xref: /minix3/external/bsd/llvm/dist/clang/lib/Format/Format.cpp (revision 5d5fbe79c1b60734f34c69330aec5496644e8651) 
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 "ContinuationIndenter.h"
17 #include "TokenAnnotator.h"
18 #include "UnwrappedLineFormatter.h"
19 #include "UnwrappedLineParser.h"
20 #include "WhitespaceManager.h"
21 #include "clang/Basic/Diagnostic.h"
22 #include "clang/Basic/DiagnosticOptions.h"
23 #include "clang/Basic/SourceManager.h"
24 #include "clang/Format/Format.h"
25 #include "clang/Lex/Lexer.h"
26 #include "llvm/ADT/STLExtras.h"
27 #include "llvm/Support/Allocator.h"
28 #include "llvm/Support/Debug.h"
29 #include "llvm/Support/Path.h"
30 #include "llvm/Support/YAMLTraits.h"
31 #include <queue>
32 #include <string>
33 
34 #define DEBUG_TYPE "format-formatter"
35 
36 using clang::format::FormatStyle;
37 
38 LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR(std::string)
39 
40 namespace llvm {
41 namespace yaml {
42 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageKind> {
43   static void enumeration(IO &IO, FormatStyle::LanguageKind &Value) {
44     IO.enumCase(Value, "Cpp", FormatStyle::LK_Cpp);
45     IO.enumCase(Value, "Java", FormatStyle::LK_Java);
46     IO.enumCase(Value, "JavaScript", FormatStyle::LK_JavaScript);
47     IO.enumCase(Value, "Proto", FormatStyle::LK_Proto);
48   }
49 };
50 
51 template <> struct ScalarEnumerationTraits<FormatStyle::LanguageStandard> {
52   static void enumeration(IO &IO, FormatStyle::LanguageStandard &Value) {
53     IO.enumCase(Value, "Cpp03", FormatStyle::LS_Cpp03);
54     IO.enumCase(Value, "C++03", FormatStyle::LS_Cpp03);
55     IO.enumCase(Value, "Cpp11", FormatStyle::LS_Cpp11);
56     IO.enumCase(Value, "C++11", FormatStyle::LS_Cpp11);
57     IO.enumCase(Value, "Auto", FormatStyle::LS_Auto);
58   }
59 };
60 
61 template <> struct ScalarEnumerationTraits<FormatStyle::UseTabStyle> {
62   static void enumeration(IO &IO, FormatStyle::UseTabStyle &Value) {
63     IO.enumCase(Value, "Never", FormatStyle::UT_Never);
64     IO.enumCase(Value, "false", FormatStyle::UT_Never);
65     IO.enumCase(Value, "Always", FormatStyle::UT_Always);
66     IO.enumCase(Value, "true", FormatStyle::UT_Always);
67     IO.enumCase(Value, "ForIndentation", FormatStyle::UT_ForIndentation);
68   }
69 };
70 
71 template <> struct ScalarEnumerationTraits<FormatStyle::ShortFunctionStyle> {
72   static void enumeration(IO &IO, FormatStyle::ShortFunctionStyle &Value) {
73     IO.enumCase(Value, "None", FormatStyle::SFS_None);
74     IO.enumCase(Value, "false", FormatStyle::SFS_None);
75     IO.enumCase(Value, "All", FormatStyle::SFS_All);
76     IO.enumCase(Value, "true", FormatStyle::SFS_All);
77     IO.enumCase(Value, "Inline", FormatStyle::SFS_Inline);
78     IO.enumCase(Value, "Empty", FormatStyle::SFS_Empty);
79   }
80 };
81 
82 template <> struct ScalarEnumerationTraits<FormatStyle::BinaryOperatorStyle> {
83   static void enumeration(IO &IO, FormatStyle::BinaryOperatorStyle &Value) {
84     IO.enumCase(Value, "All", FormatStyle::BOS_All);
85     IO.enumCase(Value, "true", FormatStyle::BOS_All);
86     IO.enumCase(Value, "None", FormatStyle::BOS_None);
87     IO.enumCase(Value, "false", FormatStyle::BOS_None);
88     IO.enumCase(Value, "NonAssignment", FormatStyle::BOS_NonAssignment);
89   }
90 };
91 
92 template <> struct ScalarEnumerationTraits<FormatStyle::BraceBreakingStyle> {
93   static void enumeration(IO &IO, FormatStyle::BraceBreakingStyle &Value) {
94     IO.enumCase(Value, "Attach", FormatStyle::BS_Attach);
95     IO.enumCase(Value, "Linux", FormatStyle::BS_Linux);
96     IO.enumCase(Value, "Stroustrup", FormatStyle::BS_Stroustrup);
97     IO.enumCase(Value, "Allman", FormatStyle::BS_Allman);
98     IO.enumCase(Value, "GNU", FormatStyle::BS_GNU);
99   }
100 };
101 
102 template <>
103 struct ScalarEnumerationTraits<FormatStyle::NamespaceIndentationKind> {
104   static void enumeration(IO &IO,
105                           FormatStyle::NamespaceIndentationKind &Value) {
106     IO.enumCase(Value, "None", FormatStyle::NI_None);
107     IO.enumCase(Value, "Inner", FormatStyle::NI_Inner);
108     IO.enumCase(Value, "All", FormatStyle::NI_All);
109   }
110 };
111 
112 template <>
113 struct ScalarEnumerationTraits<FormatStyle::PointerAlignmentStyle> {
114   static void enumeration(IO &IO,
115                           FormatStyle::PointerAlignmentStyle &Value) {
116     IO.enumCase(Value, "Middle", FormatStyle::PAS_Middle);
117     IO.enumCase(Value, "Left", FormatStyle::PAS_Left);
118     IO.enumCase(Value, "Right", FormatStyle::PAS_Right);
119 
120     // For backward compatibility.
121     IO.enumCase(Value, "true", FormatStyle::PAS_Left);
122     IO.enumCase(Value, "false", FormatStyle::PAS_Right);
123   }
124 };
125 
126 template <>
127 struct ScalarEnumerationTraits<FormatStyle::SpaceBeforeParensOptions> {
128   static void enumeration(IO &IO,
129                           FormatStyle::SpaceBeforeParensOptions &Value) {
130     IO.enumCase(Value, "Never", FormatStyle::SBPO_Never);
131     IO.enumCase(Value, "ControlStatements",
132                 FormatStyle::SBPO_ControlStatements);
133     IO.enumCase(Value, "Always", FormatStyle::SBPO_Always);
134 
135     // For backward compatibility.
136     IO.enumCase(Value, "false", FormatStyle::SBPO_Never);
137     IO.enumCase(Value, "true", FormatStyle::SBPO_ControlStatements);
138   }
139 };
140 
141 template <> struct MappingTraits<FormatStyle> {
142   static void mapping(IO &IO, FormatStyle &Style) {
143     // When reading, read the language first, we need it for getPredefinedStyle.
144     IO.mapOptional("Language", Style.Language);
145 
146     if (IO.outputting()) {
147       StringRef StylesArray[] = { "LLVM",    "Google", "Chromium",
148                                   "Mozilla", "WebKit", "GNU" };
149       ArrayRef<StringRef> Styles(StylesArray);
150       for (size_t i = 0, e = Styles.size(); i < e; ++i) {
151         StringRef StyleName(Styles[i]);
152         FormatStyle PredefinedStyle;
153         if (getPredefinedStyle(StyleName, Style.Language, &PredefinedStyle) &&
154             Style == PredefinedStyle) {
155           IO.mapOptional("# BasedOnStyle", StyleName);
156           break;
157         }
158       }
159     } else {
160       StringRef BasedOnStyle;
161       IO.mapOptional("BasedOnStyle", BasedOnStyle);
162       if (!BasedOnStyle.empty()) {
163         FormatStyle::LanguageKind OldLanguage = Style.Language;
164         FormatStyle::LanguageKind Language =
165             ((FormatStyle *)IO.getContext())->Language;
166         if (!getPredefinedStyle(BasedOnStyle, Language, &Style)) {
167           IO.setError(Twine("Unknown value for BasedOnStyle: ", BasedOnStyle));
168           return;
169         }
170         Style.Language = OldLanguage;
171       }
172     }
173 
174     IO.mapOptional("AccessModifierOffset", Style.AccessModifierOffset);
175     IO.mapOptional("AlignAfterOpenBracket", Style.AlignAfterOpenBracket);
176     IO.mapOptional("AlignEscapedNewlinesLeft", Style.AlignEscapedNewlinesLeft);
177     IO.mapOptional("AlignOperands", Style.AlignOperands);
178     IO.mapOptional("AlignTrailingComments", Style.AlignTrailingComments);
179     IO.mapOptional("AllowAllParametersOfDeclarationOnNextLine",
180                    Style.AllowAllParametersOfDeclarationOnNextLine);
181     IO.mapOptional("AllowShortBlocksOnASingleLine",
182                    Style.AllowShortBlocksOnASingleLine);
183     IO.mapOptional("AllowShortCaseLabelsOnASingleLine",
184                    Style.AllowShortCaseLabelsOnASingleLine);
185     IO.mapOptional("AllowShortIfStatementsOnASingleLine",
186                    Style.AllowShortIfStatementsOnASingleLine);
187     IO.mapOptional("AllowShortLoopsOnASingleLine",
188                    Style.AllowShortLoopsOnASingleLine);
189     IO.mapOptional("AllowShortFunctionsOnASingleLine",
190                    Style.AllowShortFunctionsOnASingleLine);
191     IO.mapOptional("AlwaysBreakAfterDefinitionReturnType",
192                    Style.AlwaysBreakAfterDefinitionReturnType);
193     IO.mapOptional("AlwaysBreakTemplateDeclarations",
194                    Style.AlwaysBreakTemplateDeclarations);
195     IO.mapOptional("AlwaysBreakBeforeMultilineStrings",
196                    Style.AlwaysBreakBeforeMultilineStrings);
197     IO.mapOptional("BreakBeforeBinaryOperators",
198                    Style.BreakBeforeBinaryOperators);
199     IO.mapOptional("BreakBeforeTernaryOperators",
200                    Style.BreakBeforeTernaryOperators);
201     IO.mapOptional("BreakConstructorInitializersBeforeComma",
202                    Style.BreakConstructorInitializersBeforeComma);
203     IO.mapOptional("BinPackParameters", Style.BinPackParameters);
204     IO.mapOptional("BinPackArguments", Style.BinPackArguments);
205     IO.mapOptional("ColumnLimit", Style.ColumnLimit);
206     IO.mapOptional("ConstructorInitializerAllOnOneLineOrOnePerLine",
207                    Style.ConstructorInitializerAllOnOneLineOrOnePerLine);
208     IO.mapOptional("ConstructorInitializerIndentWidth",
209                    Style.ConstructorInitializerIndentWidth);
210     IO.mapOptional("DerivePointerAlignment", Style.DerivePointerAlignment);
211     IO.mapOptional("ExperimentalAutoDetectBinPacking",
212                    Style.ExperimentalAutoDetectBinPacking);
213     IO.mapOptional("IndentCaseLabels", Style.IndentCaseLabels);
214     IO.mapOptional("IndentWrappedFunctionNames",
215                    Style.IndentWrappedFunctionNames);
216     IO.mapOptional("IndentFunctionDeclarationAfterType",
217                    Style.IndentWrappedFunctionNames);
218     IO.mapOptional("MaxEmptyLinesToKeep", Style.MaxEmptyLinesToKeep);
219     IO.mapOptional("KeepEmptyLinesAtTheStartOfBlocks",
220                    Style.KeepEmptyLinesAtTheStartOfBlocks);
221     IO.mapOptional("NamespaceIndentation", Style.NamespaceIndentation);
222     IO.mapOptional("ObjCBlockIndentWidth", Style.ObjCBlockIndentWidth);
223     IO.mapOptional("ObjCSpaceAfterProperty", Style.ObjCSpaceAfterProperty);
224     IO.mapOptional("ObjCSpaceBeforeProtocolList",
225                    Style.ObjCSpaceBeforeProtocolList);
226     IO.mapOptional("PenaltyBreakBeforeFirstCallParameter",
227                    Style.PenaltyBreakBeforeFirstCallParameter);
228     IO.mapOptional("PenaltyBreakComment", Style.PenaltyBreakComment);
229     IO.mapOptional("PenaltyBreakString", Style.PenaltyBreakString);
230     IO.mapOptional("PenaltyBreakFirstLessLess",
231                    Style.PenaltyBreakFirstLessLess);
232     IO.mapOptional("PenaltyExcessCharacter", Style.PenaltyExcessCharacter);
233     IO.mapOptional("PenaltyReturnTypeOnItsOwnLine",
234                    Style.PenaltyReturnTypeOnItsOwnLine);
235     IO.mapOptional("PointerAlignment", Style.PointerAlignment);
236     IO.mapOptional("SpacesBeforeTrailingComments",
237                    Style.SpacesBeforeTrailingComments);
238     IO.mapOptional("Cpp11BracedListStyle", Style.Cpp11BracedListStyle);
239     IO.mapOptional("Standard", Style.Standard);
240     IO.mapOptional("IndentWidth", Style.IndentWidth);
241     IO.mapOptional("TabWidth", Style.TabWidth);
242     IO.mapOptional("UseTab", Style.UseTab);
243     IO.mapOptional("BreakBeforeBraces", Style.BreakBeforeBraces);
244     IO.mapOptional("SpacesInParentheses", Style.SpacesInParentheses);
245     IO.mapOptional("SpacesInSquareBrackets", Style.SpacesInSquareBrackets);
246     IO.mapOptional("SpacesInAngles", Style.SpacesInAngles);
247     IO.mapOptional("SpaceInEmptyParentheses", Style.SpaceInEmptyParentheses);
248     IO.mapOptional("SpacesInCStyleCastParentheses",
249                    Style.SpacesInCStyleCastParentheses);
250     IO.mapOptional("SpaceAfterCStyleCast", Style.SpaceAfterCStyleCast);
251     IO.mapOptional("SpacesInContainerLiterals",
252                    Style.SpacesInContainerLiterals);
253     IO.mapOptional("SpaceBeforeAssignmentOperators",
254                    Style.SpaceBeforeAssignmentOperators);
255     IO.mapOptional("ContinuationIndentWidth", Style.ContinuationIndentWidth);
256     IO.mapOptional("CommentPragmas", Style.CommentPragmas);
257     IO.mapOptional("ForEachMacros", Style.ForEachMacros);
258 
259     // For backward compatibility.
260     if (!IO.outputting()) {
261       IO.mapOptional("SpaceAfterControlStatementKeyword",
262                      Style.SpaceBeforeParens);
263       IO.mapOptional("PointerBindsToType", Style.PointerAlignment);
264       IO.mapOptional("DerivePointerBinding", Style.DerivePointerAlignment);
265     }
266     IO.mapOptional("SpaceBeforeParens", Style.SpaceBeforeParens);
267     IO.mapOptional("DisableFormat", Style.DisableFormat);
268   }
269 };
270 
271 // Allows to read vector<FormatStyle> while keeping default values.
272 // IO.getContext() should contain a pointer to the FormatStyle structure, that
273 // will be used to get default values for missing keys.
274 // If the first element has no Language specified, it will be treated as the
275 // default one for the following elements.
276 template <> struct DocumentListTraits<std::vector<FormatStyle> > {
277   static size_t size(IO &IO, std::vector<FormatStyle> &Seq) {
278     return Seq.size();
279   }
280   static FormatStyle &element(IO &IO, std::vector<FormatStyle> &Seq,
281                               size_t Index) {
282     if (Index >= Seq.size()) {
283       assert(Index == Seq.size());
284       FormatStyle Template;
285       if (Seq.size() > 0 && Seq[0].Language == FormatStyle::LK_None) {
286         Template = Seq[0];
287       } else {
288         Template = *((const FormatStyle *)IO.getContext());
289         Template.Language = FormatStyle::LK_None;
290       }
291       Seq.resize(Index + 1, Template);
292     }
293     return Seq[Index];
294   }
295 };
296 }
297 }
298 
299 namespace clang {
300 namespace format {
301 
302 const std::error_category &getParseCategory() {
303   static ParseErrorCategory C;
304   return C;
305 }
306 std::error_code make_error_code(ParseError e) {
307   return std::error_code(static_cast<int>(e), getParseCategory());
308 }
309 
310 const char *ParseErrorCategory::name() const LLVM_NOEXCEPT {
311   return "clang-format.parse_error";
312 }
313 
314 std::string ParseErrorCategory::message(int EV) const {
315   switch (static_cast<ParseError>(EV)) {
316   case ParseError::Success:
317     return "Success";
318   case ParseError::Error:
319     return "Invalid argument";
320   case ParseError::Unsuitable:
321     return "Unsuitable";
322   }
323   llvm_unreachable("unexpected parse error");
324 }
325 
326 FormatStyle getLLVMStyle() {
327   FormatStyle LLVMStyle;
328   LLVMStyle.Language = FormatStyle::LK_Cpp;
329   LLVMStyle.AccessModifierOffset = -2;
330   LLVMStyle.AlignEscapedNewlinesLeft = false;
331   LLVMStyle.AlignAfterOpenBracket = true;
332   LLVMStyle.AlignOperands = true;
333   LLVMStyle.AlignTrailingComments = true;
334   LLVMStyle.AllowAllParametersOfDeclarationOnNextLine = true;
335   LLVMStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_All;
336   LLVMStyle.AllowShortBlocksOnASingleLine = false;
337   LLVMStyle.AllowShortCaseLabelsOnASingleLine = false;
338   LLVMStyle.AllowShortIfStatementsOnASingleLine = false;
339   LLVMStyle.AllowShortLoopsOnASingleLine = false;
340   LLVMStyle.AlwaysBreakAfterDefinitionReturnType = false;
341   LLVMStyle.AlwaysBreakBeforeMultilineStrings = false;
342   LLVMStyle.AlwaysBreakTemplateDeclarations = false;
343   LLVMStyle.BinPackParameters = true;
344   LLVMStyle.BinPackArguments = true;
345   LLVMStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_None;
346   LLVMStyle.BreakBeforeTernaryOperators = true;
347   LLVMStyle.BreakBeforeBraces = FormatStyle::BS_Attach;
348   LLVMStyle.BreakConstructorInitializersBeforeComma = false;
349   LLVMStyle.ColumnLimit = 80;
350   LLVMStyle.CommentPragmas = "^ IWYU pragma:";
351   LLVMStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = false;
352   LLVMStyle.ConstructorInitializerIndentWidth = 4;
353   LLVMStyle.ContinuationIndentWidth = 4;
354   LLVMStyle.Cpp11BracedListStyle = true;
355   LLVMStyle.DerivePointerAlignment = false;
356   LLVMStyle.ExperimentalAutoDetectBinPacking = false;
357   LLVMStyle.ForEachMacros.push_back("foreach");
358   LLVMStyle.ForEachMacros.push_back("Q_FOREACH");
359   LLVMStyle.ForEachMacros.push_back("BOOST_FOREACH");
360   LLVMStyle.IndentCaseLabels = false;
361   LLVMStyle.IndentWrappedFunctionNames = false;
362   LLVMStyle.IndentWidth = 2;
363   LLVMStyle.TabWidth = 8;
364   LLVMStyle.MaxEmptyLinesToKeep = 1;
365   LLVMStyle.KeepEmptyLinesAtTheStartOfBlocks = true;
366   LLVMStyle.NamespaceIndentation = FormatStyle::NI_None;
367   LLVMStyle.ObjCBlockIndentWidth = 2;
368   LLVMStyle.ObjCSpaceAfterProperty = false;
369   LLVMStyle.ObjCSpaceBeforeProtocolList = true;
370   LLVMStyle.PointerAlignment = FormatStyle::PAS_Right;
371   LLVMStyle.SpacesBeforeTrailingComments = 1;
372   LLVMStyle.Standard = FormatStyle::LS_Cpp11;
373   LLVMStyle.UseTab = FormatStyle::UT_Never;
374   LLVMStyle.SpacesInParentheses = false;
375   LLVMStyle.SpacesInSquareBrackets = false;
376   LLVMStyle.SpaceInEmptyParentheses = false;
377   LLVMStyle.SpacesInContainerLiterals = true;
378   LLVMStyle.SpacesInCStyleCastParentheses = false;
379   LLVMStyle.SpaceAfterCStyleCast = false;
380   LLVMStyle.SpaceBeforeParens = FormatStyle::SBPO_ControlStatements;
381   LLVMStyle.SpaceBeforeAssignmentOperators = true;
382   LLVMStyle.SpacesInAngles = false;
383 
384   LLVMStyle.PenaltyBreakComment = 300;
385   LLVMStyle.PenaltyBreakFirstLessLess = 120;
386   LLVMStyle.PenaltyBreakString = 1000;
387   LLVMStyle.PenaltyExcessCharacter = 1000000;
388   LLVMStyle.PenaltyReturnTypeOnItsOwnLine = 60;
389   LLVMStyle.PenaltyBreakBeforeFirstCallParameter = 19;
390 
391   LLVMStyle.DisableFormat = false;
392 
393   return LLVMStyle;
394 }
395 
396 FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
397   FormatStyle GoogleStyle = getLLVMStyle();
398   GoogleStyle.Language = Language;
399 
400   GoogleStyle.AccessModifierOffset = -1;
401   GoogleStyle.AlignEscapedNewlinesLeft = true;
402   GoogleStyle.AllowShortIfStatementsOnASingleLine = true;
403   GoogleStyle.AllowShortLoopsOnASingleLine = true;
404   GoogleStyle.AlwaysBreakBeforeMultilineStrings = true;
405   GoogleStyle.AlwaysBreakTemplateDeclarations = true;
406   GoogleStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
407   GoogleStyle.DerivePointerAlignment = true;
408   GoogleStyle.IndentCaseLabels = true;
409   GoogleStyle.KeepEmptyLinesAtTheStartOfBlocks = false;
410   GoogleStyle.ObjCSpaceAfterProperty = false;
411   GoogleStyle.ObjCSpaceBeforeProtocolList = false;
412   GoogleStyle.PointerAlignment = FormatStyle::PAS_Left;
413   GoogleStyle.SpacesBeforeTrailingComments = 2;
414   GoogleStyle.Standard = FormatStyle::LS_Auto;
415 
416   GoogleStyle.PenaltyReturnTypeOnItsOwnLine = 200;
417   GoogleStyle.PenaltyBreakBeforeFirstCallParameter = 1;
418 
419   if (Language == FormatStyle::LK_Java) {
420     GoogleStyle.AlignAfterOpenBracket = false;
421     GoogleStyle.AlignOperands = false;
422     GoogleStyle.AlignTrailingComments = false;
423     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Empty;
424     GoogleStyle.AllowShortIfStatementsOnASingleLine = false;
425     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
426     GoogleStyle.BreakBeforeBinaryOperators = FormatStyle::BOS_NonAssignment;
427     GoogleStyle.ColumnLimit = 100;
428     GoogleStyle.SpaceAfterCStyleCast = true;
429     GoogleStyle.SpacesBeforeTrailingComments = 1;
430   } else if (Language == FormatStyle::LK_JavaScript) {
431     GoogleStyle.BreakBeforeTernaryOperators = false;
432     GoogleStyle.MaxEmptyLinesToKeep = 3;
433     GoogleStyle.SpacesInContainerLiterals = false;
434     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
435     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
436   } else if (Language == FormatStyle::LK_Proto) {
437     GoogleStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_None;
438     GoogleStyle.SpacesInContainerLiterals = false;
439   }
440 
441   return GoogleStyle;
442 }
443 
444 FormatStyle getChromiumStyle(FormatStyle::LanguageKind Language) {
445   FormatStyle ChromiumStyle = getGoogleStyle(Language);
446   if (Language == FormatStyle::LK_Java) {
447     ChromiumStyle.AllowShortIfStatementsOnASingleLine = true;
448     ChromiumStyle.IndentWidth = 4;
449     ChromiumStyle.ContinuationIndentWidth = 8;
450   } else {
451     ChromiumStyle.AllowAllParametersOfDeclarationOnNextLine = false;
452     ChromiumStyle.AllowShortFunctionsOnASingleLine = FormatStyle::SFS_Inline;
453     ChromiumStyle.AllowShortIfStatementsOnASingleLine = false;
454     ChromiumStyle.AllowShortLoopsOnASingleLine = false;
455     ChromiumStyle.BinPackParameters = false;
456     ChromiumStyle.DerivePointerAlignment = false;
457   }
458   return ChromiumStyle;
459 }
460 
461 FormatStyle getMozillaStyle() {
462   FormatStyle MozillaStyle = getLLVMStyle();
463   MozillaStyle.AllowAllParametersOfDeclarationOnNextLine = false;
464   MozillaStyle.Cpp11BracedListStyle = false;
465   MozillaStyle.ConstructorInitializerAllOnOneLineOrOnePerLine = true;
466   MozillaStyle.DerivePointerAlignment = true;
467   MozillaStyle.IndentCaseLabels = true;
468   MozillaStyle.ObjCSpaceAfterProperty = true;
469   MozillaStyle.ObjCSpaceBeforeProtocolList = false;
470   MozillaStyle.PenaltyReturnTypeOnItsOwnLine = 200;
471   MozillaStyle.PointerAlignment = FormatStyle::PAS_Left;
472   MozillaStyle.Standard = FormatStyle::LS_Cpp03;
473   return MozillaStyle;
474 }
475 
476 FormatStyle getWebKitStyle() {
477   FormatStyle Style = getLLVMStyle();
478   Style.AccessModifierOffset = -4;
479   Style.AlignAfterOpenBracket = false;
480   Style.AlignOperands = false;
481   Style.AlignTrailingComments = false;
482   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
483   Style.BreakBeforeBraces = FormatStyle::BS_Stroustrup;
484   Style.BreakConstructorInitializersBeforeComma = true;
485   Style.Cpp11BracedListStyle = false;
486   Style.ColumnLimit = 0;
487   Style.IndentWidth = 4;
488   Style.NamespaceIndentation = FormatStyle::NI_Inner;
489   Style.ObjCBlockIndentWidth = 4;
490   Style.ObjCSpaceAfterProperty = true;
491   Style.PointerAlignment = FormatStyle::PAS_Left;
492   Style.Standard = FormatStyle::LS_Cpp03;
493   return Style;
494 }
495 
496 FormatStyle getGNUStyle() {
497   FormatStyle Style = getLLVMStyle();
498   Style.AlwaysBreakAfterDefinitionReturnType = true;
499   Style.BreakBeforeBinaryOperators = FormatStyle::BOS_All;
500   Style.BreakBeforeBraces = FormatStyle::BS_GNU;
501   Style.BreakBeforeTernaryOperators = true;
502   Style.Cpp11BracedListStyle = false;
503   Style.ColumnLimit = 79;
504   Style.SpaceBeforeParens = FormatStyle::SBPO_Always;
505   Style.Standard = FormatStyle::LS_Cpp03;
506   return Style;
507 }
508 
509 FormatStyle getNoStyle() {
510   FormatStyle NoStyle = getLLVMStyle();
511   NoStyle.DisableFormat = true;
512   return NoStyle;
513 }
514 
515 bool getPredefinedStyle(StringRef Name, FormatStyle::LanguageKind Language,
516                         FormatStyle *Style) {
517   if (Name.equals_lower("llvm")) {
518     *Style = getLLVMStyle();
519   } else if (Name.equals_lower("chromium")) {
520     *Style = getChromiumStyle(Language);
521   } else if (Name.equals_lower("mozilla")) {
522     *Style = getMozillaStyle();
523   } else if (Name.equals_lower("google")) {
524     *Style = getGoogleStyle(Language);
525   } else if (Name.equals_lower("webkit")) {
526     *Style = getWebKitStyle();
527   } else if (Name.equals_lower("gnu")) {
528     *Style = getGNUStyle();
529   } else if (Name.equals_lower("none")) {
530     *Style = getNoStyle();
531   } else {
532     return false;
533   }
534 
535   Style->Language = Language;
536   return true;
537 }
538 
539 std::error_code parseConfiguration(StringRef Text, FormatStyle *Style) {
540   assert(Style);
541   FormatStyle::LanguageKind Language = Style->Language;
542   assert(Language != FormatStyle::LK_None);
543   if (Text.trim().empty())
544     return make_error_code(ParseError::Error);
545 
546   std::vector<FormatStyle> Styles;
547   llvm::yaml::Input Input(Text);
548   // DocumentListTraits<vector<FormatStyle>> uses the context to get default
549   // values for the fields, keys for which are missing from the configuration.
550   // Mapping also uses the context to get the language to find the correct
551   // base style.
552   Input.setContext(Style);
553   Input >> Styles;
554   if (Input.error())
555     return Input.error();
556 
557   for (unsigned i = 0; i < Styles.size(); ++i) {
558     // Ensures that only the first configuration can skip the Language option.
559     if (Styles[i].Language == FormatStyle::LK_None && i != 0)
560       return make_error_code(ParseError::Error);
561     // Ensure that each language is configured at most once.
562     for (unsigned j = 0; j < i; ++j) {
563       if (Styles[i].Language == Styles[j].Language) {
564         DEBUG(llvm::dbgs()
565               << "Duplicate languages in the config file on positions " << j
566               << " and " << i << "\n");
567         return make_error_code(ParseError::Error);
568       }
569     }
570   }
571   // Look for a suitable configuration starting from the end, so we can
572   // find the configuration for the specific language first, and the default
573   // configuration (which can only be at slot 0) after it.
574   for (int i = Styles.size() - 1; i >= 0; --i) {
575     if (Styles[i].Language == Language ||
576         Styles[i].Language == FormatStyle::LK_None) {
577       *Style = Styles[i];
578       Style->Language = Language;
579       return make_error_code(ParseError::Success);
580     }
581   }
582   return make_error_code(ParseError::Unsuitable);
583 }
584 
585 std::string configurationAsText(const FormatStyle &Style) {
586   std::string Text;
587   llvm::raw_string_ostream Stream(Text);
588   llvm::yaml::Output Output(Stream);
589   // We use the same mapping method for input and output, so we need a non-const
590   // reference here.
591   FormatStyle NonConstStyle = Style;
592   Output << NonConstStyle;
593   return Stream.str();
594 }
595 
596 namespace {
597 
598 class FormatTokenLexer {
599 public:
600   FormatTokenLexer(SourceManager &SourceMgr, FileID ID, FormatStyle &Style,
601                    encoding::Encoding Encoding)
602       : FormatTok(nullptr), IsFirstToken(true), GreaterStashed(false),
603         Column(0), TrailingWhitespace(0), SourceMgr(SourceMgr), ID(ID),
604         Style(Style), IdentTable(getFormattingLangOpts(Style)),
605         Keywords(IdentTable), Encoding(Encoding), FirstInLineIndex(0),
606         FormattingDisabled(false) {
607     Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr,
608                         getFormattingLangOpts(Style)));
609     Lex->SetKeepWhitespaceMode(true);
610 
611     for (const std::string &ForEachMacro : Style.ForEachMacros)
612       ForEachMacros.push_back(&IdentTable.get(ForEachMacro));
613     std::sort(ForEachMacros.begin(), ForEachMacros.end());
614   }
615 
616   ArrayRef<FormatToken *> lex() {
617     assert(Tokens.empty());
618     assert(FirstInLineIndex == 0);
619     do {
620       Tokens.push_back(getNextToken());
621       tryMergePreviousTokens();
622       if (Tokens.back()->NewlinesBefore > 0)
623         FirstInLineIndex = Tokens.size() - 1;
624     } while (Tokens.back()->Tok.isNot(tok::eof));
625     return Tokens;
626   }
627 
628   const AdditionalKeywords &getKeywords() { return Keywords; }
629 
630 private:
631   void tryMergePreviousTokens() {
632     if (tryMerge_TMacro())
633       return;
634     if (tryMergeConflictMarkers())
635       return;
636 
637     if (Style.Language == FormatStyle::LK_JavaScript) {
638       if (tryMergeJSRegexLiteral())
639         return;
640       if (tryMergeEscapeSequence())
641         return;
642 
643       static tok::TokenKind JSIdentity[] = { tok::equalequal, tok::equal };
644       static tok::TokenKind JSNotIdentity[] = { tok::exclaimequal, tok::equal };
645       static tok::TokenKind JSShiftEqual[] = { tok::greater, tok::greater,
646                                                tok::greaterequal };
647       static tok::TokenKind JSRightArrow[] = { tok::equal, tok::greater };
648       // FIXME: We probably need to change token type to mimic operator with the
649       // correct priority.
650       if (tryMergeTokens(JSIdentity))
651         return;
652       if (tryMergeTokens(JSNotIdentity))
653         return;
654       if (tryMergeTokens(JSShiftEqual))
655         return;
656       if (tryMergeTokens(JSRightArrow))
657         return;
658     }
659   }
660 
661   bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds) {
662     if (Tokens.size() < Kinds.size())
663       return false;
664 
665     SmallVectorImpl<FormatToken *>::const_iterator First =
666         Tokens.end() - Kinds.size();
667     if (!First[0]->is(Kinds[0]))
668       return false;
669     unsigned AddLength = 0;
670     for (unsigned i = 1; i < Kinds.size(); ++i) {
671       if (!First[i]->is(Kinds[i]) || First[i]->WhitespaceRange.getBegin() !=
672                                          First[i]->WhitespaceRange.getEnd())
673         return false;
674       AddLength += First[i]->TokenText.size();
675     }
676     Tokens.resize(Tokens.size() - Kinds.size() + 1);
677     First[0]->TokenText = StringRef(First[0]->TokenText.data(),
678                                     First[0]->TokenText.size() + AddLength);
679     First[0]->ColumnWidth += AddLength;
680     return true;
681   }
682 
683   // Tries to merge an escape sequence, i.e. a "\\" and the following
684   // character. Use e.g. inside JavaScript regex literals.
685   bool tryMergeEscapeSequence() {
686     if (Tokens.size() < 2)
687       return false;
688     FormatToken *Previous = Tokens[Tokens.size() - 2];
689     if (Previous->isNot(tok::unknown) || Previous->TokenText != "\\")
690       return false;
691     ++Previous->ColumnWidth;
692     StringRef Text = Previous->TokenText;
693     Previous->TokenText = StringRef(Text.data(), Text.size() + 1);
694     resetLexer(SourceMgr.getFileOffset(Tokens.back()->Tok.getLocation()) + 1);
695     Tokens.resize(Tokens.size() - 1);
696     Column = Previous->OriginalColumn + Previous->ColumnWidth;
697     return true;
698   }
699 
700   // Try to determine whether the current token ends a JavaScript regex literal.
701   // We heuristically assume that this is a regex literal if we find two
702   // unescaped slashes on a line and the token before the first slash is one of
703   // "(;,{}![:?", a binary operator or 'return', as those cannot be followed by
704   // a division.
705   bool tryMergeJSRegexLiteral() {
706     if (Tokens.size() < 2)
707       return false;
708     // If a regex literal ends in "\//", this gets represented by an unknown
709     // token "\" and a comment.
710     bool MightEndWithEscapedSlash =
711         Tokens.back()->is(tok::comment) &&
712         Tokens.back()->TokenText.startswith("//") &&
713         Tokens[Tokens.size() - 2]->TokenText == "\\";
714     if (!MightEndWithEscapedSlash &&
715         (Tokens.back()->isNot(tok::slash) ||
716          (Tokens[Tokens.size() - 2]->is(tok::unknown) &&
717           Tokens[Tokens.size() - 2]->TokenText == "\\")))
718       return false;
719     unsigned TokenCount = 0;
720     unsigned LastColumn = Tokens.back()->OriginalColumn;
721     for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) {
722       ++TokenCount;
723       if (I[0]->is(tok::slash) && I + 1 != E &&
724           (I[1]->isOneOf(tok::l_paren, tok::semi, tok::l_brace, tok::r_brace,
725                          tok::exclaim, tok::l_square, tok::colon, tok::comma,
726                          tok::question, tok::kw_return) ||
727            I[1]->isBinaryOperator())) {
728         if (MightEndWithEscapedSlash) {
729           // This regex literal ends in '\//'. Skip past the '//' of the last
730           // token and re-start lexing from there.
731           SourceLocation Loc = Tokens.back()->Tok.getLocation();
732           resetLexer(SourceMgr.getFileOffset(Loc) + 2);
733         }
734         Tokens.resize(Tokens.size() - TokenCount);
735         Tokens.back()->Tok.setKind(tok::unknown);
736         Tokens.back()->Type = TT_RegexLiteral;
737         Tokens.back()->ColumnWidth += LastColumn - I[0]->OriginalColumn;
738         return true;
739       }
740 
741       // There can't be a newline inside a regex literal.
742       if (I[0]->NewlinesBefore > 0)
743         return false;
744     }
745     return false;
746   }
747 
748   bool tryMerge_TMacro() {
749     if (Tokens.size() < 4)
750       return false;
751     FormatToken *Last = Tokens.back();
752     if (!Last->is(tok::r_paren))
753       return false;
754 
755     FormatToken *String = Tokens[Tokens.size() - 2];
756     if (!String->is(tok::string_literal) || String->IsMultiline)
757       return false;
758 
759     if (!Tokens[Tokens.size() - 3]->is(tok::l_paren))
760       return false;
761 
762     FormatToken *Macro = Tokens[Tokens.size() - 4];
763     if (Macro->TokenText != "_T")
764       return false;
765 
766     const char *Start = Macro->TokenText.data();
767     const char *End = Last->TokenText.data() + Last->TokenText.size();
768     String->TokenText = StringRef(Start, End - Start);
769     String->IsFirst = Macro->IsFirst;
770     String->LastNewlineOffset = Macro->LastNewlineOffset;
771     String->WhitespaceRange = Macro->WhitespaceRange;
772     String->OriginalColumn = Macro->OriginalColumn;
773     String->ColumnWidth = encoding::columnWidthWithTabs(
774         String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding);
775 
776     Tokens.pop_back();
777     Tokens.pop_back();
778     Tokens.pop_back();
779     Tokens.back() = String;
780     return true;
781   }
782 
783   bool tryMergeConflictMarkers() {
784     if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof))
785       return false;
786 
787     // Conflict lines look like:
788     // <marker> <text from the vcs>
789     // For example:
790     // >>>>>>> /file/in/file/system at revision 1234
791     //
792     // We merge all tokens in a line that starts with a conflict marker
793     // into a single token with a special token type that the unwrapped line
794     // parser will use to correctly rebuild the underlying code.
795 
796     FileID ID;
797     // Get the position of the first token in the line.
798     unsigned FirstInLineOffset;
799     std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc(
800         Tokens[FirstInLineIndex]->getStartOfNonWhitespace());
801     StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer();
802     // Calculate the offset of the start of the current line.
803     auto LineOffset = Buffer.rfind('\n', FirstInLineOffset);
804     if (LineOffset == StringRef::npos) {
805       LineOffset = 0;
806     } else {
807       ++LineOffset;
808     }
809 
810     auto FirstSpace = Buffer.find_first_of(" \n", LineOffset);
811     StringRef LineStart;
812     if (FirstSpace == StringRef::npos) {
813       LineStart = Buffer.substr(LineOffset);
814     } else {
815       LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset);
816     }
817 
818     TokenType Type = TT_Unknown;
819     if (LineStart == "<<<<<<<" || LineStart == ">>>>") {
820       Type = TT_ConflictStart;
821     } else if (LineStart == "|||||||" || LineStart == "=======" ||
822                LineStart == "====") {
823       Type = TT_ConflictAlternative;
824     } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") {
825       Type = TT_ConflictEnd;
826     }
827 
828     if (Type != TT_Unknown) {
829       FormatToken *Next = Tokens.back();
830 
831       Tokens.resize(FirstInLineIndex + 1);
832       // We do not need to build a complete token here, as we will skip it
833       // during parsing anyway (as we must not touch whitespace around conflict
834       // markers).
835       Tokens.back()->Type = Type;
836       Tokens.back()->Tok.setKind(tok::kw___unknown_anytype);
837 
838       Tokens.push_back(Next);
839       return true;
840     }
841 
842     return false;
843   }
844 
845   FormatToken *getNextToken() {
846     if (GreaterStashed) {
847       // Create a synthesized second '>' token.
848       // FIXME: Increment Column and set OriginalColumn.
849       Token Greater = FormatTok->Tok;
850       FormatTok = new (Allocator.Allocate()) FormatToken;
851       FormatTok->Tok = Greater;
852       SourceLocation GreaterLocation =
853           FormatTok->Tok.getLocation().getLocWithOffset(1);
854       FormatTok->WhitespaceRange =
855           SourceRange(GreaterLocation, GreaterLocation);
856       FormatTok->TokenText = ">";
857       FormatTok->ColumnWidth = 1;
858       GreaterStashed = false;
859       return FormatTok;
860     }
861 
862     FormatTok = new (Allocator.Allocate()) FormatToken;
863     readRawToken(*FormatTok);
864     SourceLocation WhitespaceStart =
865         FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace);
866     FormatTok->IsFirst = IsFirstToken;
867     IsFirstToken = false;
868 
869     // Consume and record whitespace until we find a significant token.
870     unsigned WhitespaceLength = TrailingWhitespace;
871     while (FormatTok->Tok.is(tok::unknown)) {
872       for (int i = 0, e = FormatTok->TokenText.size(); i != e; ++i) {
873         switch (FormatTok->TokenText[i]) {
874         case '\n':
875           ++FormatTok->NewlinesBefore;
876           // FIXME: This is technically incorrect, as it could also
877           // be a literal backslash at the end of the line.
878           if (i == 0 || (FormatTok->TokenText[i - 1] != '\\' &&
879                          (FormatTok->TokenText[i - 1] != '\r' || i == 1 ||
880                           FormatTok->TokenText[i - 2] != '\\')))
881             FormatTok->HasUnescapedNewline = true;
882           FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
883           Column = 0;
884           break;
885         case '\r':
886         case '\f':
887         case '\v':
888           Column = 0;
889           break;
890         case ' ':
891           ++Column;
892           break;
893         case '\t':
894           Column += Style.TabWidth - Column % Style.TabWidth;
895           break;
896         case '\\':
897           if (i + 1 == e || (FormatTok->TokenText[i + 1] != '\r' &&
898                              FormatTok->TokenText[i + 1] != '\n'))
899             FormatTok->Type = TT_ImplicitStringLiteral;
900           break;
901         default:
902           FormatTok->Type = TT_ImplicitStringLiteral;
903           ++Column;
904           break;
905         }
906       }
907 
908       if (FormatTok->is(TT_ImplicitStringLiteral))
909         break;
910       WhitespaceLength += FormatTok->Tok.getLength();
911 
912       readRawToken(*FormatTok);
913     }
914 
915     // In case the token starts with escaped newlines, we want to
916     // take them into account as whitespace - this pattern is quite frequent
917     // in macro definitions.
918     // FIXME: Add a more explicit test.
919     while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' &&
920            FormatTok->TokenText[1] == '\n') {
921       ++FormatTok->NewlinesBefore;
922       WhitespaceLength += 2;
923       Column = 0;
924       FormatTok->TokenText = FormatTok->TokenText.substr(2);
925     }
926 
927     FormatTok->WhitespaceRange = SourceRange(
928         WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength));
929 
930     FormatTok->OriginalColumn = Column;
931 
932     TrailingWhitespace = 0;
933     if (FormatTok->Tok.is(tok::comment)) {
934       // FIXME: Add the trimmed whitespace to Column.
935       StringRef UntrimmedText = FormatTok->TokenText;
936       FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f");
937       TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size();
938     } else if (FormatTok->Tok.is(tok::raw_identifier)) {
939       IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText);
940       FormatTok->Tok.setIdentifierInfo(&Info);
941       FormatTok->Tok.setKind(Info.getTokenID());
942       if (Style.Language == FormatStyle::LK_Java &&
943           FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete)) {
944         FormatTok->Tok.setKind(tok::identifier);
945         FormatTok->Tok.setIdentifierInfo(nullptr);
946       }
947     } else if (FormatTok->Tok.is(tok::greatergreater)) {
948       FormatTok->Tok.setKind(tok::greater);
949       FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
950       GreaterStashed = true;
951     }
952 
953     // Now FormatTok is the next non-whitespace token.
954 
955     StringRef Text = FormatTok->TokenText;
956     size_t FirstNewlinePos = Text.find('\n');
957     if (FirstNewlinePos == StringRef::npos) {
958       // FIXME: ColumnWidth actually depends on the start column, we need to
959       // take this into account when the token is moved.
960       FormatTok->ColumnWidth =
961           encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding);
962       Column += FormatTok->ColumnWidth;
963     } else {
964       FormatTok->IsMultiline = true;
965       // FIXME: ColumnWidth actually depends on the start column, we need to
966       // take this into account when the token is moved.
967       FormatTok->ColumnWidth = encoding::columnWidthWithTabs(
968           Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding);
969 
970       // The last line of the token always starts in column 0.
971       // Thus, the length can be precomputed even in the presence of tabs.
972       FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs(
973           Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth,
974           Encoding);
975       Column = FormatTok->LastLineColumnWidth;
976     }
977 
978     FormatTok->IsForEachMacro =
979         std::binary_search(ForEachMacros.begin(), ForEachMacros.end(),
980                            FormatTok->Tok.getIdentifierInfo());
981 
982     return FormatTok;
983   }
984 
985   FormatToken *FormatTok;
986   bool IsFirstToken;
987   bool GreaterStashed;
988   unsigned Column;
989   unsigned TrailingWhitespace;
990   std::unique_ptr<Lexer> Lex;
991   SourceManager &SourceMgr;
992   FileID ID;
993   FormatStyle &Style;
994   IdentifierTable IdentTable;
995   AdditionalKeywords Keywords;
996   encoding::Encoding Encoding;
997   llvm::SpecificBumpPtrAllocator<FormatToken> Allocator;
998   // Index (in 'Tokens') of the last token that starts a new line.
999   unsigned FirstInLineIndex;
1000   SmallVector<FormatToken *, 16> Tokens;
1001   SmallVector<IdentifierInfo *, 8> ForEachMacros;
1002 
1003   bool FormattingDisabled;
1004 
1005   void readRawToken(FormatToken &Tok) {
1006     Lex->LexFromRawLexer(Tok.Tok);
1007     Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()),
1008                               Tok.Tok.getLength());
1009     // For formatting, treat unterminated string literals like normal string
1010     // literals.
1011     if (Tok.is(tok::unknown)) {
1012       if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') {
1013         Tok.Tok.setKind(tok::string_literal);
1014         Tok.IsUnterminatedLiteral = true;
1015       } else if (Style.Language == FormatStyle::LK_JavaScript &&
1016                  Tok.TokenText == "''") {
1017         Tok.Tok.setKind(tok::char_constant);
1018       }
1019     }
1020 
1021     if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" ||
1022                                  Tok.TokenText == "/* clang-format on */")) {
1023       FormattingDisabled = false;
1024     }
1025 
1026     Tok.Finalized = FormattingDisabled;
1027 
1028     if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" ||
1029                                  Tok.TokenText == "/* clang-format off */")) {
1030       FormattingDisabled = true;
1031     }
1032   }
1033 
1034   void resetLexer(unsigned Offset) {
1035     StringRef Buffer = SourceMgr.getBufferData(ID);
1036     Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID),
1037                         getFormattingLangOpts(Style), Buffer.begin(),
1038                         Buffer.begin() + Offset, Buffer.end()));
1039     Lex->SetKeepWhitespaceMode(true);
1040   }
1041 };
1042 
1043 static StringRef getLanguageName(FormatStyle::LanguageKind Language) {
1044   switch (Language) {
1045   case FormatStyle::LK_Cpp:
1046     return "C++";
1047   case FormatStyle::LK_Java:
1048     return "Java";
1049   case FormatStyle::LK_JavaScript:
1050     return "JavaScript";
1051   case FormatStyle::LK_Proto:
1052     return "Proto";
1053   default:
1054     return "Unknown";
1055   }
1056 }
1057 
1058 class Formatter : public UnwrappedLineConsumer {
1059 public:
1060   Formatter(const FormatStyle &Style, SourceManager &SourceMgr, FileID ID,
1061             ArrayRef<CharSourceRange> Ranges)
1062       : Style(Style), ID(ID), SourceMgr(SourceMgr),
1063         Whitespaces(SourceMgr, Style,
1064                     inputUsesCRLF(SourceMgr.getBufferData(ID))),
1065         Ranges(Ranges.begin(), Ranges.end()), UnwrappedLines(1),
1066         Encoding(encoding::detectEncoding(SourceMgr.getBufferData(ID))) {
1067     DEBUG(llvm::dbgs() << "File encoding: "
1068                        << (Encoding == encoding::Encoding_UTF8 ? "UTF8"
1069                                                                : "unknown")
1070                        << "\n");
1071     DEBUG(llvm::dbgs() << "Language: " << getLanguageName(Style.Language)
1072                        << "\n");
1073   }
1074 
1075   tooling::Replacements format() {
1076     tooling::Replacements Result;
1077     FormatTokenLexer Tokens(SourceMgr, ID, Style, Encoding);
1078 
1079     UnwrappedLineParser Parser(Style, Tokens.getKeywords(), Tokens.lex(),
1080                                *this);
1081     bool StructuralError = Parser.parse();
1082     assert(UnwrappedLines.rbegin()->empty());
1083     for (unsigned Run = 0, RunE = UnwrappedLines.size(); Run + 1 != RunE;
1084          ++Run) {
1085       DEBUG(llvm::dbgs() << "Run " << Run << "...\n");
1086       SmallVector<AnnotatedLine *, 16> AnnotatedLines;
1087       for (unsigned i = 0, e = UnwrappedLines[Run].size(); i != e; ++i) {
1088         AnnotatedLines.push_back(new AnnotatedLine(UnwrappedLines[Run][i]));
1089       }
1090       tooling::Replacements RunResult =
1091           format(AnnotatedLines, StructuralError, Tokens);
1092       DEBUG({
1093         llvm::dbgs() << "Replacements for run " << Run << ":\n";
1094         for (tooling::Replacements::iterator I = RunResult.begin(),
1095                                              E = RunResult.end();
1096              I != E; ++I) {
1097           llvm::dbgs() << I->toString() << "\n";
1098         }
1099       });
1100       for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1101         delete AnnotatedLines[i];
1102       }
1103       Result.insert(RunResult.begin(), RunResult.end());
1104       Whitespaces.reset();
1105     }
1106     return Result;
1107   }
1108 
1109   tooling::Replacements format(SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
1110                                bool StructuralError, FormatTokenLexer &Tokens) {
1111     TokenAnnotator Annotator(Style, Tokens.getKeywords());
1112     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1113       Annotator.annotate(*AnnotatedLines[i]);
1114     }
1115     deriveLocalStyle(AnnotatedLines);
1116     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1117       Annotator.calculateFormattingInformation(*AnnotatedLines[i]);
1118     }
1119     computeAffectedLines(AnnotatedLines.begin(), AnnotatedLines.end());
1120 
1121     Annotator.setCommentLineLevels(AnnotatedLines);
1122     ContinuationIndenter Indenter(Style, Tokens.getKeywords(), SourceMgr,
1123                                   Whitespaces, Encoding,
1124                                   BinPackInconclusiveFunctions);
1125     UnwrappedLineFormatter Formatter(&Indenter, &Whitespaces, Style);
1126     Formatter.format(AnnotatedLines, /*DryRun=*/false);
1127     return Whitespaces.generateReplacements();
1128   }
1129 
1130 private:
1131   // Determines which lines are affected by the SourceRanges given as input.
1132   // Returns \c true if at least one line between I and E or one of their
1133   // children is affected.
1134   bool computeAffectedLines(SmallVectorImpl<AnnotatedLine *>::iterator I,
1135                             SmallVectorImpl<AnnotatedLine *>::iterator E) {
1136     bool SomeLineAffected = false;
1137     const AnnotatedLine *PreviousLine = nullptr;
1138     while (I != E) {
1139       AnnotatedLine *Line = *I;
1140       Line->LeadingEmptyLinesAffected = affectsLeadingEmptyLines(*Line->First);
1141 
1142       // If a line is part of a preprocessor directive, it needs to be formatted
1143       // if any token within the directive is affected.
1144       if (Line->InPPDirective) {
1145         FormatToken *Last = Line->Last;
1146         SmallVectorImpl<AnnotatedLine *>::iterator PPEnd = I + 1;
1147         while (PPEnd != E && !(*PPEnd)->First->HasUnescapedNewline) {
1148           Last = (*PPEnd)->Last;
1149           ++PPEnd;
1150         }
1151 
1152         if (affectsTokenRange(*Line->First, *Last,
1153                               /*IncludeLeadingNewlines=*/false)) {
1154           SomeLineAffected = true;
1155           markAllAsAffected(I, PPEnd);
1156         }
1157         I = PPEnd;
1158         continue;
1159       }
1160 
1161       if (nonPPLineAffected(Line, PreviousLine))
1162         SomeLineAffected = true;
1163 
1164       PreviousLine = Line;
1165       ++I;
1166     }
1167     return SomeLineAffected;
1168   }
1169 
1170   // Determines whether 'Line' is affected by the SourceRanges given as input.
1171   // Returns \c true if line or one if its children is affected.
1172   bool nonPPLineAffected(AnnotatedLine *Line,
1173                          const AnnotatedLine *PreviousLine) {
1174     bool SomeLineAffected = false;
1175     Line->ChildrenAffected =
1176         computeAffectedLines(Line->Children.begin(), Line->Children.end());
1177     if (Line->ChildrenAffected)
1178       SomeLineAffected = true;
1179 
1180     // Stores whether one of the line's tokens is directly affected.
1181     bool SomeTokenAffected = false;
1182     // Stores whether we need to look at the leading newlines of the next token
1183     // in order to determine whether it was affected.
1184     bool IncludeLeadingNewlines = false;
1185 
1186     // Stores whether the first child line of any of this line's tokens is
1187     // affected.
1188     bool SomeFirstChildAffected = false;
1189 
1190     for (FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) {
1191       // Determine whether 'Tok' was affected.
1192       if (affectsTokenRange(*Tok, *Tok, IncludeLeadingNewlines))
1193         SomeTokenAffected = true;
1194 
1195       // Determine whether the first child of 'Tok' was affected.
1196       if (!Tok->Children.empty() && Tok->Children.front()->Affected)
1197         SomeFirstChildAffected = true;
1198 
1199       IncludeLeadingNewlines = Tok->Children.empty();
1200     }
1201 
1202     // Was this line moved, i.e. has it previously been on the same line as an
1203     // affected line?
1204     bool LineMoved = PreviousLine && PreviousLine->Affected &&
1205                      Line->First->NewlinesBefore == 0;
1206 
1207     bool IsContinuedComment =
1208         Line->First->is(tok::comment) && Line->First->Next == nullptr &&
1209         Line->First->NewlinesBefore < 2 && PreviousLine &&
1210         PreviousLine->Affected && PreviousLine->Last->is(tok::comment);
1211 
1212     if (SomeTokenAffected || SomeFirstChildAffected || LineMoved ||
1213         IsContinuedComment) {
1214       Line->Affected = true;
1215       SomeLineAffected = true;
1216     }
1217     return SomeLineAffected;
1218   }
1219 
1220   // Marks all lines between I and E as well as all their children as affected.
1221   void markAllAsAffected(SmallVectorImpl<AnnotatedLine *>::iterator I,
1222                          SmallVectorImpl<AnnotatedLine *>::iterator E) {
1223     while (I != E) {
1224       (*I)->Affected = true;
1225       markAllAsAffected((*I)->Children.begin(), (*I)->Children.end());
1226       ++I;
1227     }
1228   }
1229 
1230   // Returns true if the range from 'First' to 'Last' intersects with one of the
1231   // input ranges.
1232   bool affectsTokenRange(const FormatToken &First, const FormatToken &Last,
1233                          bool IncludeLeadingNewlines) {
1234     SourceLocation Start = First.WhitespaceRange.getBegin();
1235     if (!IncludeLeadingNewlines)
1236       Start = Start.getLocWithOffset(First.LastNewlineOffset);
1237     SourceLocation End = Last.getStartOfNonWhitespace();
1238     End = End.getLocWithOffset(Last.TokenText.size());
1239     CharSourceRange Range = CharSourceRange::getCharRange(Start, End);
1240     return affectsCharSourceRange(Range);
1241   }
1242 
1243   // Returns true if one of the input ranges intersect the leading empty lines
1244   // before 'Tok'.
1245   bool affectsLeadingEmptyLines(const FormatToken &Tok) {
1246     CharSourceRange EmptyLineRange = CharSourceRange::getCharRange(
1247         Tok.WhitespaceRange.getBegin(),
1248         Tok.WhitespaceRange.getBegin().getLocWithOffset(Tok.LastNewlineOffset));
1249     return affectsCharSourceRange(EmptyLineRange);
1250   }
1251 
1252   // Returns true if 'Range' intersects with one of the input ranges.
1253   bool affectsCharSourceRange(const CharSourceRange &Range) {
1254     for (SmallVectorImpl<CharSourceRange>::const_iterator I = Ranges.begin(),
1255                                                           E = Ranges.end();
1256          I != E; ++I) {
1257       if (!SourceMgr.isBeforeInTranslationUnit(Range.getEnd(), I->getBegin()) &&
1258           !SourceMgr.isBeforeInTranslationUnit(I->getEnd(), Range.getBegin()))
1259         return true;
1260     }
1261     return false;
1262   }
1263 
1264   static bool inputUsesCRLF(StringRef Text) {
1265     return Text.count('\r') * 2 > Text.count('\n');
1266   }
1267 
1268   void
1269   deriveLocalStyle(const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
1270     unsigned CountBoundToVariable = 0;
1271     unsigned CountBoundToType = 0;
1272     bool HasCpp03IncompatibleFormat = false;
1273     bool HasBinPackedFunction = false;
1274     bool HasOnePerLineFunction = false;
1275     for (unsigned i = 0, e = AnnotatedLines.size(); i != e; ++i) {
1276       if (!AnnotatedLines[i]->First->Next)
1277         continue;
1278       FormatToken *Tok = AnnotatedLines[i]->First->Next;
1279       while (Tok->Next) {
1280         if (Tok->is(TT_PointerOrReference)) {
1281           bool SpacesBefore =
1282               Tok->WhitespaceRange.getBegin() != Tok->WhitespaceRange.getEnd();
1283           bool SpacesAfter = Tok->Next->WhitespaceRange.getBegin() !=
1284                              Tok->Next->WhitespaceRange.getEnd();
1285           if (SpacesBefore && !SpacesAfter)
1286             ++CountBoundToVariable;
1287           else if (!SpacesBefore && SpacesAfter)
1288             ++CountBoundToType;
1289         }
1290 
1291         if (Tok->WhitespaceRange.getBegin() == Tok->WhitespaceRange.getEnd()) {
1292           if (Tok->is(tok::coloncolon) && Tok->Previous->is(TT_TemplateOpener))
1293             HasCpp03IncompatibleFormat = true;
1294           if (Tok->is(TT_TemplateCloser) &&
1295               Tok->Previous->is(TT_TemplateCloser))
1296             HasCpp03IncompatibleFormat = true;
1297         }
1298 
1299         if (Tok->PackingKind == PPK_BinPacked)
1300           HasBinPackedFunction = true;
1301         if (Tok->PackingKind == PPK_OnePerLine)
1302           HasOnePerLineFunction = true;
1303 
1304         Tok = Tok->Next;
1305       }
1306     }
1307     if (Style.DerivePointerAlignment) {
1308       if (CountBoundToType > CountBoundToVariable)
1309         Style.PointerAlignment = FormatStyle::PAS_Left;
1310       else if (CountBoundToType < CountBoundToVariable)
1311         Style.PointerAlignment = FormatStyle::PAS_Right;
1312     }
1313     if (Style.Standard == FormatStyle::LS_Auto) {
1314       Style.Standard = HasCpp03IncompatibleFormat ? FormatStyle::LS_Cpp11
1315                                                   : FormatStyle::LS_Cpp03;
1316     }
1317     BinPackInconclusiveFunctions =
1318         HasBinPackedFunction || !HasOnePerLineFunction;
1319   }
1320 
1321   void consumeUnwrappedLine(const UnwrappedLine &TheLine) override {
1322     assert(!UnwrappedLines.empty());
1323     UnwrappedLines.back().push_back(TheLine);
1324   }
1325 
1326   void finishRun() override {
1327     UnwrappedLines.push_back(SmallVector<UnwrappedLine, 16>());
1328   }
1329 
1330   FormatStyle Style;
1331   FileID ID;
1332   SourceManager &SourceMgr;
1333   WhitespaceManager Whitespaces;
1334   SmallVector<CharSourceRange, 8> Ranges;
1335   SmallVector<SmallVector<UnwrappedLine, 16>, 2> UnwrappedLines;
1336 
1337   encoding::Encoding Encoding;
1338   bool BinPackInconclusiveFunctions;
1339 };
1340 
1341 } // end anonymous namespace
1342 
1343 tooling::Replacements reformat(const FormatStyle &Style, Lexer &Lex,
1344                                SourceManager &SourceMgr,
1345                                ArrayRef<CharSourceRange> Ranges) {
1346   if (Style.DisableFormat)
1347     return tooling::Replacements();
1348   return reformat(Style, SourceMgr,
1349                   SourceMgr.getFileID(Lex.getSourceLocation()), Ranges);
1350 }
1351 
1352 tooling::Replacements reformat(const FormatStyle &Style,
1353                                SourceManager &SourceMgr, FileID ID,
1354                                ArrayRef<CharSourceRange> Ranges) {
1355   if (Style.DisableFormat)
1356     return tooling::Replacements();
1357   Formatter formatter(Style, SourceMgr, ID, Ranges);
1358   return formatter.format();
1359 }
1360 
1361 tooling::Replacements reformat(const FormatStyle &Style, StringRef Code,
1362                                ArrayRef<tooling::Range> Ranges,
1363                                StringRef FileName) {
1364   if (Style.DisableFormat)
1365     return tooling::Replacements();
1366 
1367   FileManager Files((FileSystemOptions()));
1368   DiagnosticsEngine Diagnostics(
1369       IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs),
1370       new DiagnosticOptions);
1371   SourceManager SourceMgr(Diagnostics, Files);
1372   std::unique_ptr<llvm::MemoryBuffer> Buf =
1373       llvm::MemoryBuffer::getMemBuffer(Code, FileName);
1374   const clang::FileEntry *Entry =
1375       Files.getVirtualFile(FileName, Buf->getBufferSize(), 0);
1376   SourceMgr.overrideFileContents(Entry, std::move(Buf));
1377   FileID ID =
1378       SourceMgr.createFileID(Entry, SourceLocation(), clang::SrcMgr::C_User);
1379   SourceLocation StartOfFile = SourceMgr.getLocForStartOfFile(ID);
1380   std::vector<CharSourceRange> CharRanges;
1381   for (const tooling::Range &Range : Ranges) {
1382     SourceLocation Start = StartOfFile.getLocWithOffset(Range.getOffset());
1383     SourceLocation End = Start.getLocWithOffset(Range.getLength());
1384     CharRanges.push_back(CharSourceRange::getCharRange(Start, End));
1385   }
1386   return reformat(Style, SourceMgr, ID, CharRanges);
1387 }
1388 
1389 LangOptions getFormattingLangOpts(const FormatStyle &Style) {
1390   LangOptions LangOpts;
1391   LangOpts.CPlusPlus = 1;
1392   LangOpts.CPlusPlus11 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1393   LangOpts.CPlusPlus14 = Style.Standard == FormatStyle::LS_Cpp03 ? 0 : 1;
1394   LangOpts.LineComment = 1;
1395   bool AlternativeOperators = Style.Language != FormatStyle::LK_JavaScript &&
1396                               Style.Language != FormatStyle::LK_Java;
1397   LangOpts.CXXOperatorNames = AlternativeOperators ? 1 : 0;
1398   LangOpts.Bool = 1;
1399   LangOpts.ObjC1 = 1;
1400   LangOpts.ObjC2 = 1;
1401   return LangOpts;
1402 }
1403 
1404 const char *StyleOptionHelpDescription =
1405     "Coding style, currently supports:\n"
1406     "  LLVM, Google, Chromium, Mozilla, WebKit.\n"
1407     "Use -style=file to load style configuration from\n"
1408     ".clang-format file located in one of the parent\n"
1409     "directories of the source file (or current\n"
1410     "directory for stdin).\n"
1411     "Use -style=\"{key: value, ...}\" to set specific\n"
1412     "parameters, e.g.:\n"
1413     "  -style=\"{BasedOnStyle: llvm, IndentWidth: 8}\"";
1414 
1415 static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
1416   if (FileName.endswith(".java")) {
1417     return FormatStyle::LK_Java;
1418   } else if (FileName.endswith_lower(".js")) {
1419     return FormatStyle::LK_JavaScript;
1420   } else if (FileName.endswith_lower(".proto") ||
1421              FileName.endswith_lower(".protodevel")) {
1422     return FormatStyle::LK_Proto;
1423   }
1424   return FormatStyle::LK_Cpp;
1425 }
1426 
1427 FormatStyle getStyle(StringRef StyleName, StringRef FileName,
1428                      StringRef FallbackStyle) {
1429   FormatStyle Style = getLLVMStyle();
1430   Style.Language = getLanguageByFileName(FileName);
1431   if (!getPredefinedStyle(FallbackStyle, Style.Language, &Style)) {
1432     llvm::errs() << "Invalid fallback style \"" << FallbackStyle
1433                  << "\" using LLVM style\n";
1434     return Style;
1435   }
1436 
1437   if (StyleName.startswith("{")) {
1438     // Parse YAML/JSON style from the command line.
1439     if (std::error_code ec = parseConfiguration(StyleName, &Style)) {
1440       llvm::errs() << "Error parsing -style: " << ec.message() << ", using "
1441                    << FallbackStyle << " style\n";
1442     }
1443     return Style;
1444   }
1445 
1446   if (!StyleName.equals_lower("file")) {
1447     if (!getPredefinedStyle(StyleName, Style.Language, &Style))
1448       llvm::errs() << "Invalid value for -style, using " << FallbackStyle
1449                    << " style\n";
1450     return Style;
1451   }
1452 
1453   // Look for .clang-format/_clang-format file in the file's parent directories.
1454   SmallString<128> UnsuitableConfigFiles;
1455   SmallString<128> Path(FileName);
1456   llvm::sys::fs::make_absolute(Path);
1457   for (StringRef Directory = Path; !Directory.empty();
1458        Directory = llvm::sys::path::parent_path(Directory)) {
1459     if (!llvm::sys::fs::is_directory(Directory))
1460       continue;
1461     SmallString<128> ConfigFile(Directory);
1462 
1463     llvm::sys::path::append(ConfigFile, ".clang-format");
1464     DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1465     bool IsFile = false;
1466     // Ignore errors from is_regular_file: we only need to know if we can read
1467     // the file or not.
1468     llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile);
1469 
1470     if (!IsFile) {
1471       // Try _clang-format too, since dotfiles are not commonly used on Windows.
1472       ConfigFile = Directory;
1473       llvm::sys::path::append(ConfigFile, "_clang-format");
1474       DEBUG(llvm::dbgs() << "Trying " << ConfigFile << "...\n");
1475       llvm::sys::fs::is_regular_file(Twine(ConfigFile), IsFile);
1476     }
1477 
1478     if (IsFile) {
1479       llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Text =
1480           llvm::MemoryBuffer::getFile(ConfigFile.c_str());
1481       if (std::error_code EC = Text.getError()) {
1482         llvm::errs() << EC.message() << "\n";
1483         break;
1484       }
1485       if (std::error_code ec =
1486               parseConfiguration(Text.get()->getBuffer(), &Style)) {
1487         if (ec == ParseError::Unsuitable) {
1488           if (!UnsuitableConfigFiles.empty())
1489             UnsuitableConfigFiles.append(", ");
1490           UnsuitableConfigFiles.append(ConfigFile);
1491           continue;
1492         }
1493         llvm::errs() << "Error reading " << ConfigFile << ": " << ec.message()
1494                      << "\n";
1495         break;
1496       }
1497       DEBUG(llvm::dbgs() << "Using configuration file " << ConfigFile << "\n");
1498       return Style;
1499     }
1500   }
1501   llvm::errs() << "Can't find usable .clang-format, using " << FallbackStyle
1502                << " style\n";
1503   if (!UnsuitableConfigFiles.empty()) {
1504     llvm::errs() << "Configuration file(s) do(es) not support "
1505                  << getLanguageName(Style.Language) << ": "
1506                  << UnsuitableConfigFiles << "\n";
1507   }
1508   return Style;
1509 }
1510 
1511 } // namespace format
1512 } // namespace clang
1513