonConstStyle;
804 class JavaScriptRequoter : public TokenAnalyzer {
806 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
807 : TokenAnalyzer(Env, Style) {}
809 tooling::Replacements
810 analyze(TokenAnnotator &Annotator,
811 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
812 FormatTokenLexer &Tokens) override {
813 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
814 AnnotatedLines.end());
815 tooling::Replacements Result;
816 requoteJSStringLiteral(AnnotatedLines, Result);
821 // Replaces double/single-quoted string literal as appropriate, re-escaping
822 // the contents in the process.
823 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
824 tooling::Replacements &Result) {
825 for (AnnotatedLine *Line : Lines) {
826 requoteJSStringLiteral(Line->Children, Result);
829 for (FormatToken *FormatTok = Line->First; FormatTok;
830 FormatTok = FormatTok->Next) {
831 StringRef Input = FormatTok->TokenText;
832 if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
833 // NB: testing for not starting with a double quote to avoid
834 // breaking `template strings`.
835 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
836 !Input.startswith("\"")) ||
837 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
838 !Input.startswith("\'")))
841 // Change start and end quote.
842 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
843 SourceLocation Start = FormatTok->Tok.getLocation();
844 auto Replace = [&](SourceLocation Start, unsigned Length,
845 StringRef ReplacementText) {
846 auto Err = Result.add(tooling::Replacement(
847 Env.getSourceManager(), Start, Length, ReplacementText));
848 // FIXME: handle error. For now, print error message and skip the
849 // replacement for release version.
851 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
855 Replace(Start, 1, IsSingle ? "'" : "\"");
856 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
857 IsSingle ? "'" : "\"");
859 // Escape internal quotes.
860 bool Escaped = false;
861 for (size_t i = 1; i < Input.size() - 1; i++) {
864 if (!Escaped && i + 1 < Input.size() &&
865 ((IsSingle && Input[i + 1] == '"') ||
866 (!IsSingle && Input[i + 1] == '\''))) {
867 // Remove this \, it's escaping a " or ' that no longer needs
869 Replace(Start.getLocWithOffset(i), 1, "");
876 if (!Escaped && IsSingle == (Input[i] == '\'')) {
878 Replace(Start.getLocWithOffset(i), 0, "\\");
892 class Formatter : public TokenAnalyzer {
894 Formatter(const Environment &Env, const FormatStyle &Style,
895 bool *IncompleteFormat)
896 : TokenAnalyzer(Env, Style), IncompleteFormat(IncompleteFormat) {}
898 tooling::Replacements
899 analyze(TokenAnnotator &Annotator,
900 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
901 FormatTokenLexer &Tokens) override {
902 tooling::Replacements Result;
903 deriveLocalStyle(AnnotatedLines);
904 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
905 AnnotatedLines.end());
906 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
907 Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
909 Annotator.setCommentLineLevels(AnnotatedLines);
911 WhitespaceManager Whitespaces(
912 Env.getSourceManager(), Style,
913 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
914 ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
915 Env.getSourceManager(), Whitespaces, Encoding,
916 BinPackInconclusiveFunctions);
917 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
919 .format(AnnotatedLines);
920 for (const auto &R : Whitespaces.generateReplacements())
928 static bool inputUsesCRLF(StringRef Text) {
929 return Text.count('\r') * 2 > Text.count('\n');
933 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
934 for (const AnnotatedLine *Line : Lines) {
935 if (hasCpp03IncompatibleFormat(Line->Children))
937 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
938 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
939 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
941 if (Tok->is(TT_TemplateCloser) &&
942 Tok->Previous->is(TT_TemplateCloser))
950 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
951 int AlignmentDiff = 0;
952 for (const AnnotatedLine *Line : Lines) {
953 AlignmentDiff += countVariableAlignments(Line->Children);
954 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
955 if (!Tok->is(TT_PointerOrReference))
958 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
959 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
960 Tok->Next->WhitespaceRange.getEnd();
961 if (SpaceBefore && !SpaceAfter)
963 if (!SpaceBefore && SpaceAfter)
967 return AlignmentDiff;
971 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
972 bool HasBinPackedFunction = false;
973 bool HasOnePerLineFunction = false;
974 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
975 if (!AnnotatedLines[i]->First->Next)
977 FormatToken *Tok = AnnotatedLines[i]->First->Next;
979 if (Tok->PackingKind == PPK_BinPacked)
980 HasBinPackedFunction = true;
981 if (Tok->PackingKind == PPK_OnePerLine)
982 HasOnePerLineFunction = true;
987 if (Style.DerivePointerAlignment)
988 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
989 ? FormatStyle::PAS_Left
990 : FormatStyle::PAS_Right;
991 if (Style.Standard == FormatStyle::LS_Auto)
992 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
993 ? FormatStyle::LS_Cpp11
994 : FormatStyle::LS_Cpp03;
995 BinPackInconclusiveFunctions =
996 HasBinPackedFunction || !HasOnePerLineFunction;
999 bool BinPackInconclusiveFunctions;
1000 bool *IncompleteFormat;
1003 // This class clean up the erroneous/redundant code around the given ranges in
1005 class Cleaner : public TokenAnalyzer {
1007 Cleaner(const Environment &Env, const FormatStyle &Style)
1008 : TokenAnalyzer(Env, Style),
1009 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1011 // FIXME: eliminate unused parameters.
1012 tooling::Replacements
1013 analyze(TokenAnnotator &Annotator,
1014 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1015 FormatTokenLexer &Tokens) override {
1016 // FIXME: in the current implementation the granularity of affected range
1017 // is an annotated line. However, this is not sufficient. Furthermore,
1018 // redundant code introduced by replacements does not necessarily
1019 // intercept with ranges of replacements that result in the redundancy.
1020 // To determine if some redundant code is actually introduced by
1021 // replacements(e.g. deletions), we need to come up with a more
1022 // sophisticated way of computing affected ranges.
1023 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1024 AnnotatedLines.end());
1026 checkEmptyNamespace(AnnotatedLines);
1028 for (auto &Line : AnnotatedLines) {
1029 if (Line->Affected) {
1030 cleanupRight(Line->First, tok::comma, tok::comma);
1031 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1032 cleanupRight(Line->First, tok::l_paren, tok::comma);
1033 cleanupLeft(Line->First, tok::comma, tok::r_paren);
1034 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1035 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1036 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1040 return generateFixes();
1044 bool containsOnlyComments(const AnnotatedLine &Line) {
1045 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1046 if (Tok->isNot(tok::comment))
1052 // Iterate through all lines and remove any empty (nested) namespaces.
1053 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1054 std::set<unsigned> DeletedLines;
1055 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1056 auto &Line = *AnnotatedLines[i];
1057 if (Line.startsWith(tok::kw_namespace) ||
1058 Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1059 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1063 for (auto Line : DeletedLines) {
1064 FormatToken *Tok = AnnotatedLines[Line]->First;
1072 // The function checks if the namespace, which starts from \p CurrentLine, and
1073 // its nested namespaces are empty and delete them if they are empty. It also
1074 // sets \p NewLine to the last line checked.
1075 // Returns true if the current namespace is empty.
1076 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1077 unsigned CurrentLine, unsigned &NewLine,
1078 std::set<unsigned> &DeletedLines) {
1079 unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1080 if (Style.BraceWrapping.AfterNamespace) {
1081 // If the left brace is in a new line, we should consume it first so that
1082 // it does not make the namespace non-empty.
1083 // FIXME: error handling if there is no left brace.
1084 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1085 NewLine = CurrentLine;
1088 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1091 while (++CurrentLine < End) {
1092 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1095 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1096 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1097 tok::kw_namespace)) {
1098 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1101 CurrentLine = NewLine;
1105 if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1108 // If there is anything other than comments or nested namespaces in the
1109 // current namespace, the namespace cannot be empty.
1110 NewLine = CurrentLine;
1114 NewLine = CurrentLine;
1115 if (CurrentLine >= End)
1118 // Check if the empty namespace is actually affected by changed ranges.
1119 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1120 AnnotatedLines[InitLine]->First->Tok.getLocation(),
1121 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1124 for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1125 DeletedLines.insert(i);
1131 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1132 // of the token in the pair if the left token has \p LK token kind and the
1133 // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1134 // is deleted on match; otherwise, the right token is deleted.
1135 template <typename LeftKind, typename RightKind>
1136 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1138 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1139 for (auto *Res = Tok.Next; Res; Res = Res->Next)
1140 if (!Res->is(tok::comment) &&
1141 DeletedTokens.find(Res) == DeletedTokens.end())
1145 for (auto *Left = Start; Left;) {
1146 auto *Right = NextNotDeleted(*Left);
1149 if (Left->is(LK) && Right->is(RK)) {
1150 deleteToken(DeleteLeft ? Left : Right);
1151 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1153 // If the right token is deleted, we should keep the left token
1154 // unchanged and pair it with the new right token.
1162 template <typename LeftKind, typename RightKind>
1163 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1164 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1167 template <typename LeftKind, typename RightKind>
1168 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1169 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1172 // Delete the given token.
1173 inline void deleteToken(FormatToken *Tok) {
1175 DeletedTokens.insert(Tok);
1178 tooling::Replacements generateFixes() {
1179 tooling::Replacements Fixes;
1180 std::vector<FormatToken *> Tokens;
1181 std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1182 std::back_inserter(Tokens));
1184 // Merge multiple continuous token deletions into one big deletion so that
1185 // the number of replacements can be reduced. This makes computing affected
1186 // ranges more efficient when we run reformat on the changed code.
1188 while (Idx < Tokens.size()) {
1189 unsigned St = Idx, End = Idx;
1190 while ((End + 1) < Tokens.size() &&
1191 Tokens[End]->Next == Tokens[End + 1]) {
1194 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1195 Tokens[End]->Tok.getEndLoc());
1197 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1198 // FIXME: better error handling. for now just print error message and skip
1199 // for the release version.
1201 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1202 assert(false && "Fixes must not conflict!");
1210 // Class for less-than inequality comparason for the set `RedundantTokens`.
1211 // We store tokens in the order they appear in the translation unit so that
1212 // we do not need to sort them in `generateFixes()`.
1213 struct FormatTokenLess {
1214 FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1216 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1217 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1218 RHS->Tok.getLocation());
1220 const SourceManager &SM;
1223 // Tokens to be deleted.
1224 std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1227 struct IncludeDirective {
1234 } // end anonymous namespace
1236 // Determines whether 'Ranges' intersects with ('Start', 'End').
1237 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1239 for (auto Range : Ranges) {
1240 if (Range.getOffset() < End &&
1241 Range.getOffset() + Range.getLength() > Start)
1247 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1248 // before sorting/deduplicating. Index is the index of the include under the
1249 // cursor in the original set of includes. If this include has duplicates, it is
1250 // the index of the first of the duplicates as the others are going to be
1251 // removed. OffsetToEOL describes the cursor's position relative to the end of
1252 // its current line.
1253 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1254 static std::pair<unsigned, unsigned>
1255 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1256 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1257 unsigned CursorIndex = UINT_MAX;
1258 unsigned OffsetToEOL = 0;
1259 for (int i = 0, e = Includes.size(); i != e; ++i) {
1260 unsigned Start = Includes[Indices[i]].Offset;
1261 unsigned End = Start + Includes[Indices[i]].Text.size();
1262 if (!(Cursor >= Start && Cursor < End))
1264 CursorIndex = Indices[i];
1265 OffsetToEOL = End - Cursor;
1266 // Put the cursor on the only remaining #include among the duplicate
1268 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1272 return std::make_pair(CursorIndex, OffsetToEOL);
1275 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1276 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1278 // #include directives with the same text will be deduplicated, and only the
1279 // first #include in the duplicate #includes remains. If the `Cursor` is
1280 // provided and put on a deleted #include, it will be moved to the remaining
1281 // #include in the duplicate #includes.
1282 static void sortCppIncludes(const FormatStyle &Style,
1283 const SmallVectorImpl<IncludeDirective> &Includes,
1284 ArrayRef<tooling::Range> Ranges, StringRef FileName,
1285 tooling::Replacements &Replaces, unsigned *Cursor) {
1286 unsigned IncludesBeginOffset = Includes.front().Offset;
1287 unsigned IncludesEndOffset =
1288 Includes.back().Offset + Includes.back().Text.size();
1289 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1290 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1292 SmallVector<unsigned, 16> Indices;
1293 for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1294 Indices.push_back(i);
1296 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1297 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1298 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1300 // The index of the include on which the cursor will be put after
1301 // sorting/deduplicating.
1302 unsigned CursorIndex;
1303 // The offset from cursor to the end of line.
1304 unsigned CursorToEOLOffset;
1306 std::tie(CursorIndex, CursorToEOLOffset) =
1307 FindCursorIndex(Includes, Indices, *Cursor);
1309 // Deduplicate #includes.
1310 Indices.erase(std::unique(Indices.begin(), Indices.end(),
1311 [&](unsigned LHSI, unsigned RHSI) {
1312 return Includes[LHSI].Text == Includes[RHSI].Text;
1316 // If the #includes are out of order, we generate a single replacement fixing
1317 // the entire block. Otherwise, no replacement is generated.
1318 if (Indices.size() == Includes.size() &&
1319 std::is_sorted(Indices.begin(), Indices.end()))
1323 for (unsigned Index : Indices) {
1324 if (!result.empty())
1326 result += Includes[Index].Text;
1327 if (Cursor && CursorIndex == Index)
1328 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1331 auto Err = Replaces.add(tooling::Replacement(
1332 FileName, Includes.front().Offset, IncludesBlockSize, result));
1333 // FIXME: better error handling. For now, just skip the replacement for the
1336 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1343 // This class manages priorities of #include categories and calculates
1344 // priorities for headers.
1345 class IncludeCategoryManager {
1347 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1348 : Style(Style), FileName(FileName) {
1349 FileStem = llvm::sys::path::stem(FileName);
1350 for (const auto &Category : Style.IncludeCategories)
1351 CategoryRegexs.emplace_back(Category.Regex);
1352 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1353 FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1354 FileName.endswith(".cxx") || FileName.endswith(".m") ||
1355 FileName.endswith(".mm");
1358 // Returns the priority of the category which \p IncludeName belongs to.
1359 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1360 // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1361 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1363 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1364 if (CategoryRegexs[i].match(IncludeName)) {
1365 Ret = Style.IncludeCategories[i].Priority;
1368 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1374 bool isMainHeader(StringRef IncludeName) const {
1375 if (!IncludeName.startswith("\""))
1377 StringRef HeaderStem =
1378 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1379 if (FileStem.startswith(HeaderStem)) {
1380 llvm::Regex MainIncludeRegex(
1381 (HeaderStem + Style.IncludeIsMainRegex).str());
1382 if (MainIncludeRegex.match(FileStem))
1388 const FormatStyle &Style;
1392 SmallVector<llvm::Regex, 4> CategoryRegexs;
1395 const char IncludeRegexPattern[] =
1396 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1398 } // anonymous namespace
1400 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1401 ArrayRef<tooling::Range> Ranges,
1403 tooling::Replacements &Replaces,
1406 unsigned SearchFrom = 0;
1407 llvm::Regex IncludeRegex(IncludeRegexPattern);
1408 SmallVector<StringRef, 4> Matches;
1409 SmallVector<IncludeDirective, 16> IncludesInBlock;
1411 // In compiled files, consider the first #include to be the main #include of
1412 // the file if it is not a system #include. This ensures that the header
1413 // doesn't have hidden dependencies
1414 // (http://llvm.org/docs/CodingStandards.html#include-style).
1416 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1417 // cases where the first #include is unlikely to be the main header.
1418 IncludeCategoryManager Categories(Style, FileName);
1419 bool FirstIncludeBlock = true;
1420 bool MainIncludeFound = false;
1421 bool FormattingOff = false;
1424 auto Pos = Code.find('\n', SearchFrom);
1426 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1428 StringRef Trimmed = Line.trim();
1429 if (Trimmed == "// clang-format off")
1430 FormattingOff = true;
1431 else if (Trimmed == "// clang-format on")
1432 FormattingOff = false;
1434 if (!FormattingOff && !Line.endswith("\\")) {
1435 if (IncludeRegex.match(Line, &Matches)) {
1436 StringRef IncludeName = Matches[2];
1437 int Category = Categories.getIncludePriority(
1439 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1441 MainIncludeFound = true;
1442 IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1443 } else if (!IncludesInBlock.empty()) {
1444 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1446 IncludesInBlock.clear();
1447 FirstIncludeBlock = false;
1451 if (Pos == StringRef::npos || Pos + 1 == Code.size())
1453 SearchFrom = Pos + 1;
1455 if (!IncludesInBlock.empty())
1456 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1460 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1461 ArrayRef<tooling::Range> Ranges,
1462 StringRef FileName, unsigned *Cursor) {
1463 tooling::Replacements Replaces;
1464 if (!Style.SortIncludes)
1466 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1467 return sortJavaScriptImports(Style, Code, Ranges, FileName);
1468 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1472 template <typename T>
1473 static llvm::Expected<tooling::Replacements>
1474 processReplacements(T ProcessFunc, StringRef Code,
1475 const tooling::Replacements &Replaces,
1476 const FormatStyle &Style) {
1477 if (Replaces.empty())
1478 return tooling::Replacements();
1480 auto NewCode = applyAllReplacements(Code, Replaces);
1482 return NewCode.takeError();
1483 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1484 StringRef FileName = Replaces.begin()->getFilePath();
1486 tooling::Replacements FormatReplaces =
1487 ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1489 return Replaces.merge(FormatReplaces);
1492 llvm::Expected<tooling::Replacements>
1493 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1494 const FormatStyle &Style) {
1495 // We need to use lambda function here since there are two versions of
1497 auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1498 std::vector<tooling::Range> Ranges,
1499 StringRef FileName) -> tooling::Replacements {
1500 return sortIncludes(Style, Code, Ranges, FileName);
1502 auto SortedReplaces =
1503 processReplacements(SortIncludes, Code, Replaces, Style);
1504 if (!SortedReplaces)
1505 return SortedReplaces.takeError();
1507 // We need to use lambda function here since there are two versions of
1509 auto Reformat = [](const FormatStyle &Style, StringRef Code,
1510 std::vector<tooling::Range> Ranges,
1511 StringRef FileName) -> tooling::Replacements {
1512 return reformat(Style, Code, Ranges, FileName);
1514 return processReplacements(Reformat, Code, *SortedReplaces, Style);
1519 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1520 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1521 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1524 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1525 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1528 // Returns the offset after skipping a sequence of tokens, matched by \p
1529 // GetOffsetAfterSequence, from the start of the code.
1530 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1531 // tokens and returns an offset after the sequence.
1532 unsigned getOffsetAfterTokenSequence(
1533 StringRef FileName, StringRef Code, const FormatStyle &Style,
1534 std::function<unsigned(const SourceManager &, Lexer &, Token &)>
1535 GetOffsetAfterSequense) {
1536 std::unique_ptr<Environment> Env =
1537 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1538 const SourceManager &SourceMgr = Env->getSourceManager();
1539 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1540 getFormattingLangOpts(Style));
1542 // Get the first token.
1543 Lex.LexFromRawLexer(Tok);
1544 return GetOffsetAfterSequense(SourceMgr, Lex, Tok);
1547 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1548 // \p Tok will be the token after this directive; otherwise, it can be any token
1549 // after the given \p Tok (including \p Tok).
1550 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1551 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1552 Tok.is(tok::raw_identifier) &&
1553 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1554 Tok.is(tok::raw_identifier);
1556 Lex.LexFromRawLexer(Tok);
1560 void skipComments(Lexer &Lex, Token &Tok) {
1561 while (Tok.is(tok::comment))
1562 if (Lex.LexFromRawLexer(Tok))
1566 // Returns the offset after header guard directives and any comments
1567 // before/after header guards. If no header guard presents in the code, this
1568 // will returns the offset after skipping all comments from the start of the
1570 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1572 const FormatStyle &Style) {
1573 return getOffsetAfterTokenSequence(
1574 FileName, Code, Style,
1575 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1576 skipComments(Lex, Tok);
1577 unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1578 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1579 skipComments(Lex, Tok);
1580 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1581 return SM.getFileOffset(Tok.getLocation());
1583 return InitialOffset;
1587 // Check if a sequence of tokens is like
1588 // "#include ("header.h" | <header.h>)".
1589 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1590 // any token after the given \p Tok (including \p Tok).
1591 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1592 auto Matched = [&]() {
1593 Lex.LexFromRawLexer(Tok);
1596 if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1597 Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1598 if (Lex.LexFromRawLexer(Tok))
1600 if (Tok.is(tok::string_literal))
1602 if (Tok.is(tok::less)) {
1603 while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1605 if (Tok.is(tok::greater))
1612 // Returns the offset of the last #include directive after which a new
1613 // #include can be inserted. This ignores #include's after the #include block(s)
1614 // in the beginning of a file to avoid inserting headers into code sections
1615 // where new #include's should not be added by default.
1616 // These code sections include:
1617 // - raw string literals (containing #include).
1619 // - Special #include's among declarations (e.g. functions).
1621 // If no #include after which a new #include can be inserted, this returns the
1622 // offset after skipping all comments from the start of the code.
1623 // Inserting after an #include is not allowed if it comes after code that is not
1624 // #include (e.g. pre-processing directive that is not #include, declarations).
1625 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1626 const FormatStyle &Style) {
1627 return getOffsetAfterTokenSequence(
1628 FileName, Code, Style,
1629 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1630 skipComments(Lex, Tok);
1631 unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1632 while (checkAndConsumeInclusiveDirective(Lex, Tok))
1633 MaxOffset = SM.getFileOffset(Tok.getLocation());
1638 bool isDeletedHeader(llvm::StringRef HeaderName,
1639 const std::set<llvm::StringRef> &HeadersToDelete) {
1640 return HeadersToDelete.count(HeaderName) ||
1641 HeadersToDelete.count(HeaderName.trim("\"<>"));
1644 // FIXME: insert empty lines between newly created blocks.
1645 tooling::Replacements
1646 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1647 const FormatStyle &Style) {
1648 if (Style.Language != FormatStyle::LanguageKind::LK_Cpp)
1651 tooling::Replacements HeaderInsertions;
1652 std::set<llvm::StringRef> HeadersToDelete;
1653 tooling::Replacements Result;
1654 for (const auto &R : Replaces) {
1655 if (isHeaderInsertion(R)) {
1656 // Replacements from \p Replaces must be conflict-free already, so we can
1657 // simply consume the error.
1658 llvm::consumeError(HeaderInsertions.add(R));
1659 } else if (isHeaderDeletion(R)) {
1660 HeadersToDelete.insert(R.getReplacementText());
1661 } else if (R.getOffset() == UINT_MAX) {
1662 llvm::errs() << "Insertions other than header #include insertion are "
1664 << R.getReplacementText() << "\n";
1666 llvm::consumeError(Result.add(R));
1669 if (HeaderInsertions.empty() && HeadersToDelete.empty())
1672 llvm::Regex IncludeRegex(IncludeRegexPattern);
1673 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1674 SmallVector<StringRef, 4> Matches;
1676 StringRef FileName = Replaces.begin()->getFilePath();
1677 IncludeCategoryManager Categories(Style, FileName);
1679 // Record the offset of the end of the last include in each category.
1680 std::map<int, int> CategoryEndOffsets;
1681 // All possible priorities.
1682 // Add 0 for main header and INT_MAX for headers that are not in any category.
1683 std::set<int> Priorities = {0, INT_MAX};
1684 for (const auto &Category : Style.IncludeCategories)
1685 Priorities.insert(Category.Priority);
1686 int FirstIncludeOffset = -1;
1687 // All new headers should be inserted after this offset.
1688 unsigned MinInsertOffset =
1689 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1690 StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1691 // Max insertion offset in the original code.
1692 unsigned MaxInsertOffset =
1694 getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1695 SmallVector<StringRef, 32> Lines;
1696 TrimmedCode.split(Lines, '\n');
1697 unsigned Offset = MinInsertOffset;
1698 unsigned NextLineOffset;
1699 std::set<StringRef> ExistingIncludes;
1700 for (auto Line : Lines) {
1701 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1702 if (IncludeRegex.match(Line, &Matches)) {
1703 // The header name with quotes or angle brackets.
1704 StringRef IncludeName = Matches[2];
1705 ExistingIncludes.insert(IncludeName);
1706 // Only record the offset of current #include if we can insert after it.
1707 if (Offset <= MaxInsertOffset) {
1708 int Category = Categories.getIncludePriority(
1709 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1710 CategoryEndOffsets[Category] = NextLineOffset;
1711 if (FirstIncludeOffset < 0)
1712 FirstIncludeOffset = Offset;
1714 if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1715 // If this is the last line without trailing newline, we need to make
1716 // sure we don't delete across the file boundary.
1717 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1719 Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1721 // Ignore the deletion on conflict.
1722 llvm::errs() << "Failed to add header deletion replacement for "
1723 << IncludeName << ": " << llvm::toString(std::move(Err))
1728 Offset = NextLineOffset;
1731 // Populate CategoryEndOfssets:
1732 // - Ensure that CategoryEndOffset[Highest] is always populated.
1733 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1734 // is set, up to CategoryEndOffset[Highest].
1735 auto Highest = Priorities.begin();
1736 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1737 if (FirstIncludeOffset >= 0)
1738 CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1740 CategoryEndOffsets[*Highest] = MinInsertOffset;
1742 // By this point, CategoryEndOffset[Highest] is always set appropriately:
1743 // - to an appropriate location before/after existing #includes, or
1744 // - to right after the header guard, or
1745 // - to the beginning of the file.
1746 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1747 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1748 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1750 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1751 for (const auto &R : HeaderInsertions) {
1752 auto IncludeDirective = R.getReplacementText();
1753 bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1754 assert(Matched && "Header insertion replacement must have replacement text "
1757 auto IncludeName = Matches[2];
1758 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1759 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1764 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1765 Offset = CategoryEndOffsets[Category];
1766 std::string NewInclude = !IncludeDirective.endswith("\n")
1767 ? (IncludeDirective + "\n").str()
1768 : IncludeDirective.str();
1769 // When inserting headers at end of the code, also append '\n' to the code
1770 // if it does not end with '\n'.
1771 if (NeedNewLineAtEnd && Offset == Code.size()) {
1772 NewInclude = "\n" + NewInclude;
1773 NeedNewLineAtEnd = false;
1775 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1776 auto Err = Result.add(NewReplace);
1778 llvm::consumeError(std::move(Err));
1779 unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1780 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1781 Result = Result.merge(tooling::Replacements(NewReplace));
1787 } // anonymous namespace
1789 llvm::Expected<tooling::Replacements>
1790 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1791 const FormatStyle &Style) {
1792 // We need to use lambda function here since there are two versions of
1794 auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1795 std::vector<tooling::Range> Ranges,
1796 StringRef FileName) -> tooling::Replacements {
1797 return cleanup(Style, Code, Ranges, FileName);
1799 // Make header insertion replacements insert new headers into correct blocks.
1800 tooling::Replacements NewReplaces =
1801 fixCppIncludeInsertions(Code, Replaces, Style);
1802 return processReplacements(Cleanup, Code, NewReplaces, Style);
1805 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1806 ArrayRef<tooling::Range> Ranges,
1807 StringRef FileName, bool *IncompleteFormat) {
1808 FormatStyle Expanded = expandPresets(Style);
1809 if (Expanded.DisableFormat)
1810 return tooling::Replacements();
1812 auto Env = Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1814 if (Style.Language == FormatStyle::LK_JavaScript &&
1815 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) {
1816 JavaScriptRequoter Requoter(*Env, Expanded);
1817 tooling::Replacements Requotes = Requoter.process();
1818 if (!Requotes.empty()) {
1819 auto NewCode = applyAllReplacements(Code, Requotes);
1821 auto NewEnv = Environment::CreateVirtualEnvironment(
1823 tooling::calculateRangesAfterReplacements(Requotes, Ranges));
1824 Formatter Format(*NewEnv, Expanded, IncompleteFormat);
1825 return Requotes.merge(Format.process());
1830 Formatter Format(*Env, Expanded, IncompleteFormat);
1831 return Format.process();
1834 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
1835 ArrayRef<tooling::Range> Ranges,
1836 StringRef FileName) {
1837 std::unique_ptr<Environment> Env =
1838 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1839 Cleaner Clean(*Env, Style);
1840 return Clean.process();
1843 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1844 LangOptions LangOpts;
1845 LangOpts.CPlusPlus = 1;
1846 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1847 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1848 LangOpts.LineComment = 1;
1849 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp;
1850 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1854 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally.
1855 LangOpts.DeclSpecKeyword = 1; // To get __declspec.
1859 const char *StyleOptionHelpDescription =
1860 "Coding style, currently supports:\n"
1861 " LLVM, Google, Chromium, Mozilla, WebKit.\n"
1862 "Use -style=file to load style configuration from\n"
1863 ".clang-format file located in one of the parent\n"
1864 "directories of the source file (or current\n"
1865 "directory for stdin).\n"
1866 "Use -style=\"{key: value, ...}\" to set specific\n"
1867 "parameters, e.g.:\n"
1868 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1870 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1871 if (FileName.endswith(".java"))
1872 return FormatStyle::LK_Java;
1873 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
1874 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
1875 if (FileName.endswith(".m") || FileName.endswith(".mm"))
1876 return FormatStyle::LK_ObjC;
1877 if (FileName.endswith_lower(".proto") ||
1878 FileName.endswith_lower(".protodevel"))
1879 return FormatStyle::LK_Proto;
1880 if (FileName.endswith_lower(".td"))
1881 return FormatStyle::LK_TableGen;
1882 return FormatStyle::LK_Cpp;
1885 FormatStyle getStyle(StringRef StyleName, StringRef FileName,
1886 StringRef FallbackStyle, StringRef Code,
1887 vfs::FileSystem *FS) {
1889 FS = vfs::getRealFileSystem().get();
1891 FormatStyle Style = getLLVMStyle();
1892 Style.Language = getLanguageByFileName(FileName);
1894 // This is a very crude detection of whether a header contains ObjC code that
1895 // should be improved over time and probably be done on tokens, not one the
1896 // bare content of the file.
1897 if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
1898 (Code.contains("\n- (") || Code.contains("\n+ (")))
1899 Style.Language = FormatStyle::LK_ObjC;
1901 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) {
1902 llvm::errs() << "Invalid fallback style \"" << FallbackStyle
1903 << "\" using LLVM style\n";
1907 if (StyleName.startswith("{")) {
1908 // Parse YAML/JSON style from the command line.
1909 if (std::error_code ec = parseConfiguration(StyleName, &Style)) {
1910 llvm::errs() << "Error parsing -style: " << ec.message() << ", using "
1911 << FallbackStyle << " style\n";
1916 if (!StyleName.equals_lower("file")) {
1917 if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1918 llvm::errs() << "Invalid value for -style, using " << FallbackStyle
1923 // Look for .clang-format/_clang-format file in the file's parent directories.
1924 SmallString<128> UnsuitableConfigFiles;
1925 SmallString<128> Path(FileName);
1926 if (std::error_code EC = FS->makeAbsolute(Path)) {
1927 llvm::errs() << EC.message() << "\n";
1931 for (StringRef Directory = Path; !Directory.empty();
1932 Directory = llvm::sys::path::parent_path(Directory)) {
1934 auto Status = FS->status(Directory);
1936 Status->getType() != llvm::sys::fs::file_type::directory_file) {
1940 SmallString<128> ConfigFile(Directory);
1942 llvm::sys::path::append(ConfigFile, ".clang-format");
1943 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1945 Status = FS->status(ConfigFile.str());
1947 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
1949 // Try _clang-format too, since dotfiles are not commonly used on Windows.
1950 ConfigFile = Directory;
1951 llvm::sys::path::append(ConfigFile, "_clang-format");
1952 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1953 Status = FS->status(ConfigFile.str());
1955 (Status->getType() == llvm::sys::fs::file_type::regular_file);
1959 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
1960 FS->getBufferForFile(ConfigFile.str());
1961 if (std::error_code EC = Text.getError()) {
1962 llvm::errs() << EC.message() << "\n";
1965 if (std::error_code ec =
1966 parseConfiguration(Text.get()->getBuffer(), &Style)) {
1967 if (ec == ParseError::Unsuitable) {
1968 if (!UnsuitableConfigFiles.empty())
1969 UnsuitableConfigFiles.append(", ");
1970 UnsuitableConfigFiles.append(ConfigFile);
1973 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message()
1977 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
1981 if (!UnsuitableConfigFiles.empty()) {
1982 llvm::errs() << "Configuration file(s) do(es) not support "
1983 << getLanguageName(Style.Language) << ": "
1984 << UnsuitableConfigFiles << "\n";
1989 } // namespace format
1990 } // namespace clang