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 "FormatInternal.h"
20 #include "FormatTokenLexer.h"
21 #include "NamespaceEndCommentsFixer.h"
22 #include "SortJavaScriptImports.h"
23 #include "TokenAnalyzer.h"
24 #include "TokenAnnotator.h"
25 #include "UnwrappedLineFormatter.h"
26 #include "UnwrappedLineParser.h"
27 #include "UsingDeclarationsSorter.h"
28 #include "WhitespaceManager.h"
29 #include "clang/Basic/Diagnostic.h"
30 #include "clang/Basic/DiagnosticOptions.h"
31 #include "clang/Basic/SourceManager.h"
32 #include "clang/Basic/VirtualFileSystem.h"
33 #include "clang/Lex/Lexer.h"
34 #include "llvm/ADT/STLExtras.h"
35 #include "llvm/Support/Allocator.h"
36 #include "llvm/Support/Debug.h"
37 #include "llvm/Support/Path.h"
38 #include "llvm/Support/Regex.h"
39 #include "llvm/Support/YAMLTraits.h"
40 #include <algorithm>
41 #include <memory>
42 #include <string>
43 
44 #define DEBUG_TYPE "format-formatter"
45 
46 using clang::format::FormatStyle;
47 
48 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::IncludeCategory)
49 LLVM_YAML_IS_SEQUENCE_VECTOR(clang::format::FormatStyle::RawStringFormat)
50 
51 namespace llvm {
52 namespace yaml {
53 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
54   static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
55     IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
56     IO.enumCase(Value, "Java", FormatStyle::LK_Java);
57     IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
58     IO.enumCase(Value, "ObjC", FormatStyle::LK_ObjC);
59     IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
60     IO.enumCase(Value, "TableGen", FormatStyle::LK_TableGen);
61     IO.enumCase(Value, "TextProto", FormatStyle::LK_TextProto);
62   }
63 };
64 
65 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
66   static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
67     IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
68     IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
69     IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
70     IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
71     IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
72   }
73 };
74 
75 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
76   static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
77     IO.enumCase(Value, "Never", FormatStyle::UT_Never);
78     IO.enumCase(Value, "false", FormatStyle::UT_Never);
79     IO.enumCase(Value, "Always", FormatStyle::UT_Always);
80     IO.enumCase(Value, "true", FormatStyle::UT_Always);
81     IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
82     IO.enumCase(Value, "ForContinuationAndIndentation",
83                 FormatStyle::UT_ForContinuationAndIndentation);
84   }
85 };
86 
87 template <> struct ScalarEnumerationTraits<FormatStyle::JavaScriptQuoteStyle> {
88   static void enumeration(IO &IO, FormatStyle::JavaScriptQuoteStyle &Value) {
89     IO.enumCase(Value, "Leave", FormatStyle::JSQS_Leave);
90     IO.enumCase(Value, "Single", FormatStyle::JSQS_Single);
91     IO.enumCase(Value, "Double", FormatStyle::JSQS_Double);
92   }
93 };
94 
95 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
96   static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
97     IO.enumCase(Value, "None", FormatStyle::SFS_None);
98     IO.enumCase(Value, "false", FormatStyle::SFS_None);
99     IO.enumCase(Value, "All", FormatStyle::SFS_All);
100     IO.enumCase(Value, "true", FormatStyle::SFS_All);
101     IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
102     IO.enumCase(Value, "InlineOnly", FormatStyle::SFS_InlineOnly);
103     IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
104   }
105 };
106 
107 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
108   static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
109     IO.enumCase(Value, "All", FormatStyle::BOS_All);
110     IO.enumCase(Value, "true", FormatStyle::BOS_All);
111     IO.enumCase(Value, "None", FormatStyle::BOS_None);
112     IO.enumCase(Value, "false", FormatStyle::BOS_None);
113     IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
114   }
115 };
116 
117 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
118   static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
119     IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
120     IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
121     IO.enumCase(Value, "Mozilla", FormatStyle::BS_Mozilla);
122     IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
123     IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
124     IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
125     IO.enumCase(Value, "WebKit", FormatStyle::BS_WebKit);
126     IO.enumCase(Value, "Custom", FormatStyle::BS_Custom);
127   }
128 };
129 
130 template <>
131 struct ScalarEnumerationTraits<FormatStyle::BreakConstructorInitializersStyle> {
132   static void
133   enumeration(IO &IO, FormatStyle::BreakConstructorInitializersStyle &Value) {
134     IO.enumCase(Value, "BeforeColon", FormatStyle::BCIS_BeforeColon);
135     IO.enumCase(Value, "BeforeComma", FormatStyle::BCIS_BeforeComma);
136     IO.enumCase(Value, "AfterColon", FormatStyle::BCIS_AfterColon);
137   }
138 };
139 
140 template <>
141 struct ScalarEnumerationTraits<FormatStyle::PPDirectiveIndentStyle> {
142   static void enumeration(IO &IO, FormatStyle::PPDirectiveIndentStyle &Value) {
143     IO.enumCase(Value, "None", FormatStyle::PPDIS_None);
144     IO.enumCase(Value, "AfterHash", FormatStyle::PPDIS_AfterHash);
145   }
146 };
147 
148 template <>
149 struct ScalarEnumerationTraits<FormatStyle::ReturnTypeBreakingStyle> {
150   static void enumeration(IO &IO, FormatStyle::ReturnTypeBreakingStyle &Value) {
151     IO.enumCase(Value, "None", FormatStyle::RTBS_None);
152     IO.enumCase(Value, "All", FormatStyle::RTBS_All);
153     IO.enumCase(Value, "TopLevel", FormatStyle::RTBS_TopLevel);
154     IO.enumCase(Value, "TopLevelDefinitions",
155                 FormatStyle::RTBS_TopLevelDefinitions);
156     IO.enumCase(Value, "AllDefinitions", FormatStyle::RTBS_AllDefinitions);
157   }
158 };
159 
160 template <>
161 struct ScalarEnumerationTraits<FormatStyle::DefinitionReturnTypeBreakingStyle> {
162   static void
163   enumeration(IO &IO, FormatStyle::DefinitionReturnTypeBreakingStyle &Value) {
164     IO.enumCase(Value, "None", FormatStyle::DRTBS_None);
165     IO.enumCase(Value, "All", FormatStyle::DRTBS_All);
166     IO.enumCase(Value, "TopLevel", FormatStyle::DRTBS_TopLevel);
167 
168     // For backward compatibility.
169     IO.enumCase(Value, "false", FormatStyle::DRTBS_None);
170     IO.enumCase(Value, "true", FormatStyle::DRTBS_All);
171   }
172 };
173 
174 template <>
175 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
176   static void enumeration(IO &IO,
177                           FormatStyle::NamespaceIndentationKind &Value) {
178     IO.enumCase(Value, "None", FormatStyle::NI_None);
179     IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
180     IO.enumCase(Value, "All", FormatStyle::NI_All);
181   }
182 };
183 
184 template <> struct ScalarEnumerationTraits<FormatStyle::BracketAlignmentStyle> {
185   static void enumeration(IO &IO, FormatStyle::BracketAlignmentStyle &Value) {
186     IO.enumCase(Value, "Align", FormatStyle::BAS_Align);
187     IO.enumCase(Value, "DontAlign", FormatStyle::BAS_DontAlign);
188     IO.enumCase(Value, "AlwaysBreak", FormatStyle::BAS_AlwaysBreak);
189 
190     // For backward compatibility.
191     IO.enumCase(Value, "true", FormatStyle::BAS_Align);
192     IO.enumCase(Value, "false", FormatStyle::BAS_DontAlign);
193   }
194 };
195 
196 template <>
197 struct ScalarEnumerationTraits<FormatStyle::EscapedNewlineAlignmentStyle> {
198   static void enumeration(IO &IO,
199                           FormatStyle::EscapedNewlineAlignmentStyle &Value) {
200     IO.enumCase(Value, "DontAlign", FormatStyle::ENAS_DontAlign);
201     IO.enumCase(Value, "Left", FormatStyle::ENAS_Left);
202     IO.enumCase(Value, "Right", FormatStyle::ENAS_Right);
203 
204     // For backward compatibility.
205     IO.enumCase(Value, "true", FormatStyle::ENAS_Left);
206     IO.enumCase(Value, "false", FormatStyle::ENAS_Right);
207   }
208 };
209 
210 template <> struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
211   static void enumeration(IO &IO, FormatStyle::PointerAlignmentStyle &Value) {
212     IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
213     IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
214     IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
215 
216     // For backward compatibility.
217     IO.enumCase(Value, "true", FormatStyle::PAS_Left);
218     IO.enumCase(Value, "false", FormatStyle::PAS_Right);
219   }
220 };
221 
222 template <>
223 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
224   static void enumeration(IO &IO,
225                           FormatStyle::SpaceBeforeParensOptions &Value) {
226     IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
227     IO.enumCase(Value, "ControlStatements",
228                 FormatStyle::SBPO_ControlStatements);
229     IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
230 
231     // For backward compatibility.
232     IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
233     IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
234   }
235 };
236 
237 template <> struct MappingTraits<FormatStyle> {
238   static void mapping(IO &IO, FormatStyle &Style) {
239     // When reading, read the language first, we need it for getPredefinedStyle.
240     IO.mapOptional("Language", Style.Language);
241 
242     if (IO.outputting()) {
243       StringRef StylesArray[] = {"LLVM",    "Google", "Chromium",
244                                  "Mozilla", "WebKit", "GNU"};
245       ArrayRef<StringRef> Styles(StylesArray);
246       for (size_t i = 0, e = Styles.size(); i < e; ++i) {
247         StringRef StyleName(Styles[i]);
248         FormatStyle PredefinedStyle;
249         if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
250             Style == PredefinedStyle) {
251           IO.mapOptional("# BasedOnStyle", StyleName);
252           break;
253         }
254       }
255     } else {
256       StringRef BasedOnStyle;
257       IO.mapOptional("BasedOnStyle", BasedOnStyle);
258       if (!BasedOnStyle.empty()) {
259         FormatStyle::LanguageKind OldLanguage = Style.Language;
260         FormatStyle::LanguageKind Language =
261             ((FormatStyle *)IO.getContext())->Language;
262         if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
263           IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
264           return;
265         }
266         Style.Language = OldLanguage;
267       }
268     }
269 
270     // For backward compatibility.
271     if (!IO.outputting()) {
272       IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlines);
273       IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
274       IO.mapOptional("IndentFunctionDeclarationAfterType",
275                      Style.IndentWrappedFunctionNames);
276       IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
277       IO.mapOptional("SpaceAfterControlStatementKeyword",
278                      Style.SpaceBeforeParens);
279     }
280 
281     IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
282     IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
283     IO.mapOptional("AlignConsecutiveAssignments",
284                    Style.AlignConsecutiveAssignments);
285     IO.mapOptional("AlignConsecutiveDeclarations",
286                    Style.AlignConsecutiveDeclarations);
287     IO.mapOptional("AlignEscapedNewlines", Style.AlignEscapedNewlines);
288     IO.mapOptional("AlignOperands", Style.AlignOperands);
289     IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
290     IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
291                    Style.AllowAllParametersOfDeclarationOnNextLine);
292     IO.mapOptional("AllowShortBlocksOnASingleLine",
293                    Style.AllowShortBlocksOnASingleLine);
294     IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
295                    Style.AllowShortCaseLabelsOnASingleLine);
296     IO.mapOptional("AllowShortFunctionsOnASingleLine",
297                    Style.AllowShortFunctionsOnASingleLine);
298     IO.mapOptional("AllowShortIfStatementsOnASingleLine",
299                    Style.AllowShortIfStatementsOnASingleLine);
300     IO.mapOptional("AllowShortLoopsOnASingleLine",
301                    Style.AllowShortLoopsOnASingleLine);
302     IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
303                    Style.AlwaysBreakAfterDefinitionReturnType);
304     IO.mapOptional("AlwaysBreakAfterReturnType",
305                    Style.AlwaysBreakAfterReturnType);
306     // If AlwaysBreakAfterDefinitionReturnType was specified but
307     // AlwaysBreakAfterReturnType was not, initialize the latter from the
308     // former for backwards compatibility.
309     if (Style.AlwaysBreakAfterDefinitionReturnType != FormatStyle::DRTBS_None &&
310         Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None) {
311       if (Style.AlwaysBreakAfterDefinitionReturnType == FormatStyle::DRTBS_All)
312         Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
313       else if (Style.AlwaysBreakAfterDefinitionReturnType ==
314                FormatStyle::DRTBS_TopLevel)
315         Style.AlwaysBreakAfterReturnType =
316             FormatStyle::RTBS_TopLevelDefinitions;
317     }
318 
319     IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
320                    Style.AlwaysBreakBeforeMultilineStrings);
321     IO.mapOptional("AlwaysBreakTemplateDeclarations",
322                    Style.AlwaysBreakTemplateDeclarations);
323     IO.mapOptional("BinPackArguments", Style.BinPackArguments);
324     IO.mapOptional("BinPackParameters", Style.BinPackParameters);
325     IO.mapOptional("BraceWrapping", Style.BraceWrapping);
326     IO.mapOptional("BreakBeforeBinaryOperators",
327                    Style.BreakBeforeBinaryOperators);
328     IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
329     IO.mapOptional("BreakBeforeInheritanceComma",
330                    Style.BreakBeforeInheritanceComma);
331     IO.mapOptional("BreakBeforeTernaryOperators",
332                    Style.BreakBeforeTernaryOperators);
333 
334     bool BreakConstructorInitializersBeforeComma = false;
335     IO.mapOptional("BreakConstructorInitializersBeforeComma",
336                    BreakConstructorInitializersBeforeComma);
337     IO.mapOptional("BreakConstructorInitializers",
338                    Style.BreakConstructorInitializers);
339     // If BreakConstructorInitializersBeforeComma was specified but
340     // BreakConstructorInitializers was not, initialize the latter from the
341     // former for backwards compatibility.
342     if (BreakConstructorInitializersBeforeComma &&
343         Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeColon)
344       Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
345 
346     IO.mapOptional("BreakAfterJavaFieldAnnotations",
347                    Style.BreakAfterJavaFieldAnnotations);
348     IO.mapOptional("BreakStringLiterals", Style.BreakStringLiterals);
349     IO.mapOptional("ColumnLimit", Style.ColumnLimit);
350     IO.mapOptional("CommentPragmas", Style.CommentPragmas);
351     IO.mapOptional("CompactNamespaces", Style.CompactNamespaces);
352     IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
353                    Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
354     IO.mapOptional("ConstructorInitializerIndentWidth",
355                    Style.ConstructorInitializerIndentWidth);
356     IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
357     IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
358     IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
359     IO.mapOptional("DisableFormat", Style.DisableFormat);
360     IO.mapOptional("ExperimentalAutoDetectBinPacking",
361                    Style.ExperimentalAutoDetectBinPacking);
362     IO.mapOptional("FixNamespaceComments", Style.FixNamespaceComments);
363     IO.mapOptional("ForEachMacros", Style.ForEachMacros);
364     IO.mapOptional("IncludeBlocks", Style.IncludeBlocks);
365     IO.mapOptional("IncludeCategories", Style.IncludeCategories);
366     IO.mapOptional("IncludeIsMainRegex", Style.IncludeIsMainRegex);
367     IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
368     IO.mapOptional("IndentPPDirectives", Style.IndentPPDirectives);
369     IO.mapOptional("IndentWidth", Style.IndentWidth);
370     IO.mapOptional("IndentWrappedFunctionNames",
371                    Style.IndentWrappedFunctionNames);
372     IO.mapOptional("JavaScriptQuotes", Style.JavaScriptQuotes);
373     IO.mapOptional("JavaScriptWrapImports", Style.JavaScriptWrapImports);
374     IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
375                    Style.KeepEmptyLinesAtTheStartOfBlocks);
376     IO.mapOptional("MacroBlockBegin", Style.MacroBlockBegin);
377     IO.mapOptional("MacroBlockEnd", Style.MacroBlockEnd);
378     IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
379     IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
380     IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
381     IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
382     IO.mapOptional("ObjCSpaceBeforeProtocolList",
383                    Style.ObjCSpaceBeforeProtocolList);
384     IO.mapOptional("PenaltyBreakAssignment", Style.PenaltyBreakAssignment);
385     IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
386                    Style.PenaltyBreakBeforeFirstCallParameter);
387     IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
388     IO.mapOptional("PenaltyBreakFirstLessLess",
389                    Style.PenaltyBreakFirstLessLess);
390     IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
391     IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
392     IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
393                    Style.PenaltyReturnTypeOnItsOwnLine);
394     IO.mapOptional("PointerAlignment", Style.PointerAlignment);
395     IO.mapOptional("RawStringFormats", Style.RawStringFormats);
396     IO.mapOptional("ReflowComments", Style.ReflowComments);
397     IO.mapOptional("SortIncludes", Style.SortIncludes);
398     IO.mapOptional("SortUsingDeclarations", Style.SortUsingDeclarations);
399     IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
400     IO.mapOptional("SpaceAfterTemplateKeyword",
401                    Style.SpaceAfterTemplateKeyword);
402     IO.mapOptional("SpaceBeforeAssignmentOperators",
403                    Style.SpaceBeforeAssignmentOperators);
404     IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
405     IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
406     IO.mapOptional("SpacesBeforeTrailingComments",
407                    Style.SpacesBeforeTrailingComments);
408     IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
409     IO.mapOptional("SpacesInContainerLiterals",
410                    Style.SpacesInContainerLiterals);
411     IO.mapOptional("SpacesInCStyleCastParentheses",
412                    Style.SpacesInCStyleCastParentheses);
413     IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
414     IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
415     IO.mapOptional("Standard", Style.Standard);
416     IO.mapOptional("TabWidth", Style.TabWidth);
417     IO.mapOptional("UseTab", Style.UseTab);
418   }
419 };
420 
421 template <> struct MappingTraits<FormatStyle::BraceWrappingFlags> {
422   static void mapping(IO &IO, FormatStyle::BraceWrappingFlags &Wrapping) {
423     IO.mapOptional("AfterClass", Wrapping.AfterClass);
424     IO.mapOptional("AfterControlStatement", Wrapping.AfterControlStatement);
425     IO.mapOptional("AfterEnum", Wrapping.AfterEnum);
426     IO.mapOptional("AfterFunction", Wrapping.AfterFunction);
427     IO.mapOptional("AfterNamespace", Wrapping.AfterNamespace);
428     IO.mapOptional("AfterObjCDeclaration", Wrapping.AfterObjCDeclaration);
429     IO.mapOptional("AfterStruct", Wrapping.AfterStruct);
430     IO.mapOptional("AfterUnion", Wrapping.AfterUnion);
431     IO.mapOptional("AfterExternBlock", Wrapping.AfterExternBlock);
432     IO.mapOptional("BeforeCatch", Wrapping.BeforeCatch);
433     IO.mapOptional("BeforeElse", Wrapping.BeforeElse);
434     IO.mapOptional("IndentBraces", Wrapping.IndentBraces);
435     IO.mapOptional("SplitEmptyFunction", Wrapping.SplitEmptyFunction);
436     IO.mapOptional("SplitEmptyRecord", Wrapping.SplitEmptyRecord);
437     IO.mapOptional("SplitEmptyNamespace", Wrapping.SplitEmptyNamespace);
438   }
439 };
440 
441 template <> struct MappingTraits<FormatStyle::IncludeCategory> {
442   static void mapping(IO &IO, FormatStyle::IncludeCategory &Category) {
443     IO.mapOptional("Regex", Category.Regex);
444     IO.mapOptional("Priority", Category.Priority);
445   }
446 };
447 
448 template <> struct ScalarEnumerationTraits<FormatStyle::IncludeBlocksStyle> {
449   static void enumeration(IO &IO, FormatStyle::IncludeBlocksStyle &Value) {
450     IO.enumCase(Value, "Preserve", FormatStyle::IBS_Preserve);
451     IO.enumCase(Value, "Merge", FormatStyle::IBS_Merge);
452     IO.enumCase(Value, "Regroup", FormatStyle::IBS_Regroup);
453   }
454 };
455 
456 template <> struct MappingTraits<FormatStyle::RawStringFormat> {
457   static void mapping(IO &IO, FormatStyle::RawStringFormat &Format) {
458     IO.mapOptional("Delimiter", Format.Delimiter);
459     IO.mapOptional("Language", Format.Language);
460     IO.mapOptional("BasedOnStyle", Format.BasedOnStyle);
461   }
462 };
463 
464 // Allows to read vector<FormatStyle> while keeping default values.
465 // IO.getContext() should contain a pointer to the FormatStyle structure, that
466 // will be used to get default values for missing keys.
467 // If the first element has no Language specified, it will be treated as the
468 // default one for the following elements.
469 template <> struct DocumentListTraits<std::vector<FormatStyle>> {
470   static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
471     return Seq.size();
472   }
473   static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
474                               size_t Index) {
475     if (Index >= Seq.size()) {
476       assert(Index == Seq.size());
477       FormatStyle Template;
478       if (!Seq.empty() && Seq[0].Language == FormatStyle::LK_None) {
479         Template = Seq[0];
480       } else {
481         Template = *((const FormatStyle *)IO.getContext());
482         Template.Language = FormatStyle::LK_None;
483       }
484       Seq.resize(Index + 1, Template);
485     }
486     return Seq[Index];
487   }
488 };
489 } // namespace yaml
490 } // namespace llvm
491 
492 namespace clang {
493 namespace format {
494 
495 const std::error_category &getParseCategory() {
496   static ParseErrorCategory C;
497   return C;
498 }
499 std::error_code make_error_code(ParseError e) {
500   return std::error_code(static_cast<int>(e), getParseCategory());
501 }
502 
503 inline llvm::Error make_string_error(const llvm::Twine &Message) {
504   return llvm::make_error<llvm::StringError>(Message,
505                                              llvm::inconvertibleErrorCode());
506 }
507 
508 const char *ParseErrorCategory::name() const noexcept {
509   return "clang-format.parse_error";
510 }
511 
512 std::string ParseErrorCategory::message(int EV) const {
513   switch (static_cast<ParseError>(EV)) {
514   case ParseError::Success:
515     return "Success";
516   case ParseError::Error:
517     return "Invalid argument";
518   case ParseError::Unsuitable:
519     return "Unsuitable";
520   }
521   llvm_unreachable("unexpected parse error");
522 }
523 
524 static FormatStyle expandPresets(const FormatStyle &Style) {
525   if (Style.BreakBeforeBraces == FormatStyle::BS_Custom)
526     return Style;
527   FormatStyle Expanded = Style;
528   Expanded.BraceWrapping = {false, false, false, false, false,
529                             false, false, false, false, false,
530                             false, false, true,  true,  true};
531   switch (Style.BreakBeforeBraces) {
532   case FormatStyle::BS_Linux:
533     Expanded.BraceWrapping.AfterClass = true;
534     Expanded.BraceWrapping.AfterFunction = true;
535     Expanded.BraceWrapping.AfterNamespace = true;
536     break;
537   case FormatStyle::BS_Mozilla:
538     Expanded.BraceWrapping.AfterClass = true;
539     Expanded.BraceWrapping.AfterEnum = true;
540     Expanded.BraceWrapping.AfterFunction = true;
541     Expanded.BraceWrapping.AfterStruct = true;
542     Expanded.BraceWrapping.AfterUnion = true;
543     Expanded.BraceWrapping.AfterExternBlock = true;
544     Expanded.BraceWrapping.SplitEmptyFunction = true;
545     Expanded.BraceWrapping.SplitEmptyRecord = false;
546     break;
547   case FormatStyle::BS_Stroustrup:
548     Expanded.BraceWrapping.AfterFunction = true;
549     Expanded.BraceWrapping.BeforeCatch = true;
550     Expanded.BraceWrapping.BeforeElse = true;
551     break;
552   case FormatStyle::BS_Allman:
553     Expanded.BraceWrapping.AfterClass = true;
554     Expanded.BraceWrapping.AfterControlStatement = true;
555     Expanded.BraceWrapping.AfterEnum = true;
556     Expanded.BraceWrapping.AfterFunction = true;
557     Expanded.BraceWrapping.AfterNamespace = true;
558     Expanded.BraceWrapping.AfterObjCDeclaration = true;
559     Expanded.BraceWrapping.AfterStruct = true;
560     Expanded.BraceWrapping.AfterExternBlock = true;
561     Expanded.BraceWrapping.BeforeCatch = true;
562     Expanded.BraceWrapping.BeforeElse = true;
563     break;
564   case FormatStyle::BS_GNU:
565     Expanded.BraceWrapping = {true, true, true, true, true, true, true, true,
566                               true, true, true, true, true, true, true};
567     break;
568   case FormatStyle::BS_WebKit:
569     Expanded.BraceWrapping.AfterFunction = true;
570     break;
571   default:
572     break;
573   }
574   return Expanded;
575 }
576 
577 FormatStyle getLLVMStyle() {
578   FormatStyle LLVMStyle;
579   LLVMStyle.Language = FormatStyle::LK_Cpp;
580   LLVMStyle.AccessModifierOffset = -2;
581   LLVMStyle.AlignEscapedNewlines = FormatStyle::ENAS_Right;
582   LLVMStyle.AlignAfterOpenBracket = FormatStyle::BAS_Align;
583   LLVMStyle.AlignOperands = true;
584   LLVMStyle.AlignTrailingComments = true;
585   LLVMStyle.AlignConsecutiveAssignments = false;
586   LLVMStyle.AlignConsecutiveDeclarations = false;
587   LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
588   LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
589   LLVMStyle.AllowShortBlocksOnASingleLine = false;
590   LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
591   LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
592   LLVMStyle.AllowShortLoopsOnASingleLine = false;
593   LLVMStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
594   LLVMStyle.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_None;
595   LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
596   LLVMStyle.AlwaysBreakTemplateDeclarations = false;
597   LLVMStyle.BinPackArguments = true;
598   LLVMStyle.BinPackParameters = true;
599   LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
600   LLVMStyle.BreakBeforeTernaryOperators = true;
601   LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
602   LLVMStyle.BraceWrapping = {false, false, false, false, false,
603                              false, false, false, false, false,
604                              false, false, true,  true,  true};
605   LLVMStyle.BreakAfterJavaFieldAnnotations = false;
606   LLVMStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeColon;
607   LLVMStyle.BreakBeforeInheritanceComma = false;
608   LLVMStyle.BreakStringLiterals = true;
609   LLVMStyle.ColumnLimit = 80;
610   LLVMStyle.CommentPragmas = "^ IWYU pragma:";
611   LLVMStyle.CompactNamespaces = false;
612   LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
613   LLVMStyle.ConstructorInitializerIndentWidth = 4;
614   LLVMStyle.ContinuationIndentWidth = 4;
615   LLVMStyle.Cpp11BracedListStyle = true;
616   LLVMStyle.DerivePointerAlignment = false;
617   LLVMStyle.ExperimentalAutoDetectBinPacking = false;
618   LLVMStyle.FixNamespaceComments = true;
619   LLVMStyle.ForEachMacros.push_back("foreach");
620   LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
621   LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
622   LLVMStyle.IncludeCategories = {{"^\"(llvm|llvm-c|clang|clang-c)/", 2},
623                                  {"^(<|\"(gtest|gmock|isl|json)/)", 3},
624                                  {".*", 1}};
625   LLVMStyle.IncludeIsMainRegex = "(Test)?$";
626   LLVMStyle.IncludeBlocks = FormatStyle::IBS_Preserve;
627   LLVMStyle.IndentCaseLabels = false;
628   LLVMStyle.IndentPPDirectives = FormatStyle::PPDIS_None;
629   LLVMStyle.IndentWrappedFunctionNames = false;
630   LLVMStyle.IndentWidth = 2;
631   LLVMStyle.JavaScriptQuotes = FormatStyle::JSQS_Leave;
632   LLVMStyle.JavaScriptWrapImports = true;
633   LLVMStyle.TabWidth = 8;
634   LLVMStyle.MaxEmptyLinesToKeep = 1;
635   LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
636   LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
637   LLVMStyle.ObjCBlockIndentWidth = 2;
638   LLVMStyle.ObjCSpaceAfterProperty = false;
639   LLVMStyle.ObjCSpaceBeforeProtocolList = true;
640   LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
641   LLVMStyle.SpacesBeforeTrailingComments = 1;
642   LLVMStyle.Standard = FormatStyle::LS_Cpp11;
643   LLVMStyle.UseTab = FormatStyle::UT_Never;
644   LLVMStyle.RawStringFormats = {{"pb", FormatStyle::LK_TextProto, "google"}};
645   LLVMStyle.ReflowComments = true;
646   LLVMStyle.SpacesInParentheses = false;
647   LLVMStyle.SpacesInSquareBrackets = false;
648   LLVMStyle.SpaceInEmptyParentheses = false;
649   LLVMStyle.SpacesInContainerLiterals = true;
650   LLVMStyle.SpacesInCStyleCastParentheses = false;
651   LLVMStyle.SpaceAfterCStyleCast = false;
652   LLVMStyle.SpaceAfterTemplateKeyword = true;
653   LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
654   LLVMStyle.SpaceBeforeAssignmentOperators = true;
655   LLVMStyle.SpacesInAngles = false;
656 
657   LLVMStyle.PenaltyBreakAssignment = prec::Assignment;
658   LLVMStyle.PenaltyBreakComment = 300;
659   LLVMStyle.PenaltyBreakFirstLessLess = 120;
660   LLVMStyle.PenaltyBreakString = 1000;
661   LLVMStyle.PenaltyExcessCharacter = 1000000;
662   LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
663   LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
664 
665   LLVMStyle.DisableFormat = false;
666   LLVMStyle.SortIncludes = true;
667   LLVMStyle.SortUsingDeclarations = true;
668 
669   return LLVMStyle;
670 }
671 
672 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
673   if (Language == FormatStyle::LK_TextProto) {
674     FormatStyle GoogleStyle = getGoogleStyle(FormatStyle::LK_Proto);
675     GoogleStyle.Language = FormatStyle::LK_TextProto;
676     return GoogleStyle;
677   }
678 
679   FormatStyle GoogleStyle = getLLVMStyle();
680   GoogleStyle.Language = Language;
681 
682   GoogleStyle.AccessModifierOffset = -1;
683   GoogleStyle.AlignEscapedNewlines = FormatStyle::ENAS_Left;
684   GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
685   GoogleStyle.AllowShortLoopsOnASingleLine = true;
686   GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
687   GoogleStyle.AlwaysBreakTemplateDeclarations = true;
688   GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
689   GoogleStyle.DerivePointerAlignment = true;
690   GoogleStyle.IncludeCategories = {
691       {"^<ext/.*\\.h>", 2}, {"^<.*\\.h>", 1}, {"^<.*", 2}, {".*", 3}};
692   GoogleStyle.IncludeIsMainRegex = "([-_](test|unittest))?$";
693   GoogleStyle.IndentCaseLabels = true;
694   GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
695   GoogleStyle.ObjCSpaceAfterProperty = false;
696   GoogleStyle.ObjCSpaceBeforeProtocolList = false;
697   GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
698   GoogleStyle.SpacesBeforeTrailingComments = 2;
699   GoogleStyle.Standard = FormatStyle::LS_Auto;
700 
701   GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
702   GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
703 
704   if (Language == FormatStyle::LK_Java) {
705     GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
706     GoogleStyle.AlignOperands = false;
707     GoogleStyle.AlignTrailingComments = false;
708     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
709     GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
710     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
711     GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
712     GoogleStyle.ColumnLimit = 100;
713     GoogleStyle.SpaceAfterCStyleCast = true;
714     GoogleStyle.SpacesBeforeTrailingComments = 1;
715   } else if (Language == FormatStyle::LK_JavaScript) {
716     GoogleStyle.AlignAfterOpenBracket = FormatStyle::BAS_AlwaysBreak;
717     GoogleStyle.AlignOperands = false;
718     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
719     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
720     GoogleStyle.BreakBeforeTernaryOperators = false;
721     // taze:, triple slash directives (`/// <...`), @tag followed by { for a lot
722     // of JSDoc tags, and @see, which is commonly followed by overlong URLs.
723     GoogleStyle.CommentPragmas =
724         "(taze:|^/[ \t]*<|(@[A-Za-z_0-9-]+[ \\t]*{)|@see)";
725     GoogleStyle.MaxEmptyLinesToKeep = 3;
726     GoogleStyle.NamespaceIndentation = FormatStyle::NI_All;
727     GoogleStyle.SpacesInContainerLiterals = false;
728     GoogleStyle.JavaScriptQuotes = FormatStyle::JSQS_Single;
729     GoogleStyle.JavaScriptWrapImports = false;
730   } else if (Language == FormatStyle::LK_Proto) {
731     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
732     GoogleStyle.SpacesInContainerLiterals = false;
733   } else if (Language == FormatStyle::LK_ObjC) {
734     GoogleStyle.ColumnLimit = 100;
735   }
736 
737   return GoogleStyle;
738 }
739 
740 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
741   FormatStyle ChromiumStyle = getGoogleStyle(Language);
742   if (Language == FormatStyle::LK_Java) {
743     ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
744     ChromiumStyle.BreakAfterJavaFieldAnnotations = true;
745     ChromiumStyle.ContinuationIndentWidth = 8;
746     ChromiumStyle.IndentWidth = 4;
747   } else if (Language == FormatStyle::LK_JavaScript) {
748     ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
749     ChromiumStyle.AllowShortLoopsOnASingleLine = false;
750   } else {
751     ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
752     ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
753     ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
754     ChromiumStyle.AllowShortLoopsOnASingleLine = false;
755     ChromiumStyle.BinPackParameters = false;
756     ChromiumStyle.DerivePointerAlignment = false;
757     if (Language == FormatStyle::LK_ObjC)
758       ChromiumStyle.ColumnLimit = 80;
759   }
760   return ChromiumStyle;
761 }
762 
763 FormatStyle getMozillaStyle() {
764   FormatStyle MozillaStyle = getLLVMStyle();
765   MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
766   MozillaStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
767   MozillaStyle.AlwaysBreakAfterReturnType = FormatStyle::RTBS_TopLevel;
768   MozillaStyle.AlwaysBreakAfterDefinitionReturnType =
769       FormatStyle::DRTBS_TopLevel;
770   MozillaStyle.AlwaysBreakTemplateDeclarations = true;
771   MozillaStyle.BinPackParameters = false;
772   MozillaStyle.BinPackArguments = false;
773   MozillaStyle.BreakBeforeBraces = FormatStyle::BS_Mozilla;
774   MozillaStyle.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
775   MozillaStyle.BreakBeforeInheritanceComma = true;
776   MozillaStyle.ConstructorInitializerIndentWidth = 2;
777   MozillaStyle.ContinuationIndentWidth = 2;
778   MozillaStyle.Cpp11BracedListStyle = false;
779   MozillaStyle.FixNamespaceComments = false;
780   MozillaStyle.IndentCaseLabels = true;
781   MozillaStyle.ObjCSpaceAfterProperty = true;
782   MozillaStyle.ObjCSpaceBeforeProtocolList = false;
783   MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
784   MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
785   MozillaStyle.SpaceAfterTemplateKeyword = false;
786   return MozillaStyle;
787 }
788 
789 FormatStyle getWebKitStyle() {
790   FormatStyle Style = getLLVMStyle();
791   Style.AccessModifierOffset = -4;
792   Style.AlignAfterOpenBracket = FormatStyle::BAS_DontAlign;
793   Style.AlignOperands = false;
794   Style.AlignTrailingComments = false;
795   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
796   Style.BreakBeforeBraces = FormatStyle::BS_WebKit;
797   Style.BreakConstructorInitializers = FormatStyle::BCIS_BeforeComma;
798   Style.Cpp11BracedListStyle = false;
799   Style.ColumnLimit = 0;
800   Style.FixNamespaceComments = false;
801   Style.IndentWidth = 4;
802   Style.NamespaceIndentation = FormatStyle::NI_Inner;
803   Style.ObjCBlockIndentWidth = 4;
804   Style.ObjCSpaceAfterProperty = true;
805   Style.PointerAlignment = FormatStyle::PAS_Left;
806   return Style;
807 }
808 
809 FormatStyle getGNUStyle() {
810   FormatStyle Style = getLLVMStyle();
811   Style.AlwaysBreakAfterDefinitionReturnType = FormatStyle::DRTBS_All;
812   Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_AllDefinitions;
813   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
814   Style.BreakBeforeBraces = FormatStyle::BS_GNU;
815   Style.BreakBeforeTernaryOperators = true;
816   Style.Cpp11BracedListStyle = false;
817   Style.ColumnLimit = 79;
818   Style.FixNamespaceComments = false;
819   Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
820   Style.Standard = FormatStyle::LS_Cpp03;
821   return Style;
822 }
823 
824 FormatStyle getNoStyle() {
825   FormatStyle NoStyle = getLLVMStyle();
826   NoStyle.DisableFormat = true;
827   NoStyle.SortIncludes = false;
828   NoStyle.SortUsingDeclarations = false;
829   return NoStyle;
830 }
831 
832 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
833                         FormatStyle *Style) {
834   if (Name.equals_lower("llvm")) {
835     *Style = getLLVMStyle();
836   } else if (Name.equals_lower("chromium")) {
837     *Style = getChromiumStyle(Language);
838   } else if (Name.equals_lower("mozilla")) {
839     *Style = getMozillaStyle();
840   } else if (Name.equals_lower("google")) {
841     *Style = getGoogleStyle(Language);
842   } else if (Name.equals_lower("webkit")) {
843     *Style = getWebKitStyle();
844   } else if (Name.equals_lower("gnu")) {
845     *Style = getGNUStyle();
846   } else if (Name.equals_lower("none")) {
847     *Style = getNoStyle();
848   } else {
849     return false;
850   }
851 
852   Style->Language = Language;
853   return true;
854 }
855 
856 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
857   assert(Style);
858   FormatStyle::LanguageKind Language = Style->Language;
859   assert(Language != FormatStyle::LK_None);
860   if (Text.trim().empty())
861     return make_error_code(ParseError::Error);
862   Style->StyleSet.Clear();
863   std::vector<FormatStyle> Styles;
864   llvm::yaml::Input Input(Text);
865   // DocumentListTraits<vector<FormatStyle>> uses the context to get default
866   // values for the fields, keys for which are missing from the configuration.
867   // Mapping also uses the context to get the language to find the correct
868   // base style.
869   Input.setContext(Style);
870   Input >> Styles;
871   if (Input.error())
872     return Input.error();
873 
874   for (unsigned i = 0; i < Styles.size(); ++i) {
875     // Ensures that only the first configuration can skip the Language option.
876     if (Styles[i].Language == FormatStyle::LK_None && i != 0)
877       return make_error_code(ParseError::Error);
878     // Ensure that each language is configured at most once.
879     for (unsigned j = 0; j < i; ++j) {
880       if (Styles[i].Language == Styles[j].Language) {
881         DEBUG(llvm::dbgs()
882               << "Duplicate languages in the config file on positions " << j
883               << " and " << i << "\n");
884         return make_error_code(ParseError::Error);
885       }
886     }
887   }
888   // Look for a suitable configuration starting from the end, so we can
889   // find the configuration for the specific language first, and the default
890   // configuration (which can only be at slot 0) after it.
891   FormatStyle::FormatStyleSet StyleSet;
892   bool LanguageFound = false;
893   for (int i = Styles.size() - 1; i >= 0; --i) {
894     if (Styles[i].Language != FormatStyle::LK_None)
895       StyleSet.Add(Styles[i]);
896     if (Styles[i].Language == Language)
897       LanguageFound = true;
898   }
899   if (!LanguageFound) {
900     if (Styles.empty() || Styles[0].Language != FormatStyle::LK_None)
901       return make_error_code(ParseError::Unsuitable);
902     FormatStyle DefaultStyle = Styles[0];
903     DefaultStyle.Language = Language;
904     StyleSet.Add(std::move(DefaultStyle));
905   }
906   *Style = *StyleSet.Get(Language);
907   return make_error_code(ParseError::Success);
908 }
909 
910 std::string configurationAsText(const FormatStyle &Style) {
911   std::string Text;
912   llvm::raw_string_ostream Stream(Text);
913   llvm::yaml::Output Output(Stream);
914   // We use the same mapping method for input and output, so we need a non-const
915   // reference here.
916   FormatStyle NonConstStyle = expandPresets(Style);
917   Output << NonConstStyle;
918   return Stream.str();
919 }
920 
921 llvm::Optional<FormatStyle>
922 FormatStyle::FormatStyleSet::Get(FormatStyle::LanguageKind Language) const {
923   if (!Styles)
924     return None;
925   auto It = Styles->find(Language);
926   if (It == Styles->end())
927     return None;
928   FormatStyle Style = It->second;
929   Style.StyleSet = *this;
930   return Style;
931 }
932 
933 void FormatStyle::FormatStyleSet::Add(FormatStyle Style) {
934   assert(Style.Language != LK_None &&
935          "Cannot add a style for LK_None to a StyleSet");
936   assert(
937       !Style.StyleSet.Styles &&
938       "Cannot add a style associated with an existing StyleSet to a StyleSet");
939   if (!Styles)
940     Styles = std::make_shared<MapType>();
941   (*Styles)[Style.Language] = std::move(Style);
942 }
943 
944 void FormatStyle::FormatStyleSet::Clear() {
945   Styles.reset();
946 }
947 
948 llvm::Optional<FormatStyle>
949 FormatStyle::GetLanguageStyle(FormatStyle::LanguageKind Language) const {
950   return StyleSet.Get(Language);
951 }
952 
953 namespace {
954 
955 class JavaScriptRequoter : public TokenAnalyzer {
956 public:
957   JavaScriptRequoter(const Environment &Env, const FormatStyle &Style)
958       : TokenAnalyzer(Env, Style) {}
959 
960   std::pair<tooling::Replacements, unsigned>
961   analyze(TokenAnnotator &Annotator,
962           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
963           FormatTokenLexer &Tokens) override {
964     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
965                                           AnnotatedLines.end());
966     tooling::Replacements Result;
967     requoteJSStringLiteral(AnnotatedLines, Result);
968     return {Result, 0};
969   }
970 
971 private:
972   // Replaces double/single-quoted string literal as appropriate, re-escaping
973   // the contents in the process.
974   void requoteJSStringLiteral(SmallVectorImpl<AnnotatedLine *> &Lines,
975                               tooling::Replacements &Result) {
976     for (AnnotatedLine *Line : Lines) {
977       requoteJSStringLiteral(Line->Children, Result);
978       if (!Line->Affected)
979         continue;
980       for (FormatToken *FormatTok = Line->First; FormatTok;
981            FormatTok = FormatTok->Next) {
982         StringRef Input = FormatTok->TokenText;
983         if (FormatTok->Finalized || !FormatTok->isStringLiteral() ||
984             // NB: testing for not starting with a double quote to avoid
985             // breaking `template strings`.
986             (Style.JavaScriptQuotes == FormatStyle::JSQS_Single &&
987              !Input.startswith("\"")) ||
988             (Style.JavaScriptQuotes == FormatStyle::JSQS_Double &&
989              !Input.startswith("\'")))
990           continue;
991 
992         // Change start and end quote.
993         bool IsSingle = Style.JavaScriptQuotes == FormatStyle::JSQS_Single;
994         SourceLocation Start = FormatTok->Tok.getLocation();
995         auto Replace = [&](SourceLocation Start, unsigned Length,
996                            StringRef ReplacementText) {
997           auto Err = Result.add(tooling::Replacement(
998               Env.getSourceManager(), Start, Length, ReplacementText));
999           // FIXME: handle error. For now, print error message and skip the
1000           // replacement for release version.
1001           if (Err) {
1002             llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1003             assert(false);
1004           }
1005         };
1006         Replace(Start, 1, IsSingle ? "'" : "\"");
1007         Replace(FormatTok->Tok.getEndLoc().getLocWithOffset(-1), 1,
1008                 IsSingle ? "'" : "\"");
1009 
1010         // Escape internal quotes.
1011         bool Escaped = false;
1012         for (size_t i = 1; i < Input.size() - 1; i++) {
1013           switch (Input[i]) {
1014           case '\\':
1015             if (!Escaped && i + 1 < Input.size() &&
1016                 ((IsSingle && Input[i + 1] == '"') ||
1017                  (!IsSingle && Input[i + 1] == '\''))) {
1018               // Remove this \, it's escaping a " or ' that no longer needs
1019               // escaping
1020               Replace(Start.getLocWithOffset(i), 1, "");
1021               continue;
1022             }
1023             Escaped = !Escaped;
1024             break;
1025           case '\"':
1026           case '\'':
1027             if (!Escaped && IsSingle == (Input[i] == '\'')) {
1028               // Escape the quote.
1029               Replace(Start.getLocWithOffset(i), 0, "\\");
1030             }
1031             Escaped = false;
1032             break;
1033           default:
1034             Escaped = false;
1035             break;
1036           }
1037         }
1038       }
1039     }
1040   }
1041 };
1042 
1043 class Formatter : public TokenAnalyzer {
1044 public:
1045   Formatter(const Environment &Env, const FormatStyle &Style,
1046             FormattingAttemptStatus *Status)
1047       : TokenAnalyzer(Env, Style), Status(Status) {}
1048 
1049   std::pair<tooling::Replacements, unsigned>
1050   analyze(TokenAnnotator &Annotator,
1051           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1052           FormatTokenLexer &Tokens) override {
1053     tooling::Replacements Result;
1054     deriveLocalStyle(AnnotatedLines);
1055     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1056                                           AnnotatedLines.end());
1057     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1058       Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
1059     }
1060     Annotator.setCommentLineLevels(AnnotatedLines);
1061 
1062     WhitespaceManager Whitespaces(
1063         Env.getSourceManager(), Style,
1064         inputUsesCRLF(Env.getSourceManager().getBufferData(Env.getFileID())));
1065     ContinuationIndenter Indenter(Style, Tokens.getKeywords(),
1066                                   Env.getSourceManager(), Whitespaces, Encoding,
1067                                   BinPackInconclusiveFunctions);
1068     unsigned Penalty =
1069         UnwrappedLineFormatter(&Indenter, &Whitespaces, Style,
1070                                Tokens.getKeywords(), Env.getSourceManager(),
1071                                Status)
1072             .format(AnnotatedLines, /*DryRun=*/false,
1073                     /*AdditionalIndent=*/0,
1074                     /*FixBadIndentation=*/false,
1075                     /*FirstStartColumn=*/Env.getFirstStartColumn(),
1076                     /*NextStartColumn=*/Env.getNextStartColumn(),
1077                     /*LastStartColumn=*/Env.getLastStartColumn());
1078     for (const auto &R : Whitespaces.generateReplacements())
1079       if (Result.add(R))
1080         return std::make_pair(Result, 0);
1081     return std::make_pair(Result, Penalty);
1082   }
1083 
1084 private:
1085   static bool inputUsesCRLF(StringRef Text) {
1086     return Text.count('\r') * 2 > Text.count('\n');
1087   }
1088 
1089   bool
1090   hasCpp03IncompatibleFormat(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1091     for (const AnnotatedLine *Line : Lines) {
1092       if (hasCpp03IncompatibleFormat(Line->Children))
1093         return true;
1094       for (FormatToken *Tok = Line->First->Next; Tok; Tok = Tok->Next) {
1095         if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
1096           if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
1097             return true;
1098           if (Tok->is(TT_TemplateCloser) &&
1099               Tok->Previous->is(TT_TemplateCloser))
1100             return true;
1101         }
1102       }
1103     }
1104     return false;
1105   }
1106 
1107   int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
1108     int AlignmentDiff = 0;
1109     for (const AnnotatedLine *Line : Lines) {
1110       AlignmentDiff += countVariableAlignments(Line->Children);
1111       for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
1112         if (!Tok->is(TT_PointerOrReference))
1113           continue;
1114         bool SpaceBefore =
1115             Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
1116         bool SpaceAfter = Tok->Next->WhitespaceRange.getBegin() !=
1117                           Tok->Next->WhitespaceRange.getEnd();
1118         if (SpaceBefore && !SpaceAfter)
1119           ++AlignmentDiff;
1120         if (!SpaceBefore && SpaceAfter)
1121           --AlignmentDiff;
1122       }
1123     }
1124     return AlignmentDiff;
1125   }
1126 
1127   void
1128   deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1129     bool HasBinPackedFunction = false;
1130     bool HasOnePerLineFunction = false;
1131     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1132       if (!AnnotatedLines[i]->First->Next)
1133         continue;
1134       FormatToken *Tok = AnnotatedLines[i]->First->Next;
1135       while (Tok->Next) {
1136         if (Tok->PackingKind == PPK_BinPacked)
1137           HasBinPackedFunction = true;
1138         if (Tok->PackingKind == PPK_OnePerLine)
1139           HasOnePerLineFunction = true;
1140 
1141         Tok = Tok->Next;
1142       }
1143     }
1144     if (Style.DerivePointerAlignment)
1145       Style.PointerAlignment = countVariableAlignments(AnnotatedLines) <= 0
1146                                    ? FormatStyle::PAS_Left
1147                                    : FormatStyle::PAS_Right;
1148     if (Style.Standard == FormatStyle::LS_Auto)
1149       Style.Standard = hasCpp03IncompatibleFormat(AnnotatedLines)
1150                            ? FormatStyle::LS_Cpp11
1151                            : FormatStyle::LS_Cpp03;
1152     BinPackInconclusiveFunctions =
1153         HasBinPackedFunction || !HasOnePerLineFunction;
1154   }
1155 
1156   bool BinPackInconclusiveFunctions;
1157   FormattingAttemptStatus *Status;
1158 };
1159 
1160 // This class clean up the erroneous/redundant code around the given ranges in
1161 // file.
1162 class Cleaner : public TokenAnalyzer {
1163 public:
1164   Cleaner(const Environment &Env, const FormatStyle &Style)
1165       : TokenAnalyzer(Env, Style),
1166         DeletedTokens(FormatTokenLess(Env.getSourceManager())) {}
1167 
1168   // FIXME: eliminate unused parameters.
1169   std::pair<tooling::Replacements, unsigned>
1170   analyze(TokenAnnotator &Annotator,
1171           SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1172           FormatTokenLexer &Tokens) override {
1173     // FIXME: in the current implementation the granularity of affected range
1174     // is an annotated line. However, this is not sufficient. Furthermore,
1175     // redundant code introduced by replacements does not necessarily
1176     // intercept with ranges of replacements that result in the redundancy.
1177     // To determine if some redundant code is actually introduced by
1178     // replacements(e.g. deletions), we need to come up with a more
1179     // sophisticated way of computing affected ranges.
1180     AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
1181                                           AnnotatedLines.end());
1182 
1183     checkEmptyNamespace(AnnotatedLines);
1184 
1185     for (auto &Line : AnnotatedLines) {
1186       if (Line->Affected) {
1187         cleanupRight(Line->First, tok::comma, tok::comma);
1188         cleanupRight(Line->First, TT_CtorInitializerColon, tok::comma);
1189         cleanupRight(Line->First, tok::l_paren, tok::comma);
1190         cleanupLeft(Line->First, tok::comma, tok::r_paren);
1191         cleanupLeft(Line->First, TT_CtorInitializerComma, tok::l_brace);
1192         cleanupLeft(Line->First, TT_CtorInitializerColon, tok::l_brace);
1193         cleanupLeft(Line->First, TT_CtorInitializerColon, tok::equal);
1194       }
1195     }
1196 
1197     return {generateFixes(), 0};
1198   }
1199 
1200 private:
1201   bool containsOnlyComments(const AnnotatedLine &Line) {
1202     for (FormatToken *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) {
1203       if (Tok->isNot(tok::comment))
1204         return false;
1205     }
1206     return true;
1207   }
1208 
1209   // Iterate through all lines and remove any empty (nested) namespaces.
1210   void checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1211     std::set<unsigned> DeletedLines;
1212     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1213       auto &Line = *AnnotatedLines[i];
1214       if (Line.startsWith(tok::kw_namespace) ||
1215           Line.startsWith(tok::kw_inline, tok::kw_namespace)) {
1216         checkEmptyNamespace(AnnotatedLines, i, i, DeletedLines);
1217       }
1218     }
1219 
1220     for (auto Line : DeletedLines) {
1221       FormatToken *Tok = AnnotatedLines[Line]->First;
1222       while (Tok) {
1223         deleteToken(Tok);
1224         Tok = Tok->Next;
1225       }
1226     }
1227   }
1228 
1229   // The function checks if the namespace, which starts from \p CurrentLine, and
1230   // its nested namespaces are empty and delete them if they are empty. It also
1231   // sets \p NewLine to the last line checked.
1232   // Returns true if the current namespace is empty.
1233   bool checkEmptyNamespace(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1234                            unsigned CurrentLine, unsigned &NewLine,
1235                            std::set<unsigned> &DeletedLines) {
1236     unsigned InitLine = CurrentLine, End = AnnotatedLines.size();
1237     if (Style.BraceWrapping.AfterNamespace) {
1238       // If the left brace is in a new line, we should consume it first so that
1239       // it does not make the namespace non-empty.
1240       // FIXME: error handling if there is no left brace.
1241       if (!AnnotatedLines[++CurrentLine]->startsWith(tok::l_brace)) {
1242         NewLine = CurrentLine;
1243         return false;
1244       }
1245     } else if (!AnnotatedLines[CurrentLine]->endsWith(tok::l_brace)) {
1246       return false;
1247     }
1248     while (++CurrentLine < End) {
1249       if (AnnotatedLines[CurrentLine]->startsWith(tok::r_brace))
1250         break;
1251 
1252       if (AnnotatedLines[CurrentLine]->startsWith(tok::kw_namespace) ||
1253           AnnotatedLines[CurrentLine]->startsWith(tok::kw_inline,
1254                                                   tok::kw_namespace)) {
1255         if (!checkEmptyNamespace(AnnotatedLines, CurrentLine, NewLine,
1256                                  DeletedLines))
1257           return false;
1258         CurrentLine = NewLine;
1259         continue;
1260       }
1261 
1262       if (containsOnlyComments(*AnnotatedLines[CurrentLine]))
1263         continue;
1264 
1265       // If there is anything other than comments or nested namespaces in the
1266       // current namespace, the namespace cannot be empty.
1267       NewLine = CurrentLine;
1268       return false;
1269     }
1270 
1271     NewLine = CurrentLine;
1272     if (CurrentLine >= End)
1273       return false;
1274 
1275     // Check if the empty namespace is actually affected by changed ranges.
1276     if (!AffectedRangeMgr.affectsCharSourceRange(CharSourceRange::getCharRange(
1277             AnnotatedLines[InitLine]->First->Tok.getLocation(),
1278             AnnotatedLines[CurrentLine]->Last->Tok.getEndLoc())))
1279       return false;
1280 
1281     for (unsigned i = InitLine; i <= CurrentLine; ++i) {
1282       DeletedLines.insert(i);
1283     }
1284 
1285     return true;
1286   }
1287 
1288   // Checks pairs {start, start->next},..., {end->previous, end} and deletes one
1289   // of the token in the pair if the left token has \p LK token kind and the
1290   // right token has \p RK token kind. If \p DeleteLeft is true, the left token
1291   // is deleted on match; otherwise, the right token is deleted.
1292   template <typename LeftKind, typename RightKind>
1293   void cleanupPair(FormatToken *Start, LeftKind LK, RightKind RK,
1294                    bool DeleteLeft) {
1295     auto NextNotDeleted = [this](const FormatToken &Tok) -> FormatToken * {
1296       for (auto *Res = Tok.Next; Res; Res = Res->Next)
1297         if (!Res->is(tok::comment) &&
1298             DeletedTokens.find(Res) == DeletedTokens.end())
1299           return Res;
1300       return nullptr;
1301     };
1302     for (auto *Left = Start; Left;) {
1303       auto *Right = NextNotDeleted(*Left);
1304       if (!Right)
1305         break;
1306       if (Left->is(LK) && Right->is(RK)) {
1307         deleteToken(DeleteLeft ? Left : Right);
1308         for (auto *Tok = Left->Next; Tok && Tok != Right; Tok = Tok->Next)
1309           deleteToken(Tok);
1310         // If the right token is deleted, we should keep the left token
1311         // unchanged and pair it with the new right token.
1312         if (!DeleteLeft)
1313           continue;
1314       }
1315       Left = Right;
1316     }
1317   }
1318 
1319   template <typename LeftKind, typename RightKind>
1320   void cleanupLeft(FormatToken *Start, LeftKind LK, RightKind RK) {
1321     cleanupPair(Start, LK, RK, /*DeleteLeft=*/true);
1322   }
1323 
1324   template <typename LeftKind, typename RightKind>
1325   void cleanupRight(FormatToken *Start, LeftKind LK, RightKind RK) {
1326     cleanupPair(Start, LK, RK, /*DeleteLeft=*/false);
1327   }
1328 
1329   // Delete the given token.
1330   inline void deleteToken(FormatToken *Tok) {
1331     if (Tok)
1332       DeletedTokens.insert(Tok);
1333   }
1334 
1335   tooling::Replacements generateFixes() {
1336     tooling::Replacements Fixes;
1337     std::vector<FormatToken *> Tokens;
1338     std::copy(DeletedTokens.begin(), DeletedTokens.end(),
1339               std::back_inserter(Tokens));
1340 
1341     // Merge multiple continuous token deletions into one big deletion so that
1342     // the number of replacements can be reduced. This makes computing affected
1343     // ranges more efficient when we run reformat on the changed code.
1344     unsigned Idx = 0;
1345     while (Idx < Tokens.size()) {
1346       unsigned St = Idx, End = Idx;
1347       while ((End + 1) < Tokens.size() &&
1348              Tokens[End]->Next == Tokens[End + 1]) {
1349         End++;
1350       }
1351       auto SR = CharSourceRange::getCharRange(Tokens[St]->Tok.getLocation(),
1352                                               Tokens[End]->Tok.getEndLoc());
1353       auto Err =
1354           Fixes.add(tooling::Replacement(Env.getSourceManager(), SR, ""));
1355       // FIXME: better error handling. for now just print error message and skip
1356       // for the release version.
1357       if (Err) {
1358         llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1359         assert(false && "Fixes must not conflict!");
1360       }
1361       Idx = End + 1;
1362     }
1363 
1364     return Fixes;
1365   }
1366 
1367   // Class for less-than inequality comparason for the set `RedundantTokens`.
1368   // We store tokens in the order they appear in the translation unit so that
1369   // we do not need to sort them in `generateFixes()`.
1370   struct FormatTokenLess {
1371     FormatTokenLess(const SourceManager &SM) : SM(SM) {}
1372 
1373     bool operator()(const FormatToken *LHS, const FormatToken *RHS) const {
1374       return SM.isBeforeInTranslationUnit(LHS->Tok.getLocation(),
1375                                           RHS->Tok.getLocation());
1376     }
1377     const SourceManager &SM;
1378   };
1379 
1380   // Tokens to be deleted.
1381   std::set<FormatToken *, FormatTokenLess> DeletedTokens;
1382 };
1383 
1384 struct IncludeDirective {
1385   StringRef Filename;
1386   StringRef Text;
1387   unsigned Offset;
1388   int Category;
1389 };
1390 
1391 } // end anonymous namespace
1392 
1393 // Determines whether 'Ranges' intersects with ('Start', 'End').
1394 static bool affectsRange(ArrayRef<tooling::Range> Ranges, unsigned Start,
1395                          unsigned End) {
1396   for (auto Range : Ranges) {
1397     if (Range.getOffset() < End &&
1398         Range.getOffset() + Range.getLength() > Start)
1399       return true;
1400   }
1401   return false;
1402 }
1403 
1404 // Returns a pair (Index, OffsetToEOL) describing the position of the cursor
1405 // before sorting/deduplicating. Index is the index of the include under the
1406 // cursor in the original set of includes. If this include has duplicates, it is
1407 // the index of the first of the duplicates as the others are going to be
1408 // removed. OffsetToEOL describes the cursor's position relative to the end of
1409 // its current line.
1410 // If `Cursor` is not on any #include, `Index` will be UINT_MAX.
1411 static std::pair<unsigned, unsigned>
1412 FindCursorIndex(const SmallVectorImpl<IncludeDirective> &Includes,
1413                 const SmallVectorImpl<unsigned> &Indices, unsigned Cursor) {
1414   unsigned CursorIndex = UINT_MAX;
1415   unsigned OffsetToEOL = 0;
1416   for (int i = 0, e = Includes.size(); i != e; ++i) {
1417     unsigned Start = Includes[Indices[i]].Offset;
1418     unsigned End = Start + Includes[Indices[i]].Text.size();
1419     if (!(Cursor >= Start && Cursor < End))
1420       continue;
1421     CursorIndex = Indices[i];
1422     OffsetToEOL = End - Cursor;
1423     // Put the cursor on the only remaining #include among the duplicate
1424     // #includes.
1425     while (--i >= 0 && Includes[CursorIndex].Text == Includes[Indices[i]].Text)
1426       CursorIndex = i;
1427     break;
1428   }
1429   return std::make_pair(CursorIndex, OffsetToEOL);
1430 }
1431 
1432 // Sorts and deduplicate a block of includes given by 'Includes' alphabetically
1433 // adding the necessary replacement to 'Replaces'. 'Includes' must be in strict
1434 // source order.
1435 // #include directives with the same text will be deduplicated, and only the
1436 // first #include in the duplicate #includes remains. If the `Cursor` is
1437 // provided and put on a deleted #include, it will be moved to the remaining
1438 // #include in the duplicate #includes.
1439 static void sortCppIncludes(const FormatStyle &Style,
1440                             const SmallVectorImpl<IncludeDirective> &Includes,
1441                             ArrayRef<tooling::Range> Ranges, StringRef FileName,
1442                             tooling::Replacements &Replaces, unsigned *Cursor) {
1443   unsigned IncludesBeginOffset = Includes.front().Offset;
1444   unsigned IncludesEndOffset =
1445       Includes.back().Offset + Includes.back().Text.size();
1446   unsigned IncludesBlockSize = IncludesEndOffset - IncludesBeginOffset;
1447   if (!affectsRange(Ranges, IncludesBeginOffset, IncludesEndOffset))
1448     return;
1449   SmallVector<unsigned, 16> Indices;
1450   for (unsigned i = 0, e = Includes.size(); i != e; ++i)
1451     Indices.push_back(i);
1452   std::stable_sort(
1453       Indices.begin(), Indices.end(), [&](unsigned LHSI, unsigned RHSI) {
1454         return std::tie(Includes[LHSI].Category, Includes[LHSI].Filename) <
1455                std::tie(Includes[RHSI].Category, Includes[RHSI].Filename);
1456       });
1457   // The index of the include on which the cursor will be put after
1458   // sorting/deduplicating.
1459   unsigned CursorIndex;
1460   // The offset from cursor to the end of line.
1461   unsigned CursorToEOLOffset;
1462   if (Cursor)
1463     std::tie(CursorIndex, CursorToEOLOffset) =
1464         FindCursorIndex(Includes, Indices, *Cursor);
1465 
1466   // Deduplicate #includes.
1467   Indices.erase(std::unique(Indices.begin(), Indices.end(),
1468                             [&](unsigned LHSI, unsigned RHSI) {
1469                               return Includes[LHSI].Text == Includes[RHSI].Text;
1470                             }),
1471                 Indices.end());
1472 
1473   int CurrentCategory = Includes.front().Category;
1474 
1475   // If the #includes are out of order, we generate a single replacement fixing
1476   // the entire block. Otherwise, no replacement is generated.
1477   if (Indices.size() == Includes.size() &&
1478       std::is_sorted(Indices.begin(), Indices.end()) &&
1479       Style.IncludeBlocks == FormatStyle::IBS_Preserve)
1480     return;
1481 
1482   std::string result;
1483   for (unsigned Index : Indices) {
1484     if (!result.empty()) {
1485       result += "\n";
1486       if (Style.IncludeBlocks == FormatStyle::IBS_Regroup &&
1487           CurrentCategory != Includes[Index].Category)
1488         result += "\n";
1489     }
1490     result += Includes[Index].Text;
1491     if (Cursor && CursorIndex == Index)
1492       *Cursor = IncludesBeginOffset + result.size() - CursorToEOLOffset;
1493     CurrentCategory = Includes[Index].Category;
1494   }
1495 
1496   auto Err = Replaces.add(tooling::Replacement(
1497       FileName, Includes.front().Offset, IncludesBlockSize, result));
1498   // FIXME: better error handling. For now, just skip the replacement for the
1499   // release version.
1500   if (Err) {
1501     llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1502     assert(false);
1503   }
1504 }
1505 
1506 namespace {
1507 
1508 // This class manages priorities of #include categories and calculates
1509 // priorities for headers.
1510 class IncludeCategoryManager {
1511 public:
1512   IncludeCategoryManager(const FormatStyle &Style, StringRef FileName)
1513       : Style(Style), FileName(FileName) {
1514     FileStem = llvm::sys::path::stem(FileName);
1515     for (const auto &Category : Style.IncludeCategories)
1516       CategoryRegexs.emplace_back(Category.Regex, llvm::Regex::IgnoreCase);
1517     IsMainFile = FileName.endswith(".c") || FileName.endswith(".cc") ||
1518                  FileName.endswith(".cpp") || FileName.endswith(".c++") ||
1519                  FileName.endswith(".cxx") || FileName.endswith(".m") ||
1520                  FileName.endswith(".mm");
1521   }
1522 
1523   // Returns the priority of the category which \p IncludeName belongs to.
1524   // If \p CheckMainHeader is true and \p IncludeName is a main header, returns
1525   // 0. Otherwise, returns the priority of the matching category or INT_MAX.
1526   int getIncludePriority(StringRef IncludeName, bool CheckMainHeader) {
1527     int Ret = INT_MAX;
1528     for (unsigned i = 0, e = CategoryRegexs.size(); i != e; ++i)
1529       if (CategoryRegexs[i].match(IncludeName)) {
1530         Ret = Style.IncludeCategories[i].Priority;
1531         break;
1532       }
1533     if (CheckMainHeader && IsMainFile && Ret > 0 && isMainHeader(IncludeName))
1534       Ret = 0;
1535     return Ret;
1536   }
1537 
1538 private:
1539   bool isMainHeader(StringRef IncludeName) const {
1540     if (!IncludeName.startswith("\""))
1541       return false;
1542     StringRef HeaderStem =
1543         llvm::sys::path::stem(IncludeName.drop_front(1).drop_back(1));
1544     if (FileStem.startswith(HeaderStem) ||
1545         FileStem.startswith_lower(HeaderStem)) {
1546       llvm::Regex MainIncludeRegex(
1547           (HeaderStem + Style.IncludeIsMainRegex).str(),
1548           llvm::Regex::IgnoreCase);
1549       if (MainIncludeRegex.match(FileStem))
1550         return true;
1551     }
1552     return false;
1553   }
1554 
1555   const FormatStyle &Style;
1556   bool IsMainFile;
1557   StringRef FileName;
1558   StringRef FileStem;
1559   SmallVector<llvm::Regex, 4> CategoryRegexs;
1560 };
1561 
1562 const char IncludeRegexPattern[] =
1563     R"(^[\t\ ]*#[\t\ ]*(import|include)[^"<]*(["<][^">]*[">]))";
1564 
1565 } // anonymous namespace
1566 
1567 tooling::Replacements sortCppIncludes(const FormatStyle &Style, StringRef Code,
1568                                       ArrayRef<tooling::Range> Ranges,
1569                                       StringRef FileName,
1570                                       tooling::Replacements &Replaces,
1571                                       unsigned *Cursor) {
1572   unsigned Prev = 0;
1573   unsigned SearchFrom = 0;
1574   llvm::Regex IncludeRegex(IncludeRegexPattern);
1575   SmallVector<StringRef, 4> Matches;
1576   SmallVector<IncludeDirective, 16> IncludesInBlock;
1577 
1578   // In compiled files, consider the first #include to be the main #include of
1579   // the file if it is not a system #include. This ensures that the header
1580   // doesn't have hidden dependencies
1581   // (http://llvm.org/docs/CodingStandards.html#include-style).
1582   //
1583   // FIXME: Do some sanity checking, e.g. edit distance of the base name, to fix
1584   // cases where the first #include is unlikely to be the main header.
1585   IncludeCategoryManager Categories(Style, FileName);
1586   bool FirstIncludeBlock = true;
1587   bool MainIncludeFound = false;
1588   bool FormattingOff = false;
1589 
1590   for (;;) {
1591     auto Pos = Code.find('\n', SearchFrom);
1592     StringRef Line =
1593         Code.substr(Prev, (Pos != StringRef::npos ? Pos : Code.size()) - Prev);
1594 
1595     StringRef Trimmed = Line.trim();
1596     if (Trimmed == "// clang-format off")
1597       FormattingOff = true;
1598     else if (Trimmed == "// clang-format on")
1599       FormattingOff = false;
1600 
1601     const bool EmptyLineSkipped =
1602         Trimmed.empty() && (Style.IncludeBlocks == FormatStyle::IBS_Merge ||
1603                             Style.IncludeBlocks == FormatStyle::IBS_Regroup);
1604 
1605     if (!FormattingOff && !Line.endswith("\\")) {
1606       if (IncludeRegex.match(Line, &Matches)) {
1607         StringRef IncludeName = Matches[2];
1608         int Category = Categories.getIncludePriority(
1609             IncludeName,
1610             /*CheckMainHeader=*/!MainIncludeFound && FirstIncludeBlock);
1611         if (Category == 0)
1612           MainIncludeFound = true;
1613         IncludesInBlock.push_back({IncludeName, Line, Prev, Category});
1614       } else if (!IncludesInBlock.empty() && !EmptyLineSkipped) {
1615         sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces,
1616                         Cursor);
1617         IncludesInBlock.clear();
1618         FirstIncludeBlock = false;
1619       }
1620       Prev = Pos + 1;
1621     }
1622     if (Pos == StringRef::npos || Pos + 1 == Code.size())
1623       break;
1624     SearchFrom = Pos + 1;
1625   }
1626   if (!IncludesInBlock.empty())
1627     sortCppIncludes(Style, IncludesInBlock, Ranges, FileName, Replaces, Cursor);
1628   return Replaces;
1629 }
1630 
1631 bool isMpegTS(StringRef Code) {
1632   // MPEG transport streams use the ".ts" file extension. clang-format should
1633   // not attempt to format those. MPEG TS' frame format starts with 0x47 every
1634   // 189 bytes - detect that and return.
1635   return Code.size() > 188 && Code[0] == 0x47 && Code[188] == 0x47;
1636 }
1637 
1638 bool isLikelyXml(StringRef Code) { return Code.ltrim().startswith("<"); }
1639 
1640 tooling::Replacements sortIncludes(const FormatStyle &Style, StringRef Code,
1641                                    ArrayRef<tooling::Range> Ranges,
1642                                    StringRef FileName, unsigned *Cursor) {
1643   tooling::Replacements Replaces;
1644   if (!Style.SortIncludes)
1645     return Replaces;
1646   if (isLikelyXml(Code))
1647     return Replaces;
1648   if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript &&
1649       isMpegTS(Code))
1650     return Replaces;
1651   if (Style.Language == FormatStyle::LanguageKind::LK_JavaScript)
1652     return sortJavaScriptImports(Style, Code, Ranges, FileName);
1653   sortCppIncludes(Style, Code, Ranges, FileName, Replaces, Cursor);
1654   return Replaces;
1655 }
1656 
1657 template <typename T>
1658 static llvm::Expected<tooling::Replacements>
1659 processReplacements(T ProcessFunc, StringRef Code,
1660                     const tooling::Replacements &Replaces,
1661                     const FormatStyle &Style) {
1662   if (Replaces.empty())
1663     return tooling::Replacements();
1664 
1665   auto NewCode = applyAllReplacements(Code, Replaces);
1666   if (!NewCode)
1667     return NewCode.takeError();
1668   std::vector<tooling::Range> ChangedRanges = Replaces.getAffectedRanges();
1669   StringRef FileName = Replaces.begin()->getFilePath();
1670 
1671   tooling::Replacements FormatReplaces =
1672       ProcessFunc(Style, *NewCode, ChangedRanges, FileName);
1673 
1674   return Replaces.merge(FormatReplaces);
1675 }
1676 
1677 llvm::Expected<tooling::Replacements>
1678 formatReplacements(StringRef Code, const tooling::Replacements &Replaces,
1679                    const FormatStyle &Style) {
1680   // We need to use lambda function here since there are two versions of
1681   // `sortIncludes`.
1682   auto SortIncludes = [](const FormatStyle &Style, StringRef Code,
1683                          std::vector<tooling::Range> Ranges,
1684                          StringRef FileName) -> tooling::Replacements {
1685     return sortIncludes(Style, Code, Ranges, FileName);
1686   };
1687   auto SortedReplaces =
1688       processReplacements(SortIncludes, Code, Replaces, Style);
1689   if (!SortedReplaces)
1690     return SortedReplaces.takeError();
1691 
1692   // We need to use lambda function here since there are two versions of
1693   // `reformat`.
1694   auto Reformat = [](const FormatStyle &Style, StringRef Code,
1695                      std::vector<tooling::Range> Ranges,
1696                      StringRef FileName) -> tooling::Replacements {
1697     return reformat(Style, Code, Ranges, FileName);
1698   };
1699   return processReplacements(Reformat, Code, *SortedReplaces, Style);
1700 }
1701 
1702 namespace {
1703 
1704 inline bool isHeaderInsertion(const tooling::Replacement &Replace) {
1705   return Replace.getOffset() == UINT_MAX && Replace.getLength() == 0 &&
1706          llvm::Regex(IncludeRegexPattern).match(Replace.getReplacementText());
1707 }
1708 
1709 inline bool isHeaderDeletion(const tooling::Replacement &Replace) {
1710   return Replace.getOffset() == UINT_MAX && Replace.getLength() == 1;
1711 }
1712 
1713 // Returns the offset after skipping a sequence of tokens, matched by \p
1714 // GetOffsetAfterSequence, from the start of the code.
1715 // \p GetOffsetAfterSequence should be a function that matches a sequence of
1716 // tokens and returns an offset after the sequence.
1717 unsigned getOffsetAfterTokenSequence(
1718     StringRef FileName, StringRef Code, const FormatStyle &Style,
1719     llvm::function_ref<unsigned(const SourceManager &, Lexer &, Token &)>
1720         GetOffsetAfterSequence) {
1721   std::unique_ptr<Environment> Env =
1722       Environment::CreateVirtualEnvironment(Code, FileName, /*Ranges=*/{});
1723   const SourceManager &SourceMgr = Env->getSourceManager();
1724   Lexer Lex(Env->getFileID(), SourceMgr.getBuffer(Env->getFileID()), SourceMgr,
1725             getFormattingLangOpts(Style));
1726   Token Tok;
1727   // Get the first token.
1728   Lex.LexFromRawLexer(Tok);
1729   return GetOffsetAfterSequence(SourceMgr, Lex, Tok);
1730 }
1731 
1732 // Check if a sequence of tokens is like "#<Name> <raw_identifier>". If it is,
1733 // \p Tok will be the token after this directive; otherwise, it can be any token
1734 // after the given \p Tok (including \p Tok).
1735 bool checkAndConsumeDirectiveWithName(Lexer &Lex, StringRef Name, Token &Tok) {
1736   bool Matched = Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1737                  Tok.is(tok::raw_identifier) &&
1738                  Tok.getRawIdentifier() == Name && !Lex.LexFromRawLexer(Tok) &&
1739                  Tok.is(tok::raw_identifier);
1740   if (Matched)
1741     Lex.LexFromRawLexer(Tok);
1742   return Matched;
1743 }
1744 
1745 void skipComments(Lexer &Lex, Token &Tok) {
1746   while (Tok.is(tok::comment))
1747     if (Lex.LexFromRawLexer(Tok))
1748       return;
1749 }
1750 
1751 // Returns the offset after header guard directives and any comments
1752 // before/after header guards. If no header guard presents in the code, this
1753 // will returns the offset after skipping all comments from the start of the
1754 // code.
1755 unsigned getOffsetAfterHeaderGuardsAndComments(StringRef FileName,
1756                                                StringRef Code,
1757                                                const FormatStyle &Style) {
1758   return getOffsetAfterTokenSequence(
1759       FileName, Code, Style,
1760       [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1761         skipComments(Lex, Tok);
1762         unsigned InitialOffset = SM.getFileOffset(Tok.getLocation());
1763         if (checkAndConsumeDirectiveWithName(Lex, "ifndef", Tok)) {
1764           skipComments(Lex, Tok);
1765           if (checkAndConsumeDirectiveWithName(Lex, "define", Tok))
1766             return SM.getFileOffset(Tok.getLocation());
1767         }
1768         return InitialOffset;
1769       });
1770 }
1771 
1772 // Check if a sequence of tokens is like
1773 //    "#include ("header.h" | <header.h>)".
1774 // If it is, \p Tok will be the token after this directive; otherwise, it can be
1775 // any token after the given \p Tok (including \p Tok).
1776 bool checkAndConsumeInclusiveDirective(Lexer &Lex, Token &Tok) {
1777   auto Matched = [&]() {
1778     Lex.LexFromRawLexer(Tok);
1779     return true;
1780   };
1781   if (Tok.is(tok::hash) && !Lex.LexFromRawLexer(Tok) &&
1782       Tok.is(tok::raw_identifier) && Tok.getRawIdentifier() == "include") {
1783     if (Lex.LexFromRawLexer(Tok))
1784       return false;
1785     if (Tok.is(tok::string_literal))
1786       return Matched();
1787     if (Tok.is(tok::less)) {
1788       while (!Lex.LexFromRawLexer(Tok) && Tok.isNot(tok::greater)) {
1789       }
1790       if (Tok.is(tok::greater))
1791         return Matched();
1792     }
1793   }
1794   return false;
1795 }
1796 
1797 // Returns the offset of the last #include directive after which a new
1798 // #include can be inserted. This ignores #include's after the #include block(s)
1799 // in the beginning of a file to avoid inserting headers into code sections
1800 // where new #include's should not be added by default.
1801 // These code sections include:
1802 //      - raw string literals (containing #include).
1803 //      - #if blocks.
1804 //      - Special #include's among declarations (e.g. functions).
1805 //
1806 // If no #include after which a new #include can be inserted, this returns the
1807 // offset after skipping all comments from the start of the code.
1808 // Inserting after an #include is not allowed if it comes after code that is not
1809 // #include (e.g. pre-processing directive that is not #include, declarations).
1810 unsigned getMaxHeaderInsertionOffset(StringRef FileName, StringRef Code,
1811                                      const FormatStyle &Style) {
1812   return getOffsetAfterTokenSequence(
1813       FileName, Code, Style,
1814       [](const SourceManager &SM, Lexer &Lex, Token Tok) {
1815         skipComments(Lex, Tok);
1816         unsigned MaxOffset = SM.getFileOffset(Tok.getLocation());
1817         while (checkAndConsumeInclusiveDirective(Lex, Tok))
1818           MaxOffset = SM.getFileOffset(Tok.getLocation());
1819         return MaxOffset;
1820       });
1821 }
1822 
1823 bool isDeletedHeader(llvm::StringRef HeaderName,
1824                      const std::set<llvm::StringRef> &HeadersToDelete) {
1825   return HeadersToDelete.count(HeaderName) ||
1826          HeadersToDelete.count(HeaderName.trim("\"<>"));
1827 }
1828 
1829 // FIXME: insert empty lines between newly created blocks.
1830 tooling::Replacements
1831 fixCppIncludeInsertions(StringRef Code, const tooling::Replacements &Replaces,
1832                         const FormatStyle &Style) {
1833   if (!Style.isCpp())
1834     return Replaces;
1835 
1836   tooling::Replacements HeaderInsertions;
1837   std::set<llvm::StringRef> HeadersToDelete;
1838   tooling::Replacements Result;
1839   for (const auto &R : Replaces) {
1840     if (isHeaderInsertion(R)) {
1841       // Replacements from \p Replaces must be conflict-free already, so we can
1842       // simply consume the error.
1843       llvm::consumeError(HeaderInsertions.add(R));
1844     } else if (isHeaderDeletion(R)) {
1845       HeadersToDelete.insert(R.getReplacementText());
1846     } else if (R.getOffset() == UINT_MAX) {
1847       llvm::errs() << "Insertions other than header #include insertion are "
1848                       "not supported! "
1849                    << R.getReplacementText() << "\n";
1850     } else {
1851       llvm::consumeError(Result.add(R));
1852     }
1853   }
1854   if (HeaderInsertions.empty() && HeadersToDelete.empty())
1855     return Replaces;
1856 
1857   llvm::Regex IncludeRegex(IncludeRegexPattern);
1858   llvm::Regex DefineRegex(R"(^[\t\ ]*#[\t\ ]*define[\t\ ]*[^\\]*$)");
1859   SmallVector<StringRef, 4> Matches;
1860 
1861   StringRef FileName = Replaces.begin()->getFilePath();
1862   IncludeCategoryManager Categories(Style, FileName);
1863 
1864   // Record the offset of the end of the last include in each category.
1865   std::map<int, int> CategoryEndOffsets;
1866   // All possible priorities.
1867   // Add 0 for main header and INT_MAX for headers that are not in any category.
1868   std::set<int> Priorities = {0, INT_MAX};
1869   for (const auto &Category : Style.IncludeCategories)
1870     Priorities.insert(Category.Priority);
1871   int FirstIncludeOffset = -1;
1872   // All new headers should be inserted after this offset.
1873   unsigned MinInsertOffset =
1874       getOffsetAfterHeaderGuardsAndComments(FileName, Code, Style);
1875   StringRef TrimmedCode = Code.drop_front(MinInsertOffset);
1876   // Max insertion offset in the original code.
1877   unsigned MaxInsertOffset =
1878       MinInsertOffset +
1879       getMaxHeaderInsertionOffset(FileName, TrimmedCode, Style);
1880   SmallVector<StringRef, 32> Lines;
1881   TrimmedCode.split(Lines, '\n');
1882   unsigned Offset = MinInsertOffset;
1883   unsigned NextLineOffset;
1884   std::set<StringRef> ExistingIncludes;
1885   for (auto Line : Lines) {
1886     NextLineOffset = std::min(Code.size(), Offset + Line.size() + 1);
1887     if (IncludeRegex.match(Line, &Matches)) {
1888       // The header name with quotes or angle brackets.
1889       StringRef IncludeName = Matches[2];
1890       ExistingIncludes.insert(IncludeName);
1891       // Only record the offset of current #include if we can insert after it.
1892       if (Offset <= MaxInsertOffset) {
1893         int Category = Categories.getIncludePriority(
1894             IncludeName, /*CheckMainHeader=*/FirstIncludeOffset < 0);
1895         CategoryEndOffsets[Category] = NextLineOffset;
1896         if (FirstIncludeOffset < 0)
1897           FirstIncludeOffset = Offset;
1898       }
1899       if (isDeletedHeader(IncludeName, HeadersToDelete)) {
1900         // If this is the last line without trailing newline, we need to make
1901         // sure we don't delete across the file boundary.
1902         unsigned Length = std::min(Line.size() + 1, Code.size() - Offset);
1903         llvm::Error Err =
1904             Result.add(tooling::Replacement(FileName, Offset, Length, ""));
1905         if (Err) {
1906           // Ignore the deletion on conflict.
1907           llvm::errs() << "Failed to add header deletion replacement for "
1908                        << IncludeName << ": " << llvm::toString(std::move(Err))
1909                        << "\n";
1910         }
1911       }
1912     }
1913     Offset = NextLineOffset;
1914   }
1915 
1916   // Populate CategoryEndOfssets:
1917   // - Ensure that CategoryEndOffset[Highest] is always populated.
1918   // - If CategoryEndOffset[Priority] isn't set, use the next higher value that
1919   //   is set, up to CategoryEndOffset[Highest].
1920   auto Highest = Priorities.begin();
1921   if (CategoryEndOffsets.find(*Highest) == CategoryEndOffsets.end()) {
1922     if (FirstIncludeOffset >= 0)
1923       CategoryEndOffsets[*Highest] = FirstIncludeOffset;
1924     else
1925       CategoryEndOffsets[*Highest] = MinInsertOffset;
1926   }
1927   // By this point, CategoryEndOffset[Highest] is always set appropriately:
1928   //  - to an appropriate location before/after existing #includes, or
1929   //  - to right after the header guard, or
1930   //  - to the beginning of the file.
1931   for (auto I = ++Priorities.begin(), E = Priorities.end(); I != E; ++I)
1932     if (CategoryEndOffsets.find(*I) == CategoryEndOffsets.end())
1933       CategoryEndOffsets[*I] = CategoryEndOffsets[*std::prev(I)];
1934 
1935   bool NeedNewLineAtEnd = !Code.empty() && Code.back() != '\n';
1936   for (const auto &R : HeaderInsertions) {
1937     auto IncludeDirective = R.getReplacementText();
1938     bool Matched = IncludeRegex.match(IncludeDirective, &Matches);
1939     assert(Matched && "Header insertion replacement must have replacement text "
1940                       "'#include ...'");
1941     (void)Matched;
1942     auto IncludeName = Matches[2];
1943     if (ExistingIncludes.find(IncludeName) != ExistingIncludes.end()) {
1944       DEBUG(llvm::dbgs() << "Skip adding existing include : " << IncludeName
1945                          << "\n");
1946       continue;
1947     }
1948     int Category =
1949         Categories.getIncludePriority(IncludeName, /*CheckMainHeader=*/true);
1950     Offset = CategoryEndOffsets[Category];
1951     std::string NewInclude = !IncludeDirective.endswith("\n")
1952                                  ? (IncludeDirective + "\n").str()
1953                                  : IncludeDirective.str();
1954     // When inserting headers at end of the code, also append '\n' to the code
1955     // if it does not end with '\n'.
1956     if (NeedNewLineAtEnd && Offset == Code.size()) {
1957       NewInclude = "\n" + NewInclude;
1958       NeedNewLineAtEnd = false;
1959     }
1960     auto NewReplace = tooling::Replacement(FileName, Offset, 0, NewInclude);
1961     auto Err = Result.add(NewReplace);
1962     if (Err) {
1963       llvm::consumeError(std::move(Err));
1964       unsigned NewOffset = Result.getShiftedCodePosition(Offset);
1965       NewReplace = tooling::Replacement(FileName, NewOffset, 0, NewInclude);
1966       Result = Result.merge(tooling::Replacements(NewReplace));
1967     }
1968   }
1969   return Result;
1970 }
1971 
1972 } // anonymous namespace
1973 
1974 llvm::Expected<tooling::Replacements>
1975 cleanupAroundReplacements(StringRef Code, const tooling::Replacements &Replaces,
1976                           const FormatStyle &Style) {
1977   // We need to use lambda function here since there are two versions of
1978   // `cleanup`.
1979   auto Cleanup = [](const FormatStyle &Style, StringRef Code,
1980                     std::vector<tooling::Range> Ranges,
1981                     StringRef FileName) -> tooling::Replacements {
1982     return cleanup(Style, Code, Ranges, FileName);
1983   };
1984   // Make header insertion replacements insert new headers into correct blocks.
1985   tooling::Replacements NewReplaces =
1986       fixCppIncludeInsertions(Code, Replaces, Style);
1987   return processReplacements(Cleanup, Code, NewReplaces, Style);
1988 }
1989 
1990 namespace internal {
1991 std::pair<tooling::Replacements, unsigned>
1992 reformat(const FormatStyle &Style, StringRef Code,
1993          ArrayRef<tooling::Range> Ranges, unsigned FirstStartColumn,
1994          unsigned NextStartColumn, unsigned LastStartColumn, StringRef FileName,
1995          FormattingAttemptStatus *Status) {
1996   FormatStyle Expanded = expandPresets(Style);
1997   if (Expanded.DisableFormat)
1998     return {tooling::Replacements(), 0};
1999   if (isLikelyXml(Code))
2000     return {tooling::Replacements(), 0};
2001   if (Expanded.Language == FormatStyle::LK_JavaScript && isMpegTS(Code))
2002     return {tooling::Replacements(), 0};
2003 
2004   typedef std::function<std::pair<tooling::Replacements, unsigned>(
2005       const Environment &)>
2006       AnalyzerPass;
2007   SmallVector<AnalyzerPass, 4> Passes;
2008 
2009   if (Style.Language == FormatStyle::LK_Cpp) {
2010     if (Style.FixNamespaceComments)
2011       Passes.emplace_back([&](const Environment &Env) {
2012         return NamespaceEndCommentsFixer(Env, Expanded).process();
2013       });
2014 
2015     if (Style.SortUsingDeclarations)
2016       Passes.emplace_back([&](const Environment &Env) {
2017         return UsingDeclarationsSorter(Env, Expanded).process();
2018       });
2019   }
2020 
2021   if (Style.Language == FormatStyle::LK_JavaScript &&
2022       Style.JavaScriptQuotes != FormatStyle::JSQS_Leave)
2023     Passes.emplace_back([&](const Environment &Env) {
2024       return JavaScriptRequoter(Env, Expanded).process();
2025     });
2026 
2027   Passes.emplace_back([&](const Environment &Env) {
2028     return Formatter(Env, Expanded, Status).process();
2029   });
2030 
2031   std::unique_ptr<Environment> Env = Environment::CreateVirtualEnvironment(
2032       Code, FileName, Ranges, FirstStartColumn, NextStartColumn,
2033       LastStartColumn);
2034   llvm::Optional<std::string> CurrentCode = None;
2035   tooling::Replacements Fixes;
2036   unsigned Penalty = 0;
2037   for (size_t I = 0, E = Passes.size(); I < E; ++I) {
2038     std::pair<tooling::Replacements, unsigned> PassFixes = Passes[I](*Env);
2039     auto NewCode = applyAllReplacements(
2040         CurrentCode ? StringRef(*CurrentCode) : Code, PassFixes.first);
2041     if (NewCode) {
2042       Fixes = Fixes.merge(PassFixes.first);
2043       Penalty += PassFixes.second;
2044       if (I + 1 < E) {
2045         CurrentCode = std::move(*NewCode);
2046         Env = Environment::CreateVirtualEnvironment(
2047             *CurrentCode, FileName,
2048             tooling::calculateRangesAfterReplacements(Fixes, Ranges),
2049             FirstStartColumn, NextStartColumn, LastStartColumn);
2050       }
2051     }
2052   }
2053 
2054   return {Fixes, Penalty};
2055 }
2056 } // namespace internal
2057 
2058 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
2059                                ArrayRef<tooling::Range> Ranges,
2060                                StringRef FileName,
2061                                FormattingAttemptStatus *Status) {
2062   return internal::reformat(Style, Code, Ranges,
2063                             /*FirstStartColumn=*/0,
2064                             /*NextStartColumn=*/0,
2065                             /*LastStartColumn=*/0, FileName, Status)
2066       .first;
2067 }
2068 
2069 tooling::Replacements cleanup(const FormatStyle &Style, StringRef Code,
2070                               ArrayRef<tooling::Range> Ranges,
2071                               StringRef FileName) {
2072   // cleanups only apply to C++ (they mostly concern ctor commas etc.)
2073   if (Style.Language != FormatStyle::LK_Cpp)
2074     return tooling::Replacements();
2075   std::unique_ptr<Environment> Env =
2076       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
2077   Cleaner Clean(*Env, Style);
2078   return Clean.process().first;
2079 }
2080 
2081 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
2082                                ArrayRef<tooling::Range> Ranges,
2083                                StringRef FileName, bool *IncompleteFormat) {
2084   FormattingAttemptStatus Status;
2085   auto Result = reformat(Style, Code, Ranges, FileName, &Status);
2086   if (!Status.FormatComplete)
2087     *IncompleteFormat = true;
2088   return Result;
2089 }
2090 
2091 tooling::Replacements fixNamespaceEndComments(const FormatStyle &Style,
2092                                               StringRef Code,
2093                                               ArrayRef<tooling::Range> Ranges,
2094                                               StringRef FileName) {
2095   std::unique_ptr<Environment> Env =
2096       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
2097   NamespaceEndCommentsFixer Fix(*Env, Style);
2098   return Fix.process().first;
2099 }
2100 
2101 tooling::Replacements sortUsingDeclarations(const FormatStyle &Style,
2102                                             StringRef Code,
2103                                             ArrayRef<tooling::Range> Ranges,
2104                                             StringRef FileName) {
2105   std::unique_ptr<Environment> Env =
2106       Environment::CreateVirtualEnvironment(Code, FileName, Ranges);
2107   UsingDeclarationsSorter Sorter(*Env, Style);
2108   return Sorter.process().first;
2109 }
2110 
2111 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
2112   LangOptions LangOpts;
2113   LangOpts.CPlusPlus = 1;
2114   LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2115   LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2116   LangOpts.CPlusPlus17 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2117   LangOpts.CPlusPlus2a = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
2118   LangOpts.LineComment = 1;
2119   bool AlternativeOperators = Style.isCpp();
2120   LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
2121   LangOpts.Bool = 1;
2122   LangOpts.ObjC1 = 1;
2123   LangOpts.ObjC2 = 1;
2124   LangOpts.MicrosoftExt = 1;    // To get kw___try, kw___finally.
2125   LangOpts.DeclSpecKeyword = 1; // To get __declspec.
2126   return LangOpts;
2127 }
2128 
2129 const char *StyleOptionHelpDescription =
2130     "Coding style, currently supports:\n"
2131     "  LLVM, Google, Chromium, Mozilla, WebKit.\n"
2132     "Use -style=file to load style configuration from\n"
2133     ".clang-format file located in one of the parent\n"
2134     "directories of the source file (or current\n"
2135     "directory for stdin).\n"
2136     "Use -style=\"{key: value, ...}\" to set specific\n"
2137     "parameters, e.g.:\n"
2138     "  -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
2139 
2140 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
2141   if (FileName.endswith(".java"))
2142     return FormatStyle::LK_Java;
2143   if (FileName.endswith_lower(".js") || FileName.endswith_lower(".ts"))
2144     return FormatStyle::LK_JavaScript; // JavaScript or TypeScript.
2145   if (FileName.endswith(".m") || FileName.endswith(".mm"))
2146     return FormatStyle::LK_ObjC;
2147   if (FileName.endswith_lower(".proto") ||
2148       FileName.endswith_lower(".protodevel"))
2149     return FormatStyle::LK_Proto;
2150   if (FileName.endswith_lower(".textpb") ||
2151       FileName.endswith_lower(".pb.txt") ||
2152       FileName.endswith_lower(".textproto") ||
2153       FileName.endswith_lower(".asciipb"))
2154     return FormatStyle::LK_TextProto;
2155   if (FileName.endswith_lower(".td"))
2156     return FormatStyle::LK_TableGen;
2157   return FormatStyle::LK_Cpp;
2158 }
2159 
2160 llvm::Expected<FormatStyle> getStyle(StringRef StyleName, StringRef FileName,
2161                                      StringRef FallbackStyleName,
2162                                      StringRef Code, vfs::FileSystem *FS) {
2163   if (!FS) {
2164     FS = vfs::getRealFileSystem().get();
2165   }
2166   FormatStyle Style = getLLVMStyle();
2167   Style.Language = getLanguageByFileName(FileName);
2168 
2169   // This is a very crude detection of whether a header contains ObjC code that
2170   // should be improved over time and probably be done on tokens, not one the
2171   // bare content of the file.
2172   if (Style.Language == FormatStyle::LK_Cpp && FileName.endswith(".h") &&
2173       (Code.contains("\n- (") || Code.contains("\n+ (") ||
2174        Code.contains("\n@end\n") || Code.contains("\n@end ") ||
2175        Code.endswith("@end")))
2176     Style.Language = FormatStyle::LK_ObjC;
2177 
2178   FormatStyle FallbackStyle = getNoStyle();
2179   if (!getPredefinedStyle(FallbackStyleName, Style.Language, &FallbackStyle))
2180     return make_string_error("Invalid fallback style \"" + FallbackStyleName);
2181 
2182   if (StyleName.startswith("{")) {
2183     // Parse YAML/JSON style from the command line.
2184     if (std::error_code ec = parseConfiguration(StyleName, &Style))
2185       return make_string_error("Error parsing -style: " + ec.message());
2186     return Style;
2187   }
2188 
2189   if (!StyleName.equals_lower("file")) {
2190     if (!getPredefinedStyle(StyleName, Style.Language, &Style))
2191       return make_string_error("Invalid value for -style");
2192     return Style;
2193   }
2194 
2195   // Look for .clang-format/_clang-format file in the file's parent directories.
2196   SmallString<128> UnsuitableConfigFiles;
2197   SmallString<128> Path(FileName);
2198   if (std::error_code EC = FS->makeAbsolute(Path))
2199     return make_string_error(EC.message());
2200 
2201   for (StringRef Directory = Path; !Directory.empty();
2202        Directory = llvm::sys::path::parent_path(Directory)) {
2203 
2204     auto Status = FS->status(Directory);
2205     if (!Status ||
2206         Status->getType() != llvm::sys::fs::file_type::directory_file) {
2207       continue;
2208     }
2209 
2210     SmallString<128> ConfigFile(Directory);
2211 
2212     llvm::sys::path::append(ConfigFile, ".clang-format");
2213     DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2214 
2215     Status = FS->status(ConfigFile.str());
2216     bool FoundConfigFile =
2217         Status && (Status->getType() == llvm::sys::fs::file_type::regular_file);
2218     if (!FoundConfigFile) {
2219       // Try _clang-format too, since dotfiles are not commonly used on Windows.
2220       ConfigFile = Directory;
2221       llvm::sys::path::append(ConfigFile, "_clang-format");
2222       DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
2223       Status = FS->status(ConfigFile.str());
2224       FoundConfigFile = Status && (Status->getType() ==
2225                                    llvm::sys::fs::file_type::regular_file);
2226     }
2227 
2228     if (FoundConfigFile) {
2229       llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
2230           FS->getBufferForFile(ConfigFile.str());
2231       if (std::error_code EC = Text.getError())
2232         return make_string_error(EC.message());
2233       if (std::error_code ec =
2234               parseConfiguration(Text.get()->getBuffer(), &Style)) {
2235         if (ec == ParseError::Unsuitable) {
2236           if (!UnsuitableConfigFiles.empty())
2237             UnsuitableConfigFiles.append(", ");
2238           UnsuitableConfigFiles.append(ConfigFile);
2239           continue;
2240         }
2241         return make_string_error("Error reading " + ConfigFile + ": " +
2242                                  ec.message());
2243       }
2244       DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
2245       return Style;
2246     }
2247   }
2248   if (!UnsuitableConfigFiles.empty())
2249     return make_string_error("Configuration file(s) do(es) not support " +
2250                              getLanguageName(Style.Language) + ": " +
2251                              UnsuitableConfigFiles);
2252   return FallbackStyle;
2253 }
2254 
2255 } // namespace format
2256 } // namespace clang
2257