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