1 //===--- Format.cpp - Format C++ code -------------------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// \brief This file implements functions declared in Format.h. This will be
12 /// split into separate files as we go.
13 ///
14 //===----------------------------------------------------------------------===//
15 
16 #include "clang/Format/Format.h"
17 #include "AffectedRangeManager.h"
18 #include "ContinuationIndenter.h"
19 #include "FormatTokenLexer.h"
20 #include "NamespaceEndCommentsFixer.h"
21 #include "SortJavaScriptImports.h"
22 #include "TokenAnalyzer.h"
23 #include "TokenAnnotator.h"
24 #include "UnwrappedLineFormatter.h"
25 #include "UnwrappedLineParser.h"
26 #include "UsingDeclarationsSorter.h"
27 #include "WhitespaceManager.h"
28 #include "clang/Basic/Diagnostic.h"
29 #include "clang/Basic/DiagnosticOptions.h"
30 #include "clang/Basic/SourceManager.h"
31 #include "clang/Basic/VirtualFileSystem.h"
32 #include "clang/Lex/Lexer.h"
33 #include "llvm/ADT/STLExtras.h"
34 #include "llvm/Support/Allocator.h"
35 #include "llvm/Support/Debug.h"
36 #include "llvm/Support/Path.h"
37 #include "llvm/Support/Regex.h"
38 #include "llvm/Support/YAMLTraits.h"
39 #include <algorithm>
40 #include <memory>
41 #include <string>
42 
43 #define DEBUG_TYPE "format-formatter"
44 
45 using clang::format::FormatStyle;
46 
47 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory)
48 
49 namespace llvm {
50 namespace yaml {
51 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
52   static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
53     IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
54     IO.enumCase(Value, "Java", FormatStyle::LK_Java);
55     IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
56     IO.enumCase(Value, "ObjC", FormatStyle::LK_ObjC);
57     IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
58     IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen);
59     IO.enumCase(Value, "TextProto", FormatStyle::LK_TextProto);
60   }
61 };
62 
63 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
64   static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
65     IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
66     IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
67     IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
68     IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
69     IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
70   }
71 };
72 
73 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
74   static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
75     IO.enumCase(Value, "Never", FormatStyle::UT_Never);
76     IO.enumCase(Value, "false", FormatStyle::UT_Never);
77     IO.enumCase(Value, "Always", FormatStyle::UT_Always);
78     IO.enumCase(Value, "true", FormatStyle::UT_Always);
79     IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
80     IO.enumCase(Value, "ForContinuationAndIndentation",
81                 FormatStyle::UT_ForContinuationAndIndentation);
82   }
83 };
84 
85 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> {
86   static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) {
87     IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave);
88     IO.enumCase(Value, "Single", FormatStyle::JSQS_Single);
89     IO.enumCase(Value, "Double", FormatStyle::JSQS_Double);
90   }
91 };
92 
93 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
94   static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
95     IO.enumCase(Value, "None", FormatStyle::SFS_None);
96     IO.enumCase(Value, "false", FormatStyle::SFS_None);
97     IO.enumCase(Value, "All", FormatStyle::SFS_All);
98     IO.enumCase(Value, "true", FormatStyle::SFS_All);
99     IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
100     IO.enumCase(Value, "InlineOnly", FormatStyle::SFS_InlineOnly);
101     IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
102   }
103 };
104 
105 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
106   static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
107     IO.enumCase(Value, "All", FormatStyle::BOS_All);
108     IO.enumCase(Value, "true", FormatStyle::BOS_All);
109     IO.enumCase(Value, "None", FormatStyle::BOS_None);
110     IO.enumCase(Value, "false", FormatStyle::BOS_None);
111     IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
112   }
113 };
114 
115 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
116   static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
117     IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
118     IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
119     IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla);
120     IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
121     IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
122     IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
123     IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit);
124     IO.enumCase(Value, "Custom", FormatStyle::BS_Custom);
125   }
126 };
127 
128 template <> struct ScalarEnumerationTraits<FormatStyle::BreakConstructorInitializersStyle> {
129   static void enumeration(IO &IO, FormatStyle::BreakConstructorInitializersStyle &Value) {
130     IO.enumCase(Value, "BeforeColon", FormatStyle::BCIS_BeforeColon);
131     IO.enumCase(Value, "BeforeComma", FormatStyle::BCIS_BeforeComma);
132     IO.enumCase(Value, "AfterColon", FormatStyle::BCIS_AfterColon);
133   }
134 };
135 
136 template <>
137 struct ScalarEnumerationTraits<FormatStyle::PPDirectiveIndentStyle> {
138   static void enumeration(IO &IO, FormatStyle::PPDirectiveIndentStyle &Value) {
139     IO.enumCase(Value, "None", FormatStyle::PPDIS_None);
140     IO.enumCase(Value, "AfterHash", FormatStyle::PPDIS_AfterHash);
141   }
142 };
143 
144 template <>
145 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> {
146   static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) {
147     IO.enumCase(Value, "None", FormatStyle::RTBS_None);
148     IO.enumCase(Value, "All", FormatStyle::RTBS_All);
149     IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel);
150     IO.enumCase(Value, "TopLevelDefinitions",
151                 FormatStyle::RTBS_TopLevelDefinitions);
152     IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions);
153   }
154 };
155 
156 template <>
157 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> {
158   static void
159   enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) {
160     IO.enumCase(Value, "None", FormatStyle::DRTBS_None);
161     IO.enumCase(Value, "All", FormatStyle::DRTBS_All);
162     IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel);
163 
164     // For backward compatibility.
165     IO.enumCase(Value, "false", FormatStyle::DRTBS_None);
166     IO.enumCase(Value, "true", FormatStyle::DRTBS_All);
167   }
168 };
169 
170 template <>
171 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
172   static void enumeration(IO &IO,
173                           FormatStyle::NamespaceIndentationKind &Value) {
174     IO.enumCase(Value, "None", FormatStyle::NI_None);
175     IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
176     IO.enumCase(Value, "All", FormatStyle::NI_All);
177   }
178 };
179 
180 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> {
181   static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) {
182     IO.enumCase(Value, "Align", FormatStyle::BAS_Align);
183     IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign);
184     IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak);
185 
186     // For backward compatibility.
187     IO.enumCase(Value, "true", FormatStyle::BAS_Align);
188     IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign);
189   }
190 };
191 
192 template <> struct ScalarEnumerationTraits<FormatStyle::EscapedNewlineAlignmentStyle> {
193   static void enumeration(IO &IO, FormatStyle::EscapedNewlineAlignmentStyle &Value) {
194     IO.enumCase(Value, "DontAlign", FormatStyle::ENAS_DontAlign);
195     IO.enumCase(Value, "Left", FormatStyle::ENAS_Left);
196     IO.enumCase(Value, "Right", FormatStyle::ENAS_Right);
197 
198     // For backward compatibility.
199     IO.enumCase(Value, "true", FormatStyle::ENAS_Left);
200     IO.enumCase(Value, "false", FormatStyle::ENAS_Right);
201   }
202 };
203 
204 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
205   static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
206     IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
207     IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
208     IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
209 
210     // For backward compatibility.
211     IO.enumCase(Value, "true", FormatStyle::PAS_Left);
212     IO.enumCase(Value, "false", FormatStyle::PAS_Right);
213   }
214 };
215 
216 template <>
217 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
218   static void enumeration(IO &IO,
219                           FormatStyle::SpaceBeforeParensOptions &Value) {
220     IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
221     IO.enumCase(Value, "ControlStatements",
222                 FormatStyle::SBPO_ControlStatements);
223     IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
224 
225     // For backward compatibility.
226     IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
227     IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
228   }
229 };
230 
231 template <> struct MappingTraits<FormatStyle> {
232   static void mapping(IO &IO, FormatStyle &Style) {
233     // When reading, read the language first, we need it for getPredefinedStyle.
234     IO.mapOptional("Language", Style.Language);
235 
236     if (IO.outputting()) {
237       StringRef StylesArray[] = {"LLVM",    "Google", "Chromium",
238                                  "Mozilla", "WebKit", "GNU"};
239       ArrayRef<StringRef> Styles(StylesArray);
240       for (size_t i = 0, e = Styles.size(); i < e; ++i) {
241         StringRef StyleName(Styles[i]);
242         FormatStyle PredefinedStyle;
243         if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
244             Style == PredefinedStyle) {
245           IO.mapOptional("# BasedOnStyle", StyleName);
246           break;
247         }
248       }
249     } else {
250       StringRef BasedOnStyle;
251       IO.mapOptional("BasedOnStyle", BasedOnStyle);
252       if (!BasedOnStyle.empty()) {
253         FormatStyle::LanguageKind OldLanguage = Style.Language;
254         FormatStyle::LanguageKind Language =
255             ((FormatStyle *)IO.getContext())->Language;
256         if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
257           IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
258           return;
259         }
260         Style.Language = OldLanguage;
261       }
262     }
263 
264     // For backward compatibility.
265     if (!IO.outputting()) {
266       IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlines);
267       IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
268       IO.mapOptional("IndentFunctionDeclarationAfterType",
269                      Style.IndentWrappedFunctionNames);
270       IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
271       IO.mapOptional("SpaceAfterControlStatementKeyword",
272                      Style.SpaceBeforeParens);
273     }
274 
275     IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
276     IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
277     IO.mapOptional("AlignConsecutiveAssignments",
278                    Style.AlignConsecutiveAssignments);
279     IO.mapOptional("AlignConsecutiveDeclarations",
280                    Style.AlignConsecutiveDeclarations);
281     IO.mapOptional("AlignEscapedNewlines", Style.AlignEscapedNewlines);
282     IO.mapOptional("AlignOperands", Style.AlignOperands);
283     IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
284     IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
285                    Style.AllowAllParametersOfDeclarationOnNextLine);
286     IO.mapOptional("AllowShortBlocksOnASingleLine",
287                    Style.AllowShortBlocksOnASingleLine);
288     IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
289                    Style.AllowShortCaseLabelsOnASingleLine);
290     IO.mapOptional("AllowShortFunctionsOnASingleLine",
291                    Style.AllowShortFunctionsOnASingleLine);
292     IO.mapOptional("AllowShortIfStatementsOnASingleLine",
293                    Style.AllowShortIfStatementsOnASingleLine);
294     IO.mapOptional("AllowShortLoopsOnASingleLine",
295                    Style.AllowShortLoopsOnASingleLine);
296     IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
297                    Style.AlwaysBreakAfterDefinitionReturnType);
298     IO.mapOptional("AlwaysBreakAfterReturnType",
299                    Style.AlwaysBreakAfterReturnType);
300     // If AlwaysBreakAfterDefinitionReturnType was specified but
301     // AlwaysBreakAfterReturnType was not, initialize the latter from the
302     // former for backwards compatibility.
303     if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None &&
304         Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) {
305       if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All)
306         Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
307       else if (Style.AlwaysBreakAfterDefinitionReturnType ==
308                FormatStyle::DRTBS_TopLevel)
309         Style.AlwaysBreakAfterReturnType =
310             FormatStyle::RTBS_TopLevelDefinitions;
311     }
312 
313     IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
314                    Style.AlwaysBreakBeforeMultilineStrings);
315     IO.mapOptional("AlwaysBreakTemplateDeclarations",
316                    Style.AlwaysBreakTemplateDeclarations);
317     IO.mapOptional("BinPackArguments", Style.BinPackArguments);
318     IO.mapOptional("BinPackParameters", Style.BinPackParameters);
319     IO.mapOptional("BraceWrapping", Style.BraceWrapping);
320     IO.mapOptional("BreakBeforeBinaryOperators",
321                    Style.BreakBeforeBinaryOperators);
322     IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
323     IO.mapOptional("BreakBeforeInheritanceComma",
324                    Style.BreakBeforeInheritanceComma);
325     IO.mapOptional("BreakBeforeTernaryOperators",
326                    Style.BreakBeforeTernaryOperators);
327 
328     bool BreakConstructorInitializersBeforeComma = false;
329     IO.mapOptional("BreakConstructorInitializersBeforeComma",
330                    BreakConstructorInitializersBeforeComma);
331     IO.mapOptional("BreakConstructorInitializers",
332                    Style.BreakConstructorInitializers);
333     // If BreakConstructorInitializersBeforeComma was specified but
334     // BreakConstructorInitializers was not, initialize the latter from the
335     // former for backwards compatibility.
336     if (BreakConstructorInitializersBeforeComma &&
337         Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeColon)
338       Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
339 
340     IO.mapOptional("BreakAfterJavaFieldAnnotations",
341                    Style.BreakAfterJavaFieldAnnotations);
342     IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals);
343     IO.mapOptional("ColumnLimit", Style.ColumnLimit);
344     IO.mapOptional("CommentPragmas", Style.CommentPragmas);
345     IO.mapOptional("CompactNamespaces", Style.CompactNamespaces);
346     IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
347                    Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
348     IO.mapOptional("ConstructorInitializerIndentWidth",
349                    Style.ConstructorInitializerIndentWidth);
350     IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
351     IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
352     IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
353     IO.mapOptional("DisableFormat", Style.DisableFormat);
354     IO.mapOptional("ExperimentalAutoDetectBinPacking",
355                    Style.ExperimentalAutoDetectBinPacking);
356     IO.mapOptional("FixNamespaceComments", Style.FixNamespaceComments);
357     IO.mapOptional("ForEachMacros", Style.ForEachMacros);
358     IO.mapOptional("IncludeCategories", Style.IncludeCategories);
359     IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex);
360     IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
361     IO.mapOptional("IndentPPDirectives", Style.IndentPPDirectives);
362     IO.mapOptional("IndentWidth", Style.IndentWidth);
363     IO.mapOptional("IndentWrappedFunctionNames",
364                    Style.IndentWrappedFunctionNames);
365     IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes);
366     IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports);
367     IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
368                    Style.KeepEmptyLinesAtTheStartOfBlocks);
369     IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
370     IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
371     IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
372     IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
373     IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
374     IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
375     IO.mapOptional("ObjCSpaceBeforeProtocolList",
376                    Style.ObjCSpaceBeforeProtocolList);
377     IO.mapOptional("PenaltyBreakAssignment",
378                    Style.PenaltyBreakAssignment);
379     IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
380                    Style.PenaltyBreakBeforeFirstCallParameter);
381     IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
382     IO.mapOptional("PenaltyBreakFirstLessLess",
383                    Style.PenaltyBreakFirstLessLess);
384     IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
385     IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
386     IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
387                    Style.PenaltyReturnTypeOnItsOwnLine);
388     IO.mapOptional("PointerAlignment", Style.PointerAlignment);
389     IO.mapOptional("ReflowComments", Style.ReflowComments);
390     IO.mapOptional("SortIncludes", Style.SortIncludes);
391     IO.mapOptional("SortUsingDeclarations", Style.SortUsingDeclarations);
392     IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
393     IO.mapOptional("SpaceAfterTemplateKeyword", Style.SpaceAfterTemplateKeyword);
394     IO.mapOptional("SpaceBeforeAssignmentOperators",
395                    Style.SpaceBeforeAssignmentOperators);
396     IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
397     IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
398     IO.mapOptional("SpacesBeforeTrailingComments",
399                    Style.SpacesBeforeTrailingComments);
400     IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
401     IO.mapOptional("SpacesInContainerLiterals",
402                    Style.SpacesInContainerLiterals);
403     IO.mapOptional("SpacesInCStyleCastParentheses",
404                    Style.SpacesInCStyleCastParentheses);
405     IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
406     IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
407     IO.mapOptional("Standard", Style.Standard);
408     IO.mapOptional("TabWidth", Style.TabWidth);
409     IO.mapOptional("UseTab", Style.UseTab);
410   }
411 };
412 
413 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
414   static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
415     IO.mapOptional("AfterClass", Wrapping.AfterClass);
416     IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
417     IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
418     IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
419     IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
420     IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
421     IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
422     IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
423     IO.mapOptional("AfterExternBlock", Wrapping.AfterExternBlock);
424     IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
425     IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
426     IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
427     IO.mapOptional("SplitEmptyFunction", Wrapping.SplitEmptyFunction);
428     IO.mapOptional("SplitEmptyRecord", Wrapping.SplitEmptyRecord);
429     IO.mapOptional("SplitEmptyNamespace", Wrapping.SplitEmptyNamespace);
430   }
431 };
432 
433 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
434   static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
435     IO.mapOptional("Regex", Category.Regex);
436     IO.mapOptional("Priority", Category.Priority);
437   }
438 };
439 
440 // Allows to read vector<FormatStyle> while keeping default values.
441 // IO.getContext() should contain a pointer to the FormatStyle structure, that
442 // will be used to get default values for missing keys.
443 // If the first element has no Language specified, it will be treated as the
444 // default one for the following elements.
445 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
446   static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
447     return Seq.size();
448   }
449   static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
450                               size_t Index) {
451     if (Index >= Seq.size()) {
452       assert(Index == Seq.size());
453       FormatStyle Template;
454       if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
455         Template = Seq[0];
456       } else {
457         Template = *((const FormatStyle *)IO.getContext());
458         Template.Language = FormatStyle::LK_None;
459       }
460       Seq.resize(Index + 1, Template);
461     }
462     return Seq[Index];
463   }
464 };
465 } // namespace yaml
466 } // namespace llvm
467 
468 namespace clang {
469 namespace format {
470 
471 const std::error_category &getParseCategory() {
472   static ParseErrorCategory C;
473   return C;
474 }
475 std::error_code make_error_code(ParseError e) {
476   return std::error_code(static_cast<int>(e), getParseCategory());
477 }
478 
479 inline llvm::Error make_string_error(const llvm::Twine &Message) {
480   return llvm::make_error<llvm::StringError>(Message,
481                                              llvm::inconvertibleErrorCode());
482 }
483 
484 const char *ParseErrorCategory::name() const noexcept {
485   return "clang-format.parse_error";
486 }
487 
488 std::string ParseErrorCategory::message(int EV) const {
489   switch (static_cast<ParseError>(EV)) {
490   case ParseError::Success:
491     return "Success";
492   case ParseError::Error:
493     return "Invalid argument";
494   case ParseError::Unsuitable:
495     return "Unsuitable";
496   }
497   llvm_unreachable("unexpected parse error");
498 }
499 
500 static FormatStyle expandPresets(const FormatStyle &Style) {
501   if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
502     return Style;
503   FormatStyle Expanded = Style;
504   Expanded.BraceWrapping = {false, false, false, false, false, false,
505                             false, false, false, false, false, false,
506                             true,  true,  true};
507   switch (Style.BreakBeforeBraces) {
508   case FormatStyle::BS_Linux:
509     Expanded.BraceWrapping.AfterClass = true;
510     Expanded.BraceWrapping.AfterFunction = true;
511     Expanded.BraceWrapping.AfterNamespace = true;
512     break;
513   case FormatStyle::BS_Mozilla:
514     Expanded.BraceWrapping.AfterClass = true;
515     Expanded.BraceWrapping.AfterEnum = true;
516     Expanded.BraceWrapping.AfterFunction = true;
517     Expanded.BraceWrapping.AfterStruct = true;
518     Expanded.BraceWrapping.AfterUnion = true;
519     Expanded.BraceWrapping.AfterExternBlock = true;
520     Expanded.BraceWrapping.SplitEmptyFunction = true;
521     Expanded.BraceWrapping.SplitEmptyRecord = false;
522     break;
523   case FormatStyle::BS_Stroustrup:
524     Expanded.BraceWrapping.AfterFunction = true;
525     Expanded.BraceWrapping.BeforeCatch = true;
526     Expanded.BraceWrapping.BeforeElse = true;
527     break;
528   case FormatStyle::BS_Allman:
529     Expanded.BraceWrapping.AfterClass = true;
530     Expanded.BraceWrapping.AfterControlStatement = true;
531     Expanded.BraceWrapping.AfterEnum = true;
532     Expanded.BraceWrapping.AfterFunction = true;
533     Expanded.BraceWrapping.AfterNamespace = true;
534     Expanded.BraceWrapping.AfterObjCDeclaration = true;
535     Expanded.BraceWrapping.AfterStruct = true;
536     Expanded.BraceWrapping.AfterExternBlock = true;
537     Expanded.BraceWrapping.BeforeCatch = true;
538     Expanded.BraceWrapping.BeforeElse = true;
539     break;
540   case FormatStyle::BS_GNU:
541     Expanded.BraceWrapping = {true, true, true, true, true, true,
542                               true, true, true, true, true, true,
543                               true, true, true};
544     break;
545   case FormatStyle::BS_WebKit:
546     Expanded.BraceWrapping.AfterFunction = true;
547     break;
548   default:
549     break;
550   }
551   return Expanded;
552 }
553 
554 FormatStyle getLLVMStyle() {
555   FormatStyle LLVMStyle;
556   LLVMStyle.Language = FormatStyle::LK_Cpp;
557   LLVMStyle.AccessModifierOffset = -2;
558   LLVMStyle.AlignEscapedNewlines = FormatStyle::ENAS_Right;
559   LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align;
560   LLVMStyle.AlignOperands = true;
561   LLVMStyle.AlignTrailingComments = true;
562   LLVMStyle.AlignConsecutiveAssignments = false;
563   LLVMStyle.AlignConsecutiveDeclarations = false;
564   LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
565   LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
566   LLVMStyle.AllowShortBlocksOnASingleLine = false;
567   LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
568   LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
569   LLVMStyle.AllowShortLoopsOnASingleLine = false;
570   LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
571   LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
572   LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
573   LLVMStyle.AlwaysBreakTemplateDeclarations = false;
574   LLVMStyle.BinPackArguments = true;
575   LLVMStyle.BinPackParameters = true;
576   LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
577   LLVMStyle.BreakBeforeTernaryOperators = true;
578   LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
579   LLVMStyle.BraceWrapping = {false, false, false, false, false, false,
580                              false, false, false, false, false, false,
581                              true,  true,  true};
582   LLVMStyle.BreakAfterJavaFieldAnnotations = false;
583   LLVMStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeColon;
584   LLVMStyle.BreakBeforeInheritanceComma = false;
585   LLVMStyle.BreakStringLiterals = true;
586   LLVMStyle.ColumnLimit = 80;
587   LLVMStyle.CommentPragmas = "^ IWYU pragma:";
588   LLVMStyle.CompactNamespaces = false;
589   LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
590   LLVMStyle.ConstructorInitializerIndentWidth = 4;
591   LLVMStyle.ContinuationIndentWidth = 4;
592   LLVMStyle.Cpp11BracedListStyle = true;
593   LLVMStyle.DerivePointerAlignment = false;
594   LLVMStyle.ExperimentalAutoDetectBinPacking = false;
595   LLVMStyle.FixNamespaceComments = true;
596   LLVMStyle.ForEachMacros.push_back("foreach");
597   LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
598   LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
599   LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
600                                  {"^(<|\"(gtest|gmock|isl|json)/)", 3},
601                                  {".*", 1}};
602   LLVMStyle.IncludeIsMainRegex = "(Test)?$";
603   LLVMStyle.IndentCaseLabels = false;
604   LLVMStyle.IndentPPDirectives = FormatStyle::PPDIS_None;
605   LLVMStyle.IndentWrappedFunctionNames = false;
606   LLVMStyle.IndentWidth = 2;
607   LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
608   LLVMStyle.JavaScriptWrapImports = true;
609   LLVMStyle.TabWidth = 8;
610   LLVMStyle.MaxEmptyLinesToKeep = 1;
611   LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
612   LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
613   LLVMStyle.ObjCBlockIndentWidth = 2;
614   LLVMStyle.ObjCSpaceAfterProperty = false;
615   LLVMStyle.ObjCSpaceBeforeProtocolList = true;
616   LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
617   LLVMStyle.SpacesBeforeTrailingComments = 1;
618   LLVMStyle.Standard = FormatStyle::LS_Cpp11;
619   LLVMStyle.UseTab = FormatStyle::UT_Never;
620   LLVMStyle.ReflowComments = true;
621   LLVMStyle.SpacesInParentheses = false;
622   LLVMStyle.SpacesInSquareBrackets = false;
623   LLVMStyle.SpaceInEmptyParentheses = false;
624   LLVMStyle.SpacesInContainerLiterals = true;
625   LLVMStyle.SpacesInCStyleCastParentheses = false;
626   LLVMStyle.SpaceAfterCStyleCast = false;
627   LLVMStyle.SpaceAfterTemplateKeyword = true;
628   LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
629   LLVMStyle.SpaceBeforeAssignmentOperators = true;
630   LLVMStyle.SpacesInAngles = false;
631 
632   LLVMStyle.PenaltyBreakAssignment = prec::Assignment;
633   LLVMStyle.PenaltyBreakComment = 300;
634   LLVMStyle.PenaltyBreakFirstLessLess = 120;
635   LLVMStyle.PenaltyBreakString = 1000;
636   LLVMStyle.PenaltyExcessCharacter = 1000000;
637   LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
638   LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
639 
640   LLVMStyle.DisableFormat = false;
641   LLVMStyle.SortIncludes = true;
642   LLVMStyle.SortUsingDeclarations = true;
643 
644   return LLVMStyle;
645 }
646 
647 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
648   if (Language == FormatStyle::LK_TextProto) {
649     FormatStyle GoogleStyle = getGoogleStyle(FormatStyle::LK_Proto);
650     GoogleStyle.Language = FormatStyle::LK_TextProto;
651     return GoogleStyle;
652   }
653 
654   FormatStyle GoogleStyle = getLLVMStyle();
655   GoogleStyle.Language = Language;
656 
657   GoogleStyle.AccessModifierOffset = -1;
658   GoogleStyle.AlignEscapedNewlines = FormatStyle::ENAS_Left;
659   GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
660   GoogleStyle.AllowShortLoopsOnASingleLine = true;
661   GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
662   GoogleStyle.AlwaysBreakTemplateDeclarations = true;
663   GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
664   GoogleStyle.DerivePointerAlignment = true;
665   GoogleStyle.IncludeCategories = {{"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
666   GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$";
667   GoogleStyle.IndentCaseLabels = true;
668   GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
669   GoogleStyle.ObjCSpaceAfterProperty = false;
670   GoogleStyle.ObjCSpaceBeforeProtocolList = false;
671   GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
672   GoogleStyle.SpacesBeforeTrailingComments = 2;
673   GoogleStyle.Standard = FormatStyle::LS_Auto;
674 
675   GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
676   GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
677 
678   if (Language == FormatStyle::LK_Java) {
679     GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
680     GoogleStyle.AlignOperands = false;
681     GoogleStyle.AlignTrailingComments = false;
682     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
683     GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
684     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
685     GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
686     GoogleStyle.ColumnLimit = 100;
687     GoogleStyle.SpaceAfterCStyleCast = true;
688     GoogleStyle.SpacesBeforeTrailingComments = 1;
689   } else if (Language == FormatStyle::LK_JavaScript) {
690     GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak;
691     GoogleStyle.AlignOperands = false;
692     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
693     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
694     GoogleStyle.BreakBeforeTernaryOperators = false;
695     // taze:, triple slash directives (`/// <...`), @tag followed by { for a lot
696     // of JSDoc tags, and @see, which is commonly followed by overlong URLs.
697     GoogleStyle.CommentPragmas =
698         "(taze:|^/[ \t]*<|(@[A-Za-z_0-9-]+[ \\t]*{)|@see)";
699     GoogleStyle.MaxEmptyLinesToKeep = 3;
700     GoogleStyle.NamespaceIndentation = FormatStyle::NI_All;
701     GoogleStyle.SpacesInContainerLiterals = false;
702     GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single;
703     GoogleStyle.JavaScriptWrapImports = false;
704   } else if (Language == FormatStyle::LK_Proto) {
705     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
706     GoogleStyle.SpacesInContainerLiterals = false;
707   } else if (Language == FormatStyle::LK_ObjC) {
708     GoogleStyle.ColumnLimit = 100;
709   }
710 
711   return GoogleStyle;
712 }
713 
714 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
715   FormatStyle ChromiumStyle = getGoogleStyle(Language);
716   if (Language == FormatStyle::LK_Java) {
717     ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
718     ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
719     ChromiumStyle.ContinuationIndentWidth = 8;
720     ChromiumStyle.IndentWidth = 4;
721   } else if (Language == FormatStyle::LK_JavaScript) {
722     ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
723     ChromiumStyle.AllowShortLoopsOnASingleLine = false;
724   } else {
725     ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
726     ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
727     ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
728     ChromiumStyle.AllowShortLoopsOnASingleLine = false;
729     ChromiumStyle.BinPackParameters = false;
730     ChromiumStyle.DerivePointerAlignment = false;
731     if (Language == FormatStyle::LK_ObjC)
732       ChromiumStyle.ColumnLimit = 80;
733   }
734   return ChromiumStyle;
735 }
736 
737 FormatStyle getMozillaStyle() {
738   FormatStyle MozillaStyle = getLLVMStyle();
739   MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
740   MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
741   MozillaStyle.AlwaysBreakAfterReturnType =
742       FormatStyle::RTBS_TopLevel;
743   MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
744       FormatStyle::DRTBS_TopLevel;
745   MozillaStyle.AlwaysBreakTemplateDeclarations = true;
746   MozillaStyle.BinPackParameters = false;
747   MozillaStyle.BinPackArguments = false;
748   MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
749   MozillaStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
750   MozillaStyle.BreakBeforeInheritanceComma = true;
751   MozillaStyle.ConstructorInitializerIndentWidth = 2;
752   MozillaStyle.ContinuationIndentWidth = 2;
753   MozillaStyle.Cpp11BracedListStyle = false;
754   MozillaStyle.FixNamespaceComments = false;
755   MozillaStyle.IndentCaseLabels = true;
756   MozillaStyle.ObjCSpaceAfterProperty = true;
757   MozillaStyle.ObjCSpaceBeforeProtocolList = false;
758   MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
759   MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
760   MozillaStyle.SpaceAfterTemplateKeyword = false;
761   return MozillaStyle;
762 }
763 
764 FormatStyle getWebKitStyle() {
765   FormatStyle Style = getLLVMStyle();
766   Style.AccessModifierOffset = -4;
767   Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
768   Style.AlignOperands = false;
769   Style.AlignTrailingComments = false;
770   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
771   Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
772   Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
773   Style.Cpp11BracedListStyle = false;
774   Style.ColumnLimit = 0;
775   Style.FixNamespaceComments = false;
776   Style.IndentWidth = 4;
777   Style.NamespaceIndentation = FormatStyle::NI_Inner;
778   Style.ObjCBlockIndentWidth = 4;
779   Style.ObjCSpaceAfterProperty = true;
780   Style.PointerAlignment = FormatStyle::PAS_Left;
781   return Style;
782 }
783 
784 FormatStyle getGNUStyle() {
785   FormatStyle Style = getLLVMStyle();
786   Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
787   Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
788   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
789   Style.BreakBeforeBraces = FormatStyle::BS_GNU;
790   Style.BreakBeforeTernaryOperators = true;
791   Style.Cpp11BracedListStyle = false;
792   Style.ColumnLimit = 79;
793   Style.FixNamespaceComments = false;
794   Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
795   Style.Standard = FormatStyle::LS_Cpp03;
796   return Style;
797 }
798 
799 FormatStyle getNoStyle() {
800   FormatStyle NoStyle = getLLVMStyle();
801   NoStyle.DisableFormat = true;
802   NoStyle.SortIncludes = false;
803   NoStyle.SortUsingDeclarations = false;
804   return NoStyle;
805 }
806 
807 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
808                         FormatStyle *Style) {
809   if (Name.equals_lower("llvm")) {
810     *Style = getLLVMStyle();
811   } else if (Name.equals_lower("chromium")) {
812     *Style = getChromiumStyle(Language);
813   } else if (Name.equals_lower("mozilla")) {
814     *Style = getMozillaStyle();
815   } else if (Name.equals_lower("google")) {
816     *Style = getGoogleStyle(Language);
817   } else if (Name.equals_lower("webkit")) {
818     *Style = getWebKitStyle();
819   } else if (Name.equals_lower("gnu")) {
820     *Style = getGNUStyle();
821   } else if (Name.equals_lower("none")) {
822     *Style = getNoStyle();
823   } else {
824     return false;
825   }
826 
827   Style->Language = Language;
828   return true;
829 }
830 
831 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
832   assert(Style);
833   FormatStyle::LanguageKind Language = Style->Language;
834   assert(Language != FormatStyle::LK_None);
835   if (Text.trim().empty())
836     return make_error_code(ParseError::Error);
837 
838   std::vector<FormatStyle> Styles;
839   llvm::yaml::Input Input(Text);
840   // DocumentListTraits<vector<FormatStyle>> uses the context to get default
841   // values for the fields, keys for which are missing from the configuration.
842   // Mapping also uses the context to get the language to find the correct
843   // base style.
844   Input.setContext(Style);
845   Input >> Styles;
846   if (Input.error())
847     return Input.error();
848 
849   for (unsigned i = 0; i < Styles.size(); ++i) {
850     // Ensures that only the first configuration can skip the Language option.
851     if (Styles[i].Language == FormatStyle::LK_None && i != 0)
852       return make_error_code(ParseError::Error);
853     // Ensure that each language is configured at most once.
854     for (unsigned j = 0; j < i; ++j) {
855       if (Styles[i].Language == Styles[j].Language) {
856         DEBUG(llvm::dbgs()
857               << "Duplicate languages in the config file on positions " << j
858               << " and " << i << "\n");
859         return make_error_code(ParseError::Error);
860       }
861     }
862   }
863   // Look for a suitable configuration starting from the end, so we can
864   // find the configuration for the specific language first, and the default
865   // configuration (which can only be at slot 0) after it.
866   for (int i = Styles.size() - 1; i >= 0; --i) {
867     if (Styles[i].Language == Language ||
868         Styles[i].Language == FormatStyle::LK_None) {
869       *Style = Styles[i];
870       Style->Language = Language;
871       return make_error_code(ParseError::Success);
872     }
873   }
874   return make_error_code(ParseError::Unsuitable);
875 }
876 
877 std::string configurationAsText(const FormatStyle &Style) {
878   std::string Text;
879   llvm::raw_string_ostream Stream(Text);
880   llvm::yaml::Output Output(Stream);
881   // We use the same mapping method for input and output, so we need a non-const
882   // reference here.
883   FormatStyle NonConstStyle = expandPresets(Style);
884   Output << NonConstStyle;
885   return Stream.str();
886 }
887 
888 namespace {
889 
890 class JavaScriptRequoter : public TokenAnalyzer {
891 public:
892   JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
893       : TokenAnalyzer(Env, Style) {}
894 
895   tooling::Replacements
896   analyze(TokenAnnotator &Annotator,
897           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
898           FormatTokenLexer &Tokens) override {
899     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
900                                           AnnotatedLines.end());
901     tooling::Replacements Result;
902     requoteJSStringLiteral(AnnotatedLines, Result);
903     return Result;
904   }
905 
906 private:
907   // Replaces double/single-quoted string literal as appropriate, re-escaping
908   // the contents in the process.
909   void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
910                               tooling::Replacements &Result) {
911     for (AnnotatedLine *Line : Lines) {
912       requoteJSStringLiteral(Line->Children, Result);
913       if (!Line->Affected)
914         continue;
915       for (FormatToken *FormatTok = Line->First; FormatTok;
916            FormatTok = FormatTok->Next) {
917         StringRef Input = FormatTok->TokenText;
918         if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
919             // NB: testing for not starting with a double quote to avoid
920             // breaking `template strings`.
921             (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
922              !Input.startswith("\"")) ||
923             (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
924              !Input.startswith("\'")))
925           continue;
926 
927         // Change start and end quote.
928         bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
929         SourceLocation Start = FormatTok->Tok.getLocation();
930         auto Replace = [&](SourceLocation Start, unsigned Length,
931                            StringRef ReplacementText) {
932           auto Err = Result.add(tooling::Replacement(
933               Env.getSourceManager(), Start, Length, ReplacementText));
934           // FIXME: handle error. For now, print error message and skip the
935           // replacement for release version.
936           if (Err) {
937             llvm::errs() << llvm::toString(std::move(Err)) << "\n";
938             assert(false);
939           }
940         };
941         Replace(Start, 1, IsSingle ? "'" : "\"");
942         Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
943                 IsSingle ? "'" : "\"");
944 
945         // Escape internal quotes.
946         bool Escaped = false;
947         for (size_t i = 1; i < Input.size() - 1; i++) {
948           switch (Input[i]) {
949           case '\\':
950             if (!Escaped && i + 1 < Input.size() &&
951                 ((IsSingle && Input[i + 1] == '"') ||
952                  (!IsSingle && Input[i + 1] == '\''))) {
953               // Remove this \, it's escaping a " or ' that no longer needs
954               // escaping
955               Replace(Start.getLocWithOffset(i), 1, "");
956               continue;
957             }
958             Escaped = !Escaped;
959             break;
960           case '\"':
961           case '\'':
962             if (!Escaped && IsSingle == (Input[i] == '\'')) {
963               // Escape the quote.
964               Replace(Start.getLocWithOffset(i), 0, "\\");
965             }
966             Escaped = false;
967             break;
968           default:
969             Escaped = false;
970             break;
971           }
972         }
973       }
974     }
975   }
976 };
977 
978 class Formatter : public TokenAnalyzer {
979 public:
980   Formatter(const Environment &Env, const FormatStyle &Style,
981             FormattingAttemptStatus *Status)
982       : TokenAnalyzer(Env, Style), Status(Status) {}
983 
984   tooling::Replacements
985   analyze(TokenAnnotator &Annotator,
986           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
987           FormatTokenLexer &Tokens) override {
988     tooling::Replacements Result;
989     deriveLocalStyle(AnnotatedLines);
990     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
991                                           AnnotatedLines.end());
992     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
993       Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
994     }
995     Annotator.setCommentLineLevels(AnnotatedLines);
996 
997     WhitespaceManager Whitespaces(
998         Env.getSourceManager(), Style,
999         inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
1000     ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
1001                                   Env.getSourceManager(), Whitespaces, Encoding,
1002                                   BinPackInconclusiveFunctions);
1003     UnwrappedLineFormatter(&Indenter, &Whitespaces, Style, Tokens.getKeywords(),
1004                            Env.getSourceManager(), Status)
1005         .format(AnnotatedLines);
1006     for (const auto &R : Whitespaces.generateReplacements())
1007       if (Result.add(R))
1008         return Result;
1009     return Result;
1010   }
1011 
1012 private:
1013 
1014   static bool inputUsesCRLF(StringRef Text) {
1015     return Text.count('\r') * 2 > Text.count('\n');
1016   }
1017 
1018   bool
1019   hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1020     for (const AnnotatedLine *Line : Lines) {
1021       if (hasCpp03IncompatibleFormat(Line->Children))
1022         return true;
1023       for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
1024         if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
1025           if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
1026             return true;
1027           if (Tok->is(TT_TemplateCloser) &&
1028               Tok->Previous->is(TT_TemplateCloser))
1029             return true;
1030         }
1031       }
1032     }
1033     return false;
1034   }
1035 
1036   int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1037     int AlignmentDiff = 0;
1038     for (const AnnotatedLine *Line : Lines) {
1039       AlignmentDiff += countVariableAlignments(Line->Children);
1040       for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
1041         if (!Tok->is(TT_PointerOrReference))
1042           continue;
1043         bool SpaceBefore =
1044             Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
1045         bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
1046                           Tok->Next->WhitespaceRange.getEnd();
1047         if (SpaceBefore && !SpaceAfter)
1048           ++AlignmentDiff;
1049         if (!SpaceBefore && SpaceAfter)
1050           --AlignmentDiff;
1051       }
1052     }
1053     return AlignmentDiff;
1054   }
1055 
1056   void
1057   deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1058     bool HasBinPackedFunction = false;
1059     bool HasOnePerLineFunction = false;
1060     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1061       if (!AnnotatedLines[i]->First->Next)
1062         continue;
1063       FormatToken *Tok = AnnotatedLines[i]->First->Next;
1064       while (Tok->Next) {
1065         if (Tok->PackingKind == PPK_BinPacked)
1066           HasBinPackedFunction = true;
1067         if (Tok->PackingKind == PPK_OnePerLine)
1068           HasOnePerLineFunction = true;
1069 
1070         Tok = Tok->Next;
1071       }
1072     }
1073     if (Style.DerivePointerAlignment)
1074       Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
1075                                    ? FormatStyle::PAS_Left
1076                                    : FormatStyle::PAS_Right;
1077     if (Style.Standard == FormatStyle::LS_Auto)
1078       Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
1079                            ? FormatStyle::LS_Cpp11
1080                            : FormatStyle::LS_Cpp03;
1081     BinPackInconclusiveFunctions =
1082         HasBinPackedFunction || !HasOnePerLineFunction;
1083   }
1084 
1085   bool BinPackInconclusiveFunctions;
1086   FormattingAttemptStatus *Status;
1087 };
1088 
1089 // This class clean up the erroneous/redundant code around the given ranges in
1090 // file.
1091 class Cleaner : public TokenAnalyzer {
1092 public:
1093   Cleaner(const Environment &Env, const FormatStyle &Style)
1094       : TokenAnalyzer(Env, Style),
1095         DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1096 
1097   // FIXME: eliminate unused parameters.
1098   tooling::Replacements
1099   analyze(TokenAnnotator &Annotator,
1100           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1101           FormatTokenLexer &Tokens) override {
1102     // FIXME: in the current implementation the granularity of affected range
1103     // is an annotated line. However, this is not sufficient. Furthermore,
1104     // redundant code introduced by replacements does not necessarily
1105     // intercept with ranges of replacements that result in the redundancy.
1106     // To determine if some redundant code is actually introduced by
1107     // replacements(e.g. deletions), we need to come up with a more
1108     // sophisticated way of computing affected ranges.
1109     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1110                                           AnnotatedLines.end());
1111 
1112     checkEmptyNamespace(AnnotatedLines);
1113 
1114     for (auto &Line : AnnotatedLines) {
1115       if (Line->Affected) {
1116         cleanupRight(Line->First, tok::comma, tok::comma);
1117         cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1118         cleanupRight(Line->First, tok::l_paren, tok::comma);
1119         cleanupLeft(Line->First, tok::comma, tok::r_paren);
1120         cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1121         cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1122         cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1123       }
1124     }
1125 
1126     return generateFixes();
1127   }
1128 
1129 private:
1130   bool containsOnlyComments(const AnnotatedLine &Line) {
1131     for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1132       if (Tok->isNot(tok::comment))
1133         return false;
1134     }
1135     return true;
1136   }
1137 
1138   // Iterate through all lines and remove any empty (nested) namespaces.
1139   void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1140     std::set<unsigned> DeletedLines;
1141     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1142       auto &Line = *AnnotatedLines[i];
1143       if (Line.startsWith(tok::kw_namespace) ||
1144           Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1145         checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1146       }
1147     }
1148 
1149     for (auto Line : DeletedLines) {
1150       FormatToken *Tok = AnnotatedLines[Line]->First;
1151       while (Tok) {
1152         deleteToken(Tok);
1153         Tok = Tok->Next;
1154       }
1155     }
1156   }
1157 
1158   // The function checks if the namespace, which starts from \p CurrentLine, and
1159   // its nested namespaces are empty and delete them if they are empty. It also
1160   // sets \p NewLine to the last line checked.
1161   // Returns true if the current namespace is empty.
1162   bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1163                            unsigned CurrentLine, unsigned &NewLine,
1164                            std::set<unsigned> &DeletedLines) {
1165     unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1166     if (Style.BraceWrapping.AfterNamespace) {
1167       // If the left brace is in a new line, we should consume it first so that
1168       // it does not make the namespace non-empty.
1169       // FIXME: error handling if there is no left brace.
1170       if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1171         NewLine = CurrentLine;
1172         return false;
1173       }
1174     } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1175       return false;
1176     }
1177     while (++CurrentLine < End) {
1178       if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1179         break;
1180 
1181       if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1182           AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1183                                                   tok::kw_namespace)) {
1184         if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1185                                  DeletedLines))
1186           return false;
1187         CurrentLine = NewLine;
1188         continue;
1189       }
1190 
1191       if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1192         continue;
1193 
1194       // If there is anything other than comments or nested namespaces in the
1195       // current namespace, the namespace cannot be empty.
1196       NewLine = CurrentLine;
1197       return false;
1198     }
1199 
1200     NewLine = CurrentLine;
1201     if (CurrentLine >= End)
1202       return false;
1203 
1204     // Check if the empty namespace is actually affected by changed ranges.
1205     if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1206             AnnotatedLines[InitLine]->First->Tok.getLocation(),
1207             AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1208       return false;
1209 
1210     for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1211       DeletedLines.insert(i);
1212     }
1213 
1214     return true;
1215   }
1216 
1217   // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1218   // of the token in the pair if the left token has \p LK token kind and the
1219   // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1220   // is deleted on match; otherwise, the right token is deleted.
1221   template <typename LeftKind, typename RightKind>
1222   void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1223                    bool DeleteLeft) {
1224     auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1225       for (auto *Res = Tok.Next; Res; Res = Res->Next)
1226         if (!Res->is(tok::comment) &&
1227             DeletedTokens.find(Res) == DeletedTokens.end())
1228           return Res;
1229       return nullptr;
1230     };
1231     for (auto *Left = Start; Left;) {
1232       auto *Right = NextNotDeleted(*Left);
1233       if (!Right)
1234         break;
1235       if (Left->is(LK) && Right->is(RK)) {
1236         deleteToken(DeleteLeft ? Left : Right);
1237         for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1238           deleteToken(Tok);
1239         // If the right token is deleted, we should keep the left token
1240         // unchanged and pair it with the new right token.
1241         if (!DeleteLeft)
1242           continue;
1243       }
1244       Left = Right;
1245     }
1246   }
1247 
1248   template <typename LeftKind, typename RightKind>
1249   void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1250     cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1251   }
1252 
1253   template <typename LeftKind, typename RightKind>
1254   void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1255     cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1256   }
1257 
1258   // Delete the given token.
1259   inline void deleteToken(FormatToken *Tok) {
1260     if (Tok)
1261       DeletedTokens.insert(Tok);
1262   }
1263 
1264   tooling::Replacements generateFixes() {
1265     tooling::Replacements Fixes;
1266     std::vector<FormatToken *> Tokens;
1267     std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1268               std::back_inserter(Tokens));
1269 
1270     // Merge multiple continuous token deletions into one big deletion so that
1271     // the number of replacements can be reduced. This makes computing affected
1272     // ranges more efficient when we run reformat on the changed code.
1273     unsigned Idx = 0;
1274     while (Idx < Tokens.size()) {
1275       unsigned St = Idx, End = Idx;
1276       while ((End + 1) < Tokens.size() &&
1277              Tokens[End]->Next == Tokens[End + 1]) {
1278         End++;
1279       }
1280       auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1281                                               Tokens[End]->Tok.getEndLoc());
1282       auto Err =
1283           Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1284       // FIXME: better error handling. for now just print error message and skip
1285       // for the release version.
1286       if (Err) {
1287         llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1288         assert(false && "Fixes must not conflict!");
1289       }
1290       Idx = End + 1;
1291     }
1292 
1293     return Fixes;
1294   }
1295 
1296   // Class for less-than inequality comparason for the set `RedundantTokens`.
1297   // We store tokens in the order they appear in the translation unit so that
1298   // we do not need to sort them in `generateFixes()`.
1299   struct FormatTokenLess {
1300     FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1301 
1302     bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1303       return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1304                                           RHS->Tok.getLocation());
1305     }
1306     const SourceManager &SM;
1307   };
1308 
1309   // Tokens to be deleted.
1310   std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1311 };
1312 
1313 struct IncludeDirective {
1314   StringRef Filename;
1315   StringRef Text;
1316   unsigned Offset;
1317   int Category;
1318 };
1319 
1320 } // end anonymous namespace
1321 
1322 // Determines whether 'Ranges' intersects with ('Start', 'End').
1323 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1324                          unsigned End) {
1325   for (auto Range : Ranges) {
1326     if (Range.getOffset() < End &&
1327         Range.getOffset() + Range.getLength() > Start)
1328       return true;
1329   }
1330   return false;
1331 }
1332 
1333 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1334 // before sorting/deduplicating. Index is the index of the include under the
1335 // cursor in the original set of includes. If this include has duplicates, it is
1336 // the index of the first of the duplicates as the others are going to be
1337 // removed. OffsetToEOL describes the cursor's position relative to the end of
1338 // its current line.
1339 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1340 static std::pair<unsigned, unsigned>
1341 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1342                 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1343   unsigned CursorIndex = UINT_MAX;
1344   unsigned OffsetToEOL = 0;
1345   for (int i = 0, e = Includes.size(); i != e; ++i) {
1346     unsigned Start = Includes[Indices[i]].Offset;
1347     unsigned End = Start + Includes[Indices[i]].Text.size();
1348     if (!(Cursor >= Start && Cursor < End))
1349       continue;
1350     CursorIndex = Indices[i];
1351     OffsetToEOL = End - Cursor;
1352     // Put the cursor on the only remaining #include among the duplicate
1353     // #includes.
1354     while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1355       CursorIndex = i;
1356     break;
1357   }
1358   return std::make_pair(CursorIndex, OffsetToEOL);
1359 }
1360 
1361 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1362 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1363 // source order.
1364 // #include directives with the same text will be deduplicated, and only the
1365 // first #include in the duplicate #includes remains. If the `Cursor` is
1366 // provided and put on a deleted #include, it will be moved to the remaining
1367 // #include in the duplicate #includes.
1368 static void sortCppIncludes(const FormatStyle &Style,
1369                             const SmallVectorImpl<IncludeDirective> &Includes,
1370                             ArrayRef<tooling::Range> Ranges, StringRef FileName,
1371                             tooling::Replacements &Replaces, unsigned *Cursor) {
1372   unsigned IncludesBeginOffset = Includes.front().Offset;
1373   unsigned IncludesEndOffset =
1374       Includes.back().Offset + Includes.back().Text.size();
1375   unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1376   if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1377     return;
1378   SmallVector<unsigned, 16> Indices;
1379   for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1380     Indices.push_back(i);
1381   std::stable_sort(
1382       Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1383         return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1384                std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1385       });
1386   // The index of the include on which the cursor will be put after
1387   // sorting/deduplicating.
1388   unsigned CursorIndex;
1389   // The offset from cursor to the end of line.
1390   unsigned CursorToEOLOffset;
1391   if (Cursor)
1392     std::tie(CursorIndex, CursorToEOLOffset) =
1393         FindCursorIndex(Includes, Indices, *Cursor);
1394 
1395   // Deduplicate #includes.
1396   Indices.erase(std::unique(Indices.begin(), Indices.end(),
1397                             [&](unsigned LHSI, unsigned RHSI) {
1398                               return Includes[LHSI].Text == Includes[RHSI].Text;
1399                             }),
1400                 Indices.end());
1401 
1402   // If the #includes are out of order, we generate a single replacement fixing
1403   // the entire block. Otherwise, no replacement is generated.
1404   if (Indices.size() == Includes.size() &&
1405       std::is_sorted(Indices.begin(), Indices.end()))
1406     return;
1407 
1408   std::string result;
1409   for (unsigned Index : Indices) {
1410     if (!result.empty())
1411       result += "\n";
1412     result += Includes[Index].Text;
1413     if (Cursor && CursorIndex == Index)
1414       *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1415   }
1416 
1417   auto Err = Replaces.add(tooling::Replacement(
1418       FileName, Includes.front().Offset, IncludesBlockSize, result));
1419   // FIXME: better error handling. For now, just skip the replacement for the
1420   // release version.
1421   if (Err) {
1422     llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1423     assert(false);
1424   }
1425 }
1426 
1427 namespace {
1428 
1429 // This class manages priorities of #include categories and calculates
1430 // priorities for headers.
1431 class IncludeCategoryManager {
1432 public:
1433   IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1434       : Style(Style), FileName(FileName) {
1435     FileStem = llvm::sys::path::stem(FileName);
1436     for (const auto &Category : Style.IncludeCategories)
1437       CategoryRegexs.emplace_back(Category.Regex, llvm::Regex::IgnoreCase);
1438     IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1439                  FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1440                  FileName.endswith(".cxx") || FileName.endswith(".m") ||
1441                  FileName.endswith(".mm");
1442   }
1443 
1444   // Returns the priority of the category which \p IncludeName belongs to.
1445   // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1446   // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1447   int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1448     int Ret = INT_MAX;
1449     for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1450       if (CategoryRegexs[i].match(IncludeName)) {
1451         Ret = Style.IncludeCategories[i].Priority;
1452         break;
1453       }
1454     if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1455       Ret = 0;
1456     return Ret;
1457   }
1458 
1459 private:
1460   bool isMainHeader(StringRef IncludeName) const {
1461     if (!IncludeName.startswith("\""))
1462       return false;
1463     StringRef HeaderStem =
1464         llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1465     if (FileStem.startswith(HeaderStem) ||
1466         FileStem.startswith_lower(HeaderStem)) {
1467       llvm::Regex MainIncludeRegex(
1468           (HeaderStem + Style.IncludeIsMainRegex).str(),
1469           llvm::Regex::IgnoreCase);
1470       if (MainIncludeRegex.match(FileStem))
1471         return true;
1472     }
1473     return false;
1474   }
1475 
1476   const FormatStyle &Style;
1477   bool IsMainFile;
1478   StringRef FileName;
1479   StringRef FileStem;
1480   SmallVector<llvm::Regex, 4> CategoryRegexs;
1481 };
1482 
1483 const char IncludeRegexPattern[] =
1484     R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1485 
1486 } // anonymous namespace
1487 
1488 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1489                                       ArrayRef<tooling::Range> Ranges,
1490                                       StringRef FileName,
1491                                       tooling::Replacements &Replaces,
1492                                       unsigned *Cursor) {
1493   unsigned Prev = 0;
1494   unsigned SearchFrom = 0;
1495   llvm::Regex IncludeRegex(IncludeRegexPattern);
1496   SmallVector<StringRef, 4> Matches;
1497   SmallVector<IncludeDirective, 16> IncludesInBlock;
1498 
1499   // In compiled files, consider the first #include to be the main #include of
1500   // the file if it is not a system #include. This ensures that the header
1501   // doesn't have hidden dependencies
1502   // (http://llvm.org/docs/CodingStandards.html#include-style).
1503   //
1504   // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1505   // cases where the first #include is unlikely to be the main header.
1506   IncludeCategoryManager Categories(Style, FileName);
1507   bool FirstIncludeBlock = true;
1508   bool MainIncludeFound = false;
1509   bool FormattingOff = false;
1510 
1511   for (;;) {
1512     auto Pos = Code.find('\n', SearchFrom);
1513     StringRef Line =
1514         Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1515 
1516     StringRef Trimmed = Line.trim();
1517     if (Trimmed == "// clang-format off")
1518       FormattingOff = true;
1519     else if (Trimmed == "// clang-format on")
1520       FormattingOff = false;
1521 
1522     if (!FormattingOff && !Line.endswith("\\")) {
1523       if (IncludeRegex.match(Line, &Matches)) {
1524         StringRef IncludeName = Matches[2];
1525         int Category = Categories.getIncludePriority(
1526             IncludeName,
1527             /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1528         if (Category == 0)
1529           MainIncludeFound = true;
1530         IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1531       } else if (!IncludesInBlock.empty()) {
1532         sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1533                         Cursor);
1534         IncludesInBlock.clear();
1535         FirstIncludeBlock = false;
1536       }
1537       Prev = Pos + 1;
1538     }
1539     if (Pos == StringRef::npos || Pos + 1 == Code.size())
1540       break;
1541     SearchFrom = Pos + 1;
1542   }
1543   if (!IncludesInBlock.empty())
1544     sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1545   return Replaces;
1546 }
1547 
1548 bool isMpegTS(StringRef Code) {
1549   // MPEG transport streams use the ".ts" file extension. clang-format should
1550   // not attempt to format those. MPEG TS' frame format starts with 0x47 every
1551   // 189 bytes - detect that and return.
1552   return Code.size() > 188 && Code[0] == 0x47 && Code[188] == 0x47;
1553 }
1554 
1555 bool isLikelyXml(StringRef Code) {
1556   return Code.ltrim().startswith("<");
1557 }
1558 
1559 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1560                                    ArrayRef<tooling::Range> Ranges,
1561                                    StringRef FileName, unsigned *Cursor) {
1562   tooling::Replacements Replaces;
1563   if (!Style.SortIncludes)
1564     return Replaces;
1565   if (isLikelyXml(Code))
1566     return Replaces;
1567   if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript &&
1568       isMpegTS(Code))
1569     return Replaces;
1570   if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1571     return sortJavaScriptImports(Style, Code, Ranges, FileName);
1572   sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1573   return Replaces;
1574 }
1575 
1576 template <typename T>
1577 static llvm::Expected<tooling::Replacements>
1578 processReplacements(T ProcessFunc, StringRef Code,
1579                     const tooling::Replacements &Replaces,
1580                     const FormatStyle &Style) {
1581   if (Replaces.empty())
1582     return tooling::Replacements();
1583 
1584   auto NewCode = applyAllReplacements(Code, Replaces);
1585   if (!NewCode)
1586     return NewCode.takeError();
1587   std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1588   StringRef FileName = Replaces.begin()->getFilePath();
1589 
1590   tooling::Replacements FormatReplaces =
1591       ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1592 
1593   return Replaces.merge(FormatReplaces);
1594 }
1595 
1596 llvm::Expected<tooling::Replacements>
1597 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1598                    const FormatStyle &Style) {
1599   // We need to use lambda function here since there are two versions of
1600   // `sortIncludes`.
1601   auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1602                          std::vector<tooling::Range> Ranges,
1603                          StringRef FileName) -> tooling::Replacements {
1604     return sortIncludes(Style, Code, Ranges, FileName);
1605   };
1606   auto SortedReplaces =
1607       processReplacements(SortIncludes, Code, Replaces, Style);
1608   if (!SortedReplaces)
1609     return SortedReplaces.takeError();
1610 
1611   // We need to use lambda function here since there are two versions of
1612   // `reformat`.
1613   auto Reformat = [](const FormatStyle &Style, StringRef Code,
1614                      std::vector<tooling::Range> Ranges,
1615                      StringRef FileName) -> tooling::Replacements {
1616     return reformat(Style, Code, Ranges, FileName);
1617   };
1618   return processReplacements(Reformat, Code, *SortedReplaces, Style);
1619 }
1620 
1621 namespace {
1622 
1623 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1624   return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1625          llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1626 }
1627 
1628 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1629   return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1630 }
1631 
1632 // Returns the offset after skipping a sequence of tokens, matched by \p
1633 // GetOffsetAfterSequence, from the start of the code.
1634 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1635 // tokens and returns an offset after the sequence.
1636 unsigned getOffsetAfterTokenSequence(
1637     StringRef FileName, StringRef Code, const FormatStyle &Style,
1638     llvm::function_ref<unsigned(const SourceManager &, Lexer &, Token &)>
1639         GetOffsetAfterSequence) {
1640   std::unique_ptr<Environment> Env =
1641       Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1642   const SourceManager &SourceMgr = Env->getSourceManager();
1643   Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1644             getFormattingLangOpts(Style));
1645   Token Tok;
1646   // Get the first token.
1647   Lex.LexFromRawLexer(Tok);
1648   return GetOffsetAfterSequence(SourceMgr, Lex, Tok);
1649 }
1650 
1651 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1652 // \p Tok will be the token after this directive; otherwise, it can be any token
1653 // after the given \p Tok (including \p Tok).
1654 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1655   bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1656                  Tok.is(tok::raw_identifier) &&
1657                  Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1658                  Tok.is(tok::raw_identifier);
1659   if (Matched)
1660     Lex.LexFromRawLexer(Tok);
1661   return Matched;
1662 }
1663 
1664 void skipComments(Lexer &Lex, Token &Tok) {
1665   while (Tok.is(tok::comment))
1666     if (Lex.LexFromRawLexer(Tok))
1667       return;
1668 }
1669 
1670 // Returns the offset after header guard directives and any comments
1671 // before/after header guards. If no header guard presents in the code, this
1672 // will returns the offset after skipping all comments from the start of the
1673 // code.
1674 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1675                                                StringRef Code,
1676                                                const FormatStyle &Style) {
1677   return getOffsetAfterTokenSequence(
1678       FileName, Code, Style,
1679       [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1680         skipComments(Lex, Tok);
1681         unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1682         if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1683           skipComments(Lex, Tok);
1684           if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1685             return SM.getFileOffset(Tok.getLocation());
1686         }
1687         return InitialOffset;
1688       });
1689 }
1690 
1691 // Check if a sequence of tokens is like
1692 //    "#include ("header.h" | <header.h>)".
1693 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1694 // any token after the given \p Tok (including \p Tok).
1695 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1696   auto Matched = [&]() {
1697     Lex.LexFromRawLexer(Tok);
1698     return true;
1699   };
1700   if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1701       Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1702     if (Lex.LexFromRawLexer(Tok))
1703       return false;
1704     if (Tok.is(tok::string_literal))
1705       return Matched();
1706     if (Tok.is(tok::less)) {
1707       while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1708       }
1709       if (Tok.is(tok::greater))
1710         return Matched();
1711     }
1712   }
1713   return false;
1714 }
1715 
1716 // Returns the offset of the last #include directive after which a new
1717 // #include can be inserted. This ignores #include's after the #include block(s)
1718 // in the beginning of a file to avoid inserting headers into code sections
1719 // where new #include's should not be added by default.
1720 // These code sections include:
1721 //      - raw string literals (containing #include).
1722 //      - #if blocks.
1723 //      - Special #include's among declarations (e.g. functions).
1724 //
1725 // If no #include after which a new #include can be inserted, this returns the
1726 // offset after skipping all comments from the start of the code.
1727 // Inserting after an #include is not allowed if it comes after code that is not
1728 // #include (e.g. pre-processing directive that is not #include, declarations).
1729 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1730                                      const FormatStyle &Style) {
1731   return getOffsetAfterTokenSequence(
1732       FileName, Code, Style,
1733       [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1734         skipComments(Lex, Tok);
1735         unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1736         while (checkAndConsumeInclusiveDirective(Lex, Tok))
1737           MaxOffset = SM.getFileOffset(Tok.getLocation());
1738         return MaxOffset;
1739       });
1740 }
1741 
1742 bool isDeletedHeader(llvm::StringRef HeaderName,
1743                      const std::set<llvm::StringRef> &HeadersToDelete) {
1744   return HeadersToDelete.count(HeaderName) ||
1745          HeadersToDelete.count(HeaderName.trim("\"<>"));
1746 }
1747 
1748 // FIXME: insert empty lines between newly created blocks.
1749 tooling::Replacements
1750 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1751                         const FormatStyle &Style) {
1752   if (!Style.isCpp())
1753     return Replaces;
1754 
1755   tooling::Replacements HeaderInsertions;
1756   std::set<llvm::StringRef> HeadersToDelete;
1757   tooling::Replacements Result;
1758   for (const auto &R : Replaces) {
1759     if (isHeaderInsertion(R)) {
1760       // Replacements from \p Replaces must be conflict-free already, so we can
1761       // simply consume the error.
1762       llvm::consumeError(HeaderInsertions.add(R));
1763     } else if (isHeaderDeletion(R)) {
1764       HeadersToDelete.insert(R.getReplacementText());
1765     } else if (R.getOffset() == UINT_MAX) {
1766       llvm::errs() << "Insertions other than header #include insertion are "
1767                       "not supported! "
1768                    << R.getReplacementText() << "\n";
1769     } else {
1770       llvm::consumeError(Result.add(R));
1771     }
1772   }
1773   if (HeaderInsertions.empty() && HeadersToDelete.empty())
1774     return Replaces;
1775 
1776   llvm::Regex IncludeRegex(IncludeRegexPattern);
1777   llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1778   SmallVector<StringRef, 4> Matches;
1779 
1780   StringRef FileName = Replaces.begin()->getFilePath();
1781   IncludeCategoryManager Categories(Style, FileName);
1782 
1783   // Record the offset of the end of the last include in each category.
1784   std::map<int, int> CategoryEndOffsets;
1785   // All possible priorities.
1786   // Add 0 for main header and INT_MAX for headers that are not in any category.
1787   std::set<int> Priorities = {0, INT_MAX};
1788   for (const auto &Category : Style.IncludeCategories)
1789     Priorities.insert(Category.Priority);
1790   int FirstIncludeOffset = -1;
1791   // All new headers should be inserted after this offset.
1792   unsigned MinInsertOffset =
1793       getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1794   StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1795   // Max insertion offset in the original code.
1796   unsigned MaxInsertOffset =
1797       MinInsertOffset +
1798       getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1799   SmallVector<StringRef, 32> Lines;
1800   TrimmedCode.split(Lines, '\n');
1801   unsigned Offset = MinInsertOffset;
1802   unsigned NextLineOffset;
1803   std::set<StringRef> ExistingIncludes;
1804   for (auto Line : Lines) {
1805     NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1806     if (IncludeRegex.match(Line, &Matches)) {
1807       // The header name with quotes or angle brackets.
1808       StringRef IncludeName = Matches[2];
1809       ExistingIncludes.insert(IncludeName);
1810       // Only record the offset of current #include if we can insert after it.
1811       if (Offset <= MaxInsertOffset) {
1812         int Category = Categories.getIncludePriority(
1813             IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1814         CategoryEndOffsets[Category] = NextLineOffset;
1815         if (FirstIncludeOffset < 0)
1816           FirstIncludeOffset = Offset;
1817       }
1818       if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1819         // If this is the last line without trailing newline, we need to make
1820         // sure we don't delete across the file boundary.
1821         unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1822         llvm::Error Err =
1823             Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1824         if (Err) {
1825           // Ignore the deletion on conflict.
1826           llvm::errs() << "Failed to add header deletion replacement for "
1827                        << IncludeName << ": " << llvm::toString(std::move(Err))
1828                        << "\n";
1829         }
1830       }
1831     }
1832     Offset = NextLineOffset;
1833   }
1834 
1835   // Populate CategoryEndOfssets:
1836   // - Ensure that CategoryEndOffset[Highest] is always populated.
1837   // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1838   //   is set, up to CategoryEndOffset[Highest].
1839   auto Highest = Priorities.begin();
1840   if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1841     if (FirstIncludeOffset >= 0)
1842       CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1843     else
1844       CategoryEndOffsets[*Highest] = MinInsertOffset;
1845   }
1846   // By this point, CategoryEndOffset[Highest] is always set appropriately:
1847   //  - to an appropriate location before/after existing #includes, or
1848   //  - to right after the header guard, or
1849   //  - to the beginning of the file.
1850   for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1851     if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1852       CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1853 
1854   bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1855   for (const auto &R : HeaderInsertions) {
1856     auto IncludeDirective = R.getReplacementText();
1857     bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1858     assert(Matched && "Header insertion replacement must have replacement text "
1859                       "'#include ...'");
1860     (void)Matched;
1861     auto IncludeName = Matches[2];
1862     if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1863       DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1864                          << "\n");
1865       continue;
1866     }
1867     int Category =
1868         Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1869     Offset = CategoryEndOffsets[Category];
1870     std::string NewInclude = !IncludeDirective.endswith("\n")
1871                                  ? (IncludeDirective + "\n").str()
1872                                  : IncludeDirective.str();
1873     // When inserting headers at end of the code, also append '\n' to the code
1874     // if it does not end with '\n'.
1875     if (NeedNewLineAtEnd && Offset == Code.size()) {
1876       NewInclude = "\n" + NewInclude;
1877       NeedNewLineAtEnd = false;
1878     }
1879     auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1880     auto Err = Result.add(NewReplace);
1881     if (Err) {
1882       llvm::consumeError(std::move(Err));
1883       unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1884       NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1885       Result = Result.merge(tooling::Replacements(NewReplace));
1886     }
1887   }
1888   return Result;
1889 }
1890 
1891 } // anonymous namespace
1892 
1893 llvm::Expected<tooling::Replacements>
1894 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1895                           const FormatStyle &Style) {
1896   // We need to use lambda function here since there are two versions of
1897   // `cleanup`.
1898   auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1899                     std::vector<tooling::Range> Ranges,
1900                     StringRef FileName) -> tooling::Replacements {
1901     return cleanup(Style, Code, Ranges, FileName);
1902   };
1903   // Make header insertion replacements insert new headers into correct blocks.
1904   tooling::Replacements NewReplaces =
1905       fixCppIncludeInsertions(Code, Replaces, Style);
1906   return processReplacements(Cleanup, Code, NewReplaces, Style);
1907 }
1908 
1909 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1910                                ArrayRef<tooling::Range> Ranges,
1911                                StringRef FileName,
1912                                FormattingAttemptStatus *Status) {
1913   FormatStyle Expanded = expandPresets(Style);
1914   if (Expanded.DisableFormat)
1915     return tooling::Replacements();
1916   if (isLikelyXml(Code))
1917     return tooling::Replacements();
1918   if (Expanded.Language == FormatStyle::LK_JavaScript && isMpegTS(Code))
1919     return tooling::Replacements();
1920 
1921   typedef std::function<tooling::Replacements(const Environment &)>
1922       AnalyzerPass;
1923   SmallVector<AnalyzerPass, 4> Passes;
1924 
1925   if (Style.Language == FormatStyle::LK_Cpp) {
1926     if (Style.FixNamespaceComments)
1927       Passes.emplace_back([&](const Environment &Env) {
1928         return NamespaceEndCommentsFixer(Env, Expanded).process();
1929       });
1930 
1931     if (Style.SortUsingDeclarations)
1932       Passes.emplace_back([&](const Environment &Env) {
1933         return UsingDeclarationsSorter(Env, Expanded).process();
1934       });
1935   }
1936 
1937   if (Style.Language == FormatStyle::LK_JavaScript &&
1938       Style.JavaScriptQuotes != FormatStyle::JSQS_Leave)
1939     Passes.emplace_back([&](const Environment &Env) {
1940       return JavaScriptRequoter(Env, Expanded).process();
1941     });
1942 
1943   Passes.emplace_back([&](const Environment &Env) {
1944     return Formatter(Env, Expanded, Status).process();
1945   });
1946 
1947   std::unique_ptr<Environment> Env =
1948       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1949   llvm::Optional<std::string> CurrentCode = None;
1950   tooling::Replacements Fixes;
1951   for (size_t I = 0, E = Passes.size(); I < E; ++I) {
1952     tooling::Replacements PassFixes = Passes[I](*Env);
1953     auto NewCode = applyAllReplacements(
1954         CurrentCode ? StringRef(*CurrentCode) : Code, PassFixes);
1955     if (NewCode) {
1956       Fixes = Fixes.merge(PassFixes);
1957       if (I + 1 < E) {
1958         CurrentCode = std::move(*NewCode);
1959         Env = Environment::CreateVirtualEnvironment(
1960             *CurrentCode, FileName,
1961             tooling::calculateRangesAfterReplacements(Fixes, Ranges));
1962       }
1963     }
1964   }
1965 
1966   return Fixes;
1967 }
1968 
1969 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
1970                               ArrayRef<tooling::Range> Ranges,
1971                               StringRef FileName) {
1972   // cleanups only apply to C++ (they mostly concern ctor commas etc.)
1973   if (Style.Language != FormatStyle::LK_Cpp)
1974     return tooling::Replacements();
1975   std::unique_ptr<Environment> Env =
1976       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1977   Cleaner Clean(*Env, Style);
1978   return Clean.process();
1979 }
1980 
1981 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1982                                ArrayRef<tooling::Range> Ranges,
1983                                StringRef FileName, bool *IncompleteFormat) {
1984   FormattingAttemptStatus Status;
1985   auto Result = reformat(Style, Code, Ranges, FileName, &Status);
1986   if (!Status.FormatComplete)
1987     *IncompleteFormat = true;
1988   return Result;
1989 }
1990 
1991 tooling::Replacements fixNamespaceEndComments(const FormatStyle &Style,
1992                                               StringRef Code,
1993                                               ArrayRef<tooling::Range> Ranges,
1994                                               StringRef FileName) {
1995   std::unique_ptr<Environment> Env =
1996       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
1997   NamespaceEndCommentsFixer Fix(*Env, Style);
1998   return Fix.process();
1999 }
2000 
2001 tooling::Replacements sortUsingDeclarations(const FormatStyle &Style,
2002                                             StringRef Code,
2003                                             ArrayRef<tooling::Range> Ranges,
2004                                             StringRef FileName) {
2005   std::unique_ptr<Environment> Env =
2006       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
2007   UsingDeclarationsSorter Sorter(*Env, Style);
2008   return Sorter.process();
2009 }
2010 
2011 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
2012   LangOptions LangOpts;
2013   LangOpts.CPlusPlus = 1;
2014   LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2015   LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2016   LangOpts.CPlusPlus1z = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2017   LangOpts.LineComment = 1;
2018   bool AlternativeOperators = Style.isCpp();
2019   LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
2020   LangOpts.Bool = 1;
2021   LangOpts.ObjC1 = 1;
2022   LangOpts.ObjC2 = 1;
2023   LangOpts.MicrosoftExt = 1;    // To get kw___try, kw___finally.
2024   LangOpts.DeclSpecKeyword = 1; // To get __declspec.
2025   return LangOpts;
2026 }
2027 
2028 const char *StyleOptionHelpDescription =
2029     "Coding style, currently supports:\n"
2030     "  LLVM, Google, Chromium, Mozilla, WebKit.\n"
2031     "Use -style=file to load style configuration from\n"
2032     ".clang-format file located in one of the parent\n"
2033     "directories of the source file (or current\n"
2034     "directory for stdin).\n"
2035     "Use -style=\"{key: value, ...}\" to set specific\n"
2036     "parameters, e.g.:\n"
2037     "  -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
2038 
2039 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
2040   if (FileName.endswith(".java"))
2041     return FormatStyle::LK_Java;
2042   if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
2043     return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
2044   if (FileName.endswith(".m") || FileName.endswith(".mm"))
2045     return FormatStyle::LK_ObjC;
2046   if (FileName.endswith_lower(".proto") ||
2047       FileName.endswith_lower(".protodevel"))
2048     return FormatStyle::LK_Proto;
2049   if (FileName.endswith_lower(".td"))
2050     return FormatStyle::LK_TableGen;
2051   return FormatStyle::LK_Cpp;
2052 }
2053 
2054 llvm::Expected<FormatStyle> getStyle(StringRef StyleName, StringRef FileName,
2055                                      StringRef FallbackStyleName,
2056                                      StringRef Code, vfs::FileSystem *FS) {
2057   if (!FS) {
2058     FS = vfs::getRealFileSystem().get();
2059   }
2060   FormatStyle Style = getLLVMStyle();
2061   Style.Language = getLanguageByFileName(FileName);
2062 
2063   // This is a very crude detection of whether a header contains ObjC code that
2064   // should be improved over time and probably be done on tokens, not one the
2065   // bare content of the file.
2066   if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
2067       (Code.contains("\n- (") || Code.contains("\n+ (")))
2068     Style.Language = FormatStyle::LK_ObjC;
2069 
2070   FormatStyle FallbackStyle = getNoStyle();
2071   if (!getPredefinedStyle(FallbackStyleName, Style.Language, &FallbackStyle))
2072     return make_string_error("Invalid fallback style \"" + FallbackStyleName);
2073 
2074   if (StyleName.startswith("{")) {
2075     // Parse YAML/JSON style from the command line.
2076     if (std::error_code ec = parseConfiguration(StyleName, &Style))
2077       return make_string_error("Error parsing -style: " + ec.message());
2078     return Style;
2079   }
2080 
2081   if (!StyleName.equals_lower("file")) {
2082     if (!getPredefinedStyle(StyleName, Style.Language, &Style))
2083       return make_string_error("Invalid value for -style");
2084     return Style;
2085   }
2086 
2087   // Look for .clang-format/_clang-format file in the file's parent directories.
2088   SmallString<128> UnsuitableConfigFiles;
2089   SmallString<128> Path(FileName);
2090   if (std::error_code EC = FS->makeAbsolute(Path))
2091     return make_string_error(EC.message());
2092 
2093   for (StringRef Directory = Path; !Directory.empty();
2094        Directory = llvm::sys::path::parent_path(Directory)) {
2095 
2096     auto Status = FS->status(Directory);
2097     if (!Status ||
2098         Status->getType() != llvm::sys::fs::file_type::directory_file) {
2099       continue;
2100     }
2101 
2102     SmallString<128> ConfigFile(Directory);
2103 
2104     llvm::sys::path::append(ConfigFile, ".clang-format");
2105     DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2106 
2107     Status = FS->status(ConfigFile.str());
2108     bool FoundConfigFile =
2109         Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
2110     if (!FoundConfigFile) {
2111       // Try _clang-format too, since dotfiles are not commonly used on Windows.
2112       ConfigFile = Directory;
2113       llvm::sys::path::append(ConfigFile, "_clang-format");
2114       DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2115       Status = FS->status(ConfigFile.str());
2116       FoundConfigFile = Status && (Status->getType() ==
2117                                    llvm::sys::fs::file_type::regular_file);
2118     }
2119 
2120     if (FoundConfigFile) {
2121       llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
2122           FS->getBufferForFile(ConfigFile.str());
2123       if (std::error_code EC = Text.getError())
2124         return make_string_error(EC.message());
2125       if (std::error_code ec =
2126               parseConfiguration(Text.get()->getBuffer(), &Style)) {
2127         if (ec == ParseError::Unsuitable) {
2128           if (!UnsuitableConfigFiles.empty())
2129             UnsuitableConfigFiles.append(", ");
2130           UnsuitableConfigFiles.append(ConfigFile);
2131           continue;
2132         }
2133         return make_string_error("Error reading " + ConfigFile + ": " +
2134                                  ec.message());
2135       }
2136       DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
2137       return Style;
2138     }
2139   }
2140   if (!UnsuitableConfigFiles.empty())
2141     return make_string_error("Configuration file(s) do(es) not support " +
2142                              getLanguageName(Style.Language) + ": " +
2143                              UnsuitableConfigFiles);
2144   return FallbackStyle;
2145 }
2146 
2147 } // namespace format
2148 } // namespace clang
2149