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;
642 ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
643 ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
644 ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
645 ChromiumStyle.AllowShortLoopsOnASingleLine = false;
646 ChromiumStyle.BinPackParameters = false;
647 ChromiumStyle.DerivePointerAlignment = false;
649 ChromiumStyle.SortIncludes = false;
650 return ChromiumStyle;
653 FormatStyle getMozillaStyle() {
654 FormatStyle MozillaStyle = getLLVMStyle();
655 MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
656 MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
657 MozillaStyle.AlwaysBreakAfterReturnType =
658 FormatStyle::RTBS_TopLevel;
659 MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
660 FormatStyle::DRTBS_TopLevel;
661 MozillaStyle.AlwaysBreakTemplateDeclarations = true;
662 MozillaStyle.BinPackParameters = false;
663 MozillaStyle.BinPackArguments = false;
664 MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
665 MozillaStyle.BreakConstructorInitializersBeforeComma = true;
666 MozillaStyle.ConstructorInitializerIndentWidth = 2;
667 MozillaStyle.ContinuationIndentWidth = 2;
668 MozillaStyle.Cpp11BracedListStyle = false;
669 MozillaStyle.IndentCaseLabels = true;
670 MozillaStyle.ObjCSpaceAfterProperty = true;
671 MozillaStyle.ObjCSpaceBeforeProtocolList = false;
672 MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
673 MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
674 MozillaStyle.SpaceAfterTemplateKeyword = false;
678 FormatStyle getWebKitStyle() {
679 FormatStyle Style = getLLVMStyle();
680 Style.AccessModifierOffset = -4;
681 Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
682 Style.AlignOperands = false;
683 Style.AlignTrailingComments = false;
684 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
685 Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
686 Style.BreakConstructorInitializersBeforeComma = true;
687 Style.Cpp11BracedListStyle = false;
688 Style.ColumnLimit = 0;
689 Style.IndentWidth = 4;
690 Style.NamespaceIndentation = FormatStyle::NI_Inner;
691 Style.ObjCBlockIndentWidth = 4;
692 Style.ObjCSpaceAfterProperty = true;
693 Style.PointerAlignment = FormatStyle::PAS_Left;
697 FormatStyle getGNUStyle() {
698 FormatStyle Style = getLLVMStyle();
699 Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
700 Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
701 Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
702 Style.BreakBeforeBraces = FormatStyle::BS_GNU;
703 Style.BreakBeforeTernaryOperators = true;
704 Style.Cpp11BracedListStyle = false;
705 Style.ColumnLimit = 79;
706 Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
707 Style.Standard = FormatStyle::LS_Cpp03;
711 FormatStyle getNoStyle() {
712 FormatStyle NoStyle = getLLVMStyle();
713 NoStyle.DisableFormat = true;
714 NoStyle.SortIncludes = false;
718 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
719 FormatStyle *Style) {
720 if (Name.equals_lower("llvm")) {
721 *Style = getLLVMStyle();
722 } else if (Name.equals_lower("chromium")) {
723 *Style = getChromiumStyle(Language);
724 } else if (Name.equals_lower("mozilla")) {
725 *Style = getMozillaStyle();
726 } else if (Name.equals_lower("google")) {
727 *Style = getGoogleStyle(Language);
728 } else if (Name.equals_lower("webkit")) {
729 *Style = getWebKitStyle();
730 } else if (Name.equals_lower("gnu")) {
731 *Style = getGNUStyle();
732 } else if (Name.equals_lower("none")) {
733 *Style = getNoStyle();
738 Style->Language = Language;
742 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
744 FormatStyle::LanguageKind Language = Style->Language;
745 assert(Language != FormatStyle::LK_None);
746 if (Text.trim().empty())
747 return make_error_code(ParseError::Error);
749 std::vector<FormatStyle> Styles;
750 llvm::yaml::Input Input(Text);
751 // DocumentListTraits<vector<FormatStyle>> uses the context to get default
752 // values for the fields, keys for which are missing from the configuration.
753 // Mapping also uses the context to get the language to find the correct
755 Input.setContext(Style);
758 return Input.error();
760 for (unsigned i = 0; i < Styles.size(); ++i) {
761 // Ensures that only the first configuration can skip the Language option.
762 if (Styles[i].Language == FormatStyle::LK_None && i != 0)
763 return make_error_code(ParseError::Error);
764 // Ensure that each language is configured at most once.
765 for (unsigned j = 0; j < i; ++j) {
766 if (Styles[i].Language == Styles[j].Language) {
768 << "Duplicate languages in the config file on positions " << j
769 << " and " << i << "\n");
770 return make_error_code(ParseError::Error);
774 // Look for a suitable configuration starting from the end, so we can
775 // find the configuration for the specific language first, and the default
776 // configuration (which can only be at slot 0) after it.
777 for (int i = Styles.size() - 1; i >= 0; --i) {
778 if (Styles[i].Language == Language ||
779 Styles[i].Language == FormatStyle::LK_None) {
781 Style->Language = Language;
782 return make_error_code(ParseError::Success);
785 return make_error_code(ParseError::Unsuitable);
788 std::string configurationAsText(const FormatStyle &Style) {
790 llvm::raw_string_ostream Stream(Text);
791 llvm::yaml::Output Output(Stream);
792 // We use the same mapping method for input and output, so we need a non-const
794 FormatStyle NonConstStyle = expandPresets(Style);
795 Output << NonConstStyle;
801 class JavaScriptRequoter : public TokenAnalyzer {
803 JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
804 : TokenAnalyzer(Env, Style) {}
806 tooling::Replacements
807 analyze(TokenAnnotator &Annotator,
808 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
809 FormatTokenLexer &Tokens) override {
810 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
811 AnnotatedLines.end());
812 tooling::Replacements Result;
813 requoteJSStringLiteral(AnnotatedLines, Result);
818 // Replaces double/single-quoted string literal as appropriate, re-escaping
819 // the contents in the process.
820 void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
821 tooling::Replacements &Result) {
822 for (AnnotatedLine *Line : Lines) {
823 requoteJSStringLiteral(Line->Children, Result);
826 for (FormatToken *FormatTok = Line->First; FormatTok;
827 FormatTok = FormatTok->Next) {
828 StringRef Input = FormatTok->TokenText;
829 if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
830 // NB: testing for not starting with a double quote to avoid
831 // breaking `template strings`.
832 (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
833 !Input.startswith("\"")) ||
834 (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
835 !Input.startswith("\'")))
838 // Change start and end quote.
839 bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
840 SourceLocation Start = FormatTok->Tok.getLocation();
841 auto Replace = [&](SourceLocation Start, unsigned Length,
842 StringRef ReplacementText) {
843 auto Err = Result.add(tooling::Replacement(
844 Env.getSourceManager(), Start, Length, ReplacementText));
845 // FIXME: handle error. For now, print error message and skip the
846 // replacement for release version.
848 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
852 Replace(Start, 1, IsSingle ? "'" : "\"");
853 Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
854 IsSingle ? "'" : "\"");
856 // Escape internal quotes.
857 bool Escaped = false;
858 for (size_t i = 1; i < Input.size() - 1; i++) {
861 if (!Escaped && i + 1 < Input.size() &&
862 ((IsSingle && Input[i + 1] == '"') ||
863 (!IsSingle && Input[i + 1] == '\''))) {
864 // Remove this \, it's escaping a " or ' that no longer needs
866 Replace(Start.getLocWithOffset(i), 1, "");
873 if (!Escaped && IsSingle == (Input[i] == '\'')) {
875 Replace(Start.getLocWithOffset(i), 0, "\\");
889 class Formatter : public TokenAnalyzer {
891 Formatter(const Environment &Env, const FormatStyle &Style,
892 bool *IncompleteFormat)
893 : TokenAnalyzer(Env, Style), IncompleteFormat(IncompleteFormat) {}
895 tooling::Replacements
896 analyze(TokenAnnotator &Annotator,
897 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
898 FormatTokenLexer &Tokens) override {
899 tooling::Replacements Result;
900 deriveLocalStyle(AnnotatedLines);
901 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
902 AnnotatedLines.end());
903 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
904 Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
906 Annotator.setCommentLineLevels(AnnotatedLines);
908 WhitespaceManager Whitespaces(
909 Env.getSourceManager(), Style,
910 inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
911 ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
912 Env.getSourceManager(), Whitespaces, Encoding,
913 BinPackInconclusiveFunctions);
914 UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
916 .format(AnnotatedLines);
917 for (const auto &R : Whitespaces.generateReplacements())
925 static bool inputUsesCRLF(StringRef Text) {
926 return Text.count('\r') * 2 > Text.count('\n');
930 hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
931 for (const AnnotatedLine *Line : Lines) {
932 if (hasCpp03IncompatibleFormat(Line->Children))
934 for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
935 if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
936 if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
938 if (Tok->is(TT_TemplateCloser) &&
939 Tok->Previous->is(TT_TemplateCloser))
947 int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
948 int AlignmentDiff = 0;
949 for (const AnnotatedLine *Line : Lines) {
950 AlignmentDiff += countVariableAlignments(Line->Children);
951 for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
952 if (!Tok->is(TT_PointerOrReference))
955 Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
956 bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
957 Tok->Next->WhitespaceRange.getEnd();
958 if (SpaceBefore && !SpaceAfter)
960 if (!SpaceBefore && SpaceAfter)
964 return AlignmentDiff;
968 deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
969 bool HasBinPackedFunction = false;
970 bool HasOnePerLineFunction = false;
971 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
972 if (!AnnotatedLines[i]->First->Next)
974 FormatToken *Tok = AnnotatedLines[i]->First->Next;
976 if (Tok->PackingKind == PPK_BinPacked)
977 HasBinPackedFunction = true;
978 if (Tok->PackingKind == PPK_OnePerLine)
979 HasOnePerLineFunction = true;
984 if (Style.DerivePointerAlignment)
985 Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
986 ? FormatStyle::PAS_Left
987 : FormatStyle::PAS_Right;
988 if (Style.Standard == FormatStyle::LS_Auto)
989 Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
990 ? FormatStyle::LS_Cpp11
991 : FormatStyle::LS_Cpp03;
992 BinPackInconclusiveFunctions =
993 HasBinPackedFunction || !HasOnePerLineFunction;
996 bool BinPackInconclusiveFunctions;
997 bool *IncompleteFormat;
1000 // This class clean up the erroneous/redundant code around the given ranges in
1002 class Cleaner : public TokenAnalyzer {
1004 Cleaner(const Environment &Env, const FormatStyle &Style)
1005 : TokenAnalyzer(Env, Style),
1006 DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1008 // FIXME: eliminate unused parameters.
1009 tooling::Replacements
1010 analyze(TokenAnnotator &Annotator,
1011 SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1012 FormatTokenLexer &Tokens) override {
1013 // FIXME: in the current implementation the granularity of affected range
1014 // is an annotated line. However, this is not sufficient. Furthermore,
1015 // redundant code introduced by replacements does not necessarily
1016 // intercept with ranges of replacements that result in the redundancy.
1017 // To determine if some redundant code is actually introduced by
1018 // replacements(e.g. deletions), we need to come up with a more
1019 // sophisticated way of computing affected ranges.
1020 AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1021 AnnotatedLines.end());
1023 checkEmptyNamespace(AnnotatedLines);
1025 for (auto &Line : AnnotatedLines) {
1026 if (Line->Affected) {
1027 cleanupRight(Line->First, tok::comma, tok::comma);
1028 cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1029 cleanupRight(Line->First, tok::l_paren, tok::comma);
1030 cleanupLeft(Line->First, tok::comma, tok::r_paren);
1031 cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1032 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1033 cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1037 return generateFixes();
1041 bool containsOnlyComments(const AnnotatedLine &Line) {
1042 for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1043 if (Tok->isNot(tok::comment))
1049 // Iterate through all lines and remove any empty (nested) namespaces.
1050 void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1051 std::set<unsigned> DeletedLines;
1052 for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1053 auto &Line = *AnnotatedLines[i];
1054 if (Line.startsWith(tok::kw_namespace) ||
1055 Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1056 checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1060 for (auto Line : DeletedLines) {
1061 FormatToken *Tok = AnnotatedLines[Line]->First;
1069 // The function checks if the namespace, which starts from \p CurrentLine, and
1070 // its nested namespaces are empty and delete them if they are empty. It also
1071 // sets \p NewLine to the last line checked.
1072 // Returns true if the current namespace is empty.
1073 bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1074 unsigned CurrentLine, unsigned &NewLine,
1075 std::set<unsigned> &DeletedLines) {
1076 unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1077 if (Style.BraceWrapping.AfterNamespace) {
1078 // If the left brace is in a new line, we should consume it first so that
1079 // it does not make the namespace non-empty.
1080 // FIXME: error handling if there is no left brace.
1081 if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1082 NewLine = CurrentLine;
1085 } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1088 while (++CurrentLine < End) {
1089 if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1092 if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1093 AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1094 tok::kw_namespace)) {
1095 if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1098 CurrentLine = NewLine;
1102 if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1105 // If there is anything other than comments or nested namespaces in the
1106 // current namespace, the namespace cannot be empty.
1107 NewLine = CurrentLine;
1111 NewLine = CurrentLine;
1112 if (CurrentLine >= End)
1115 // Check if the empty namespace is actually affected by changed ranges.
1116 if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1117 AnnotatedLines[InitLine]->First->Tok.getLocation(),
1118 AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1121 for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1122 DeletedLines.insert(i);
1128 // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1129 // of the token in the pair if the left token has \p LK token kind and the
1130 // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1131 // is deleted on match; otherwise, the right token is deleted.
1132 template <typename LeftKind, typename RightKind>
1133 void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1135 auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1136 for (auto *Res = Tok.Next; Res; Res = Res->Next)
1137 if (!Res->is(tok::comment) &&
1138 DeletedTokens.find(Res) == DeletedTokens.end())
1142 for (auto *Left = Start; Left;) {
1143 auto *Right = NextNotDeleted(*Left);
1146 if (Left->is(LK) && Right->is(RK)) {
1147 deleteToken(DeleteLeft ? Left : Right);
1148 for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1150 // If the right token is deleted, we should keep the left token
1151 // unchanged and pair it with the new right token.
1159 template <typename LeftKind, typename RightKind>
1160 void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1161 cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1164 template <typename LeftKind, typename RightKind>
1165 void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1166 cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1169 // Delete the given token.
1170 inline void deleteToken(FormatToken *Tok) {
1172 DeletedTokens.insert(Tok);
1175 tooling::Replacements generateFixes() {
1176 tooling::Replacements Fixes;
1177 std::vector<FormatToken *> Tokens;
1178 std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1179 std::back_inserter(Tokens));
1181 // Merge multiple continuous token deletions into one big deletion so that
1182 // the number of replacements can be reduced. This makes computing affected
1183 // ranges more efficient when we run reformat on the changed code.
1185 while (Idx < Tokens.size()) {
1186 unsigned St = Idx, End = Idx;
1187 while ((End + 1) < Tokens.size() &&
1188 Tokens[End]->Next == Tokens[End + 1]) {
1191 auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1192 Tokens[End]->Tok.getEndLoc());
1194 Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1195 // FIXME: better error handling. for now just print error message and skip
1196 // for the release version.
1198 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1199 assert(false && "Fixes must not conflict!");
1207 // Class for less-than inequality comparason for the set `RedundantTokens`.
1208 // We store tokens in the order they appear in the translation unit so that
1209 // we do not need to sort them in `generateFixes()`.
1210 struct FormatTokenLess {
1211 FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1213 bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1214 return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1215 RHS->Tok.getLocation());
1217 const SourceManager &SM;
1220 // Tokens to be deleted.
1221 std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1224 struct IncludeDirective {
1231 } // end anonymous namespace
1233 // Determines whether 'Ranges' intersects with ('Start', 'End').
1234 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1236 for (auto Range : Ranges) {
1237 if (Range.getOffset() < End &&
1238 Range.getOffset() + Range.getLength() > Start)
1244 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1245 // before sorting/deduplicating. Index is the index of the include under the
1246 // cursor in the original set of includes. If this include has duplicates, it is
1247 // the index of the first of the duplicates as the others are going to be
1248 // removed. OffsetToEOL describes the cursor's position relative to the end of
1249 // its current line.
1250 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1251 static std::pair<unsigned, unsigned>
1252 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1253 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1254 unsigned CursorIndex = UINT_MAX;
1255 unsigned OffsetToEOL = 0;
1256 for (int i = 0, e = Includes.size(); i != e; ++i) {
1257 unsigned Start = Includes[Indices[i]].Offset;
1258 unsigned End = Start + Includes[Indices[i]].Text.size();
1259 if (!(Cursor >= Start && Cursor < End))
1261 CursorIndex = Indices[i];
1262 OffsetToEOL = End - Cursor;
1263 // Put the cursor on the only remaining #include among the duplicate
1265 while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1269 return std::make_pair(CursorIndex, OffsetToEOL);
1272 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1273 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1275 // #include directives with the same text will be deduplicated, and only the
1276 // first #include in the duplicate #includes remains. If the `Cursor` is
1277 // provided and put on a deleted #include, it will be moved to the remaining
1278 // #include in the duplicate #includes.
1279 static void sortCppIncludes(const FormatStyle &Style,
1280 const SmallVectorImpl<IncludeDirective> &Includes,
1281 ArrayRef<tooling::Range> Ranges, StringRef FileName,
1282 tooling::Replacements &Replaces, unsigned *Cursor) {
1283 unsigned IncludesBeginOffset = Includes.front().Offset;
1284 unsigned IncludesEndOffset =
1285 Includes.back().Offset + Includes.back().Text.size();
1286 unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1287 if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1289 SmallVector<unsigned, 16> Indices;
1290 for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1291 Indices.push_back(i);
1293 Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1294 return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1295 std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1297 // The index of the include on which the cursor will be put after
1298 // sorting/deduplicating.
1299 unsigned CursorIndex;
1300 // The offset from cursor to the end of line.
1301 unsigned CursorToEOLOffset;
1303 std::tie(CursorIndex, CursorToEOLOffset) =
1304 FindCursorIndex(Includes, Indices, *Cursor);
1306 // Deduplicate #includes.
1307 Indices.erase(std::unique(Indices.begin(), Indices.end(),
1308 [&](unsigned LHSI, unsigned RHSI) {
1309 return Includes[LHSI].Text == Includes[RHSI].Text;
1313 // If the #includes are out of order, we generate a single replacement fixing
1314 // the entire block. Otherwise, no replacement is generated.
1315 if (Indices.size() == Includes.size() &&
1316 std::is_sorted(Indices.begin(), Indices.end()))
1320 for (unsigned Index : Indices) {
1321 if (!result.empty())
1323 result += Includes[Index].Text;
1324 if (Cursor && CursorIndex == Index)
1325 *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1328 auto Err = Replaces.add(tooling::Replacement(
1329 FileName, Includes.front().Offset, IncludesBlockSize, result));
1330 // FIXME: better error handling. For now, just skip the replacement for the
1333 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1340 // This class manages priorities of #include categories and calculates
1341 // priorities for headers.
1342 class IncludeCategoryManager {
1344 IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1345 : Style(Style), FileName(FileName) {
1346 FileStem = llvm::sys::path::stem(FileName);
1347 for (const auto &Category : Style.IncludeCategories)
1348 CategoryRegexs.emplace_back(Category.Regex);
1349 IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1350 FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1351 FileName.endswith(".cxx") || FileName.endswith(".m") ||
1352 FileName.endswith(".mm");
1355 // Returns the priority of the category which \p IncludeName belongs to.
1356 // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1357 // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1358 int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1360 for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1361 if (CategoryRegexs[i].match(IncludeName)) {
1362 Ret = Style.IncludeCategories[i].Priority;
1365 if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1371 bool isMainHeader(StringRef IncludeName) const {
1372 if (!IncludeName.startswith("\""))
1374 StringRef HeaderStem =
1375 llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1376 if (FileStem.startswith(HeaderStem)) {
1377 llvm::Regex MainIncludeRegex(
1378 (HeaderStem + Style.IncludeIsMainRegex).str());
1379 if (MainIncludeRegex.match(FileStem))
1385 const FormatStyle &Style;
1389 SmallVector<llvm::Regex, 4> CategoryRegexs;
1392 const char IncludeRegexPattern[] =
1393 R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1395 } // anonymous namespace
1397 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1398 ArrayRef<tooling::Range> Ranges,
1400 tooling::Replacements &Replaces,
1403 unsigned SearchFrom = 0;
1404 llvm::Regex IncludeRegex(IncludeRegexPattern);
1405 SmallVector<StringRef, 4> Matches;
1406 SmallVector<IncludeDirective, 16> IncludesInBlock;
1408 // In compiled files, consider the first #include to be the main #include of
1409 // the file if it is not a system #include. This ensures that the header
1410 // doesn't have hidden dependencies
1411 // (http://llvm.org/docs/CodingStandards.html#include-style).
1413 // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1414 // cases where the first #include is unlikely to be the main header.
1415 IncludeCategoryManager Categories(Style, FileName);
1416 bool FirstIncludeBlock = true;
1417 bool MainIncludeFound = false;
1418 bool FormattingOff = false;
1421 auto Pos = Code.find('\n', SearchFrom);
1423 Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1425 StringRef Trimmed = Line.trim();
1426 if (Trimmed == "// clang-format off")
1427 FormattingOff = true;
1428 else if (Trimmed == "// clang-format on")
1429 FormattingOff = false;
1431 if (!FormattingOff && !Line.endswith("\\")) {
1432 if (IncludeRegex.match(Line, &Matches)) {
1433 StringRef IncludeName = Matches[2];
1434 int Category = Categories.getIncludePriority(
1436 /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1438 MainIncludeFound = true;
1439 IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1440 } else if (!IncludesInBlock.empty()) {
1441 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1443 IncludesInBlock.clear();
1444 FirstIncludeBlock = false;
1448 if (Pos == StringRef::npos || Pos + 1 == Code.size())
1450 SearchFrom = Pos + 1;
1452 if (!IncludesInBlock.empty())
1453 sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1457 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1458 ArrayRef<tooling::Range> Ranges,
1459 StringRef FileName, unsigned *Cursor) {
1460 tooling::Replacements Replaces;
1461 if (!Style.SortIncludes)
1463 if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1464 return sortJavaScriptImports(Style, Code, Ranges, FileName);
1465 sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1469 template <typename T>
1470 static llvm::Expected<tooling::Replacements>
1471 processReplacements(T ProcessFunc, StringRef Code,
1472 const tooling::Replacements &Replaces,
1473 const FormatStyle &Style) {
1474 if (Replaces.empty())
1475 return tooling::Replacements();
1477 auto NewCode = applyAllReplacements(Code, Replaces);
1479 return NewCode.takeError();
1480 std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1481 StringRef FileName = Replaces.begin()->getFilePath();
1483 tooling::Replacements FormatReplaces =
1484 ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1486 return Replaces.merge(FormatReplaces);
1489 llvm::Expected<tooling::Replacements>
1490 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1491 const FormatStyle &Style) {
1492 // We need to use lambda function here since there are two versions of
1494 auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1495 std::vector<tooling::Range> Ranges,
1496 StringRef FileName) -> tooling::Replacements {
1497 return sortIncludes(Style, Code, Ranges, FileName);
1499 auto SortedReplaces =
1500 processReplacements(SortIncludes, Code, Replaces, Style);
1501 if (!SortedReplaces)
1502 return SortedReplaces.takeError();
1504 // We need to use lambda function here since there are two versions of
1506 auto Reformat = [](const FormatStyle &Style, StringRef Code,
1507 std::vector<tooling::Range> Ranges,
1508 StringRef FileName) -> tooling::Replacements {
1509 return reformat(Style, Code, Ranges, FileName);
1511 return processReplacements(Reformat, Code, *SortedReplaces, Style);
1516 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1517 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1518 llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1521 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1522 return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1525 // Returns the offset after skipping a sequence of tokens, matched by \p
1526 // GetOffsetAfterSequence, from the start of the code.
1527 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1528 // tokens and returns an offset after the sequence.
1529 unsigned getOffsetAfterTokenSequence(
1530 StringRef FileName, StringRef Code, const FormatStyle &Style,
1531 std::function<unsigned(const SourceManager &, Lexer &, Token &)>
1532 GetOffsetAfterSequense) {
1533 std::unique_ptr<Environment> Env =
1534 Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1535 const SourceManager &SourceMgr = Env->getSourceManager();
1536 Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1537 getFormattingLangOpts(Style));
1539 // Get the first token.
1540 Lex.LexFromRawLexer(Tok);
1541 return GetOffsetAfterSequense(SourceMgr, Lex, Tok);
1544 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1545 // \p Tok will be the token after this directive; otherwise, it can be any token
1546 // after the given \p Tok (including \p Tok).
1547 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1548 bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1549 Tok.is(tok::raw_identifier) &&
1550 Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1551 Tok.is(tok::raw_identifier);
1553 Lex.LexFromRawLexer(Tok);
1557 void skipComments(Lexer &Lex, Token &Tok) {
1558 while (Tok.is(tok::comment))
1559 if (Lex.LexFromRawLexer(Tok))
1563 // Returns the offset after header guard directives and any comments
1564 // before/after header guards. If no header guard presents in the code, this
1565 // will returns the offset after skipping all comments from the start of the
1567 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1569 const FormatStyle &Style) {
1570 return getOffsetAfterTokenSequence(
1571 FileName, Code, Style,
1572 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1573 skipComments(Lex, Tok);
1574 unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1575 if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1576 skipComments(Lex, Tok);
1577 if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1578 return SM.getFileOffset(Tok.getLocation());
1580 return InitialOffset;
1584 // Check if a sequence of tokens is like
1585 // "#include ("header.h" | <header.h>)".
1586 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1587 // any token after the given \p Tok (including \p Tok).
1588 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1589 auto Matched = [&]() {
1590 Lex.LexFromRawLexer(Tok);
1593 if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1594 Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1595 if (Lex.LexFromRawLexer(Tok))
1597 if (Tok.is(tok::string_literal))
1599 if (Tok.is(tok::less)) {
1600 while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1602 if (Tok.is(tok::greater))
1609 // Returns the offset of the last #include directive after which a new
1610 // #include can be inserted. This ignores #include's after the #include block(s)
1611 // in the beginning of a file to avoid inserting headers into code sections
1612 // where new #include's should not be added by default.
1613 // These code sections include:
1614 // - raw string literals (containing #include).
1616 // - Special #include's among declarations (e.g. functions).
1618 // If no #include after which a new #include can be inserted, this returns the
1619 // offset after skipping all comments from the start of the code.
1620 // Inserting after an #include is not allowed if it comes after code that is not
1621 // #include (e.g. pre-processing directive that is not #include, declarations).
1622 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1623 const FormatStyle &Style) {
1624 return getOffsetAfterTokenSequence(
1625 FileName, Code, Style,
1626 [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1627 skipComments(Lex, Tok);
1628 unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1629 while (checkAndConsumeInclusiveDirective(Lex, Tok))
1630 MaxOffset = SM.getFileOffset(Tok.getLocation());
1635 bool isDeletedHeader(llvm::StringRef HeaderName,
1636 const std::set<llvm::StringRef> &HeadersToDelete) {
1637 return HeadersToDelete.count(HeaderName) ||
1638 HeadersToDelete.count(HeaderName.trim("\"<>"));
1641 // FIXME: insert empty lines between newly created blocks.
1642 tooling::Replacements
1643 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1644 const FormatStyle &Style) {
1645 if (Style.Language != FormatStyle::LanguageKind::LK_Cpp)
1648 tooling::Replacements HeaderInsertions;
1649 std::set<llvm::StringRef> HeadersToDelete;
1650 tooling::Replacements Result;
1651 for (const auto &R : Replaces) {
1652 if (isHeaderInsertion(R)) {
1653 // Replacements from \p Replaces must be conflict-free already, so we can
1654 // simply consume the error.
1655 llvm::consumeError(HeaderInsertions.add(R));
1656 } else if (isHeaderDeletion(R)) {
1657 HeadersToDelete.insert(R.getReplacementText());
1658 } else if (R.getOffset() == UINT_MAX) {
1659 llvm::errs() << "Insertions other than header #include insertion are "
1661 << R.getReplacementText() << "\n";
1663 llvm::consumeError(Result.add(R));
1666 if (HeaderInsertions.empty() && HeadersToDelete.empty())
1669 llvm::Regex IncludeRegex(IncludeRegexPattern);
1670 llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1671 SmallVector<StringRef, 4> Matches;
1673 StringRef FileName = Replaces.begin()->getFilePath();
1674 IncludeCategoryManager Categories(Style, FileName);
1676 // Record the offset of the end of the last include in each category.
1677 std::map<int, int> CategoryEndOffsets;
1678 // All possible priorities.
1679 // Add 0 for main header and INT_MAX for headers that are not in any category.
1680 std::set<int> Priorities = {0, INT_MAX};
1681 for (const auto &Category : Style.IncludeCategories)
1682 Priorities.insert(Category.Priority);
1683 int FirstIncludeOffset = -1;
1684 // All new headers should be inserted after this offset.
1685 unsigned MinInsertOffset =
1686 getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1687 StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1688 // Max insertion offset in the original code.
1689 unsigned MaxInsertOffset =
1691 getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1692 SmallVector<StringRef, 32> Lines;
1693 TrimmedCode.split(Lines, '\n');
1694 unsigned Offset = MinInsertOffset;
1695 unsigned NextLineOffset;
1696 std::set<StringRef> ExistingIncludes;
1697 for (auto Line : Lines) {
1698 NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1699 if (IncludeRegex.match(Line, &Matches)) {
1700 // The header name with quotes or angle brackets.
1701 StringRef IncludeName = Matches[2];
1702 ExistingIncludes.insert(IncludeName);
1703 // Only record the offset of current #include if we can insert after it.
1704 if (Offset <= MaxInsertOffset) {
1705 int Category = Categories.getIncludePriority(
1706 IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1707 CategoryEndOffsets[Category] = NextLineOffset;
1708 if (FirstIncludeOffset < 0)
1709 FirstIncludeOffset = Offset;
1711 if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1712 // If this is the last line without trailing newline, we need to make
1713 // sure we don't delete across the file boundary.
1714 unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1716 Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1718 // Ignore the deletion on conflict.
1719 llvm::errs() << "Failed to add header deletion replacement for "
1720 << IncludeName << ": " << llvm::toString(std::move(Err))
1725 Offset = NextLineOffset;
1728 // Populate CategoryEndOfssets:
1729 // - Ensure that CategoryEndOffset[Highest] is always populated.
1730 // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1731 // is set, up to CategoryEndOffset[Highest].
1732 auto Highest = Priorities.begin();
1733 if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1734 if (FirstIncludeOffset >= 0)
1735 CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1737 CategoryEndOffsets[*Highest] = MinInsertOffset;
1739 // By this point, CategoryEndOffset[Highest] is always set appropriately:
1740 // - to an appropriate location before/after existing #includes, or
1741 // - to right after the header guard, or
1742 // - to the beginning of the file.
1743 for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1744 if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1745 CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1747 bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1748 for (const auto &R : HeaderInsertions) {
1749 auto IncludeDirective = R.getReplacementText();
1750 bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1751 assert(Matched && "Header insertion replacement must have replacement text "
1754 auto IncludeName = Matches[2];
1755 if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1756 DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1761 Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1762 Offset = CategoryEndOffsets[Category];
1763 std::string NewInclude = !IncludeDirective.endswith("\n")
1764 ? (IncludeDirective + "\n").str()
1765 : IncludeDirective.str();
1766 // When inserting headers at end of the code, also append '\n' to the code
1767 // if it does not end with '\n'.
1768 if (NeedNewLineAtEnd && Offset == Code.size()) {
1769 NewInclude = "\n" + NewInclude;
1770 NeedNewLineAtEnd = false;
1772 auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1773 auto Err = Result.add(NewReplace);
1775 llvm::consumeError(std::move(Err));
1776 unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1777 NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1778 Result = Result.merge(tooling::Replacements(NewReplace));
1784 } // anonymous namespace
1786 llvm::Expected<tooling::Replacements>
1787 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1788 const FormatStyle &Style) {
1789 // We need to use lambda function here since there are two versions of
1791 auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1792 std::vector<tooling::Range> Ranges,
1793 StringRef FileName) -> tooling::Replacements {
1794 return cleanup(Style, Code, Ranges, FileName);
1796 // Make header insertion replacements insert new headers into correct blocks.
1797 tooling::Replacements NewReplaces =
1798 fixCppIncludeInsertions(Code, Replaces, Style);
1799 return processReplacements(Cleanup, Code, NewReplaces, Style);
1802 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1803 ArrayRef<tooling::Range> Ranges,
1804 StringRef FileName, bool *IncompleteFormat) {
1805 FormatStyle Expanded = expandPresets(Style);
1806 if (Expanded.DisableFormat)
1807 return tooling::Replacements();
1809 auto Env = Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1811 if (Style.Language == FormatStyle::LK_JavaScript &&
1812 Style.JavaScriptQuotes != FormatStyle::JSQS_Leave) {
1813 JavaScriptRequoter Requoter(*Env, Expanded);
1814 tooling::Replacements Requotes = Requoter.process();
1815 if (!Requotes.empty()) {
1816 auto NewCode = applyAllReplacements(Code, Requotes);
1818 auto NewEnv = Environment::CreateVirtualEnvironment(
1820 tooling::calculateRangesAfterReplacements(Requotes, Ranges));
1821 Formatter Format(*NewEnv, Expanded, IncompleteFormat);
1822 return Requotes.merge(Format.process());
1827 Formatter Format(*Env, Expanded, IncompleteFormat);
1828 return Format.process();
1831 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
1832 ArrayRef<tooling::Range> Ranges,
1833 StringRef FileName) {
1834 std::unique_ptr<Environment> Env =
1835 Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1836 Cleaner Clean(*Env, Style);
1837 return Clean.process();
1840 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1841 LangOptions LangOpts;
1842 LangOpts.CPlusPlus = 1;
1843 LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1844 LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1845 LangOpts.LineComment = 1;
1846 bool AlternativeOperators = Style.Language == FormatStyle::LK_Cpp;
1847 LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1851 LangOpts.MicrosoftExt = 1; // To get kw___try, kw___finally.
1852 LangOpts.DeclSpecKeyword = 1; // To get __declspec.
1856 const char *StyleOptionHelpDescription =
1857 "Coding style, currently supports:\n"
1858 " LLVM, Google, Chromium, Mozilla, WebKit.\n"
1859 "Use -style=file to load style configuration from\n"
1860 ".clang-format file located in one of the parent\n"
1861 "directories of the source file (or current\n"
1862 "directory for stdin).\n"
1863 "Use -style=\"{key: value, ...}\" to set specific\n"
1864 "parameters, e.g.:\n"
1865 " -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1867 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1868 if (FileName.endswith(".java"))
1869 return FormatStyle::LK_Java;
1870 if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
1871 return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
1872 if (FileName.endswith(".m") || FileName.endswith(".mm"))
1873 return FormatStyle::LK_ObjC;
1874 if (FileName.endswith_lower(".proto") ||
1875 FileName.endswith_lower(".protodevel"))
1876 return FormatStyle::LK_Proto;
1877 if (FileName.endswith_lower(".td"))
1878 return FormatStyle::LK_TableGen;
1879 return FormatStyle::LK_Cpp;
1882 FormatStyle getStyle(StringRef StyleName, StringRef FileName,
1883 StringRef FallbackStyle, StringRef Code,
1884 vfs::FileSystem *FS) {
1886 FS = vfs::getRealFileSystem().get();
1888 FormatStyle Style = getLLVMStyle();
1889 Style.Language = getLanguageByFileName(FileName);
1891 // This is a very crude detection of whether a header contains ObjC code that
1892 // should be improved over time and probably be done on tokens, not one the
1893 // bare content of the file.
1894 if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
1895 (Code.contains("\n- (") || Code.contains("\n+ (")))
1896 Style.Language = FormatStyle::LK_ObjC;
1898 if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) {
1899 llvm::errs() << "Invalid fallback style \"" << FallbackStyle
1900 << "\" using LLVM style\n";
1904 if (StyleName.startswith("{")) {
1905 // Parse YAML/JSON style from the command line.
1906 if (std::error_code ec = parseConfiguration(StyleName, &Style)) {
1907 llvm::errs() << "Error parsing -style: " << ec.message() << ", using "
1908 << FallbackStyle << " style\n";
1913 if (!StyleName.equals_lower("file")) {
1914 if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1915 llvm::errs() << "Invalid value for -style, using " << FallbackStyle
1920 // Look for .clang-format/_clang-format file in the file's parent directories.
1921 SmallString<128> UnsuitableConfigFiles;
1922 SmallString<128> Path(FileName);
1923 if (std::error_code EC = FS->makeAbsolute(Path)) {
1924 llvm::errs() << EC.message() << "\n";
1928 for (StringRef Directory = Path; !Directory.empty();
1929 Directory = llvm::sys::path::parent_path(Directory)) {
1931 auto Status = FS->status(Directory);
1933 Status->getType() != llvm::sys::fs::file_type::directory_file) {
1937 SmallString<128> ConfigFile(Directory);
1939 llvm::sys::path::append(ConfigFile, ".clang-format");
1940 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1942 Status = FS->status(ConfigFile.str());
1944 Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
1946 // Try _clang-format too, since dotfiles are not commonly used on Windows.
1947 ConfigFile = Directory;
1948 llvm::sys::path::append(ConfigFile, "_clang-format");
1949 DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1950 Status = FS->status(ConfigFile.str());
1952 (Status->getType() == llvm::sys::fs::file_type::regular_file);
1956 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
1957 FS->getBufferForFile(ConfigFile.str());
1958 if (std::error_code EC = Text.getError()) {
1959 llvm::errs() << EC.message() << "\n";
1962 if (std::error_code ec =
1963 parseConfiguration(Text.get()->getBuffer(), &Style)) {
1964 if (ec == ParseError::Unsuitable) {
1965 if (!UnsuitableConfigFiles.empty())
1966 UnsuitableConfigFiles.append(", ");
1967 UnsuitableConfigFiles.append(ConfigFile);
1970 llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message()
1974 DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
1978 if (!UnsuitableConfigFiles.empty()) {
1979 llvm::errs() << "Configuration file(s) do(es) not support "
1980 << getLanguageName(Style.Language) << ": "
1981 << UnsuitableConfigFiles << "\n";
1986 } // namespace format
1987 } // namespace clang