Bug Summary

File:clang/lib/Format/ContinuationIndenter.cpp
Warning:line 559, column 9
Access to field 'PartOfMultiVariableDeclStmt' results in a dereference of a null pointer (loaded from variable 'Previous')

Annotated Source Code

Press '?' to see keyboard shortcuts

clang -cc1 -triple x86_64-pc-linux-gnu -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name ContinuationIndenter.cpp -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -analyzer-config-compatibility-mode=true -mrelocation-model pic -pic-level 2 -mthread-model posix -mframe-pointer=none -relaxed-aliasing -fmath-errno -fno-rounding-math -masm-verbose -mconstructor-aliases -munwind-tables -fuse-init-array -target-cpu x86-64 -dwarf-column-info -debugger-tuning=gdb -ffunction-sections -fdata-sections -resource-dir /usr/lib/llvm-10/lib/clang/10.0.0 -D CLANG_VENDOR="Debian " -D _DEBUG -D _GNU_SOURCE -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D __STDC_LIMIT_MACROS -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/build-llvm/tools/clang/lib/Format -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/build-llvm/tools/clang/include -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/build-llvm/include -I /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/llvm/include -U NDEBUG -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/6.3.0/../../../../include/c++/6.3.0 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/6.3.0/../../../../include/x86_64-linux-gnu/c++/6.3.0 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/6.3.0/../../../../include/x86_64-linux-gnu/c++/6.3.0 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/6.3.0/../../../../include/c++/6.3.0/backward -internal-isystem /usr/local/include -internal-isystem /usr/lib/llvm-10/lib/clang/10.0.0/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-maybe-uninitialized -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/build-llvm/tools/clang/lib/Format -fdebug-prefix-map=/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809=. -ferror-limit 19 -fmessage-length 0 -fvisibility-inlines-hidden -stack-protector 2 -fgnuc-version=4.2.1 -fobjc-runtime=gcc -fno-common -fdiagnostics-show-option -vectorize-loops -vectorize-slp -analyzer-output=html -analyzer-config stable-report-filename=true -faddrsig -o /tmp/scan-build-2019-12-07-102640-14763-1 -x c++ /build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp

/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp

1//===--- ContinuationIndenter.cpp - Format C++ code -----------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8///
9/// \file
10/// This file implements the continuation indenter.
11///
12//===----------------------------------------------------------------------===//
13
14#include "ContinuationIndenter.h"
15#include "BreakableToken.h"
16#include "FormatInternal.h"
17#include "WhitespaceManager.h"
18#include "clang/Basic/OperatorPrecedence.h"
19#include "clang/Basic/SourceManager.h"
20#include "clang/Format/Format.h"
21#include "llvm/Support/Debug.h"
22
23#define DEBUG_TYPE"format-indenter" "format-indenter"
24
25namespace clang {
26namespace format {
27
28// Returns true if a TT_SelectorName should be indented when wrapped,
29// false otherwise.
30static bool shouldIndentWrappedSelectorName(const FormatStyle &Style,
31 LineType LineType) {
32 return Style.IndentWrappedFunctionNames || LineType == LT_ObjCMethodDecl;
33}
34
35// Returns the length of everything up to the first possible line break after
36// the ), ], } or > matching \c Tok.
37static unsigned getLengthToMatchingParen(const FormatToken &Tok,
38 const std::vector<ParenState> &Stack) {
39 // Normally whether or not a break before T is possible is calculated and
40 // stored in T.CanBreakBefore. Braces, array initializers and text proto
41 // messages like `key: < ... >` are an exception: a break is possible
42 // before a closing brace R if a break was inserted after the corresponding
43 // opening brace. The information about whether or not a break is needed
44 // before a closing brace R is stored in the ParenState field
45 // S.BreakBeforeClosingBrace where S is the state that R closes.
46 //
47 // In order to decide whether there can be a break before encountered right
48 // braces, this implementation iterates over the sequence of tokens and over
49 // the paren stack in lockstep, keeping track of the stack level which visited
50 // right braces correspond to in MatchingStackIndex.
51 //
52 // For example, consider:
53 // L. <- line number
54 // 1. {
55 // 2. {1},
56 // 3. {2},
57 // 4. {{3}}}
58 // ^ where we call this method with this token.
59 // The paren stack at this point contains 3 brace levels:
60 // 0. { at line 1, BreakBeforeClosingBrace: true
61 // 1. first { at line 4, BreakBeforeClosingBrace: false
62 // 2. second { at line 4, BreakBeforeClosingBrace: false,
63 // where there might be fake parens levels in-between these levels.
64 // The algorithm will start at the first } on line 4, which is the matching
65 // brace of the initial left brace and at level 2 of the stack. Then,
66 // examining BreakBeforeClosingBrace: false at level 2, it will continue to
67 // the second } on line 4, and will traverse the stack downwards until it
68 // finds the matching { on level 1. Then, examining BreakBeforeClosingBrace:
69 // false at level 1, it will continue to the third } on line 4 and will
70 // traverse the stack downwards until it finds the matching { on level 0.
71 // Then, examining BreakBeforeClosingBrace: true at level 0, the algorithm
72 // will stop and will use the second } on line 4 to determine the length to
73 // return, as in this example the range will include the tokens: {3}}
74 //
75 // The algorithm will only traverse the stack if it encounters braces, array
76 // initializer squares or text proto angle brackets.
77 if (!Tok.MatchingParen)
78 return 0;
79 FormatToken *End = Tok.MatchingParen;
80 // Maintains a stack level corresponding to the current End token.
81 int MatchingStackIndex = Stack.size() - 1;
82 // Traverses the stack downwards, looking for the level to which LBrace
83 // corresponds. Returns either a pointer to the matching level or nullptr if
84 // LParen is not found in the initial portion of the stack up to
85 // MatchingStackIndex.
86 auto FindParenState = [&](const FormatToken *LBrace) -> const ParenState * {
87 while (MatchingStackIndex >= 0 && Stack[MatchingStackIndex].Tok != LBrace)
88 --MatchingStackIndex;
89 return MatchingStackIndex >= 0 ? &Stack[MatchingStackIndex] : nullptr;
90 };
91 for (; End->Next; End = End->Next) {
92 if (End->Next->CanBreakBefore)
93 break;
94 if (!End->Next->closesScope())
95 continue;
96 if (End->Next->MatchingParen &&
97 End->Next->MatchingParen->isOneOf(
98 tok::l_brace, TT_ArrayInitializerLSquare, tok::less)) {
99 const ParenState *State = FindParenState(End->Next->MatchingParen);
100 if (State && State->BreakBeforeClosingBrace)
101 break;
102 }
103 }
104 return End->TotalLength - Tok.TotalLength + 1;
105}
106
107static unsigned getLengthToNextOperator(const FormatToken &Tok) {
108 if (!Tok.NextOperator)
109 return 0;
110 return Tok.NextOperator->TotalLength - Tok.TotalLength;
111}
112
113// Returns \c true if \c Tok is the "." or "->" of a call and starts the next
114// segment of a builder type call.
115static bool startsSegmentOfBuilderTypeCall(const FormatToken &Tok) {
116 return Tok.isMemberAccess() && Tok.Previous && Tok.Previous->closesScope();
117}
118
119// Returns \c true if \c Current starts a new parameter.
120static bool startsNextParameter(const FormatToken &Current,
121 const FormatStyle &Style) {
122 const FormatToken &Previous = *Current.Previous;
123 if (Current.is(TT_CtorInitializerComma) &&
124 Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma)
125 return true;
126 if (Style.Language == FormatStyle::LK_Proto && Current.is(TT_SelectorName))
127 return true;
128 return Previous.is(tok::comma) && !Current.isTrailingComment() &&
129 ((Previous.isNot(TT_CtorInitializerComma) ||
130 Style.BreakConstructorInitializers !=
131 FormatStyle::BCIS_BeforeComma) &&
132 (Previous.isNot(TT_InheritanceComma) ||
133 Style.BreakInheritanceList != FormatStyle::BILS_BeforeComma));
134}
135
136static bool opensProtoMessageField(const FormatToken &LessTok,
137 const FormatStyle &Style) {
138 if (LessTok.isNot(tok::less))
139 return false;
140 return Style.Language == FormatStyle::LK_TextProto ||
141 (Style.Language == FormatStyle::LK_Proto &&
142 (LessTok.NestingLevel > 0 ||
143 (LessTok.Previous && LessTok.Previous->is(tok::equal))));
144}
145
146// Returns the delimiter of a raw string literal, or None if TokenText is not
147// the text of a raw string literal. The delimiter could be the empty string.
148// For example, the delimiter of R"deli(cont)deli" is deli.
149static llvm::Optional<StringRef> getRawStringDelimiter(StringRef TokenText) {
150 if (TokenText.size() < 5 // The smallest raw string possible is 'R"()"'.
151 || !TokenText.startswith("R\"") || !TokenText.endswith("\""))
152 return None;
153
154 // A raw string starts with 'R"<delimiter>(' and delimiter is ascii and has
155 // size at most 16 by the standard, so the first '(' must be among the first
156 // 19 bytes.
157 size_t LParenPos = TokenText.substr(0, 19).find_first_of('(');
158 if (LParenPos == StringRef::npos)
159 return None;
160 StringRef Delimiter = TokenText.substr(2, LParenPos - 2);
161
162 // Check that the string ends in ')Delimiter"'.
163 size_t RParenPos = TokenText.size() - Delimiter.size() - 2;
164 if (TokenText[RParenPos] != ')')
165 return None;
166 if (!TokenText.substr(RParenPos + 1).startswith(Delimiter))
167 return None;
168 return Delimiter;
169}
170
171// Returns the canonical delimiter for \p Language, or the empty string if no
172// canonical delimiter is specified.
173static StringRef
174getCanonicalRawStringDelimiter(const FormatStyle &Style,
175 FormatStyle::LanguageKind Language) {
176 for (const auto &Format : Style.RawStringFormats) {
177 if (Format.Language == Language)
178 return StringRef(Format.CanonicalDelimiter);
179 }
180 return "";
181}
182
183RawStringFormatStyleManager::RawStringFormatStyleManager(
184 const FormatStyle &CodeStyle) {
185 for (const auto &RawStringFormat : CodeStyle.RawStringFormats) {
186 llvm::Optional<FormatStyle> LanguageStyle =
187 CodeStyle.GetLanguageStyle(RawStringFormat.Language);
188 if (!LanguageStyle) {
189 FormatStyle PredefinedStyle;
190 if (!getPredefinedStyle(RawStringFormat.BasedOnStyle,
191 RawStringFormat.Language, &PredefinedStyle)) {
192 PredefinedStyle = getLLVMStyle();
193 PredefinedStyle.Language = RawStringFormat.Language;
194 }
195 LanguageStyle = PredefinedStyle;
196 }
197 LanguageStyle->ColumnLimit = CodeStyle.ColumnLimit;
198 for (StringRef Delimiter : RawStringFormat.Delimiters) {
199 DelimiterStyle.insert({Delimiter, *LanguageStyle});
200 }
201 for (StringRef EnclosingFunction : RawStringFormat.EnclosingFunctions) {
202 EnclosingFunctionStyle.insert({EnclosingFunction, *LanguageStyle});
203 }
204 }
205}
206
207llvm::Optional<FormatStyle>
208RawStringFormatStyleManager::getDelimiterStyle(StringRef Delimiter) const {
209 auto It = DelimiterStyle.find(Delimiter);
210 if (It == DelimiterStyle.end())
211 return None;
212 return It->second;
213}
214
215llvm::Optional<FormatStyle>
216RawStringFormatStyleManager::getEnclosingFunctionStyle(
217 StringRef EnclosingFunction) const {
218 auto It = EnclosingFunctionStyle.find(EnclosingFunction);
219 if (It == EnclosingFunctionStyle.end())
220 return None;
221 return It->second;
222}
223
224ContinuationIndenter::ContinuationIndenter(const FormatStyle &Style,
225 const AdditionalKeywords &Keywords,
226 const SourceManager &SourceMgr,
227 WhitespaceManager &Whitespaces,
228 encoding::Encoding Encoding,
229 bool BinPackInconclusiveFunctions)
230 : Style(Style), Keywords(Keywords), SourceMgr(SourceMgr),
231 Whitespaces(Whitespaces), Encoding(Encoding),
232 BinPackInconclusiveFunctions(BinPackInconclusiveFunctions),
233 CommentPragmasRegex(Style.CommentPragmas), RawStringFormats(Style) {}
234
235LineState ContinuationIndenter::getInitialState(unsigned FirstIndent,
236 unsigned FirstStartColumn,
237 const AnnotatedLine *Line,
238 bool DryRun) {
239 LineState State;
240 State.FirstIndent = FirstIndent;
241 if (FirstStartColumn && Line->First->NewlinesBefore == 0)
242 State.Column = FirstStartColumn;
243 else
244 State.Column = FirstIndent;
245 // With preprocessor directive indentation, the line starts on column 0
246 // since it's indented after the hash, but FirstIndent is set to the
247 // preprocessor indent.
248 if (Style.IndentPPDirectives == FormatStyle::PPDIS_AfterHash &&
249 (Line->Type == LT_PreprocessorDirective ||
250 Line->Type == LT_ImportStatement))
251 State.Column = 0;
252 State.Line = Line;
253 State.NextToken = Line->First;
254 State.Stack.push_back(ParenState(/*Tok=*/nullptr, FirstIndent, FirstIndent,
255 /*AvoidBinPacking=*/false,
256 /*NoLineBreak=*/false));
257 State.LineContainsContinuedForLoopSection = false;
258 State.NoContinuation = false;
259 State.StartOfStringLiteral = 0;
260 State.StartOfLineLevel = 0;
261 State.LowestLevelOnLine = 0;
262 State.IgnoreStackForComparison = false;
263
264 if (Style.Language == FormatStyle::LK_TextProto) {
265 // We need this in order to deal with the bin packing of text fields at
266 // global scope.
267 State.Stack.back().AvoidBinPacking = true;
268 State.Stack.back().BreakBeforeParameter = true;
269 State.Stack.back().AlignColons = false;
270 }
271
272 // The first token has already been indented and thus consumed.
273 moveStateToNextToken(State, DryRun, /*Newline=*/false);
274 return State;
275}
276
277bool ContinuationIndenter::canBreak(const LineState &State) {
278 const FormatToken &Current = *State.NextToken;
279 const FormatToken &Previous = *Current.Previous;
280 assert(&Previous == Current.Previous)((&Previous == Current.Previous) ? static_cast<void>
(0) : __assert_fail ("&Previous == Current.Previous", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 280, __PRETTY_FUNCTION__))
;
281 if (!Current.CanBreakBefore && !(State.Stack.back().BreakBeforeClosingBrace &&
282 Current.closesBlockOrBlockTypeList(Style)))
283 return false;
284 // The opening "{" of a braced list has to be on the same line as the first
285 // element if it is nested in another braced init list or function call.
286 if (!Current.MustBreakBefore && Previous.is(tok::l_brace) &&
287 Previous.isNot(TT_DictLiteral) && Previous.BlockKind == BK_BracedInit &&
288 Previous.Previous &&
289 Previous.Previous->isOneOf(tok::l_brace, tok::l_paren, tok::comma))
290 return false;
291 // This prevents breaks like:
292 // ...
293 // SomeParameter, OtherParameter).DoSomething(
294 // ...
295 // As they hide "DoSomething" and are generally bad for readability.
296 if (Previous.opensScope() && Previous.isNot(tok::l_brace) &&
297 State.LowestLevelOnLine < State.StartOfLineLevel &&
298 State.LowestLevelOnLine < Current.NestingLevel)
299 return false;
300 if (Current.isMemberAccess() && State.Stack.back().ContainsUnwrappedBuilder)
301 return false;
302
303 // Don't create a 'hanging' indent if there are multiple blocks in a single
304 // statement.
305 if (Previous.is(tok::l_brace) && State.Stack.size() > 1 &&
306 State.Stack[State.Stack.size() - 2].NestedBlockInlined &&
307 State.Stack[State.Stack.size() - 2].HasMultipleNestedBlocks)
308 return false;
309
310 // Don't break after very short return types (e.g. "void") as that is often
311 // unexpected.
312 if (Current.is(TT_FunctionDeclarationName) && State.Column < 6) {
313 if (Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_None)
314 return false;
315 }
316
317 // If binary operators are moved to the next line (including commas for some
318 // styles of constructor initializers), that's always ok.
319 if (!Current.isOneOf(TT_BinaryOperator, tok::comma) &&
320 State.Stack.back().NoLineBreakInOperand)
321 return false;
322
323 if (Previous.is(tok::l_square) && Previous.is(TT_ObjCMethodExpr))
324 return false;
325
326 return !State.Stack.back().NoLineBreak;
327}
328
329bool ContinuationIndenter::mustBreak(const LineState &State) {
330 const FormatToken &Current = *State.NextToken;
331 const FormatToken &Previous = *Current.Previous;
332 if (Current.MustBreakBefore || Current.is(TT_InlineASMColon))
333 return true;
334 if (State.Stack.back().BreakBeforeClosingBrace &&
335 Current.closesBlockOrBlockTypeList(Style))
336 return true;
337 if (Previous.is(tok::semi) && State.LineContainsContinuedForLoopSection)
338 return true;
339 if (Style.Language == FormatStyle::LK_ObjC &&
340 Current.ObjCSelectorNameParts > 1 &&
341 Current.startsSequence(TT_SelectorName, tok::colon, tok::caret)) {
342 return true;
343 }
344 if ((startsNextParameter(Current, Style) || Previous.is(tok::semi) ||
345 (Previous.is(TT_TemplateCloser) && Current.is(TT_StartOfName) &&
346 Style.isCpp() &&
347 // FIXME: This is a temporary workaround for the case where clang-format
348 // sets BreakBeforeParameter to avoid bin packing and this creates a
349 // completely unnecessary line break after a template type that isn't
350 // line-wrapped.
351 (Previous.NestingLevel == 1 || Style.BinPackParameters)) ||
352 (Style.BreakBeforeTernaryOperators && Current.is(TT_ConditionalExpr) &&
353 Previous.isNot(tok::question)) ||
354 (!Style.BreakBeforeTernaryOperators &&
355 Previous.is(TT_ConditionalExpr))) &&
356 State.Stack.back().BreakBeforeParameter && !Current.isTrailingComment() &&
357 !Current.isOneOf(tok::r_paren, tok::r_brace))
358 return true;
359 if (((Previous.is(TT_DictLiteral) && Previous.is(tok::l_brace)) ||
360 (Previous.is(TT_ArrayInitializerLSquare) &&
361 Previous.ParameterCount > 1) ||
362 opensProtoMessageField(Previous, Style)) &&
363 Style.ColumnLimit > 0 &&
364 getLengthToMatchingParen(Previous, State.Stack) + State.Column - 1 >
365 getColumnLimit(State))
366 return true;
367
368 const FormatToken &BreakConstructorInitializersToken =
369 Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon
370 ? Previous
371 : Current;
372 if (BreakConstructorInitializersToken.is(TT_CtorInitializerColon) &&
373 (State.Column + State.Line->Last->TotalLength - Previous.TotalLength >
374 getColumnLimit(State) ||
375 State.Stack.back().BreakBeforeParameter) &&
376 (Style.AllowShortFunctionsOnASingleLine != FormatStyle::SFS_All ||
377 Style.BreakConstructorInitializers != FormatStyle::BCIS_BeforeColon ||
378 Style.ColumnLimit != 0))
379 return true;
380
381 if (Current.is(TT_ObjCMethodExpr) && !Previous.is(TT_SelectorName) &&
382 State.Line->startsWith(TT_ObjCMethodSpecifier))
383 return true;
384 if (Current.is(TT_SelectorName) && !Previous.is(tok::at) &&
385 State.Stack.back().ObjCSelectorNameFound &&
386 State.Stack.back().BreakBeforeParameter)
387 return true;
388
389 unsigned NewLineColumn = getNewLineColumn(State);
390 if (Current.isMemberAccess() && Style.ColumnLimit != 0 &&
391 State.Column + getLengthToNextOperator(Current) > Style.ColumnLimit &&
392 (State.Column > NewLineColumn ||
393 Current.NestingLevel < State.StartOfLineLevel))
394 return true;
395
396 if (startsSegmentOfBuilderTypeCall(Current) &&
397 (State.Stack.back().CallContinuation != 0 ||
398 State.Stack.back().BreakBeforeParameter) &&
399 // JavaScript is treated different here as there is a frequent pattern:
400 // SomeFunction(function() {
401 // ...
402 // }.bind(...));
403 // FIXME: We should find a more generic solution to this problem.
404 !(State.Column <= NewLineColumn &&
405 Style.Language == FormatStyle::LK_JavaScript) &&
406 !(Previous.closesScopeAfterBlock() && State.Column <= NewLineColumn))
407 return true;
408
409 // If the template declaration spans multiple lines, force wrap before the
410 // function/class declaration
411 if (Previous.ClosesTemplateDeclaration &&
412 State.Stack.back().BreakBeforeParameter && Current.CanBreakBefore)
413 return true;
414
415 if (State.Column <= NewLineColumn)
416 return false;
417
418 if (Style.AlwaysBreakBeforeMultilineStrings &&
419 (NewLineColumn == State.FirstIndent + Style.ContinuationIndentWidth ||
420 Previous.is(tok::comma) || Current.NestingLevel < 2) &&
421 !Previous.isOneOf(tok::kw_return, tok::lessless, tok::at,
422 Keywords.kw_dollar) &&
423 !Previous.isOneOf(TT_InlineASMColon, TT_ConditionalExpr) &&
424 nextIsMultilineString(State))
425 return true;
426
427 // Using CanBreakBefore here and below takes care of the decision whether the
428 // current style uses wrapping before or after operators for the given
429 // operator.
430 if (Previous.is(TT_BinaryOperator) && Current.CanBreakBefore) {
431 // If we need to break somewhere inside the LHS of a binary expression, we
432 // should also break after the operator. Otherwise, the formatting would
433 // hide the operator precedence, e.g. in:
434 // if (aaaaaaaaaaaaaa ==
435 // bbbbbbbbbbbbbb && c) {..
436 // For comparisons, we only apply this rule, if the LHS is a binary
437 // expression itself as otherwise, the line breaks seem superfluous.
438 // We need special cases for ">>" which we have split into two ">" while
439 // lexing in order to make template parsing easier.
440 bool IsComparison = (Previous.getPrecedence() == prec::Relational ||
441 Previous.getPrecedence() == prec::Equality ||
442 Previous.getPrecedence() == prec::Spaceship) &&
443 Previous.Previous &&
444 Previous.Previous->isNot(TT_BinaryOperator); // For >>.
445 bool LHSIsBinaryExpr =
446 Previous.Previous && Previous.Previous->EndsBinaryExpression;
447 if ((!IsComparison || LHSIsBinaryExpr) && !Current.isTrailingComment() &&
448 Previous.getPrecedence() != prec::Assignment &&
449 State.Stack.back().BreakBeforeParameter)
450 return true;
451 } else if (Current.is(TT_BinaryOperator) && Current.CanBreakBefore &&
452 State.Stack.back().BreakBeforeParameter) {
453 return true;
454 }
455
456 // Same as above, but for the first "<<" operator.
457 if (Current.is(tok::lessless) && Current.isNot(TT_OverloadedOperator) &&
458 State.Stack.back().BreakBeforeParameter &&
459 State.Stack.back().FirstLessLess == 0)
460 return true;
461
462 if (Current.NestingLevel == 0 && !Current.isTrailingComment()) {
463 // Always break after "template <...>" and leading annotations. This is only
464 // for cases where the entire line does not fit on a single line as a
465 // different LineFormatter would be used otherwise.
466 if (Previous.ClosesTemplateDeclaration)
467 return Style.AlwaysBreakTemplateDeclarations != FormatStyle::BTDS_No;
468 if (Previous.is(TT_FunctionAnnotationRParen))
469 return true;
470 if (Previous.is(TT_LeadingJavaAnnotation) && Current.isNot(tok::l_paren) &&
471 Current.isNot(TT_LeadingJavaAnnotation))
472 return true;
473 }
474
475 // If the return type spans multiple lines, wrap before the function name.
476 if (((Current.is(TT_FunctionDeclarationName) &&
477 // Don't break before a C# function when no break after return type
478 (!Style.isCSharp() ||
479 Style.AlwaysBreakAfterReturnType != FormatStyle::RTBS_None)) ||
480 (Current.is(tok::kw_operator) && !Previous.is(tok::coloncolon))) &&
481 !Previous.is(tok::kw_template) && State.Stack.back().BreakBeforeParameter)
482 return true;
483
484 // The following could be precomputed as they do not depend on the state.
485 // However, as they should take effect only if the UnwrappedLine does not fit
486 // into the ColumnLimit, they are checked here in the ContinuationIndenter.
487 if (Style.ColumnLimit != 0 && Previous.BlockKind == BK_Block &&
488 Previous.is(tok::l_brace) && !Current.isOneOf(tok::r_brace, tok::comment))
489 return true;
490
491 if (Current.is(tok::lessless) &&
492 ((Previous.is(tok::identifier) && Previous.TokenText == "endl") ||
493 (Previous.Tok.isLiteral() && (Previous.TokenText.endswith("\\n\"") ||
494 Previous.TokenText == "\'\\n\'"))))
495 return true;
496
497 if (Previous.is(TT_BlockComment) && Previous.IsMultiline)
498 return true;
499
500 if (State.NoContinuation)
501 return true;
502
503 return false;
504}
505
506unsigned ContinuationIndenter::addTokenToState(LineState &State, bool Newline,
507 bool DryRun,
508 unsigned ExtraSpaces) {
509 const FormatToken &Current = *State.NextToken;
510
511 assert(!State.Stack.empty())((!State.Stack.empty()) ? static_cast<void> (0) : __assert_fail
("!State.Stack.empty()", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 511, __PRETTY_FUNCTION__))
;
512 State.NoContinuation = false;
513
514 if ((Current.is(TT_ImplicitStringLiteral) &&
515 (Current.Previous->Tok.getIdentifierInfo() == nullptr ||
516 Current.Previous->Tok.getIdentifierInfo()->getPPKeywordID() ==
517 tok::pp_not_keyword))) {
518 unsigned EndColumn =
519 SourceMgr.getSpellingColumnNumber(Current.WhitespaceRange.getEnd());
520 if (Current.LastNewlineOffset != 0) {
521 // If there is a newline within this token, the final column will solely
522 // determined by the current end column.
523 State.Column = EndColumn;
524 } else {
525 unsigned StartColumn =
526 SourceMgr.getSpellingColumnNumber(Current.WhitespaceRange.getBegin());
527 assert(EndColumn >= StartColumn)((EndColumn >= StartColumn) ? static_cast<void> (0) :
__assert_fail ("EndColumn >= StartColumn", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 527, __PRETTY_FUNCTION__))
;
528 State.Column += EndColumn - StartColumn;
529 }
530 moveStateToNextToken(State, DryRun, /*Newline=*/false);
531 return 0;
532 }
533
534 unsigned Penalty = 0;
535 if (Newline)
536 Penalty = addTokenOnNewLine(State, DryRun);
537 else
538 addTokenOnCurrentLine(State, DryRun, ExtraSpaces);
539
540 return moveStateToNextToken(State, DryRun, Newline) + Penalty;
541}
542
543void ContinuationIndenter::addTokenOnCurrentLine(LineState &State, bool DryRun,
544 unsigned ExtraSpaces) {
545 FormatToken &Current = *State.NextToken;
546 const FormatToken &Previous = *State.NextToken->Previous;
1
'Previous' initialized here
547 if (Current.is(tok::equal) &&
2
Calling 'FormatToken::is'
8
Returning from 'FormatToken::is'
17
Taking true branch
548 (State.Line->First->is(tok::kw_for) || Current.NestingLevel == 0) &&
9
Calling 'FormatToken::is'
15
Returning from 'FormatToken::is'
549 State.Stack.back().VariablePos == 0) {
16
Assuming field 'VariablePos' is equal to 0
550 State.Stack.back().VariablePos = State.Column;
551 // Move over * and & if they are bound to the variable name.
552 const FormatToken *Tok = &Previous;
553 while (Tok && State.Stack.back().VariablePos >= Tok->ColumnWidth) {
18
Assuming 'Tok' is null
554 State.Stack.back().VariablePos -= Tok->ColumnWidth;
555 if (Tok->SpacesRequiredBefore != 0)
556 break;
557 Tok = Tok->Previous;
558 }
559 if (Previous.PartOfMultiVariableDeclStmt)
19
Access to field 'PartOfMultiVariableDeclStmt' results in a dereference of a null pointer (loaded from variable 'Previous')
560 State.Stack.back().LastSpace = State.Stack.back().VariablePos;
561 }
562
563 unsigned Spaces = Current.SpacesRequiredBefore + ExtraSpaces;
564
565 // Indent preprocessor directives after the hash if required.
566 int PPColumnCorrection = 0;
567 if (Style.IndentPPDirectives == FormatStyle::PPDIS_AfterHash &&
568 Previous.is(tok::hash) && State.FirstIndent > 0 &&
569 (State.Line->Type == LT_PreprocessorDirective ||
570 State.Line->Type == LT_ImportStatement)) {
571 Spaces += State.FirstIndent;
572
573 // For preprocessor indent with tabs, State.Column will be 1 because of the
574 // hash. This causes second-level indents onward to have an extra space
575 // after the tabs. We avoid this misalignment by subtracting 1 from the
576 // column value passed to replaceWhitespace().
577 if (Style.UseTab != FormatStyle::UT_Never)
578 PPColumnCorrection = -1;
579 }
580
581 if (!DryRun)
582 Whitespaces.replaceWhitespace(Current, /*Newlines=*/0, Spaces,
583 State.Column + Spaces + PPColumnCorrection);
584
585 // If "BreakBeforeInheritanceComma" mode, don't break within the inheritance
586 // declaration unless there is multiple inheritance.
587 if (Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma &&
588 Current.is(TT_InheritanceColon))
589 State.Stack.back().NoLineBreak = true;
590 if (Style.BreakInheritanceList == FormatStyle::BILS_AfterColon &&
591 Previous.is(TT_InheritanceColon))
592 State.Stack.back().NoLineBreak = true;
593
594 if (Current.is(TT_SelectorName) &&
595 !State.Stack.back().ObjCSelectorNameFound) {
596 unsigned MinIndent =
597 std::max(State.FirstIndent + Style.ContinuationIndentWidth,
598 State.Stack.back().Indent);
599 unsigned FirstColonPos = State.Column + Spaces + Current.ColumnWidth;
600 if (Current.LongestObjCSelectorName == 0)
601 State.Stack.back().AlignColons = false;
602 else if (MinIndent + Current.LongestObjCSelectorName > FirstColonPos)
603 State.Stack.back().ColonPos = MinIndent + Current.LongestObjCSelectorName;
604 else
605 State.Stack.back().ColonPos = FirstColonPos;
606 }
607
608 // In "AlwaysBreak" mode, enforce wrapping directly after the parenthesis by
609 // disallowing any further line breaks if there is no line break after the
610 // opening parenthesis. Don't break if it doesn't conserve columns.
611 if (Style.AlignAfterOpenBracket == FormatStyle::BAS_AlwaysBreak &&
612 (Previous.isOneOf(tok::l_paren, TT_TemplateOpener, tok::l_square) ||
613 (Previous.is(tok::l_brace) && Previous.BlockKind != BK_Block &&
614 Style.Cpp11BracedListStyle)) &&
615 State.Column > getNewLineColumn(State) &&
616 (!Previous.Previous || !Previous.Previous->isOneOf(
617 tok::kw_for, tok::kw_while, tok::kw_switch)) &&
618 // Don't do this for simple (no expressions) one-argument function calls
619 // as that feels like needlessly wasting whitespace, e.g.:
620 //
621 // caaaaaaaaaaaall(
622 // caaaaaaaaaaaall(
623 // caaaaaaaaaaaall(
624 // caaaaaaaaaaaaaaaaaaaaaaall(aaaaaaaaaaaaaa, aaaaaaaaa))));
625 Current.FakeLParens.size() > 0 &&
626 Current.FakeLParens.back() > prec::Unknown)
627 State.Stack.back().NoLineBreak = true;
628 if (Previous.is(TT_TemplateString) && Previous.opensScope())
629 State.Stack.back().NoLineBreak = true;
630
631 if (Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign &&
632 Previous.opensScope() && Previous.isNot(TT_ObjCMethodExpr) &&
633 (Current.isNot(TT_LineComment) || Previous.BlockKind == BK_BracedInit))
634 State.Stack.back().Indent = State.Column + Spaces;
635 if (State.Stack.back().AvoidBinPacking && startsNextParameter(Current, Style))
636 State.Stack.back().NoLineBreak = true;
637 if (startsSegmentOfBuilderTypeCall(Current) &&
638 State.Column > getNewLineColumn(State))
639 State.Stack.back().ContainsUnwrappedBuilder = true;
640
641 if (Current.is(TT_LambdaArrow) && Style.Language == FormatStyle::LK_Java)
642 State.Stack.back().NoLineBreak = true;
643 if (Current.isMemberAccess() && Previous.is(tok::r_paren) &&
644 (Previous.MatchingParen &&
645 (Previous.TotalLength - Previous.MatchingParen->TotalLength > 10)))
646 // If there is a function call with long parameters, break before trailing
647 // calls. This prevents things like:
648 // EXPECT_CALL(SomeLongParameter).Times(
649 // 2);
650 // We don't want to do this for short parameters as they can just be
651 // indexes.
652 State.Stack.back().NoLineBreak = true;
653
654 // Don't allow the RHS of an operator to be split over multiple lines unless
655 // there is a line-break right after the operator.
656 // Exclude relational operators, as there, it is always more desirable to
657 // have the LHS 'left' of the RHS.
658 const FormatToken *P = Current.getPreviousNonComment();
659 if (!Current.is(tok::comment) && P &&
660 (P->isOneOf(TT_BinaryOperator, tok::comma) ||
661 (P->is(TT_ConditionalExpr) && P->is(tok::colon))) &&
662 !P->isOneOf(TT_OverloadedOperator, TT_CtorInitializerComma) &&
663 P->getPrecedence() != prec::Assignment &&
664 P->getPrecedence() != prec::Relational &&
665 P->getPrecedence() != prec::Spaceship) {
666 bool BreakBeforeOperator =
667 P->MustBreakBefore || P->is(tok::lessless) ||
668 (P->is(TT_BinaryOperator) &&
669 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None) ||
670 (P->is(TT_ConditionalExpr) && Style.BreakBeforeTernaryOperators);
671 // Don't do this if there are only two operands. In these cases, there is
672 // always a nice vertical separation between them and the extra line break
673 // does not help.
674 bool HasTwoOperands =
675 P->OperatorIndex == 0 && !P->NextOperator && !P->is(TT_ConditionalExpr);
676 if ((!BreakBeforeOperator && !(HasTwoOperands && Style.AlignOperands)) ||
677 (!State.Stack.back().LastOperatorWrapped && BreakBeforeOperator))
678 State.Stack.back().NoLineBreakInOperand = true;
679 }
680
681 State.Column += Spaces;
682 if (Current.isNot(tok::comment) && Previous.is(tok::l_paren) &&
683 Previous.Previous &&
684 (Previous.Previous->is(tok::kw_for) || Previous.Previous->isIf())) {
685 // Treat the condition inside an if as if it was a second function
686 // parameter, i.e. let nested calls have a continuation indent.
687 State.Stack.back().LastSpace = State.Column;
688 State.Stack.back().NestedBlockIndent = State.Column;
689 } else if (!Current.isOneOf(tok::comment, tok::caret) &&
690 ((Previous.is(tok::comma) &&
691 !Previous.is(TT_OverloadedOperator)) ||
692 (Previous.is(tok::colon) && Previous.is(TT_ObjCMethodExpr)))) {
693 State.Stack.back().LastSpace = State.Column;
694 } else if (Previous.is(TT_CtorInitializerColon) &&
695 Style.BreakConstructorInitializers ==
696 FormatStyle::BCIS_AfterColon) {
697 State.Stack.back().Indent = State.Column;
698 State.Stack.back().LastSpace = State.Column;
699 } else if ((Previous.isOneOf(TT_BinaryOperator, TT_ConditionalExpr,
700 TT_CtorInitializerColon)) &&
701 ((Previous.getPrecedence() != prec::Assignment &&
702 (Previous.isNot(tok::lessless) || Previous.OperatorIndex != 0 ||
703 Previous.NextOperator)) ||
704 Current.StartsBinaryExpression)) {
705 // Indent relative to the RHS of the expression unless this is a simple
706 // assignment without binary expression on the RHS. Also indent relative to
707 // unary operators and the colons of constructor initializers.
708 if (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None)
709 State.Stack.back().LastSpace = State.Column;
710 } else if (Previous.is(TT_InheritanceColon)) {
711 State.Stack.back().Indent = State.Column;
712 State.Stack.back().LastSpace = State.Column;
713 } else if (Previous.opensScope()) {
714 // If a function has a trailing call, indent all parameters from the
715 // opening parenthesis. This avoids confusing indents like:
716 // OuterFunction(InnerFunctionCall( // break
717 // ParameterToInnerFunction)) // break
718 // .SecondInnerFunctionCall();
719 bool HasTrailingCall = false;
720 if (Previous.MatchingParen) {
721 const FormatToken *Next = Previous.MatchingParen->getNextNonComment();
722 HasTrailingCall = Next && Next->isMemberAccess();
723 }
724 if (HasTrailingCall && State.Stack.size() > 1 &&
725 State.Stack[State.Stack.size() - 2].CallContinuation == 0)
726 State.Stack.back().LastSpace = State.Column;
727 }
728}
729
730unsigned ContinuationIndenter::addTokenOnNewLine(LineState &State,
731 bool DryRun) {
732 FormatToken &Current = *State.NextToken;
733 const FormatToken &Previous = *State.NextToken->Previous;
734
735 // Extra penalty that needs to be added because of the way certain line
736 // breaks are chosen.
737 unsigned Penalty = 0;
738
739 const FormatToken *PreviousNonComment = Current.getPreviousNonComment();
740 const FormatToken *NextNonComment = Previous.getNextNonComment();
741 if (!NextNonComment)
742 NextNonComment = &Current;
743 // The first line break on any NestingLevel causes an extra penalty in order
744 // prefer similar line breaks.
745 if (!State.Stack.back().ContainsLineBreak)
746 Penalty += 15;
747 State.Stack.back().ContainsLineBreak = true;
748
749 Penalty += State.NextToken->SplitPenalty;
750
751 // Breaking before the first "<<" is generally not desirable if the LHS is
752 // short. Also always add the penalty if the LHS is split over multiple lines
753 // to avoid unnecessary line breaks that just work around this penalty.
754 if (NextNonComment->is(tok::lessless) &&
755 State.Stack.back().FirstLessLess == 0 &&
756 (State.Column <= Style.ColumnLimit / 3 ||
757 State.Stack.back().BreakBeforeParameter))
758 Penalty += Style.PenaltyBreakFirstLessLess;
759
760 State.Column = getNewLineColumn(State);
761
762 // Indent nested blocks relative to this column, unless in a very specific
763 // JavaScript special case where:
764 //
765 // var loooooong_name =
766 // function() {
767 // // code
768 // }
769 //
770 // is common and should be formatted like a free-standing function. The same
771 // goes for wrapping before the lambda return type arrow.
772 if (!Current.is(TT_LambdaArrow) &&
773 (Style.Language != FormatStyle::LK_JavaScript ||
774 Current.NestingLevel != 0 || !PreviousNonComment ||
775 !PreviousNonComment->is(tok::equal) ||
776 !Current.isOneOf(Keywords.kw_async, Keywords.kw_function)))
777 State.Stack.back().NestedBlockIndent = State.Column;
778
779 if (NextNonComment->isMemberAccess()) {
780 if (State.Stack.back().CallContinuation == 0)
781 State.Stack.back().CallContinuation = State.Column;
782 } else if (NextNonComment->is(TT_SelectorName)) {
783 if (!State.Stack.back().ObjCSelectorNameFound) {
784 if (NextNonComment->LongestObjCSelectorName == 0) {
785 State.Stack.back().AlignColons = false;
786 } else {
787 State.Stack.back().ColonPos =
788 (shouldIndentWrappedSelectorName(Style, State.Line->Type)
789 ? std::max(State.Stack.back().Indent,
790 State.FirstIndent + Style.ContinuationIndentWidth)
791 : State.Stack.back().Indent) +
792 std::max(NextNonComment->LongestObjCSelectorName,
793 NextNonComment->ColumnWidth);
794 }
795 } else if (State.Stack.back().AlignColons &&
796 State.Stack.back().ColonPos <= NextNonComment->ColumnWidth) {
797 State.Stack.back().ColonPos = State.Column + NextNonComment->ColumnWidth;
798 }
799 } else if (PreviousNonComment && PreviousNonComment->is(tok::colon) &&
800 PreviousNonComment->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral)) {
801 // FIXME: This is hacky, find a better way. The problem is that in an ObjC
802 // method expression, the block should be aligned to the line starting it,
803 // e.g.:
804 // [aaaaaaaaaaaaaaa aaaaaaaaa: \\ break for some reason
805 // ^(int *i) {
806 // // ...
807 // }];
808 // Thus, we set LastSpace of the next higher NestingLevel, to which we move
809 // when we consume all of the "}"'s FakeRParens at the "{".
810 if (State.Stack.size() > 1)
811 State.Stack[State.Stack.size() - 2].LastSpace =
812 std::max(State.Stack.back().LastSpace, State.Stack.back().Indent) +
813 Style.ContinuationIndentWidth;
814 }
815
816 if ((PreviousNonComment &&
817 PreviousNonComment->isOneOf(tok::comma, tok::semi) &&
818 !State.Stack.back().AvoidBinPacking) ||
819 Previous.is(TT_BinaryOperator))
820 State.Stack.back().BreakBeforeParameter = false;
821 if (PreviousNonComment &&
822 PreviousNonComment->isOneOf(TT_TemplateCloser, TT_JavaAnnotation) &&
823 Current.NestingLevel == 0)
824 State.Stack.back().BreakBeforeParameter = false;
825 if (NextNonComment->is(tok::question) ||
826 (PreviousNonComment && PreviousNonComment->is(tok::question)))
827 State.Stack.back().BreakBeforeParameter = true;
828 if (Current.is(TT_BinaryOperator) && Current.CanBreakBefore)
829 State.Stack.back().BreakBeforeParameter = false;
830
831 if (!DryRun) {
832 unsigned MaxEmptyLinesToKeep = Style.MaxEmptyLinesToKeep + 1;
833 if (Current.is(tok::r_brace) && Current.MatchingParen &&
834 // Only strip trailing empty lines for l_braces that have children, i.e.
835 // for function expressions (lambdas, arrows, etc).
836 !Current.MatchingParen->Children.empty()) {
837 // lambdas and arrow functions are expressions, thus their r_brace is not
838 // on its own line, and thus not covered by UnwrappedLineFormatter's logic
839 // about removing empty lines on closing blocks. Special case them here.
840 MaxEmptyLinesToKeep = 1;
841 }
842 unsigned Newlines =
843 std::max(1u, std::min(Current.NewlinesBefore, MaxEmptyLinesToKeep));
844 bool ContinuePPDirective =
845 State.Line->InPPDirective && State.Line->Type != LT_ImportStatement;
846 Whitespaces.replaceWhitespace(Current, Newlines, State.Column, State.Column,
847 ContinuePPDirective);
848 }
849
850 if (!Current.isTrailingComment())
851 State.Stack.back().LastSpace = State.Column;
852 if (Current.is(tok::lessless))
853 // If we are breaking before a "<<", we always want to indent relative to
854 // RHS. This is necessary only for "<<", as we special-case it and don't
855 // always indent relative to the RHS.
856 State.Stack.back().LastSpace += 3; // 3 -> width of "<< ".
857
858 State.StartOfLineLevel = Current.NestingLevel;
859 State.LowestLevelOnLine = Current.NestingLevel;
860
861 // Any break on this level means that the parent level has been broken
862 // and we need to avoid bin packing there.
863 bool NestedBlockSpecialCase =
864 !Style.isCpp() && Current.is(tok::r_brace) && State.Stack.size() > 1 &&
865 State.Stack[State.Stack.size() - 2].NestedBlockInlined;
866 if (!NestedBlockSpecialCase)
867 for (unsigned i = 0, e = State.Stack.size() - 1; i != e; ++i)
868 State.Stack[i].BreakBeforeParameter = true;
869
870 if (PreviousNonComment &&
871 !PreviousNonComment->isOneOf(tok::comma, tok::colon, tok::semi) &&
872 (PreviousNonComment->isNot(TT_TemplateCloser) ||
873 Current.NestingLevel != 0) &&
874 !PreviousNonComment->isOneOf(
875 TT_BinaryOperator, TT_FunctionAnnotationRParen, TT_JavaAnnotation,
876 TT_LeadingJavaAnnotation) &&
877 Current.isNot(TT_BinaryOperator) && !PreviousNonComment->opensScope())
878 State.Stack.back().BreakBeforeParameter = true;
879
880 // If we break after { or the [ of an array initializer, we should also break
881 // before the corresponding } or ].
882 if (PreviousNonComment &&
883 (PreviousNonComment->isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) ||
884 opensProtoMessageField(*PreviousNonComment, Style)))
885 State.Stack.back().BreakBeforeClosingBrace = true;
886
887 if (State.Stack.back().AvoidBinPacking) {
888 // If we are breaking after '(', '{', '<', or this is the break after a ':'
889 // to start a member initializater list in a constructor, this should not
890 // be considered bin packing unless the relevant AllowAll option is false or
891 // this is a dict/object literal.
892 bool PreviousIsBreakingCtorInitializerColon =
893 Previous.is(TT_CtorInitializerColon) &&
894 Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon;
895 if (!(Previous.isOneOf(tok::l_paren, tok::l_brace, TT_BinaryOperator) ||
896 PreviousIsBreakingCtorInitializerColon) ||
897 (!Style.AllowAllParametersOfDeclarationOnNextLine &&
898 State.Line->MustBeDeclaration) ||
899 (!Style.AllowAllArgumentsOnNextLine &&
900 !State.Line->MustBeDeclaration) ||
901 (!Style.AllowAllConstructorInitializersOnNextLine &&
902 PreviousIsBreakingCtorInitializerColon) ||
903 Previous.is(TT_DictLiteral))
904 State.Stack.back().BreakBeforeParameter = true;
905
906 // If we are breaking after a ':' to start a member initializer list,
907 // and we allow all arguments on the next line, we should not break
908 // before the next parameter.
909 if (PreviousIsBreakingCtorInitializerColon &&
910 Style.AllowAllConstructorInitializersOnNextLine)
911 State.Stack.back().BreakBeforeParameter = false;
912 }
913
914 return Penalty;
915}
916
917unsigned ContinuationIndenter::getNewLineColumn(const LineState &State) {
918 if (!State.NextToken || !State.NextToken->Previous)
919 return 0;
920 FormatToken &Current = *State.NextToken;
921 const FormatToken &Previous = *Current.Previous;
922 // If we are continuing an expression, we want to use the continuation indent.
923 unsigned ContinuationIndent =
924 std::max(State.Stack.back().LastSpace, State.Stack.back().Indent) +
925 Style.ContinuationIndentWidth;
926 const FormatToken *PreviousNonComment = Current.getPreviousNonComment();
927 const FormatToken *NextNonComment = Previous.getNextNonComment();
928 if (!NextNonComment)
929 NextNonComment = &Current;
930
931 // Java specific bits.
932 if (Style.Language == FormatStyle::LK_Java &&
933 Current.isOneOf(Keywords.kw_implements, Keywords.kw_extends))
934 return std::max(State.Stack.back().LastSpace,
935 State.Stack.back().Indent + Style.ContinuationIndentWidth);
936
937 if (Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths &&
938 State.Line->First->is(tok::kw_enum))
939 return (Style.IndentWidth * State.Line->First->IndentLevel) +
940 Style.IndentWidth;
941
942 if (NextNonComment->is(tok::l_brace) && NextNonComment->BlockKind == BK_Block)
943 return Current.NestingLevel == 0 ? State.FirstIndent
944 : State.Stack.back().Indent;
945 if ((Current.isOneOf(tok::r_brace, tok::r_square) ||
946 (Current.is(tok::greater) &&
947 (Style.Language == FormatStyle::LK_Proto ||
948 Style.Language == FormatStyle::LK_TextProto))) &&
949 State.Stack.size() > 1) {
950 if (Current.closesBlockOrBlockTypeList(Style))
951 return State.Stack[State.Stack.size() - 2].NestedBlockIndent;
952 if (Current.MatchingParen &&
953 Current.MatchingParen->BlockKind == BK_BracedInit)
954 return State.Stack[State.Stack.size() - 2].LastSpace;
955 return State.FirstIndent;
956 }
957 // Indent a closing parenthesis at the previous level if followed by a semi,
958 // const, or opening brace. This allows indentations such as:
959 // foo(
960 // a,
961 // );
962 // int Foo::getter(
963 // //
964 // ) const {
965 // return foo;
966 // }
967 // function foo(
968 // a,
969 // ) {
970 // code(); //
971 // }
972 if (Current.is(tok::r_paren) && State.Stack.size() > 1 &&
973 (!Current.Next ||
974 Current.Next->isOneOf(tok::semi, tok::kw_const, tok::l_brace)))
975 return State.Stack[State.Stack.size() - 2].LastSpace;
976 if (NextNonComment->is(TT_TemplateString) && NextNonComment->closesScope())
977 return State.Stack[State.Stack.size() - 2].LastSpace;
978 if (Current.is(tok::identifier) && Current.Next &&
979 (Current.Next->is(TT_DictLiteral) ||
980 ((Style.Language == FormatStyle::LK_Proto ||
981 Style.Language == FormatStyle::LK_TextProto) &&
982 Current.Next->isOneOf(tok::less, tok::l_brace))))
983 return State.Stack.back().Indent;
984 if (NextNonComment->is(TT_ObjCStringLiteral) &&
985 State.StartOfStringLiteral != 0)
986 return State.StartOfStringLiteral - 1;
987 if (NextNonComment->isStringLiteral() && State.StartOfStringLiteral != 0)
988 return State.StartOfStringLiteral;
989 if (NextNonComment->is(tok::lessless) &&
990 State.Stack.back().FirstLessLess != 0)
991 return State.Stack.back().FirstLessLess;
992 if (NextNonComment->isMemberAccess()) {
993 if (State.Stack.back().CallContinuation == 0)
994 return ContinuationIndent;
995 return State.Stack.back().CallContinuation;
996 }
997 if (State.Stack.back().QuestionColumn != 0 &&
998 ((NextNonComment->is(tok::colon) &&
999 NextNonComment->is(TT_ConditionalExpr)) ||
1000 Previous.is(TT_ConditionalExpr)))
1001 return State.Stack.back().QuestionColumn;
1002 if (Previous.is(tok::comma) && State.Stack.back().VariablePos != 0)
1003 return State.Stack.back().VariablePos;
1004 if ((PreviousNonComment &&
1005 (PreviousNonComment->ClosesTemplateDeclaration ||
1006 PreviousNonComment->isOneOf(
1007 TT_AttributeParen, TT_AttributeSquare, TT_FunctionAnnotationRParen,
1008 TT_JavaAnnotation, TT_LeadingJavaAnnotation))) ||
1009 (!Style.IndentWrappedFunctionNames &&
1010 NextNonComment->isOneOf(tok::kw_operator, TT_FunctionDeclarationName)))
1011 return std::max(State.Stack.back().LastSpace, State.Stack.back().Indent);
1012 if (NextNonComment->is(TT_SelectorName)) {
1013 if (!State.Stack.back().ObjCSelectorNameFound) {
1014 unsigned MinIndent = State.Stack.back().Indent;
1015 if (shouldIndentWrappedSelectorName(Style, State.Line->Type))
1016 MinIndent = std::max(MinIndent,
1017 State.FirstIndent + Style.ContinuationIndentWidth);
1018 // If LongestObjCSelectorName is 0, we are indenting the first
1019 // part of an ObjC selector (or a selector component which is
1020 // not colon-aligned due to block formatting).
1021 //
1022 // Otherwise, we are indenting a subsequent part of an ObjC
1023 // selector which should be colon-aligned to the longest
1024 // component of the ObjC selector.
1025 //
1026 // In either case, we want to respect Style.IndentWrappedFunctionNames.
1027 return MinIndent +
1028 std::max(NextNonComment->LongestObjCSelectorName,
1029 NextNonComment->ColumnWidth) -
1030 NextNonComment->ColumnWidth;
1031 }
1032 if (!State.Stack.back().AlignColons)
1033 return State.Stack.back().Indent;
1034 if (State.Stack.back().ColonPos > NextNonComment->ColumnWidth)
1035 return State.Stack.back().ColonPos - NextNonComment->ColumnWidth;
1036 return State.Stack.back().Indent;
1037 }
1038 if (NextNonComment->is(tok::colon) && NextNonComment->is(TT_ObjCMethodExpr))
1039 return State.Stack.back().ColonPos;
1040 if (NextNonComment->is(TT_ArraySubscriptLSquare)) {
1041 if (State.Stack.back().StartOfArraySubscripts != 0)
1042 return State.Stack.back().StartOfArraySubscripts;
1043 return ContinuationIndent;
1044 }
1045
1046 // This ensure that we correctly format ObjC methods calls without inputs,
1047 // i.e. where the last element isn't selector like: [callee method];
1048 if (NextNonComment->is(tok::identifier) && NextNonComment->FakeRParens == 0 &&
1049 NextNonComment->Next && NextNonComment->Next->is(TT_ObjCMethodExpr))
1050 return State.Stack.back().Indent;
1051
1052 if (NextNonComment->isOneOf(TT_StartOfName, TT_PointerOrReference) ||
1053 Previous.isOneOf(tok::coloncolon, tok::equal, TT_JsTypeColon))
1054 return ContinuationIndent;
1055 if (PreviousNonComment && PreviousNonComment->is(tok::colon) &&
1056 PreviousNonComment->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral))
1057 return ContinuationIndent;
1058 if (NextNonComment->is(TT_CtorInitializerComma))
1059 return State.Stack.back().Indent;
1060 if (PreviousNonComment && PreviousNonComment->is(TT_CtorInitializerColon) &&
1061 Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon)
1062 return State.Stack.back().Indent;
1063 if (PreviousNonComment && PreviousNonComment->is(TT_InheritanceColon) &&
1064 Style.BreakInheritanceList == FormatStyle::BILS_AfterColon)
1065 return State.Stack.back().Indent;
1066 if (NextNonComment->isOneOf(TT_CtorInitializerColon, TT_InheritanceColon,
1067 TT_InheritanceComma))
1068 return State.FirstIndent + Style.ConstructorInitializerIndentWidth;
1069 if (Previous.is(tok::r_paren) && !Current.isBinaryOperator() &&
1070 !Current.isOneOf(tok::colon, tok::comment))
1071 return ContinuationIndent;
1072 if (Current.is(TT_ProtoExtensionLSquare))
1073 return State.Stack.back().Indent;
1074 if (State.Stack.back().Indent == State.FirstIndent && PreviousNonComment &&
1075 !PreviousNonComment->isOneOf(tok::r_brace, TT_CtorInitializerComma))
1076 // Ensure that we fall back to the continuation indent width instead of
1077 // just flushing continuations left.
1078 return State.Stack.back().Indent + Style.ContinuationIndentWidth;
1079 return State.Stack.back().Indent;
1080}
1081
1082unsigned ContinuationIndenter::moveStateToNextToken(LineState &State,
1083 bool DryRun, bool Newline) {
1084 assert(State.Stack.size())((State.Stack.size()) ? static_cast<void> (0) : __assert_fail
("State.Stack.size()", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 1084, __PRETTY_FUNCTION__))
;
1085 const FormatToken &Current = *State.NextToken;
1086
1087 if (Current.isOneOf(tok::comma, TT_BinaryOperator))
1088 State.Stack.back().NoLineBreakInOperand = false;
1089 if (Current.is(TT_InheritanceColon))
1090 State.Stack.back().AvoidBinPacking = true;
1091 if (Current.is(tok::lessless) && Current.isNot(TT_OverloadedOperator)) {
1092 if (State.Stack.back().FirstLessLess == 0)
1093 State.Stack.back().FirstLessLess = State.Column;
1094 else
1095 State.Stack.back().LastOperatorWrapped = Newline;
1096 }
1097 if (Current.is(TT_BinaryOperator) && Current.isNot(tok::lessless))
1098 State.Stack.back().LastOperatorWrapped = Newline;
1099 if (Current.is(TT_ConditionalExpr) && Current.Previous &&
1100 !Current.Previous->is(TT_ConditionalExpr))
1101 State.Stack.back().LastOperatorWrapped = Newline;
1102 if (Current.is(TT_ArraySubscriptLSquare) &&
1103 State.Stack.back().StartOfArraySubscripts == 0)
1104 State.Stack.back().StartOfArraySubscripts = State.Column;
1105 if (Style.BreakBeforeTernaryOperators && Current.is(tok::question))
1106 State.Stack.back().QuestionColumn = State.Column;
1107 if (!Style.BreakBeforeTernaryOperators && Current.isNot(tok::colon)) {
1108 const FormatToken *Previous = Current.Previous;
1109 while (Previous && Previous->isTrailingComment())
1110 Previous = Previous->Previous;
1111 if (Previous && Previous->is(tok::question))
1112 State.Stack.back().QuestionColumn = State.Column;
1113 }
1114 if (!Current.opensScope() && !Current.closesScope() &&
1115 !Current.is(TT_PointerOrReference))
1116 State.LowestLevelOnLine =
1117 std::min(State.LowestLevelOnLine, Current.NestingLevel);
1118 if (Current.isMemberAccess())
1119 State.Stack.back().StartOfFunctionCall =
1120 !Current.NextOperator ? 0 : State.Column;
1121 if (Current.is(TT_SelectorName))
1122 State.Stack.back().ObjCSelectorNameFound = true;
1123 if (Current.is(TT_CtorInitializerColon) &&
1124 Style.BreakConstructorInitializers != FormatStyle::BCIS_AfterColon) {
1125 // Indent 2 from the column, so:
1126 // SomeClass::SomeClass()
1127 // : First(...), ...
1128 // Next(...)
1129 // ^ line up here.
1130 State.Stack.back().Indent =
1131 State.Column +
1132 (Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma
1133 ? 0
1134 : 2);
1135 State.Stack.back().NestedBlockIndent = State.Stack.back().Indent;
1136 if (Style.ConstructorInitializerAllOnOneLineOrOnePerLine) {
1137 State.Stack.back().AvoidBinPacking = true;
1138 State.Stack.back().BreakBeforeParameter =
1139 !Style.AllowAllConstructorInitializersOnNextLine;
1140 } else {
1141 State.Stack.back().BreakBeforeParameter = false;
1142 }
1143 }
1144 if (Current.is(TT_CtorInitializerColon) &&
1145 Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon) {
1146 State.Stack.back().Indent =
1147 State.FirstIndent + Style.ConstructorInitializerIndentWidth;
1148 State.Stack.back().NestedBlockIndent = State.Stack.back().Indent;
1149 if (Style.ConstructorInitializerAllOnOneLineOrOnePerLine)
1150 State.Stack.back().AvoidBinPacking = true;
1151 }
1152 if (Current.is(TT_InheritanceColon))
1153 State.Stack.back().Indent =
1154 State.FirstIndent + Style.ConstructorInitializerIndentWidth;
1155 if (Current.isOneOf(TT_BinaryOperator, TT_ConditionalExpr) && Newline)
1156 State.Stack.back().NestedBlockIndent =
1157 State.Column + Current.ColumnWidth + 1;
1158 if (Current.isOneOf(TT_LambdaLSquare, TT_LambdaArrow))
1159 State.Stack.back().LastSpace = State.Column;
1160
1161 // Insert scopes created by fake parenthesis.
1162 const FormatToken *Previous = Current.getPreviousNonComment();
1163
1164 // Add special behavior to support a format commonly used for JavaScript
1165 // closures:
1166 // SomeFunction(function() {
1167 // foo();
1168 // bar();
1169 // }, a, b, c);
1170 if (Current.isNot(tok::comment) && Previous &&
1171 Previous->isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) &&
1172 !Previous->is(TT_DictLiteral) && State.Stack.size() > 1 &&
1173 !State.Stack.back().HasMultipleNestedBlocks) {
1174 if (State.Stack[State.Stack.size() - 2].NestedBlockInlined && Newline)
1175 for (unsigned i = 0, e = State.Stack.size() - 1; i != e; ++i)
1176 State.Stack[i].NoLineBreak = true;
1177 State.Stack[State.Stack.size() - 2].NestedBlockInlined = false;
1178 }
1179 if (Previous &&
1180 (Previous->isOneOf(tok::l_paren, tok::comma, tok::colon) ||
1181 Previous->isOneOf(TT_BinaryOperator, TT_ConditionalExpr)) &&
1182 !Previous->isOneOf(TT_DictLiteral, TT_ObjCMethodExpr)) {
1183 State.Stack.back().NestedBlockInlined =
1184 !Newline &&
1185 (Previous->isNot(tok::l_paren) || Previous->ParameterCount > 1);
1186 }
1187
1188 moveStatePastFakeLParens(State, Newline);
1189 moveStatePastScopeCloser(State);
1190 bool AllowBreak = !State.Stack.back().NoLineBreak &&
1191 !State.Stack.back().NoLineBreakInOperand;
1192 moveStatePastScopeOpener(State, Newline);
1193 moveStatePastFakeRParens(State);
1194
1195 if (Current.is(TT_ObjCStringLiteral) && State.StartOfStringLiteral == 0)
1196 State.StartOfStringLiteral = State.Column + 1;
1197 if (Current.is(TT_CSharpStringLiteral) && State.StartOfStringLiteral == 0)
1198 State.StartOfStringLiteral = State.Column + 1;
1199 else if (Current.isStringLiteral() && State.StartOfStringLiteral == 0)
1200 State.StartOfStringLiteral = State.Column;
1201 else if (!Current.isOneOf(tok::comment, tok::identifier, tok::hash) &&
1202 !Current.isStringLiteral())
1203 State.StartOfStringLiteral = 0;
1204
1205 State.Column += Current.ColumnWidth;
1206 State.NextToken = State.NextToken->Next;
1207
1208 unsigned Penalty =
1209 handleEndOfLine(Current, State, DryRun, AllowBreak, Newline);
1210
1211 if (Current.Role)
1212 Current.Role->formatFromToken(State, this, DryRun);
1213 // If the previous has a special role, let it consume tokens as appropriate.
1214 // It is necessary to start at the previous token for the only implemented
1215 // role (comma separated list). That way, the decision whether or not to break
1216 // after the "{" is already done and both options are tried and evaluated.
1217 // FIXME: This is ugly, find a better way.
1218 if (Previous && Previous->Role)
1219 Penalty += Previous->Role->formatAfterToken(State, this, DryRun);
1220
1221 return Penalty;
1222}
1223
1224void ContinuationIndenter::moveStatePastFakeLParens(LineState &State,
1225 bool Newline) {
1226 const FormatToken &Current = *State.NextToken;
1227 const FormatToken *Previous = Current.getPreviousNonComment();
1228
1229 // Don't add extra indentation for the first fake parenthesis after
1230 // 'return', assignments or opening <({[. The indentation for these cases
1231 // is special cased.
1232 bool SkipFirstExtraIndent =
1233 (Previous && (Previous->opensScope() ||
1234 Previous->isOneOf(tok::semi, tok::kw_return) ||
1235 (Previous->getPrecedence() == prec::Assignment &&
1236 Style.AlignOperands) ||
1237 Previous->is(TT_ObjCMethodExpr)));
1238 for (SmallVectorImpl<prec::Level>::const_reverse_iterator
1239 I = Current.FakeLParens.rbegin(),
1240 E = Current.FakeLParens.rend();
1241 I != E; ++I) {
1242 ParenState NewParenState = State.Stack.back();
1243 NewParenState.Tok = nullptr;
1244 NewParenState.ContainsLineBreak = false;
1245 NewParenState.LastOperatorWrapped = true;
1246 NewParenState.NoLineBreak =
1247 NewParenState.NoLineBreak || State.Stack.back().NoLineBreakInOperand;
1248
1249 // Don't propagate AvoidBinPacking into subexpressions of arg/param lists.
1250 if (*I > prec::Comma)
1251 NewParenState.AvoidBinPacking = false;
1252
1253 // Indent from 'LastSpace' unless these are fake parentheses encapsulating
1254 // a builder type call after 'return' or, if the alignment after opening
1255 // brackets is disabled.
1256 if (!Current.isTrailingComment() &&
1257 (Style.AlignOperands || *I < prec::Assignment) &&
1258 (!Previous || Previous->isNot(tok::kw_return) ||
1259 (Style.Language != FormatStyle::LK_Java && *I > 0)) &&
1260 (Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign ||
1261 *I != prec::Comma || Current.NestingLevel == 0))
1262 NewParenState.Indent =
1263 std::max(std::max(State.Column, NewParenState.Indent),
1264 State.Stack.back().LastSpace);
1265
1266 // Do not indent relative to the fake parentheses inserted for "." or "->".
1267 // This is a special case to make the following to statements consistent:
1268 // OuterFunction(InnerFunctionCall( // break
1269 // ParameterToInnerFunction));
1270 // OuterFunction(SomeObject.InnerFunctionCall( // break
1271 // ParameterToInnerFunction));
1272 if (*I > prec::Unknown)
1273 NewParenState.LastSpace = std::max(NewParenState.LastSpace, State.Column);
1274 if (*I != prec::Conditional && !Current.is(TT_UnaryOperator) &&
1275 Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign)
1276 NewParenState.StartOfFunctionCall = State.Column;
1277
1278 // Always indent conditional expressions. Never indent expression where
1279 // the 'operator' is ',', ';' or an assignment (i.e. *I <=
1280 // prec::Assignment) as those have different indentation rules. Indent
1281 // other expression, unless the indentation needs to be skipped.
1282 if (*I == prec::Conditional ||
1283 (!SkipFirstExtraIndent && *I > prec::Assignment &&
1284 !Current.isTrailingComment()))
1285 NewParenState.Indent += Style.ContinuationIndentWidth;
1286 if ((Previous && !Previous->opensScope()) || *I != prec::Comma)
1287 NewParenState.BreakBeforeParameter = false;
1288 State.Stack.push_back(NewParenState);
1289 SkipFirstExtraIndent = false;
1290 }
1291}
1292
1293void ContinuationIndenter::moveStatePastFakeRParens(LineState &State) {
1294 for (unsigned i = 0, e = State.NextToken->FakeRParens; i != e; ++i) {
1295 unsigned VariablePos = State.Stack.back().VariablePos;
1296 if (State.Stack.size() == 1) {
1297 // Do not pop the last element.
1298 break;
1299 }
1300 State.Stack.pop_back();
1301 State.Stack.back().VariablePos = VariablePos;
1302 }
1303}
1304
1305void ContinuationIndenter::moveStatePastScopeOpener(LineState &State,
1306 bool Newline) {
1307 const FormatToken &Current = *State.NextToken;
1308 if (!Current.opensScope())
1309 return;
1310
1311 if (Current.MatchingParen && Current.BlockKind == BK_Block) {
1312 moveStateToNewBlock(State);
1313 return;
1314 }
1315
1316 unsigned NewIndent;
1317 unsigned LastSpace = State.Stack.back().LastSpace;
1318 bool AvoidBinPacking;
1319 bool BreakBeforeParameter = false;
1320 unsigned NestedBlockIndent = std::max(State.Stack.back().StartOfFunctionCall,
1321 State.Stack.back().NestedBlockIndent);
1322 if (Current.isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) ||
1323 opensProtoMessageField(Current, Style)) {
1324 if (Current.opensBlockOrBlockTypeList(Style)) {
1325 NewIndent = Style.IndentWidth +
1326 std::min(State.Column, State.Stack.back().NestedBlockIndent);
1327 } else {
1328 NewIndent = State.Stack.back().LastSpace + Style.ContinuationIndentWidth;
1329 }
1330 const FormatToken *NextNoComment = Current.getNextNonComment();
1331 bool EndsInComma = Current.MatchingParen &&
1332 Current.MatchingParen->Previous &&
1333 Current.MatchingParen->Previous->is(tok::comma);
1334 AvoidBinPacking = EndsInComma || Current.is(TT_DictLiteral) ||
1335 Style.Language == FormatStyle::LK_Proto ||
1336 Style.Language == FormatStyle::LK_TextProto ||
1337 !Style.BinPackArguments ||
1338 (NextNoComment &&
1339 NextNoComment->isOneOf(TT_DesignatedInitializerPeriod,
1340 TT_DesignatedInitializerLSquare));
1341 BreakBeforeParameter = EndsInComma;
1342 if (Current.ParameterCount > 1)
1343 NestedBlockIndent = std::max(NestedBlockIndent, State.Column + 1);
1344 } else {
1345 NewIndent = Style.ContinuationIndentWidth +
1346 std::max(State.Stack.back().LastSpace,
1347 State.Stack.back().StartOfFunctionCall);
1348
1349 // Ensure that different different brackets force relative alignment, e.g.:
1350 // void SomeFunction(vector< // break
1351 // int> v);
1352 // FIXME: We likely want to do this for more combinations of brackets.
1353 if (Current.is(tok::less) && Current.ParentBracket == tok::l_paren) {
1354 NewIndent = std::max(NewIndent, State.Stack.back().Indent);
1355 LastSpace = std::max(LastSpace, State.Stack.back().Indent);
1356 }
1357
1358 bool EndsInComma =
1359 Current.MatchingParen &&
1360 Current.MatchingParen->getPreviousNonComment() &&
1361 Current.MatchingParen->getPreviousNonComment()->is(tok::comma);
1362
1363 // If ObjCBinPackProtocolList is unspecified, fall back to BinPackParameters
1364 // for backwards compatibility.
1365 bool ObjCBinPackProtocolList =
1366 (Style.ObjCBinPackProtocolList == FormatStyle::BPS_Auto &&
1367 Style.BinPackParameters) ||
1368 Style.ObjCBinPackProtocolList == FormatStyle::BPS_Always;
1369
1370 bool BinPackDeclaration =
1371 (State.Line->Type != LT_ObjCDecl && Style.BinPackParameters) ||
1372 (State.Line->Type == LT_ObjCDecl && ObjCBinPackProtocolList);
1373
1374 AvoidBinPacking =
1375 (Style.Language == FormatStyle::LK_JavaScript && EndsInComma) ||
1376 (State.Line->MustBeDeclaration && !BinPackDeclaration) ||
1377 (!State.Line->MustBeDeclaration && !Style.BinPackArguments) ||
1378 (Style.ExperimentalAutoDetectBinPacking &&
1379 (Current.PackingKind == PPK_OnePerLine ||
1380 (!BinPackInconclusiveFunctions &&
1381 Current.PackingKind == PPK_Inconclusive)));
1382
1383 if (Current.is(TT_ObjCMethodExpr) && Current.MatchingParen) {
1384 if (Style.ColumnLimit) {
1385 // If this '[' opens an ObjC call, determine whether all parameters fit
1386 // into one line and put one per line if they don't.
1387 if (getLengthToMatchingParen(Current, State.Stack) + State.Column >
1388 getColumnLimit(State))
1389 BreakBeforeParameter = true;
1390 } else {
1391 // For ColumnLimit = 0, we have to figure out whether there is or has to
1392 // be a line break within this call.
1393 for (const FormatToken *Tok = &Current;
1394 Tok && Tok != Current.MatchingParen; Tok = Tok->Next) {
1395 if (Tok->MustBreakBefore ||
1396 (Tok->CanBreakBefore && Tok->NewlinesBefore > 0)) {
1397 BreakBeforeParameter = true;
1398 break;
1399 }
1400 }
1401 }
1402 }
1403
1404 if (Style.Language == FormatStyle::LK_JavaScript && EndsInComma)
1405 BreakBeforeParameter = true;
1406 }
1407 // Generally inherit NoLineBreak from the current scope to nested scope.
1408 // However, don't do this for non-empty nested blocks, dict literals and
1409 // array literals as these follow different indentation rules.
1410 bool NoLineBreak =
1411 Current.Children.empty() &&
1412 !Current.isOneOf(TT_DictLiteral, TT_ArrayInitializerLSquare) &&
1413 (State.Stack.back().NoLineBreak ||
1414 State.Stack.back().NoLineBreakInOperand ||
1415 (Current.is(TT_TemplateOpener) &&
1416 State.Stack.back().ContainsUnwrappedBuilder));
1417 State.Stack.push_back(
1418 ParenState(&Current, NewIndent, LastSpace, AvoidBinPacking, NoLineBreak));
1419 State.Stack.back().NestedBlockIndent = NestedBlockIndent;
1420 State.Stack.back().BreakBeforeParameter = BreakBeforeParameter;
1421 State.Stack.back().HasMultipleNestedBlocks = Current.BlockParameterCount > 1;
1422 State.Stack.back().IsInsideObjCArrayLiteral =
1423 Current.is(TT_ArrayInitializerLSquare) && Current.Previous &&
1424 Current.Previous->is(tok::at);
1425}
1426
1427void ContinuationIndenter::moveStatePastScopeCloser(LineState &State) {
1428 const FormatToken &Current = *State.NextToken;
1429 if (!Current.closesScope())
1430 return;
1431
1432 // If we encounter a closing ), ], } or >, we can remove a level from our
1433 // stacks.
1434 if (State.Stack.size() > 1 &&
1435 (Current.isOneOf(tok::r_paren, tok::r_square, TT_TemplateString) ||
1436 (Current.is(tok::r_brace) && State.NextToken != State.Line->First) ||
1437 State.NextToken->is(TT_TemplateCloser) ||
1438 (Current.is(tok::greater) && Current.is(TT_DictLiteral))))
1439 State.Stack.pop_back();
1440
1441 // Reevaluate whether ObjC message arguments fit into one line.
1442 // If a receiver spans multiple lines, e.g.:
1443 // [[object block:^{
1444 // return 42;
1445 // }] a:42 b:42];
1446 // BreakBeforeParameter is calculated based on an incorrect assumption
1447 // (it is checked whether the whole expression fits into one line without
1448 // considering a line break inside a message receiver).
1449 // We check whether arguements fit after receiver scope closer (into the same
1450 // line).
1451 if (State.Stack.back().BreakBeforeParameter && Current.MatchingParen &&
1452 Current.MatchingParen->Previous) {
1453 const FormatToken &CurrentScopeOpener = *Current.MatchingParen->Previous;
1454 if (CurrentScopeOpener.is(TT_ObjCMethodExpr) &&
1455 CurrentScopeOpener.MatchingParen) {
1456 int NecessarySpaceInLine =
1457 getLengthToMatchingParen(CurrentScopeOpener, State.Stack) +
1458 CurrentScopeOpener.TotalLength - Current.TotalLength - 1;
1459 if (State.Column + Current.ColumnWidth + NecessarySpaceInLine <=
1460 Style.ColumnLimit)
1461 State.Stack.back().BreakBeforeParameter = false;
1462 }
1463 }
1464
1465 if (Current.is(tok::r_square)) {
1466 // If this ends the array subscript expr, reset the corresponding value.
1467 const FormatToken *NextNonComment = Current.getNextNonComment();
1468 if (NextNonComment && NextNonComment->isNot(tok::l_square))
1469 State.Stack.back().StartOfArraySubscripts = 0;
1470 }
1471}
1472
1473void ContinuationIndenter::moveStateToNewBlock(LineState &State) {
1474 unsigned NestedBlockIndent = State.Stack.back().NestedBlockIndent;
1475 // ObjC block sometimes follow special indentation rules.
1476 unsigned NewIndent =
1477 NestedBlockIndent + (State.NextToken->is(TT_ObjCBlockLBrace)
1478 ? Style.ObjCBlockIndentWidth
1479 : Style.IndentWidth);
1480 State.Stack.push_back(ParenState(State.NextToken, NewIndent,
1481 State.Stack.back().LastSpace,
1482 /*AvoidBinPacking=*/true,
1483 /*NoLineBreak=*/false));
1484 State.Stack.back().NestedBlockIndent = NestedBlockIndent;
1485 State.Stack.back().BreakBeforeParameter = true;
1486}
1487
1488static unsigned getLastLineEndColumn(StringRef Text, unsigned StartColumn,
1489 unsigned TabWidth,
1490 encoding::Encoding Encoding) {
1491 size_t LastNewlinePos = Text.find_last_of("\n");
1492 if (LastNewlinePos == StringRef::npos) {
1493 return StartColumn +
1494 encoding::columnWidthWithTabs(Text, StartColumn, TabWidth, Encoding);
1495 } else {
1496 return encoding::columnWidthWithTabs(Text.substr(LastNewlinePos),
1497 /*StartColumn=*/0, TabWidth, Encoding);
1498 }
1499}
1500
1501unsigned ContinuationIndenter::reformatRawStringLiteral(
1502 const FormatToken &Current, LineState &State,
1503 const FormatStyle &RawStringStyle, bool DryRun, bool Newline) {
1504 unsigned StartColumn = State.Column - Current.ColumnWidth;
1505 StringRef OldDelimiter = *getRawStringDelimiter(Current.TokenText);
1506 StringRef NewDelimiter =
1507 getCanonicalRawStringDelimiter(Style, RawStringStyle.Language);
1508 if (NewDelimiter.empty() || OldDelimiter.empty())
1509 NewDelimiter = OldDelimiter;
1510 // The text of a raw string is between the leading 'R"delimiter(' and the
1511 // trailing 'delimiter)"'.
1512 unsigned OldPrefixSize = 3 + OldDelimiter.size();
1513 unsigned OldSuffixSize = 2 + OldDelimiter.size();
1514 // We create a virtual text environment which expects a null-terminated
1515 // string, so we cannot use StringRef.
1516 std::string RawText =
1517 Current.TokenText.substr(OldPrefixSize).drop_back(OldSuffixSize);
1518 if (NewDelimiter != OldDelimiter) {
1519 // Don't update to the canonical delimiter 'deli' if ')deli"' occurs in the
1520 // raw string.
1521 std::string CanonicalDelimiterSuffix = (")" + NewDelimiter + "\"").str();
1522 if (StringRef(RawText).contains(CanonicalDelimiterSuffix))
1523 NewDelimiter = OldDelimiter;
1524 }
1525
1526 unsigned NewPrefixSize = 3 + NewDelimiter.size();
1527 unsigned NewSuffixSize = 2 + NewDelimiter.size();
1528
1529 // The first start column is the column the raw text starts after formatting.
1530 unsigned FirstStartColumn = StartColumn + NewPrefixSize;
1531
1532 // The next start column is the intended indentation a line break inside
1533 // the raw string at level 0. It is determined by the following rules:
1534 // - if the content starts on newline, it is one level more than the current
1535 // indent, and
1536 // - if the content does not start on a newline, it is the first start
1537 // column.
1538 // These rules have the advantage that the formatted content both does not
1539 // violate the rectangle rule and visually flows within the surrounding
1540 // source.
1541 bool ContentStartsOnNewline = Current.TokenText[OldPrefixSize] == '\n';
1542 // If this token is the last parameter (checked by looking if it's followed by
1543 // `)` and is not on a newline, the base the indent off the line's nested
1544 // block indent. Otherwise, base the indent off the arguments indent, so we
1545 // can achieve:
1546 //
1547 // fffffffffff(1, 2, 3, R"pb(
1548 // key1: 1 #
1549 // key2: 2)pb");
1550 //
1551 // fffffffffff(1, 2, 3,
1552 // R"pb(
1553 // key1: 1 #
1554 // key2: 2
1555 // )pb");
1556 //
1557 // fffffffffff(1, 2, 3,
1558 // R"pb(
1559 // key1: 1 #
1560 // key2: 2
1561 // )pb",
1562 // 5);
1563 unsigned CurrentIndent =
1564 (!Newline && Current.Next && Current.Next->is(tok::r_paren))
1565 ? State.Stack.back().NestedBlockIndent
1566 : State.Stack.back().Indent;
1567 unsigned NextStartColumn = ContentStartsOnNewline
1568 ? CurrentIndent + Style.IndentWidth
1569 : FirstStartColumn;
1570
1571 // The last start column is the column the raw string suffix starts if it is
1572 // put on a newline.
1573 // The last start column is the intended indentation of the raw string postfix
1574 // if it is put on a newline. It is determined by the following rules:
1575 // - if the raw string prefix starts on a newline, it is the column where
1576 // that raw string prefix starts, and
1577 // - if the raw string prefix does not start on a newline, it is the current
1578 // indent.
1579 unsigned LastStartColumn =
1580 Current.NewlinesBefore ? FirstStartColumn - NewPrefixSize : CurrentIndent;
1581
1582 std::pair<tooling::Replacements, unsigned> Fixes = internal::reformat(
1583 RawStringStyle, RawText, {tooling::Range(0, RawText.size())},
1584 FirstStartColumn, NextStartColumn, LastStartColumn, "<stdin>",
1585 /*Status=*/nullptr);
1586
1587 auto NewCode = applyAllReplacements(RawText, Fixes.first);
1588 tooling::Replacements NoFixes;
1589 if (!NewCode) {
1590 return addMultilineToken(Current, State);
1591 }
1592 if (!DryRun) {
1593 if (NewDelimiter != OldDelimiter) {
1594 // In 'R"delimiter(...', the delimiter starts 2 characters after the start
1595 // of the token.
1596 SourceLocation PrefixDelimiterStart =
1597 Current.Tok.getLocation().getLocWithOffset(2);
1598 auto PrefixErr = Whitespaces.addReplacement(tooling::Replacement(
1599 SourceMgr, PrefixDelimiterStart, OldDelimiter.size(), NewDelimiter));
1600 if (PrefixErr) {
1601 llvm::errs()
1602 << "Failed to update the prefix delimiter of a raw string: "
1603 << llvm::toString(std::move(PrefixErr)) << "\n";
1604 }
1605 // In 'R"delimiter(...)delimiter"', the suffix delimiter starts at
1606 // position length - 1 - |delimiter|.
1607 SourceLocation SuffixDelimiterStart =
1608 Current.Tok.getLocation().getLocWithOffset(Current.TokenText.size() -
1609 1 - OldDelimiter.size());
1610 auto SuffixErr = Whitespaces.addReplacement(tooling::Replacement(
1611 SourceMgr, SuffixDelimiterStart, OldDelimiter.size(), NewDelimiter));
1612 if (SuffixErr) {
1613 llvm::errs()
1614 << "Failed to update the suffix delimiter of a raw string: "
1615 << llvm::toString(std::move(SuffixErr)) << "\n";
1616 }
1617 }
1618 SourceLocation OriginLoc =
1619 Current.Tok.getLocation().getLocWithOffset(OldPrefixSize);
1620 for (const tooling::Replacement &Fix : Fixes.first) {
1621 auto Err = Whitespaces.addReplacement(tooling::Replacement(
1622 SourceMgr, OriginLoc.getLocWithOffset(Fix.getOffset()),
1623 Fix.getLength(), Fix.getReplacementText()));
1624 if (Err) {
1625 llvm::errs() << "Failed to reformat raw string: "
1626 << llvm::toString(std::move(Err)) << "\n";
1627 }
1628 }
1629 }
1630 unsigned RawLastLineEndColumn = getLastLineEndColumn(
1631 *NewCode, FirstStartColumn, Style.TabWidth, Encoding);
1632 State.Column = RawLastLineEndColumn + NewSuffixSize;
1633 // Since we're updating the column to after the raw string literal here, we
1634 // have to manually add the penalty for the prefix R"delim( over the column
1635 // limit.
1636 unsigned PrefixExcessCharacters =
1637 StartColumn + NewPrefixSize > Style.ColumnLimit
1638 ? StartColumn + NewPrefixSize - Style.ColumnLimit
1639 : 0;
1640 bool IsMultiline =
1641 ContentStartsOnNewline || (NewCode->find('\n') != std::string::npos);
1642 if (IsMultiline) {
1643 // Break before further function parameters on all levels.
1644 for (unsigned i = 0, e = State.Stack.size(); i != e; ++i)
1645 State.Stack[i].BreakBeforeParameter = true;
1646 }
1647 return Fixes.second + PrefixExcessCharacters * Style.PenaltyExcessCharacter;
1648}
1649
1650unsigned ContinuationIndenter::addMultilineToken(const FormatToken &Current,
1651 LineState &State) {
1652 // Break before further function parameters on all levels.
1653 for (unsigned i = 0, e = State.Stack.size(); i != e; ++i)
1654 State.Stack[i].BreakBeforeParameter = true;
1655
1656 unsigned ColumnsUsed = State.Column;
1657 // We can only affect layout of the first and the last line, so the penalty
1658 // for all other lines is constant, and we ignore it.
1659 State.Column = Current.LastLineColumnWidth;
1660
1661 if (ColumnsUsed > getColumnLimit(State))
1662 return Style.PenaltyExcessCharacter * (ColumnsUsed - getColumnLimit(State));
1663 return 0;
1664}
1665
1666unsigned ContinuationIndenter::handleEndOfLine(const FormatToken &Current,
1667 LineState &State, bool DryRun,
1668 bool AllowBreak, bool Newline) {
1669 unsigned Penalty = 0;
1670 // Compute the raw string style to use in case this is a raw string literal
1671 // that can be reformatted.
1672 auto RawStringStyle = getRawStringStyle(Current, State);
1673 if (RawStringStyle && !Current.Finalized) {
1674 Penalty = reformatRawStringLiteral(Current, State, *RawStringStyle, DryRun,
1675 Newline);
1676 } else if (Current.IsMultiline && Current.isNot(TT_BlockComment)) {
1677 // Don't break multi-line tokens other than block comments and raw string
1678 // literals. Instead, just update the state.
1679 Penalty = addMultilineToken(Current, State);
1680 } else if (State.Line->Type != LT_ImportStatement) {
1681 // We generally don't break import statements.
1682 LineState OriginalState = State;
1683
1684 // Whether we force the reflowing algorithm to stay strictly within the
1685 // column limit.
1686 bool Strict = false;
1687 // Whether the first non-strict attempt at reflowing did intentionally
1688 // exceed the column limit.
1689 bool Exceeded = false;
1690 std::tie(Penalty, Exceeded) = breakProtrudingToken(
1691 Current, State, AllowBreak, /*DryRun=*/true, Strict);
1692 if (Exceeded) {
1693 // If non-strict reflowing exceeds the column limit, try whether strict
1694 // reflowing leads to an overall lower penalty.
1695 LineState StrictState = OriginalState;
1696 unsigned StrictPenalty =
1697 breakProtrudingToken(Current, StrictState, AllowBreak,
1698 /*DryRun=*/true, /*Strict=*/true)
1699 .first;
1700 Strict = StrictPenalty <= Penalty;
1701 if (Strict) {
1702 Penalty = StrictPenalty;
1703 State = StrictState;
1704 }
1705 }
1706 if (!DryRun) {
1707 // If we're not in dry-run mode, apply the changes with the decision on
1708 // strictness made above.
1709 breakProtrudingToken(Current, OriginalState, AllowBreak, /*DryRun=*/false,
1710 Strict);
1711 }
1712 }
1713 if (State.Column > getColumnLimit(State)) {
1714 unsigned ExcessCharacters = State.Column - getColumnLimit(State);
1715 Penalty += Style.PenaltyExcessCharacter * ExcessCharacters;
1716 }
1717 return Penalty;
1718}
1719
1720// Returns the enclosing function name of a token, or the empty string if not
1721// found.
1722static StringRef getEnclosingFunctionName(const FormatToken &Current) {
1723 // Look for: 'function(' or 'function<templates>(' before Current.
1724 auto Tok = Current.getPreviousNonComment();
1725 if (!Tok || !Tok->is(tok::l_paren))
1726 return "";
1727 Tok = Tok->getPreviousNonComment();
1728 if (!Tok)
1729 return "";
1730 if (Tok->is(TT_TemplateCloser)) {
1731 Tok = Tok->MatchingParen;
1732 if (Tok)
1733 Tok = Tok->getPreviousNonComment();
1734 }
1735 if (!Tok || !Tok->is(tok::identifier))
1736 return "";
1737 return Tok->TokenText;
1738}
1739
1740llvm::Optional<FormatStyle>
1741ContinuationIndenter::getRawStringStyle(const FormatToken &Current,
1742 const LineState &State) {
1743 if (!Current.isStringLiteral())
1744 return None;
1745 auto Delimiter = getRawStringDelimiter(Current.TokenText);
1746 if (!Delimiter)
1747 return None;
1748 auto RawStringStyle = RawStringFormats.getDelimiterStyle(*Delimiter);
1749 if (!RawStringStyle && Delimiter->empty())
1750 RawStringStyle = RawStringFormats.getEnclosingFunctionStyle(
1751 getEnclosingFunctionName(Current));
1752 if (!RawStringStyle)
1753 return None;
1754 RawStringStyle->ColumnLimit = getColumnLimit(State);
1755 return RawStringStyle;
1756}
1757
1758std::unique_ptr<BreakableToken>
1759ContinuationIndenter::createBreakableToken(const FormatToken &Current,
1760 LineState &State, bool AllowBreak) {
1761 unsigned StartColumn = State.Column - Current.ColumnWidth;
1762 if (Current.isStringLiteral()) {
1763 // FIXME: String literal breaking is currently disabled for C#,Java and
1764 // JavaScript, as it requires strings to be merged using "+" which we
1765 // don't support.
1766 if (Style.Language == FormatStyle::LK_Java ||
1767 Style.Language == FormatStyle::LK_JavaScript || Style.isCSharp() ||
1768 !Style.BreakStringLiterals || !AllowBreak)
1769 return nullptr;
1770
1771 // Don't break string literals inside preprocessor directives (except for
1772 // #define directives, as their contents are stored in separate lines and
1773 // are not affected by this check).
1774 // This way we avoid breaking code with line directives and unknown
1775 // preprocessor directives that contain long string literals.
1776 if (State.Line->Type == LT_PreprocessorDirective)
1777 return nullptr;
1778 // Exempts unterminated string literals from line breaking. The user will
1779 // likely want to terminate the string before any line breaking is done.
1780 if (Current.IsUnterminatedLiteral)
1781 return nullptr;
1782 // Don't break string literals inside Objective-C array literals (doing so
1783 // raises the warning -Wobjc-string-concatenation).
1784 if (State.Stack.back().IsInsideObjCArrayLiteral) {
1785 return nullptr;
1786 }
1787
1788 StringRef Text = Current.TokenText;
1789 StringRef Prefix;
1790 StringRef Postfix;
1791 // FIXME: Handle whitespace between '_T', '(', '"..."', and ')'.
1792 // FIXME: Store Prefix and Suffix (or PrefixLength and SuffixLength to
1793 // reduce the overhead) for each FormatToken, which is a string, so that we
1794 // don't run multiple checks here on the hot path.
1795 if ((Text.endswith(Postfix = "\"") &&
1796 (Text.startswith(Prefix = "@\"") || Text.startswith(Prefix = "\"") ||
1797 Text.startswith(Prefix = "u\"") || Text.startswith(Prefix = "U\"") ||
1798 Text.startswith(Prefix = "u8\"") ||
1799 Text.startswith(Prefix = "L\""))) ||
1800 (Text.startswith(Prefix = "_T(\"") && Text.endswith(Postfix = "\")"))) {
1801 // We need this to address the case where there is an unbreakable tail
1802 // only if certain other formatting decisions have been taken. The
1803 // UnbreakableTailLength of Current is an overapproximation is that case
1804 // and we need to be correct here.
1805 unsigned UnbreakableTailLength = (State.NextToken && canBreak(State))
1806 ? 0
1807 : Current.UnbreakableTailLength;
1808 return std::make_unique<BreakableStringLiteral>(
1809 Current, StartColumn, Prefix, Postfix, UnbreakableTailLength,
1810 State.Line->InPPDirective, Encoding, Style);
1811 }
1812 } else if (Current.is(TT_BlockComment)) {
1813 if (!Style.ReflowComments ||
1814 // If a comment token switches formatting, like
1815 // /* clang-format on */, we don't want to break it further,
1816 // but we may still want to adjust its indentation.
1817 switchesFormatting(Current)) {
1818 return nullptr;
1819 }
1820 return std::make_unique<BreakableBlockComment>(
1821 Current, StartColumn, Current.OriginalColumn, !Current.Previous,
1822 State.Line->InPPDirective, Encoding, Style, Whitespaces.useCRLF());
1823 } else if (Current.is(TT_LineComment) &&
1824 (Current.Previous == nullptr ||
1825 Current.Previous->isNot(TT_ImplicitStringLiteral))) {
1826 if (!Style.ReflowComments ||
1827 CommentPragmasRegex.match(Current.TokenText.substr(2)) ||
1828 switchesFormatting(Current))
1829 return nullptr;
1830 return std::make_unique<BreakableLineCommentSection>(
1831 Current, StartColumn, Current.OriginalColumn, !Current.Previous,
1832 /*InPPDirective=*/false, Encoding, Style);
1833 }
1834 return nullptr;
1835}
1836
1837std::pair<unsigned, bool>
1838ContinuationIndenter::breakProtrudingToken(const FormatToken &Current,
1839 LineState &State, bool AllowBreak,
1840 bool DryRun, bool Strict) {
1841 std::unique_ptr<const BreakableToken> Token =
1842 createBreakableToken(Current, State, AllowBreak);
1843 if (!Token)
1844 return {0, false};
1845 assert(Token->getLineCount() > 0)((Token->getLineCount() > 0) ? static_cast<void> (
0) : __assert_fail ("Token->getLineCount() > 0", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 1845, __PRETTY_FUNCTION__))
;
1846 unsigned ColumnLimit = getColumnLimit(State);
1847 if (Current.is(TT_LineComment)) {
1848 // We don't insert backslashes when breaking line comments.
1849 ColumnLimit = Style.ColumnLimit;
1850 }
1851 if (Current.UnbreakableTailLength >= ColumnLimit)
1852 return {0, false};
1853 // ColumnWidth was already accounted into State.Column before calling
1854 // breakProtrudingToken.
1855 unsigned StartColumn = State.Column - Current.ColumnWidth;
1856 unsigned NewBreakPenalty = Current.isStringLiteral()
1857 ? Style.PenaltyBreakString
1858 : Style.PenaltyBreakComment;
1859 // Stores whether we intentionally decide to let a line exceed the column
1860 // limit.
1861 bool Exceeded = false;
1862 // Stores whether we introduce a break anywhere in the token.
1863 bool BreakInserted = Token->introducesBreakBeforeToken();
1864 // Store whether we inserted a new line break at the end of the previous
1865 // logical line.
1866 bool NewBreakBefore = false;
1867 // We use a conservative reflowing strategy. Reflow starts after a line is
1868 // broken or the corresponding whitespace compressed. Reflow ends as soon as a
1869 // line that doesn't get reflown with the previous line is reached.
1870 bool Reflow = false;
1871 // Keep track of where we are in the token:
1872 // Where we are in the content of the current logical line.
1873 unsigned TailOffset = 0;
1874 // The column number we're currently at.
1875 unsigned ContentStartColumn =
1876 Token->getContentStartColumn(0, /*Break=*/false);
1877 // The number of columns left in the current logical line after TailOffset.
1878 unsigned RemainingTokenColumns =
1879 Token->getRemainingLength(0, TailOffset, ContentStartColumn);
1880 // Adapt the start of the token, for example indent.
1881 if (!DryRun)
1882 Token->adaptStartOfLine(0, Whitespaces);
1883
1884 unsigned ContentIndent = 0;
1885 unsigned Penalty = 0;
1886 LLVM_DEBUG(llvm::dbgs() << "Breaking protruding token at column "do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << "Breaking protruding token at column "
<< StartColumn << ".\n"; } } while (false)
1887 << StartColumn << ".\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << "Breaking protruding token at column "
<< StartColumn << ".\n"; } } while (false)
;
1888 for (unsigned LineIndex = 0, EndIndex = Token->getLineCount();
1889 LineIndex != EndIndex; ++LineIndex) {
1890 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Line: " <<
LineIndex << " (Reflow: " << Reflow << ")\n"
; } } while (false)
1891 << " Line: " << LineIndex << " (Reflow: " << Reflow << ")\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Line: " <<
LineIndex << " (Reflow: " << Reflow << ")\n"
; } } while (false)
;
1892 NewBreakBefore = false;
1893 // If we did reflow the previous line, we'll try reflowing again. Otherwise
1894 // we'll start reflowing if the current line is broken or whitespace is
1895 // compressed.
1896 bool TryReflow = Reflow;
1897 // Break the current token until we can fit the rest of the line.
1898 while (ContentStartColumn + RemainingTokenColumns > ColumnLimit) {
1899 LLVM_DEBUG(llvm::dbgs() << " Over limit, need: "do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
1900 << (ContentStartColumn + RemainingTokenColumns)do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
1901 << ", space: " << ColumnLimitdo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
1902 << ", reflown prefix: " << ContentStartColumndo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
1903 << ", offset in line: " << TailOffset << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
;
1904 // If the current token doesn't fit, find the latest possible split in the
1905 // current line so that breaking at it will be under the column limit.
1906 // FIXME: Use the earliest possible split while reflowing to correctly
1907 // compress whitespace within a line.
1908 BreakableToken::Split Split =
1909 Token->getSplit(LineIndex, TailOffset, ColumnLimit,
1910 ContentStartColumn, CommentPragmasRegex);
1911 if (Split.first == StringRef::npos) {
1912 // No break opportunity - update the penalty and continue with the next
1913 // logical line.
1914 if (LineIndex < EndIndex - 1)
1915 // The last line's penalty is handled in addNextStateToQueue() or when
1916 // calling replaceWhitespaceAfterLastLine below.
1917 Penalty += Style.PenaltyExcessCharacter *
1918 (ContentStartColumn + RemainingTokenColumns - ColumnLimit);
1919 LLVM_DEBUG(llvm::dbgs() << " No break opportunity.\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " No break opportunity.\n"
; } } while (false)
;
1920 break;
1921 }
1922 assert(Split.first != 0)((Split.first != 0) ? static_cast<void> (0) : __assert_fail
("Split.first != 0", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 1922, __PRETTY_FUNCTION__))
;
1923
1924 if (Token->supportsReflow()) {
1925 // Check whether the next natural split point after the current one can
1926 // still fit the line, either because we can compress away whitespace,
1927 // or because the penalty the excess characters introduce is lower than
1928 // the break penalty.
1929 // We only do this for tokens that support reflowing, and thus allow us
1930 // to change the whitespace arbitrarily (e.g. comments).
1931 // Other tokens, like string literals, can be broken on arbitrary
1932 // positions.
1933
1934 // First, compute the columns from TailOffset to the next possible split
1935 // position.
1936 // For example:
1937 // ColumnLimit: |
1938 // // Some text that breaks
1939 // ^ tail offset
1940 // ^-- split
1941 // ^-------- to split columns
1942 // ^--- next split
1943 // ^--------------- to next split columns
1944 unsigned ToSplitColumns = Token->getRangeLength(
1945 LineIndex, TailOffset, Split.first, ContentStartColumn);
1946 LLVM_DEBUG(llvm::dbgs() << " ToSplit: " << ToSplitColumns << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ToSplit: " <<
ToSplitColumns << "\n"; } } while (false)
;
1947
1948 BreakableToken::Split NextSplit = Token->getSplit(
1949 LineIndex, TailOffset + Split.first + Split.second, ColumnLimit,
1950 ContentStartColumn + ToSplitColumns + 1, CommentPragmasRegex);
1951 // Compute the columns necessary to fit the next non-breakable sequence
1952 // into the current line.
1953 unsigned ToNextSplitColumns = 0;
1954 if (NextSplit.first == StringRef::npos) {
1955 ToNextSplitColumns = Token->getRemainingLength(LineIndex, TailOffset,
1956 ContentStartColumn);
1957 } else {
1958 ToNextSplitColumns = Token->getRangeLength(
1959 LineIndex, TailOffset,
1960 Split.first + Split.second + NextSplit.first, ContentStartColumn);
1961 }
1962 // Compress the whitespace between the break and the start of the next
1963 // unbreakable sequence.
1964 ToNextSplitColumns =
1965 Token->getLengthAfterCompression(ToNextSplitColumns, Split);
1966 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ContentStartColumn: "
<< ContentStartColumn << "\n"; } } while (false)
1967 << " ContentStartColumn: " << ContentStartColumn << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ContentStartColumn: "
<< ContentStartColumn << "\n"; } } while (false)
;
1968 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ToNextSplit: "
<< ToNextSplitColumns << "\n"; } } while (false)
1969 << " ToNextSplit: " << ToNextSplitColumns << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ToNextSplit: "
<< ToNextSplitColumns << "\n"; } } while (false)
;
1970 // If the whitespace compression makes us fit, continue on the current
1971 // line.
1972 bool ContinueOnLine =
1973 ContentStartColumn + ToNextSplitColumns <= ColumnLimit;
1974 unsigned ExcessCharactersPenalty = 0;
1975 if (!ContinueOnLine && !Strict) {
1976 // Similarly, if the excess characters' penalty is lower than the
1977 // penalty of introducing a new break, continue on the current line.
1978 ExcessCharactersPenalty =
1979 (ContentStartColumn + ToNextSplitColumns - ColumnLimit) *
1980 Style.PenaltyExcessCharacter;
1981 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Penalty excess: "
<< ExcessCharactersPenalty << "\n break : "
<< NewBreakPenalty << "\n"; } } while (false)
1982 << " Penalty excess: " << ExcessCharactersPenaltydo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Penalty excess: "
<< ExcessCharactersPenalty << "\n break : "
<< NewBreakPenalty << "\n"; } } while (false)
1983 << "\n break : " << NewBreakPenalty << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Penalty excess: "
<< ExcessCharactersPenalty << "\n break : "
<< NewBreakPenalty << "\n"; } } while (false)
;
1984 if (ExcessCharactersPenalty < NewBreakPenalty) {
1985 Exceeded = true;
1986 ContinueOnLine = true;
1987 }
1988 }
1989 if (ContinueOnLine) {
1990 LLVM_DEBUG(llvm::dbgs() << " Continuing on line...\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Continuing on line...\n"
; } } while (false)
;
1991 // The current line fits after compressing the whitespace - reflow
1992 // the next line into it if possible.
1993 TryReflow = true;
1994 if (!DryRun)
1995 Token->compressWhitespace(LineIndex, TailOffset, Split,
1996 Whitespaces);
1997 // When we continue on the same line, leave one space between content.
1998 ContentStartColumn += ToSplitColumns + 1;
1999 Penalty += ExcessCharactersPenalty;
2000 TailOffset += Split.first + Split.second;
2001 RemainingTokenColumns = Token->getRemainingLength(
2002 LineIndex, TailOffset, ContentStartColumn);
2003 continue;
2004 }
2005 }
2006 LLVM_DEBUG(llvm::dbgs() << " Breaking...\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Breaking...\n"
; } } while (false)
;
2007 // Update the ContentIndent only if the current line was not reflown with
2008 // the previous line, since in that case the previous line should still
2009 // determine the ContentIndent. Also never intent the last line.
2010 if (!Reflow)
2011 ContentIndent = Token->getContentIndent(LineIndex);
2012 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ContentIndent: "
<< ContentIndent << "\n"; } } while (false)
2013 << " ContentIndent: " << ContentIndent << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " ContentIndent: "
<< ContentIndent << "\n"; } } while (false)
;
2014 ContentStartColumn = ContentIndent + Token->getContentStartColumn(
2015 LineIndex, /*Break=*/true);
2016
2017 unsigned NewRemainingTokenColumns = Token->getRemainingLength(
2018 LineIndex, TailOffset + Split.first + Split.second,
2019 ContentStartColumn);
2020 if (NewRemainingTokenColumns == 0) {
2021 // No content to indent.
2022 ContentIndent = 0;
2023 ContentStartColumn =
2024 Token->getContentStartColumn(LineIndex, /*Break=*/true);
2025 NewRemainingTokenColumns = Token->getRemainingLength(
2026 LineIndex, TailOffset + Split.first + Split.second,
2027 ContentStartColumn);
2028 }
2029
2030 // When breaking before a tab character, it may be moved by a few columns,
2031 // but will still be expanded to the next tab stop, so we don't save any
2032 // columns.
2033 if (NewRemainingTokenColumns == RemainingTokenColumns) {
2034 // FIXME: Do we need to adjust the penalty?
2035 break;
2036 }
2037 assert(NewRemainingTokenColumns < RemainingTokenColumns)((NewRemainingTokenColumns < RemainingTokenColumns) ? static_cast
<void> (0) : __assert_fail ("NewRemainingTokenColumns < RemainingTokenColumns"
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 2037, __PRETTY_FUNCTION__))
;
2038
2039 LLVM_DEBUG(llvm::dbgs() << " Breaking at: " << TailOffset + Split.firstdo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Breaking at: "
<< TailOffset + Split.first << ", " << Split
.second << "\n"; } } while (false)
2040 << ", " << Split.second << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Breaking at: "
<< TailOffset + Split.first << ", " << Split
.second << "\n"; } } while (false)
;
2041 if (!DryRun)
2042 Token->insertBreak(LineIndex, TailOffset, Split, ContentIndent,
2043 Whitespaces);
2044
2045 Penalty += NewBreakPenalty;
2046 TailOffset += Split.first + Split.second;
2047 RemainingTokenColumns = NewRemainingTokenColumns;
2048 BreakInserted = true;
2049 NewBreakBefore = true;
2050 }
2051 // In case there's another line, prepare the state for the start of the next
2052 // line.
2053 if (LineIndex + 1 != EndIndex) {
2054 unsigned NextLineIndex = LineIndex + 1;
2055 if (NewBreakBefore)
2056 // After breaking a line, try to reflow the next line into the current
2057 // one once RemainingTokenColumns fits.
2058 TryReflow = true;
2059 if (TryReflow) {
2060 // We decided that we want to try reflowing the next line into the
2061 // current one.
2062 // We will now adjust the state as if the reflow is successful (in
2063 // preparation for the next line), and see whether that works. If we
2064 // decide that we cannot reflow, we will later reset the state to the
2065 // start of the next line.
2066 Reflow = false;
2067 // As we did not continue breaking the line, RemainingTokenColumns is
2068 // known to fit after ContentStartColumn. Adapt ContentStartColumn to
2069 // the position at which we want to format the next line if we do
2070 // actually reflow.
2071 // When we reflow, we need to add a space between the end of the current
2072 // line and the next line's start column.
2073 ContentStartColumn += RemainingTokenColumns + 1;
2074 // Get the split that we need to reflow next logical line into the end
2075 // of the current one; the split will include any leading whitespace of
2076 // the next logical line.
2077 BreakableToken::Split SplitBeforeNext =
2078 Token->getReflowSplit(NextLineIndex, CommentPragmasRegex);
2079 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Size of reflown text: "
<< ContentStartColumn << "\n Potential reflow split: "
; } } while (false)
2080 << " Size of reflown text: " << ContentStartColumndo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Size of reflown text: "
<< ContentStartColumn << "\n Potential reflow split: "
; } } while (false)
2081 << "\n Potential reflow split: ")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Size of reflown text: "
<< ContentStartColumn << "\n Potential reflow split: "
; } } while (false)
;
2082 if (SplitBeforeNext.first != StringRef::npos) {
2083 LLVM_DEBUG(llvm::dbgs() << SplitBeforeNext.first << ", "do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << SplitBeforeNext.
first << ", " << SplitBeforeNext.second << "\n"
; } } while (false)
2084 << SplitBeforeNext.second << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << SplitBeforeNext.
first << ", " << SplitBeforeNext.second << "\n"
; } } while (false)
;
2085 TailOffset = SplitBeforeNext.first + SplitBeforeNext.second;
2086 // If the rest of the next line fits into the current line below the
2087 // column limit, we can safely reflow.
2088 RemainingTokenColumns = Token->getRemainingLength(
2089 NextLineIndex, TailOffset, ContentStartColumn);
2090 Reflow = true;
2091 if (ContentStartColumn + RemainingTokenColumns > ColumnLimit) {
2092 LLVM_DEBUG(llvm::dbgs()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
2093 << " Over limit after reflow, need: "do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
2094 << (ContentStartColumn + RemainingTokenColumns)do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
2095 << ", space: " << ColumnLimitdo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
2096 << ", reflown prefix: " << ContentStartColumndo { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
2097 << ", offset in line: " << TailOffset << "\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Over limit after reflow, need: "
<< (ContentStartColumn + RemainingTokenColumns) <<
", space: " << ColumnLimit << ", reflown prefix: "
<< ContentStartColumn << ", offset in line: " <<
TailOffset << "\n"; } } while (false)
;
2098 // If the whole next line does not fit, try to find a point in
2099 // the next line at which we can break so that attaching the part
2100 // of the next line to that break point onto the current line is
2101 // below the column limit.
2102 BreakableToken::Split Split =
2103 Token->getSplit(NextLineIndex, TailOffset, ColumnLimit,
2104 ContentStartColumn, CommentPragmasRegex);
2105 if (Split.first == StringRef::npos) {
2106 LLVM_DEBUG(llvm::dbgs() << " Did not find later break\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Did not find later break\n"
; } } while (false)
;
2107 Reflow = false;
2108 } else {
2109 // Check whether the first split point gets us below the column
2110 // limit. Note that we will execute this split below as part of
2111 // the normal token breaking and reflow logic within the line.
2112 unsigned ToSplitColumns = Token->getRangeLength(
2113 NextLineIndex, TailOffset, Split.first, ContentStartColumn);
2114 if (ContentStartColumn + ToSplitColumns > ColumnLimit) {
2115 LLVM_DEBUG(llvm::dbgs() << " Next split protrudes, need: "do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Next split protrudes, need: "
<< (ContentStartColumn + ToSplitColumns) << ", space: "
<< ColumnLimit; } } while (false)
2116 << (ContentStartColumn + ToSplitColumns)do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Next split protrudes, need: "
<< (ContentStartColumn + ToSplitColumns) << ", space: "
<< ColumnLimit; } } while (false)
2117 << ", space: " << ColumnLimit)do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << " Next split protrudes, need: "
<< (ContentStartColumn + ToSplitColumns) << ", space: "
<< ColumnLimit; } } while (false)
;
2118 unsigned ExcessCharactersPenalty =
2119 (ContentStartColumn + ToSplitColumns - ColumnLimit) *
2120 Style.PenaltyExcessCharacter;
2121 if (NewBreakPenalty < ExcessCharactersPenalty) {
2122 Reflow = false;
2123 }
2124 }
2125 }
2126 }
2127 } else {
2128 LLVM_DEBUG(llvm::dbgs() << "not found.\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << "not found.\n"; }
} while (false)
;
2129 }
2130 }
2131 if (!Reflow) {
2132 // If we didn't reflow into the next line, the only space to consider is
2133 // the next logical line. Reset our state to match the start of the next
2134 // line.
2135 TailOffset = 0;
2136 ContentStartColumn =
2137 Token->getContentStartColumn(NextLineIndex, /*Break=*/false);
2138 RemainingTokenColumns = Token->getRemainingLength(
2139 NextLineIndex, TailOffset, ContentStartColumn);
2140 // Adapt the start of the token, for example indent.
2141 if (!DryRun)
2142 Token->adaptStartOfLine(NextLineIndex, Whitespaces);
2143 } else {
2144 // If we found a reflow split and have added a new break before the next
2145 // line, we are going to remove the line break at the start of the next
2146 // logical line. For example, here we'll add a new line break after
2147 // 'text', and subsequently delete the line break between 'that' and
2148 // 'reflows'.
2149 // // some text that
2150 // // reflows
2151 // ->
2152 // // some text
2153 // // that reflows
2154 // When adding the line break, we also added the penalty for it, so we
2155 // need to subtract that penalty again when we remove the line break due
2156 // to reflowing.
2157 if (NewBreakBefore) {
2158 assert(Penalty >= NewBreakPenalty)((Penalty >= NewBreakPenalty) ? static_cast<void> (0
) : __assert_fail ("Penalty >= NewBreakPenalty", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/ContinuationIndenter.cpp"
, 2158, __PRETTY_FUNCTION__))
;
2159 Penalty -= NewBreakPenalty;
2160 }
2161 if (!DryRun)
2162 Token->reflow(NextLineIndex, Whitespaces);
2163 }
2164 }
2165 }
2166
2167 BreakableToken::Split SplitAfterLastLine =
2168 Token->getSplitAfterLastLine(TailOffset);
2169 if (SplitAfterLastLine.first != StringRef::npos) {
2170 LLVM_DEBUG(llvm::dbgs() << "Replacing whitespace after last line.\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType
("format-indenter")) { llvm::dbgs() << "Replacing whitespace after last line.\n"
; } } while (false)
;
2171
2172 // We add the last line's penalty here, since that line is going to be split
2173 // now.
2174 Penalty += Style.PenaltyExcessCharacter *
2175 (ContentStartColumn + RemainingTokenColumns - ColumnLimit);
2176
2177 if (!DryRun)
2178 Token->replaceWhitespaceAfterLastLine(TailOffset, SplitAfterLastLine,
2179 Whitespaces);
2180 ContentStartColumn =
2181 Token->getContentStartColumn(Token->getLineCount() - 1, /*Break=*/true);
2182 RemainingTokenColumns = Token->getRemainingLength(
2183 Token->getLineCount() - 1,
2184 TailOffset + SplitAfterLastLine.first + SplitAfterLastLine.second,
2185 ContentStartColumn);
2186 }
2187
2188 State.Column = ContentStartColumn + RemainingTokenColumns -
2189 Current.UnbreakableTailLength;
2190
2191 if (BreakInserted) {
2192 // If we break the token inside a parameter list, we need to break before
2193 // the next parameter on all levels, so that the next parameter is clearly
2194 // visible. Line comments already introduce a break.
2195 if (Current.isNot(TT_LineComment)) {
2196 for (unsigned i = 0, e = State.Stack.size(); i != e; ++i)
2197 State.Stack[i].BreakBeforeParameter = true;
2198 }
2199
2200 if (Current.is(TT_BlockComment))
2201 State.NoContinuation = true;
2202
2203 State.Stack.back().LastSpace = StartColumn;
2204 }
2205
2206 Token->updateNextToken(State);
2207
2208 return {Penalty, Exceeded};
2209}
2210
2211unsigned ContinuationIndenter::getColumnLimit(const LineState &State) const {
2212 // In preprocessor directives reserve two chars for trailing " \"
2213 return Style.ColumnLimit - (State.Line->InPPDirective ? 2 : 0);
2214}
2215
2216bool ContinuationIndenter::nextIsMultilineString(const LineState &State) {
2217 const FormatToken &Current = *State.NextToken;
2218 if (!Current.isStringLiteral() || Current.is(TT_ImplicitStringLiteral))
2219 return false;
2220 // We never consider raw string literals "multiline" for the purpose of
2221 // AlwaysBreakBeforeMultilineStrings implementation as they are special-cased
2222 // (see TokenAnnotator::mustBreakBefore().
2223 if (Current.TokenText.startswith("R\""))
2224 return false;
2225 if (Current.IsMultiline)
2226 return true;
2227 if (Current.getNextNonComment() &&
2228 Current.getNextNonComment()->isStringLiteral())
2229 return true; // Implicit concatenation.
2230 if (Style.ColumnLimit != 0 && Style.BreakStringLiterals &&
2231 State.Column + Current.ColumnWidth + Current.UnbreakableTailLength >
2232 Style.ColumnLimit)
2233 return true; // String will be split.
2234 return false;
2235}
2236
2237} // namespace format
2238} // namespace clang

/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/lib/Format/FormatToken.h

1//===--- FormatToken.h - Format C++ code ------------------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8///
9/// \file
10/// This file contains the declaration of the FormatToken, a wrapper
11/// around Token with additional information related to formatting.
12///
13//===----------------------------------------------------------------------===//
14
15#ifndef LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H
16#define LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H
17
18#include "clang/Basic/IdentifierTable.h"
19#include "clang/Basic/OperatorPrecedence.h"
20#include "clang/Format/Format.h"
21#include "clang/Lex/Lexer.h"
22#include <memory>
23#include <unordered_set>
24
25namespace clang {
26namespace format {
27
28#define LIST_TOKEN_TYPESTYPE(ArrayInitializerLSquare) TYPE(ArraySubscriptLSquare) TYPE
(AttributeColon) TYPE(AttributeParen) TYPE(AttributeSquare) TYPE
(BinaryOperator) TYPE(BitFieldColon) TYPE(BlockComment) TYPE(
CastRParen) TYPE(ConditionalExpr) TYPE(ConflictAlternative) TYPE
(ConflictEnd) TYPE(ConflictStart) TYPE(CtorInitializerColon) TYPE
(CtorInitializerComma) TYPE(DesignatedInitializerLSquare) TYPE
(DesignatedInitializerPeriod) TYPE(DictLiteral) TYPE(ForEachMacro
) TYPE(FunctionAnnotationRParen) TYPE(FunctionDeclarationName
) TYPE(FunctionLBrace) TYPE(FunctionTypeLParen) TYPE(ImplicitStringLiteral
) TYPE(InheritanceColon) TYPE(InheritanceComma) TYPE(InlineASMBrace
) TYPE(InlineASMColon) TYPE(JavaAnnotation) TYPE(JsComputedPropertyName
) TYPE(JsExponentiation) TYPE(JsExponentiationEqual) TYPE(JsFatArrow
) TYPE(JsNonNullAssertion) TYPE(JsNullishCoalescingOperator) TYPE
(JsNullPropagatingOperator) TYPE(JsPrivateIdentifier) TYPE(JsTypeColon
) TYPE(JsTypeOperator) TYPE(JsTypeOptionalQuestion) TYPE(LambdaArrow
) TYPE(LambdaLBrace) TYPE(LambdaLSquare) TYPE(LeadingJavaAnnotation
) TYPE(LineComment) TYPE(MacroBlockBegin) TYPE(MacroBlockEnd)
TYPE(NamespaceMacro) TYPE(ObjCBlockLBrace) TYPE(ObjCBlockLParen
) TYPE(ObjCDecl) TYPE(ObjCForIn) TYPE(ObjCMethodExpr) TYPE(ObjCMethodSpecifier
) TYPE(ObjCProperty) TYPE(ObjCStringLiteral) TYPE(OverloadedOperator
) TYPE(OverloadedOperatorLParen) TYPE(PointerOrReference) TYPE
(PureVirtualSpecifier) TYPE(RangeBasedForLoopColon) TYPE(RegexLiteral
) TYPE(SelectorName) TYPE(StartOfName) TYPE(StatementMacro) TYPE
(StructuredBindingLSquare) TYPE(TemplateCloser) TYPE(TemplateOpener
) TYPE(TemplateString) TYPE(ProtoExtensionLSquare) TYPE(TrailingAnnotation
) TYPE(TrailingReturnArrow) TYPE(TrailingUnaryOperator) TYPE(
TypenameMacro) TYPE(UnaryOperator) TYPE(CSharpStringLiteral) TYPE
(CSharpNullCoalescing) TYPE(Unknown)
\
29 TYPE(ArrayInitializerLSquare) \
30 TYPE(ArraySubscriptLSquare) \
31 TYPE(AttributeColon) \
32 TYPE(AttributeParen) \
33 TYPE(AttributeSquare) \
34 TYPE(BinaryOperator) \
35 TYPE(BitFieldColon) \
36 TYPE(BlockComment) \
37 TYPE(CastRParen) \
38 TYPE(ConditionalExpr) \
39 TYPE(ConflictAlternative) \
40 TYPE(ConflictEnd) \
41 TYPE(ConflictStart) \
42 TYPE(CtorInitializerColon) \
43 TYPE(CtorInitializerComma) \
44 TYPE(DesignatedInitializerLSquare) \
45 TYPE(DesignatedInitializerPeriod) \
46 TYPE(DictLiteral) \
47 TYPE(ForEachMacro) \
48 TYPE(FunctionAnnotationRParen) \
49 TYPE(FunctionDeclarationName) \
50 TYPE(FunctionLBrace) \
51 TYPE(FunctionTypeLParen) \
52 TYPE(ImplicitStringLiteral) \
53 TYPE(InheritanceColon) \
54 TYPE(InheritanceComma) \
55 TYPE(InlineASMBrace) \
56 TYPE(InlineASMColon) \
57 TYPE(JavaAnnotation) \
58 TYPE(JsComputedPropertyName) \
59 TYPE(JsExponentiation) \
60 TYPE(JsExponentiationEqual) \
61 TYPE(JsFatArrow) \
62 TYPE(JsNonNullAssertion) \
63 TYPE(JsNullishCoalescingOperator) \
64 TYPE(JsNullPropagatingOperator) \
65 TYPE(JsPrivateIdentifier) \
66 TYPE(JsTypeColon) \
67 TYPE(JsTypeOperator) \
68 TYPE(JsTypeOptionalQuestion) \
69 TYPE(LambdaArrow) \
70 TYPE(LambdaLBrace) \
71 TYPE(LambdaLSquare) \
72 TYPE(LeadingJavaAnnotation) \
73 TYPE(LineComment) \
74 TYPE(MacroBlockBegin) \
75 TYPE(MacroBlockEnd) \
76 TYPE(NamespaceMacro) \
77 TYPE(ObjCBlockLBrace) \
78 TYPE(ObjCBlockLParen) \
79 TYPE(ObjCDecl) \
80 TYPE(ObjCForIn) \
81 TYPE(ObjCMethodExpr) \
82 TYPE(ObjCMethodSpecifier) \
83 TYPE(ObjCProperty) \
84 TYPE(ObjCStringLiteral) \
85 TYPE(OverloadedOperator) \
86 TYPE(OverloadedOperatorLParen) \
87 TYPE(PointerOrReference) \
88 TYPE(PureVirtualSpecifier) \
89 TYPE(RangeBasedForLoopColon) \
90 TYPE(RegexLiteral) \
91 TYPE(SelectorName) \
92 TYPE(StartOfName) \
93 TYPE(StatementMacro) \
94 TYPE(StructuredBindingLSquare) \
95 TYPE(TemplateCloser) \
96 TYPE(TemplateOpener) \
97 TYPE(TemplateString) \
98 TYPE(ProtoExtensionLSquare) \
99 TYPE(TrailingAnnotation) \
100 TYPE(TrailingReturnArrow) \
101 TYPE(TrailingUnaryOperator) \
102 TYPE(TypenameMacro) \
103 TYPE(UnaryOperator) \
104 TYPE(CSharpStringLiteral) \
105 TYPE(CSharpNullCoalescing) \
106 TYPE(Unknown)
107
108enum TokenType {
109#define TYPE(X) TT_##X,
110 LIST_TOKEN_TYPESTYPE(ArrayInitializerLSquare) TYPE(ArraySubscriptLSquare) TYPE
(AttributeColon) TYPE(AttributeParen) TYPE(AttributeSquare) TYPE
(BinaryOperator) TYPE(BitFieldColon) TYPE(BlockComment) TYPE(
CastRParen) TYPE(ConditionalExpr) TYPE(ConflictAlternative) TYPE
(ConflictEnd) TYPE(ConflictStart) TYPE(CtorInitializerColon) TYPE
(CtorInitializerComma) TYPE(DesignatedInitializerLSquare) TYPE
(DesignatedInitializerPeriod) TYPE(DictLiteral) TYPE(ForEachMacro
) TYPE(FunctionAnnotationRParen) TYPE(FunctionDeclarationName
) TYPE(FunctionLBrace) TYPE(FunctionTypeLParen) TYPE(ImplicitStringLiteral
) TYPE(InheritanceColon) TYPE(InheritanceComma) TYPE(InlineASMBrace
) TYPE(InlineASMColon) TYPE(JavaAnnotation) TYPE(JsComputedPropertyName
) TYPE(JsExponentiation) TYPE(JsExponentiationEqual) TYPE(JsFatArrow
) TYPE(JsNonNullAssertion) TYPE(JsNullishCoalescingOperator) TYPE
(JsNullPropagatingOperator) TYPE(JsPrivateIdentifier) TYPE(JsTypeColon
) TYPE(JsTypeOperator) TYPE(JsTypeOptionalQuestion) TYPE(LambdaArrow
) TYPE(LambdaLBrace) TYPE(LambdaLSquare) TYPE(LeadingJavaAnnotation
) TYPE(LineComment) TYPE(MacroBlockBegin) TYPE(MacroBlockEnd)
TYPE(NamespaceMacro) TYPE(ObjCBlockLBrace) TYPE(ObjCBlockLParen
) TYPE(ObjCDecl) TYPE(ObjCForIn) TYPE(ObjCMethodExpr) TYPE(ObjCMethodSpecifier
) TYPE(ObjCProperty) TYPE(ObjCStringLiteral) TYPE(OverloadedOperator
) TYPE(OverloadedOperatorLParen) TYPE(PointerOrReference) TYPE
(PureVirtualSpecifier) TYPE(RangeBasedForLoopColon) TYPE(RegexLiteral
) TYPE(SelectorName) TYPE(StartOfName) TYPE(StatementMacro) TYPE
(StructuredBindingLSquare) TYPE(TemplateCloser) TYPE(TemplateOpener
) TYPE(TemplateString) TYPE(ProtoExtensionLSquare) TYPE(TrailingAnnotation
) TYPE(TrailingReturnArrow) TYPE(TrailingUnaryOperator) TYPE(
TypenameMacro) TYPE(UnaryOperator) TYPE(CSharpStringLiteral) TYPE
(CSharpNullCoalescing) TYPE(Unknown)
111#undef TYPE
112 NUM_TOKEN_TYPES
113};
114
115/// Determines the name of a token type.
116const char *getTokenTypeName(TokenType Type);
117
118// Represents what type of block a set of braces open.
119enum BraceBlockKind { BK_Unknown, BK_Block, BK_BracedInit };
120
121// The packing kind of a function's parameters.
122enum ParameterPackingKind { PPK_BinPacked, PPK_OnePerLine, PPK_Inconclusive };
123
124enum FormatDecision { FD_Unformatted, FD_Continue, FD_Break };
125
126class TokenRole;
127class AnnotatedLine;
128
129/// A wrapper around a \c Token storing information about the
130/// whitespace characters preceding it.
131struct FormatToken {
132 FormatToken() {}
133
134 /// The \c Token.
135 Token Tok;
136
137 /// The number of newlines immediately before the \c Token.
138 ///
139 /// This can be used to determine what the user wrote in the original code
140 /// and thereby e.g. leave an empty line between two function definitions.
141 unsigned NewlinesBefore = 0;
142
143 /// Whether there is at least one unescaped newline before the \c
144 /// Token.
145 bool HasUnescapedNewline = false;
146
147 /// The range of the whitespace immediately preceding the \c Token.
148 SourceRange WhitespaceRange;
149
150 /// The offset just past the last '\n' in this token's leading
151 /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'.
152 unsigned LastNewlineOffset = 0;
153
154 /// The width of the non-whitespace parts of the token (or its first
155 /// line for multi-line tokens) in columns.
156 /// We need this to correctly measure number of columns a token spans.
157 unsigned ColumnWidth = 0;
158
159 /// Contains the width in columns of the last line of a multi-line
160 /// token.
161 unsigned LastLineColumnWidth = 0;
162
163 /// Whether the token text contains newlines (escaped or not).
164 bool IsMultiline = false;
165
166 /// Indicates that this is the first token of the file.
167 bool IsFirst = false;
168
169 /// Whether there must be a line break before this token.
170 ///
171 /// This happens for example when a preprocessor directive ended directly
172 /// before the token.
173 bool MustBreakBefore = false;
174
175 /// The raw text of the token.
176 ///
177 /// Contains the raw token text without leading whitespace and without leading
178 /// escaped newlines.
179 StringRef TokenText;
180
181 /// Set to \c true if this token is an unterminated literal.
182 bool IsUnterminatedLiteral = 0;
183
184 /// Contains the kind of block if this token is a brace.
185 BraceBlockKind BlockKind = BK_Unknown;
186
187 TokenType Type = TT_Unknown;
188
189 /// The number of spaces that should be inserted before this token.
190 unsigned SpacesRequiredBefore = 0;
191
192 /// \c true if it is allowed to break before this token.
193 bool CanBreakBefore = false;
194
195 /// \c true if this is the ">" of "template<..>".
196 bool ClosesTemplateDeclaration = false;
197
198 /// Number of parameters, if this is "(", "[" or "<".
199 unsigned ParameterCount = 0;
200
201 /// Number of parameters that are nested blocks,
202 /// if this is "(", "[" or "<".
203 unsigned BlockParameterCount = 0;
204
205 /// If this is a bracket ("<", "(", "[" or "{"), contains the kind of
206 /// the surrounding bracket.
207 tok::TokenKind ParentBracket = tok::unknown;
208
209 /// A token can have a special role that can carry extra information
210 /// about the token's formatting.
211 std::unique_ptr<TokenRole> Role;
212
213 /// If this is an opening parenthesis, how are the parameters packed?
214 ParameterPackingKind PackingKind = PPK_Inconclusive;
215
216 /// The total length of the unwrapped line up to and including this
217 /// token.
218 unsigned TotalLength = 0;
219
220 /// The original 0-based column of this token, including expanded tabs.
221 /// The configured TabWidth is used as tab width.
222 unsigned OriginalColumn = 0;
223
224 /// The length of following tokens until the next natural split point,
225 /// or the next token that can be broken.
226 unsigned UnbreakableTailLength = 0;
227
228 // FIXME: Come up with a 'cleaner' concept.
229 /// The binding strength of a token. This is a combined value of
230 /// operator precedence, parenthesis nesting, etc.
231 unsigned BindingStrength = 0;
232
233 /// The nesting level of this token, i.e. the number of surrounding (),
234 /// [], {} or <>.
235 unsigned NestingLevel = 0;
236
237 /// The indent level of this token. Copied from the surrounding line.
238 unsigned IndentLevel = 0;
239
240 /// Penalty for inserting a line break before this token.
241 unsigned SplitPenalty = 0;
242
243 /// If this is the first ObjC selector name in an ObjC method
244 /// definition or call, this contains the length of the longest name.
245 ///
246 /// This being set to 0 means that the selectors should not be colon-aligned,
247 /// e.g. because several of them are block-type.
248 unsigned LongestObjCSelectorName = 0;
249
250 /// If this is the first ObjC selector name in an ObjC method
251 /// definition or call, this contains the number of parts that the whole
252 /// selector consist of.
253 unsigned ObjCSelectorNameParts = 0;
254
255 /// The 0-based index of the parameter/argument. For ObjC it is set
256 /// for the selector name token.
257 /// For now calculated only for ObjC.
258 unsigned ParameterIndex = 0;
259
260 /// Stores the number of required fake parentheses and the
261 /// corresponding operator precedence.
262 ///
263 /// If multiple fake parentheses start at a token, this vector stores them in
264 /// reverse order, i.e. inner fake parenthesis first.
265 SmallVector<prec::Level, 4> FakeLParens;
266 /// Insert this many fake ) after this token for correct indentation.
267 unsigned FakeRParens = 0;
268
269 /// \c true if this token starts a binary expression, i.e. has at least
270 /// one fake l_paren with a precedence greater than prec::Unknown.
271 bool StartsBinaryExpression = false;
272 /// \c true if this token ends a binary expression.
273 bool EndsBinaryExpression = false;
274
275 /// If this is an operator (or "."/"->") in a sequence of operators
276 /// with the same precedence, contains the 0-based operator index.
277 unsigned OperatorIndex = 0;
278
279 /// If this is an operator (or "."/"->") in a sequence of operators
280 /// with the same precedence, points to the next operator.
281 FormatToken *NextOperator = nullptr;
282
283 /// Is this token part of a \c DeclStmt defining multiple variables?
284 ///
285 /// Only set if \c Type == \c TT_StartOfName.
286 bool PartOfMultiVariableDeclStmt = false;
287
288 /// Does this line comment continue a line comment section?
289 ///
290 /// Only set to true if \c Type == \c TT_LineComment.
291 bool ContinuesLineCommentSection = false;
292
293 /// If this is a bracket, this points to the matching one.
294 FormatToken *MatchingParen = nullptr;
295
296 /// The previous token in the unwrapped line.
297 FormatToken *Previous = nullptr;
298
299 /// The next token in the unwrapped line.
300 FormatToken *Next = nullptr;
301
302 /// If this token starts a block, this contains all the unwrapped lines
303 /// in it.
304 SmallVector<AnnotatedLine *, 1> Children;
305
306 /// Stores the formatting decision for the token once it was made.
307 FormatDecision Decision = FD_Unformatted;
308
309 /// If \c true, this token has been fully formatted (indented and
310 /// potentially re-formatted inside), and we do not allow further formatting
311 /// changes.
312 bool Finalized = false;
313
314 bool is(tok::TokenKind Kind) const { return Tok.is(Kind); }
3
Calling 'Token::is'
6
Returning from 'Token::is'
7
Returning the value 1, which participates in a condition later
10
Calling 'Token::is'
13
Returning from 'Token::is'
14
Returning the value 1, which participates in a condition later
315 bool is(TokenType TT) const { return Type == TT; }
316 bool is(const IdentifierInfo *II) const {
317 return II && II == Tok.getIdentifierInfo();
318 }
319 bool is(tok::PPKeywordKind Kind) const {
320 return Tok.getIdentifierInfo() &&
321 Tok.getIdentifierInfo()->getPPKeywordID() == Kind;
322 }
323 template <typename A, typename B> bool isOneOf(A K1, B K2) const {
324 return is(K1) || is(K2);
325 }
326 template <typename A, typename B, typename... Ts>
327 bool isOneOf(A K1, B K2, Ts... Ks) const {
328 return is(K1) || isOneOf(K2, Ks...);
329 }
330 template <typename T> bool isNot(T Kind) const { return !is(Kind); }
331
332 bool isIf(bool AllowConstexprMacro = true) const {
333 return is(tok::kw_if) || endsSequence(tok::kw_constexpr, tok::kw_if) ||
334 (endsSequence(tok::identifier, tok::kw_if) && AllowConstexprMacro);
335 }
336
337 bool closesScopeAfterBlock() const {
338 if (BlockKind == BK_Block)
339 return true;
340 if (closesScope())
341 return Previous->closesScopeAfterBlock();
342 return false;
343 }
344
345 /// \c true if this token starts a sequence with the given tokens in order,
346 /// following the ``Next`` pointers, ignoring comments.
347 template <typename A, typename... Ts>
348 bool startsSequence(A K1, Ts... Tokens) const {
349 return startsSequenceInternal(K1, Tokens...);
350 }
351
352 /// \c true if this token ends a sequence with the given tokens in order,
353 /// following the ``Previous`` pointers, ignoring comments.
354 /// For example, given tokens [T1, T2, T3], the function returns true if
355 /// 3 tokens ending at this (ignoring comments) are [T3, T2, T1]. In other
356 /// words, the tokens passed to this function need to the reverse of the
357 /// order the tokens appear in code.
358 template <typename A, typename... Ts>
359 bool endsSequence(A K1, Ts... Tokens) const {
360 return endsSequenceInternal(K1, Tokens...);
361 }
362
363 bool isStringLiteral() const { return tok::isStringLiteral(Tok.getKind()); }
364
365 bool isObjCAtKeyword(tok::ObjCKeywordKind Kind) const {
366 return Tok.isObjCAtKeyword(Kind);
367 }
368
369 bool isAccessSpecifier(bool ColonRequired = true) const {
370 return isOneOf(tok::kw_public, tok::kw_protected, tok::kw_private) &&
371 (!ColonRequired || (Next && Next->is(tok::colon)));
372 }
373
374 /// Determine whether the token is a simple-type-specifier.
375 bool isSimpleTypeSpecifier() const;
376
377 bool isObjCAccessSpecifier() const {
378 return is(tok::at) && Next &&
379 (Next->isObjCAtKeyword(tok::objc_public) ||
380 Next->isObjCAtKeyword(tok::objc_protected) ||
381 Next->isObjCAtKeyword(tok::objc_package) ||
382 Next->isObjCAtKeyword(tok::objc_private));
383 }
384
385 /// Returns whether \p Tok is ([{ or an opening < of a template or in
386 /// protos.
387 bool opensScope() const {
388 if (is(TT_TemplateString) && TokenText.endswith("${"))
389 return true;
390 if (is(TT_DictLiteral) && is(tok::less))
391 return true;
392 return isOneOf(tok::l_paren, tok::l_brace, tok::l_square,
393 TT_TemplateOpener);
394 }
395 /// Returns whether \p Tok is )]} or a closing > of a template or in
396 /// protos.
397 bool closesScope() const {
398 if (is(TT_TemplateString) && TokenText.startswith("}"))
399 return true;
400 if (is(TT_DictLiteral) && is(tok::greater))
401 return true;
402 return isOneOf(tok::r_paren, tok::r_brace, tok::r_square,
403 TT_TemplateCloser);
404 }
405
406 /// Returns \c true if this is a "." or "->" accessing a member.
407 bool isMemberAccess() const {
408 return isOneOf(tok::arrow, tok::period, tok::arrowstar) &&
409 !isOneOf(TT_DesignatedInitializerPeriod, TT_TrailingReturnArrow,
410 TT_LambdaArrow);
411 }
412
413 bool isUnaryOperator() const {
414 switch (Tok.getKind()) {
415 case tok::plus:
416 case tok::plusplus:
417 case tok::minus:
418 case tok::minusminus:
419 case tok::exclaim:
420 case tok::tilde:
421 case tok::kw_sizeof:
422 case tok::kw_alignof:
423 return true;
424 default:
425 return false;
426 }
427 }
428
429 bool isBinaryOperator() const {
430 // Comma is a binary operator, but does not behave as such wrt. formatting.
431 return getPrecedence() > prec::Comma;
432 }
433
434 bool isTrailingComment() const {
435 return is(tok::comment) &&
436 (is(TT_LineComment) || !Next || Next->NewlinesBefore > 0);
437 }
438
439 /// Returns \c true if this is a keyword that can be used
440 /// like a function call (e.g. sizeof, typeid, ...).
441 bool isFunctionLikeKeyword() const {
442 switch (Tok.getKind()) {
443 case tok::kw_throw:
444 case tok::kw_typeid:
445 case tok::kw_return:
446 case tok::kw_sizeof:
447 case tok::kw_alignof:
448 case tok::kw_alignas:
449 case tok::kw_decltype:
450 case tok::kw_noexcept:
451 case tok::kw_static_assert:
452 case tok::kw___attribute:
453 return true;
454 default:
455 return false;
456 }
457 }
458
459 /// Returns \c true if this is a string literal that's like a label,
460 /// e.g. ends with "=" or ":".
461 bool isLabelString() const {
462 if (!is(tok::string_literal))
463 return false;
464 StringRef Content = TokenText;
465 if (Content.startswith("\"") || Content.startswith("'"))
466 Content = Content.drop_front(1);
467 if (Content.endswith("\"") || Content.endswith("'"))
468 Content = Content.drop_back(1);
469 Content = Content.trim();
470 return Content.size() > 1 &&
471 (Content.back() == ':' || Content.back() == '=');
472 }
473
474 /// Returns actual token start location without leading escaped
475 /// newlines and whitespace.
476 ///
477 /// This can be different to Tok.getLocation(), which includes leading escaped
478 /// newlines.
479 SourceLocation getStartOfNonWhitespace() const {
480 return WhitespaceRange.getEnd();
481 }
482
483 prec::Level getPrecedence() const {
484 return getBinOpPrecedence(Tok.getKind(), /*GreaterThanIsOperator=*/true,
485 /*CPlusPlus11=*/true);
486 }
487
488 /// Returns the previous token ignoring comments.
489 FormatToken *getPreviousNonComment() const {
490 FormatToken *Tok = Previous;
491 while (Tok && Tok->is(tok::comment))
492 Tok = Tok->Previous;
493 return Tok;
494 }
495
496 /// Returns the next token ignoring comments.
497 const FormatToken *getNextNonComment() const {
498 const FormatToken *Tok = Next;
499 while (Tok && Tok->is(tok::comment))
500 Tok = Tok->Next;
501 return Tok;
502 }
503
504 /// Returns \c true if this tokens starts a block-type list, i.e. a
505 /// list that should be indented with a block indent.
506 bool opensBlockOrBlockTypeList(const FormatStyle &Style) const {
507 if (is(TT_TemplateString) && opensScope())
508 return true;
509 return is(TT_ArrayInitializerLSquare) || is(TT_ProtoExtensionLSquare) ||
510 (is(tok::l_brace) &&
511 (BlockKind == BK_Block || is(TT_DictLiteral) ||
512 (!Style.Cpp11BracedListStyle && NestingLevel == 0))) ||
513 (is(tok::less) && (Style.Language == FormatStyle::LK_Proto ||
514 Style.Language == FormatStyle::LK_TextProto));
515 }
516
517 /// Returns whether the token is the left square bracket of a C++
518 /// structured binding declaration.
519 bool isCppStructuredBinding(const FormatStyle &Style) const {
520 if (!Style.isCpp() || isNot(tok::l_square))
521 return false;
522 const FormatToken *T = this;
523 do {
524 T = T->getPreviousNonComment();
525 } while (T && T->isOneOf(tok::kw_const, tok::kw_volatile, tok::amp,
526 tok::ampamp));
527 return T && T->is(tok::kw_auto);
528 }
529
530 /// Same as opensBlockOrBlockTypeList, but for the closing token.
531 bool closesBlockOrBlockTypeList(const FormatStyle &Style) const {
532 if (is(TT_TemplateString) && closesScope())
533 return true;
534 return MatchingParen && MatchingParen->opensBlockOrBlockTypeList(Style);
535 }
536
537 /// Return the actual namespace token, if this token starts a namespace
538 /// block.
539 const FormatToken *getNamespaceToken() const {
540 const FormatToken *NamespaceTok = this;
541 if (is(tok::comment))
542 NamespaceTok = NamespaceTok->getNextNonComment();
543 // Detect "(inline|export)? namespace" in the beginning of a line.
544 if (NamespaceTok && NamespaceTok->isOneOf(tok::kw_inline, tok::kw_export))
545 NamespaceTok = NamespaceTok->getNextNonComment();
546 return NamespaceTok &&
547 NamespaceTok->isOneOf(tok::kw_namespace, TT_NamespaceMacro)
548 ? NamespaceTok
549 : nullptr;
550 }
551
552private:
553 // Disallow copying.
554 FormatToken(const FormatToken &) = delete;
555 void operator=(const FormatToken &) = delete;
556
557 template <typename A, typename... Ts>
558 bool startsSequenceInternal(A K1, Ts... Tokens) const {
559 if (is(tok::comment) && Next)
560 return Next->startsSequenceInternal(K1, Tokens...);
561 return is(K1) && Next && Next->startsSequenceInternal(Tokens...);
562 }
563
564 template <typename A> bool startsSequenceInternal(A K1) const {
565 if (is(tok::comment) && Next)
566 return Next->startsSequenceInternal(K1);
567 return is(K1);
568 }
569
570 template <typename A, typename... Ts> bool endsSequenceInternal(A K1) const {
571 if (is(tok::comment) && Previous)
572 return Previous->endsSequenceInternal(K1);
573 return is(K1);
574 }
575
576 template <typename A, typename... Ts>
577 bool endsSequenceInternal(A K1, Ts... Tokens) const {
578 if (is(tok::comment) && Previous)
579 return Previous->endsSequenceInternal(K1, Tokens...);
580 return is(K1) && Previous && Previous->endsSequenceInternal(Tokens...);
581 }
582};
583
584class ContinuationIndenter;
585struct LineState;
586
587class TokenRole {
588public:
589 TokenRole(const FormatStyle &Style) : Style(Style) {}
590 virtual ~TokenRole();
591
592 /// After the \c TokenAnnotator has finished annotating all the tokens,
593 /// this function precomputes required information for formatting.
594 virtual void precomputeFormattingInfos(const FormatToken *Token);
595
596 /// Apply the special formatting that the given role demands.
597 ///
598 /// Assumes that the token having this role is already formatted.
599 ///
600 /// Continues formatting from \p State leaving indentation to \p Indenter and
601 /// returns the total penalty that this formatting incurs.
602 virtual unsigned formatFromToken(LineState &State,
603 ContinuationIndenter *Indenter,
604 bool DryRun) {
605 return 0;
606 }
607
608 /// Same as \c formatFromToken, but assumes that the first token has
609 /// already been set thereby deciding on the first line break.
610 virtual unsigned formatAfterToken(LineState &State,
611 ContinuationIndenter *Indenter,
612 bool DryRun) {
613 return 0;
614 }
615
616 /// Notifies the \c Role that a comma was found.
617 virtual void CommaFound(const FormatToken *Token) {}
618
619 virtual const FormatToken *lastComma() { return nullptr; }
620
621protected:
622 const FormatStyle &Style;
623};
624
625class CommaSeparatedList : public TokenRole {
626public:
627 CommaSeparatedList(const FormatStyle &Style)
628 : TokenRole(Style), HasNestedBracedList(false) {}
629
630 void precomputeFormattingInfos(const FormatToken *Token) override;
631
632 unsigned formatAfterToken(LineState &State, ContinuationIndenter *Indenter,
633 bool DryRun) override;
634
635 unsigned formatFromToken(LineState &State, ContinuationIndenter *Indenter,
636 bool DryRun) override;
637
638 /// Adds \p Token as the next comma to the \c CommaSeparated list.
639 void CommaFound(const FormatToken *Token) override {
640 Commas.push_back(Token);
641 }
642
643 const FormatToken *lastComma() override {
644 if (Commas.empty())
645 return nullptr;
646 return Commas.back();
647 }
648
649private:
650 /// A struct that holds information on how to format a given list with
651 /// a specific number of columns.
652 struct ColumnFormat {
653 /// The number of columns to use.
654 unsigned Columns;
655
656 /// The total width in characters.
657 unsigned TotalWidth;
658
659 /// The number of lines required for this format.
660 unsigned LineCount;
661
662 /// The size of each column in characters.
663 SmallVector<unsigned, 8> ColumnSizes;
664 };
665
666 /// Calculate which \c ColumnFormat fits best into
667 /// \p RemainingCharacters.
668 const ColumnFormat *getColumnFormat(unsigned RemainingCharacters) const;
669
670 /// The ordered \c FormatTokens making up the commas of this list.
671 SmallVector<const FormatToken *, 8> Commas;
672
673 /// The length of each of the list's items in characters including the
674 /// trailing comma.
675 SmallVector<unsigned, 8> ItemLengths;
676
677 /// Precomputed formats that can be used for this list.
678 SmallVector<ColumnFormat, 4> Formats;
679
680 bool HasNestedBracedList;
681};
682
683/// Encapsulates keywords that are context sensitive or for languages not
684/// properly supported by Clang's lexer.
685struct AdditionalKeywords {
686 AdditionalKeywords(IdentifierTable &IdentTable) {
687 kw_final = &IdentTable.get("final");
688 kw_override = &IdentTable.get("override");
689 kw_in = &IdentTable.get("in");
690 kw_of = &IdentTable.get("of");
691 kw_CF_CLOSED_ENUM = &IdentTable.get("CF_CLOSED_ENUM");
692 kw_CF_ENUM = &IdentTable.get("CF_ENUM");
693 kw_CF_OPTIONS = &IdentTable.get("CF_OPTIONS");
694 kw_NS_CLOSED_ENUM = &IdentTable.get("NS_CLOSED_ENUM");
695 kw_NS_ENUM = &IdentTable.get("NS_ENUM");
696 kw_NS_OPTIONS = &IdentTable.get("NS_OPTIONS");
697
698 kw_as = &IdentTable.get("as");
699 kw_async = &IdentTable.get("async");
700 kw_await = &IdentTable.get("await");
701 kw_declare = &IdentTable.get("declare");
702 kw_finally = &IdentTable.get("finally");
703 kw_from = &IdentTable.get("from");
704 kw_function = &IdentTable.get("function");
705 kw_get = &IdentTable.get("get");
706 kw_import = &IdentTable.get("import");
707 kw_infer = &IdentTable.get("infer");
708 kw_is = &IdentTable.get("is");
709 kw_let = &IdentTable.get("let");
710 kw_module = &IdentTable.get("module");
711 kw_readonly = &IdentTable.get("readonly");
712 kw_set = &IdentTable.get("set");
713 kw_type = &IdentTable.get("type");
714 kw_typeof = &IdentTable.get("typeof");
715 kw_var = &IdentTable.get("var");
716 kw_yield = &IdentTable.get("yield");
717
718 kw_abstract = &IdentTable.get("abstract");
719 kw_assert = &IdentTable.get("assert");
720 kw_extends = &IdentTable.get("extends");
721 kw_implements = &IdentTable.get("implements");
722 kw_instanceof = &IdentTable.get("instanceof");
723 kw_interface = &IdentTable.get("interface");
724 kw_native = &IdentTable.get("native");
725 kw_package = &IdentTable.get("package");
726 kw_synchronized = &IdentTable.get("synchronized");
727 kw_throws = &IdentTable.get("throws");
728 kw___except = &IdentTable.get("__except");
729 kw___has_include = &IdentTable.get("__has_include");
730 kw___has_include_next = &IdentTable.get("__has_include_next");
731
732 kw_mark = &IdentTable.get("mark");
733
734 kw_extend = &IdentTable.get("extend");
735 kw_option = &IdentTable.get("option");
736 kw_optional = &IdentTable.get("optional");
737 kw_repeated = &IdentTable.get("repeated");
738 kw_required = &IdentTable.get("required");
739 kw_returns = &IdentTable.get("returns");
740
741 kw_signals = &IdentTable.get("signals");
742 kw_qsignals = &IdentTable.get("Q_SIGNALS");
743 kw_slots = &IdentTable.get("slots");
744 kw_qslots = &IdentTable.get("Q_SLOTS");
745
746 // C# keywords
747 kw_dollar = &IdentTable.get("dollar");
748 kw_base = &IdentTable.get("base");
749 kw_byte = &IdentTable.get("byte");
750 kw_checked = &IdentTable.get("checked");
751 kw_decimal = &IdentTable.get("decimal");
752 kw_delegate = &IdentTable.get("delegate");
753 kw_event = &IdentTable.get("event");
754 kw_fixed = &IdentTable.get("fixed");
755 kw_foreach = &IdentTable.get("foreach");
756 kw_implicit = &IdentTable.get("implicit");
757 kw_internal = &IdentTable.get("internal");
758 kw_lock = &IdentTable.get("lock");
759 kw_null = &IdentTable.get("null");
760 kw_object = &IdentTable.get("object");
761 kw_out = &IdentTable.get("out");
762 kw_params = &IdentTable.get("params");
763 kw_ref = &IdentTable.get("ref");
764 kw_string = &IdentTable.get("string");
765 kw_stackalloc = &IdentTable.get("stackalloc");
766 kw_sbyte = &IdentTable.get("sbyte");
767 kw_sealed = &IdentTable.get("sealed");
768 kw_uint = &IdentTable.get("uint");
769 kw_ulong = &IdentTable.get("ulong");
770 kw_unchecked = &IdentTable.get("unchecked");
771 kw_unsafe = &IdentTable.get("unsafe");
772 kw_ushort = &IdentTable.get("ushort");
773
774 // Keep this at the end of the constructor to make sure everything here
775 // is
776 // already initialized.
777 JsExtraKeywords = std::unordered_set<IdentifierInfo *>(
778 {kw_as, kw_async, kw_await, kw_declare, kw_finally, kw_from,
779 kw_function, kw_get, kw_import, kw_is, kw_let, kw_module, kw_readonly,
780 kw_set, kw_type, kw_typeof, kw_var, kw_yield,
781 // Keywords from the Java section.
782 kw_abstract, kw_extends, kw_implements, kw_instanceof, kw_interface});
783
784 CSharpExtraKeywords = std::unordered_set<IdentifierInfo *>(
785 {kw_base, kw_byte, kw_checked, kw_decimal, kw_delegate, kw_event,
786 kw_fixed, kw_foreach, kw_implicit, kw_in, kw_interface, kw_internal,
787 kw_is, kw_lock, kw_null, kw_object, kw_out, kw_override, kw_params,
788 kw_readonly, kw_ref, kw_string, kw_stackalloc, kw_sbyte, kw_sealed,
789 kw_uint, kw_ulong, kw_unchecked, kw_unsafe, kw_ushort,
790 // Keywords from the JavaScript section.
791 kw_as, kw_async, kw_await, kw_declare, kw_finally, kw_from,
792 kw_function, kw_get, kw_import, kw_is, kw_let, kw_module, kw_readonly,
793 kw_set, kw_type, kw_typeof, kw_var, kw_yield,
794 // Keywords from the Java section.
795 kw_abstract, kw_extends, kw_implements, kw_instanceof, kw_interface});
796 }
797
798 // Context sensitive keywords.
799 IdentifierInfo *kw_final;
800 IdentifierInfo *kw_override;
801 IdentifierInfo *kw_in;
802 IdentifierInfo *kw_of;
803 IdentifierInfo *kw_CF_CLOSED_ENUM;
804 IdentifierInfo *kw_CF_ENUM;
805 IdentifierInfo *kw_CF_OPTIONS;
806 IdentifierInfo *kw_NS_CLOSED_ENUM;
807 IdentifierInfo *kw_NS_ENUM;
808 IdentifierInfo *kw_NS_OPTIONS;
809 IdentifierInfo *kw___except;
810 IdentifierInfo *kw___has_include;
811 IdentifierInfo *kw___has_include_next;
812
813 // JavaScript keywords.
814 IdentifierInfo *kw_as;
815 IdentifierInfo *kw_async;
816 IdentifierInfo *kw_await;
817 IdentifierInfo *kw_declare;
818 IdentifierInfo *kw_finally;
819 IdentifierInfo *kw_from;
820 IdentifierInfo *kw_function;
821 IdentifierInfo *kw_get;
822 IdentifierInfo *kw_import;
823 IdentifierInfo *kw_infer;
824 IdentifierInfo *kw_is;
825 IdentifierInfo *kw_let;
826 IdentifierInfo *kw_module;
827 IdentifierInfo *kw_readonly;
828 IdentifierInfo *kw_set;
829 IdentifierInfo *kw_type;
830 IdentifierInfo *kw_typeof;
831 IdentifierInfo *kw_var;
832 IdentifierInfo *kw_yield;
833
834 // Java keywords.
835 IdentifierInfo *kw_abstract;
836 IdentifierInfo *kw_assert;
837 IdentifierInfo *kw_extends;
838 IdentifierInfo *kw_implements;
839 IdentifierInfo *kw_instanceof;
840 IdentifierInfo *kw_interface;
841 IdentifierInfo *kw_native;
842 IdentifierInfo *kw_package;
843 IdentifierInfo *kw_synchronized;
844 IdentifierInfo *kw_throws;
845
846 // Pragma keywords.
847 IdentifierInfo *kw_mark;
848
849 // Proto keywords.
850 IdentifierInfo *kw_extend;
851 IdentifierInfo *kw_option;
852 IdentifierInfo *kw_optional;
853 IdentifierInfo *kw_repeated;
854 IdentifierInfo *kw_required;
855 IdentifierInfo *kw_returns;
856
857 // QT keywords.
858 IdentifierInfo *kw_signals;
859 IdentifierInfo *kw_qsignals;
860 IdentifierInfo *kw_slots;
861 IdentifierInfo *kw_qslots;
862
863 // C# keywords
864 IdentifierInfo *kw_dollar;
865 IdentifierInfo *kw_base;
866 IdentifierInfo *kw_byte;
867 IdentifierInfo *kw_checked;
868 IdentifierInfo *kw_decimal;
869 IdentifierInfo *kw_delegate;
870 IdentifierInfo *kw_event;
871 IdentifierInfo *kw_fixed;
872 IdentifierInfo *kw_foreach;
873 IdentifierInfo *kw_implicit;
874 IdentifierInfo *kw_internal;
875
876 IdentifierInfo *kw_lock;
877 IdentifierInfo *kw_null;
878 IdentifierInfo *kw_object;
879 IdentifierInfo *kw_out;
880
881 IdentifierInfo *kw_params;
882
883 IdentifierInfo *kw_ref;
884 IdentifierInfo *kw_string;
885 IdentifierInfo *kw_stackalloc;
886 IdentifierInfo *kw_sbyte;
887 IdentifierInfo *kw_sealed;
888 IdentifierInfo *kw_uint;
889 IdentifierInfo *kw_ulong;
890 IdentifierInfo *kw_unchecked;
891 IdentifierInfo *kw_unsafe;
892 IdentifierInfo *kw_ushort;
893
894 /// Returns \c true if \p Tok is a true JavaScript identifier, returns
895 /// \c false if it is a keyword or a pseudo keyword.
896 bool IsJavaScriptIdentifier(const FormatToken &Tok) const {
897 return Tok.is(tok::identifier) &&
898 JsExtraKeywords.find(Tok.Tok.getIdentifierInfo()) ==
899 JsExtraKeywords.end();
900 }
901
902 /// Returns \c true if \p Tok is a C# keyword, returns
903 /// \c false if it is a anything else.
904 bool isCSharpKeyword(const FormatToken &Tok) const {
905 switch (Tok.Tok.getKind()) {
906 case tok::kw_bool:
907 case tok::kw_break:
908 case tok::kw_case:
909 case tok::kw_catch:
910 case tok::kw_char:
911 case tok::kw_class:
912 case tok::kw_const:
913 case tok::kw_continue:
914 case tok::kw_default:
915 case tok::kw_do:
916 case tok::kw_double:
917 case tok::kw_else:
918 case tok::kw_enum:
919 case tok::kw_explicit:
920 case tok::kw_extern:
921 case tok::kw_false:
922 case tok::kw_float:
923 case tok::kw_for:
924 case tok::kw_goto:
925 case tok::kw_if:
926 case tok::kw_int:
927 case tok::kw_long:
928 case tok::kw_namespace:
929 case tok::kw_new:
930 case tok::kw_operator:
931 case tok::kw_private:
932 case tok::kw_protected:
933 case tok::kw_public:
934 case tok::kw_return:
935 case tok::kw_short:
936 case tok::kw_sizeof:
937 case tok::kw_static:
938 case tok::kw_struct:
939 case tok::kw_switch:
940 case tok::kw_this:
941 case tok::kw_throw:
942 case tok::kw_true:
943 case tok::kw_try:
944 case tok::kw_typeof:
945 case tok::kw_using:
946 case tok::kw_virtual:
947 case tok::kw_void:
948 case tok::kw_volatile:
949 case tok::kw_while:
950 return true;
951 default:
952 return Tok.is(tok::identifier) &&
953 CSharpExtraKeywords.find(Tok.Tok.getIdentifierInfo()) ==
954 CSharpExtraKeywords.end();
955 }
956 }
957
958private:
959 /// The JavaScript keywords beyond the C++ keyword set.
960 std::unordered_set<IdentifierInfo *> JsExtraKeywords;
961
962 /// The C# keywords beyond the C++ keyword set
963 std::unordered_set<IdentifierInfo *> CSharpExtraKeywords;
964};
965
966} // namespace format
967} // namespace clang
968
969#endif

/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h

1//===--- Token.h - Token interface ------------------------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file defines the Token interface.
10//
11//===----------------------------------------------------------------------===//
12
13#ifndef LLVM_CLANG_LEX_TOKEN_H
14#define LLVM_CLANG_LEX_TOKEN_H
15
16#include "clang/Basic/SourceLocation.h"
17#include "clang/Basic/TokenKinds.h"
18#include "llvm/ADT/StringRef.h"
19#include <cassert>
20
21namespace clang {
22
23class IdentifierInfo;
24
25/// Token - This structure provides full information about a lexed token.
26/// It is not intended to be space efficient, it is intended to return as much
27/// information as possible about each returned token. This is expected to be
28/// compressed into a smaller form if memory footprint is important.
29///
30/// The parser can create a special "annotation token" representing a stream of
31/// tokens that were parsed and semantically resolved, e.g.: "foo::MyClass<int>"
32/// can be represented by a single typename annotation token that carries
33/// information about the SourceRange of the tokens and the type object.
34class Token {
35 /// The location of the token. This is actually a SourceLocation.
36 unsigned Loc;
37
38 // Conceptually these next two fields could be in a union. However, this
39 // causes gcc 4.2 to pessimize LexTokenInternal, a very performance critical
40 // routine. Keeping as separate members with casts until a more beautiful fix
41 // presents itself.
42
43 /// UintData - This holds either the length of the token text, when
44 /// a normal token, or the end of the SourceRange when an annotation
45 /// token.
46 unsigned UintData;
47
48 /// PtrData - This is a union of four different pointer types, which depends
49 /// on what type of token this is:
50 /// Identifiers, keywords, etc:
51 /// This is an IdentifierInfo*, which contains the uniqued identifier
52 /// spelling.
53 /// Literals: isLiteral() returns true.
54 /// This is a pointer to the start of the token in a text buffer, which
55 /// may be dirty (have trigraphs / escaped newlines).
56 /// Annotations (resolved type names, C++ scopes, etc): isAnnotation().
57 /// This is a pointer to sema-specific data for the annotation token.
58 /// Eof:
59 // This is a pointer to a Decl.
60 /// Other:
61 /// This is null.
62 void *PtrData;
63
64 /// Kind - The actual flavor of token this is.
65 tok::TokenKind Kind;
66
67 /// Flags - Bits we track about this token, members of the TokenFlags enum.
68 unsigned short Flags;
69
70public:
71 // Various flags set per token:
72 enum TokenFlags {
73 StartOfLine = 0x01, // At start of line or only after whitespace
74 // (considering the line after macro expansion).
75 LeadingSpace = 0x02, // Whitespace exists before this token (considering
76 // whitespace after macro expansion).
77 DisableExpand = 0x04, // This identifier may never be macro expanded.
78 NeedsCleaning = 0x08, // Contained an escaped newline or trigraph.
79 LeadingEmptyMacro = 0x10, // Empty macro exists before this token.
80 HasUDSuffix = 0x20, // This string or character literal has a ud-suffix.
81 HasUCN = 0x40, // This identifier contains a UCN.
82 IgnoredComma = 0x80, // This comma is not a macro argument separator (MS).
83 StringifiedInMacro = 0x100, // This string or character literal is formed by
84 // macro stringizing or charizing operator.
85 CommaAfterElided = 0x200, // The comma following this token was elided (MS).
86 IsEditorPlaceholder = 0x400, // This identifier is a placeholder.
87 IsReinjected = 0x800, // A phase 4 token that was produced before and
88 // re-added, e.g. via EnterTokenStream. Annotation
89 // tokens are *not* reinjected.
90 };
91
92 tok::TokenKind getKind() const { return Kind; }
93 void setKind(tok::TokenKind K) { Kind = K; }
94
95 /// is/isNot - Predicates to check if this token is a specific kind, as in
96 /// "if (Tok.is(tok::l_brace)) {...}".
97 bool is(tok::TokenKind K) const { return Kind == K; }
4
Assuming 'K' is equal to field 'Kind'
5
Returning the value 1, which participates in a condition later
11
Assuming 'K' is equal to field 'Kind'
12
Returning the value 1, which participates in a condition later
98 bool isNot(tok::TokenKind K) const { return Kind != K; }
99 bool isOneOf(tok::TokenKind K1, tok::TokenKind K2) const {
100 return is(K1) || is(K2);
101 }
102 template <typename... Ts>
103 bool isOneOf(tok::TokenKind K1, tok::TokenKind K2, Ts... Ks) const {
104 return is(K1) || isOneOf(K2, Ks...);
105 }
106
107 /// Return true if this is a raw identifier (when lexing
108 /// in raw mode) or a non-keyword identifier (when lexing in non-raw mode).
109 bool isAnyIdentifier() const {
110 return tok::isAnyIdentifier(getKind());
111 }
112
113 /// Return true if this is a "literal", like a numeric
114 /// constant, string, etc.
115 bool isLiteral() const {
116 return tok::isLiteral(getKind());
117 }
118
119 /// Return true if this is any of tok::annot_* kind tokens.
120 bool isAnnotation() const {
121 return tok::isAnnotation(getKind());
122 }
123
124 /// Return a source location identifier for the specified
125 /// offset in the current file.
126 SourceLocation getLocation() const {
127 return SourceLocation::getFromRawEncoding(Loc);
128 }
129 unsigned getLength() const {
130 assert(!isAnnotation() && "Annotation tokens have no length field")((!isAnnotation() && "Annotation tokens have no length field"
) ? static_cast<void> (0) : __assert_fail ("!isAnnotation() && \"Annotation tokens have no length field\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 130, __PRETTY_FUNCTION__))
;
131 return UintData;
132 }
133
134 void setLocation(SourceLocation L) { Loc = L.getRawEncoding(); }
135 void setLength(unsigned Len) {
136 assert(!isAnnotation() && "Annotation tokens have no length field")((!isAnnotation() && "Annotation tokens have no length field"
) ? static_cast<void> (0) : __assert_fail ("!isAnnotation() && \"Annotation tokens have no length field\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 136, __PRETTY_FUNCTION__))
;
137 UintData = Len;
138 }
139
140 SourceLocation getAnnotationEndLoc() const {
141 assert(isAnnotation() && "Used AnnotEndLocID on non-annotation token")((isAnnotation() && "Used AnnotEndLocID on non-annotation token"
) ? static_cast<void> (0) : __assert_fail ("isAnnotation() && \"Used AnnotEndLocID on non-annotation token\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 141, __PRETTY_FUNCTION__))
;
142 return SourceLocation::getFromRawEncoding(UintData ? UintData : Loc);
143 }
144 void setAnnotationEndLoc(SourceLocation L) {
145 assert(isAnnotation() && "Used AnnotEndLocID on non-annotation token")((isAnnotation() && "Used AnnotEndLocID on non-annotation token"
) ? static_cast<void> (0) : __assert_fail ("isAnnotation() && \"Used AnnotEndLocID on non-annotation token\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 145, __PRETTY_FUNCTION__))
;
146 UintData = L.getRawEncoding();
147 }
148
149 SourceLocation getLastLoc() const {
150 return isAnnotation() ? getAnnotationEndLoc() : getLocation();
151 }
152
153 SourceLocation getEndLoc() const {
154 return isAnnotation() ? getAnnotationEndLoc()
155 : getLocation().getLocWithOffset(getLength());
156 }
157
158 /// SourceRange of the group of tokens that this annotation token
159 /// represents.
160 SourceRange getAnnotationRange() const {
161 return SourceRange(getLocation(), getAnnotationEndLoc());
162 }
163 void setAnnotationRange(SourceRange R) {
164 setLocation(R.getBegin());
165 setAnnotationEndLoc(R.getEnd());
166 }
167
168 const char *getName() const { return tok::getTokenName(Kind); }
169
170 /// Reset all flags to cleared.
171 void startToken() {
172 Kind = tok::unknown;
173 Flags = 0;
174 PtrData = nullptr;
175 UintData = 0;
176 Loc = SourceLocation().getRawEncoding();
177 }
178
179 IdentifierInfo *getIdentifierInfo() const {
180 assert(isNot(tok::raw_identifier) &&((isNot(tok::raw_identifier) && "getIdentifierInfo() on a tok::raw_identifier token!"
) ? static_cast<void> (0) : __assert_fail ("isNot(tok::raw_identifier) && \"getIdentifierInfo() on a tok::raw_identifier token!\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 181, __PRETTY_FUNCTION__))
181 "getIdentifierInfo() on a tok::raw_identifier token!")((isNot(tok::raw_identifier) && "getIdentifierInfo() on a tok::raw_identifier token!"
) ? static_cast<void> (0) : __assert_fail ("isNot(tok::raw_identifier) && \"getIdentifierInfo() on a tok::raw_identifier token!\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 181, __PRETTY_FUNCTION__))
;
182 assert(!isAnnotation() &&((!isAnnotation() && "getIdentifierInfo() on an annotation token!"
) ? static_cast<void> (0) : __assert_fail ("!isAnnotation() && \"getIdentifierInfo() on an annotation token!\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 183, __PRETTY_FUNCTION__))
183 "getIdentifierInfo() on an annotation token!")((!isAnnotation() && "getIdentifierInfo() on an annotation token!"
) ? static_cast<void> (0) : __assert_fail ("!isAnnotation() && \"getIdentifierInfo() on an annotation token!\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 183, __PRETTY_FUNCTION__))
;
184 if (isLiteral()) return nullptr;
185 if (is(tok::eof)) return nullptr;
186 return (IdentifierInfo*) PtrData;
187 }
188 void setIdentifierInfo(IdentifierInfo *II) {
189 PtrData = (void*) II;
190 }
191
192 const void *getEofData() const {
193 assert(is(tok::eof))((is(tok::eof)) ? static_cast<void> (0) : __assert_fail
("is(tok::eof)", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 193, __PRETTY_FUNCTION__))
;
194 return reinterpret_cast<const void *>(PtrData);
195 }
196 void setEofData(const void *D) {
197 assert(is(tok::eof))((is(tok::eof)) ? static_cast<void> (0) : __assert_fail
("is(tok::eof)", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 197, __PRETTY_FUNCTION__))
;
198 assert(!PtrData)((!PtrData) ? static_cast<void> (0) : __assert_fail ("!PtrData"
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 198, __PRETTY_FUNCTION__))
;
199 PtrData = const_cast<void *>(D);
200 }
201
202 /// getRawIdentifier - For a raw identifier token (i.e., an identifier
203 /// lexed in raw mode), returns a reference to the text substring in the
204 /// buffer if known.
205 StringRef getRawIdentifier() const {
206 assert(is(tok::raw_identifier))((is(tok::raw_identifier)) ? static_cast<void> (0) : __assert_fail
("is(tok::raw_identifier)", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 206, __PRETTY_FUNCTION__))
;
207 return StringRef(reinterpret_cast<const char *>(PtrData), getLength());
208 }
209 void setRawIdentifierData(const char *Ptr) {
210 assert(is(tok::raw_identifier))((is(tok::raw_identifier)) ? static_cast<void> (0) : __assert_fail
("is(tok::raw_identifier)", "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 210, __PRETTY_FUNCTION__))
;
211 PtrData = const_cast<char*>(Ptr);
212 }
213
214 /// getLiteralData - For a literal token (numeric constant, string, etc), this
215 /// returns a pointer to the start of it in the text buffer if known, null
216 /// otherwise.
217 const char *getLiteralData() const {
218 assert(isLiteral() && "Cannot get literal data of non-literal")((isLiteral() && "Cannot get literal data of non-literal"
) ? static_cast<void> (0) : __assert_fail ("isLiteral() && \"Cannot get literal data of non-literal\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 218, __PRETTY_FUNCTION__))
;
219 return reinterpret_cast<const char*>(PtrData);
220 }
221 void setLiteralData(const char *Ptr) {
222 assert(isLiteral() && "Cannot set literal data of non-literal")((isLiteral() && "Cannot set literal data of non-literal"
) ? static_cast<void> (0) : __assert_fail ("isLiteral() && \"Cannot set literal data of non-literal\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 222, __PRETTY_FUNCTION__))
;
223 PtrData = const_cast<char*>(Ptr);
224 }
225
226 void *getAnnotationValue() const {
227 assert(isAnnotation() && "Used AnnotVal on non-annotation token")((isAnnotation() && "Used AnnotVal on non-annotation token"
) ? static_cast<void> (0) : __assert_fail ("isAnnotation() && \"Used AnnotVal on non-annotation token\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 227, __PRETTY_FUNCTION__))
;
228 return PtrData;
229 }
230 void setAnnotationValue(void *val) {
231 assert(isAnnotation() && "Used AnnotVal on non-annotation token")((isAnnotation() && "Used AnnotVal on non-annotation token"
) ? static_cast<void> (0) : __assert_fail ("isAnnotation() && \"Used AnnotVal on non-annotation token\""
, "/build/llvm-toolchain-snapshot-10~+201911111502510600c19528f1809/clang/include/clang/Lex/Token.h"
, 231, __PRETTY_FUNCTION__))
;
232 PtrData = val;
233 }
234
235 /// Set the specified flag.
236 void setFlag(TokenFlags Flag) {
237 Flags |= Flag;
238 }
239
240 /// Get the specified flag.
241 bool getFlag(TokenFlags Flag) const {
242 return (Flags & Flag) != 0;
243 }
244
245 /// Unset the specified flag.
246 void clearFlag(TokenFlags Flag) {
247 Flags &= ~Flag;
248 }
249
250 /// Return the internal represtation of the flags.
251 ///
252 /// This is only intended for low-level operations such as writing tokens to
253 /// disk.
254 unsigned getFlags() const {
255 return Flags;
256 }
257
258 /// Set a flag to either true or false.
259 void setFlagValue(TokenFlags Flag, bool Val) {
260 if (Val)
261 setFlag(Flag);
262 else
263 clearFlag(Flag);
264 }
265
266 /// isAtStartOfLine - Return true if this token is at the start of a line.
267 ///
268 bool isAtStartOfLine() const { return getFlag(StartOfLine); }
269
270 /// Return true if this token has whitespace before it.
271 ///
272 bool hasLeadingSpace() const { return getFlag(LeadingSpace); }
273
274 /// Return true if this identifier token should never
275 /// be expanded in the future, due to C99 6.10.3.4p2.
276 bool isExpandDisabled() const { return getFlag(DisableExpand); }
277
278 /// Return true if we have an ObjC keyword identifier.
279 bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const;
280
281 /// Return the ObjC keyword kind.
282 tok::ObjCKeywordKind getObjCKeywordID() const;
283
284 /// Return true if this token has trigraphs or escaped newlines in it.
285 bool needsCleaning() const { return getFlag(NeedsCleaning); }
286
287 /// Return true if this token has an empty macro before it.
288 ///
289 bool hasLeadingEmptyMacro() const { return getFlag(LeadingEmptyMacro); }
290
291 /// Return true if this token is a string or character literal which
292 /// has a ud-suffix.
293 bool hasUDSuffix() const { return getFlag(HasUDSuffix); }
294
295 /// Returns true if this token contains a universal character name.
296 bool hasUCN() const { return getFlag(HasUCN); }
297
298 /// Returns true if this token is formed by macro by stringizing or charizing
299 /// operator.
300 bool stringifiedInMacro() const { return getFlag(StringifiedInMacro); }
301
302 /// Returns true if the comma after this token was elided.
303 bool commaAfterElided() const { return getFlag(CommaAfterElided); }
304
305 /// Returns true if this token is an editor placeholder.
306 ///
307 /// Editor placeholders are produced by the code-completion engine and are
308 /// represented as characters between '<#' and '#>' in the source code. The
309 /// lexer uses identifier tokens to represent placeholders.
310 bool isEditorPlaceholder() const { return getFlag(IsEditorPlaceholder); }
311};
312
313/// Information about the conditional stack (\#if directives)
314/// currently active.
315struct PPConditionalInfo {
316 /// Location where the conditional started.
317 SourceLocation IfLoc;
318
319 /// True if this was contained in a skipping directive, e.g.,
320 /// in a "\#if 0" block.
321 bool WasSkipping;
322
323 /// True if we have emitted tokens already, and now we're in
324 /// an \#else block or something. Only useful in Skipping blocks.
325 bool FoundNonSkip;
326
327 /// True if we've seen a \#else in this block. If so,
328 /// \#elif/\#else directives are not allowed.
329 bool FoundElse;
330};
331
332} // end namespace clang
333
334#endif // LLVM_CLANG_LEX_TOKEN_H