/src/llvm-project/clang/lib/Format/UnwrappedLineParser.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | //===--- UnwrappedLineParser.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 contains the implementation of the UnwrappedLineParser, |
11 | | /// which turns a stream of tokens into UnwrappedLines. |
12 | | /// |
13 | | //===----------------------------------------------------------------------===// |
14 | | |
15 | | #include "UnwrappedLineParser.h" |
16 | | #include "FormatToken.h" |
17 | | #include "FormatTokenLexer.h" |
18 | | #include "FormatTokenSource.h" |
19 | | #include "Macros.h" |
20 | | #include "TokenAnnotator.h" |
21 | | #include "clang/Basic/TokenKinds.h" |
22 | | #include "llvm/ADT/STLExtras.h" |
23 | | #include "llvm/ADT/StringRef.h" |
24 | | #include "llvm/Support/Debug.h" |
25 | | #include "llvm/Support/raw_os_ostream.h" |
26 | | #include "llvm/Support/raw_ostream.h" |
27 | | |
28 | | #include <algorithm> |
29 | | #include <utility> |
30 | | |
31 | | #define DEBUG_TYPE "format-parser" |
32 | | |
33 | | namespace clang { |
34 | | namespace format { |
35 | | |
36 | | namespace { |
37 | | |
38 | | void printLine(llvm::raw_ostream &OS, const UnwrappedLine &Line, |
39 | 0 | StringRef Prefix = "", bool PrintText = false) { |
40 | 0 | OS << Prefix << "Line(" << Line.Level << ", FSC=" << Line.FirstStartColumn |
41 | 0 | << ")" << (Line.InPPDirective ? " MACRO" : "") << ": "; |
42 | 0 | bool NewLine = false; |
43 | 0 | for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), |
44 | 0 | E = Line.Tokens.end(); |
45 | 0 | I != E; ++I) { |
46 | 0 | if (NewLine) { |
47 | 0 | OS << Prefix; |
48 | 0 | NewLine = false; |
49 | 0 | } |
50 | 0 | OS << I->Tok->Tok.getName() << "[" << "T=" << (unsigned)I->Tok->getType() |
51 | 0 | << ", OC=" << I->Tok->OriginalColumn << ", \"" << I->Tok->TokenText |
52 | 0 | << "\"] "; |
53 | 0 | for (SmallVectorImpl<UnwrappedLine>::const_iterator |
54 | 0 | CI = I->Children.begin(), |
55 | 0 | CE = I->Children.end(); |
56 | 0 | CI != CE; ++CI) { |
57 | 0 | OS << "\n"; |
58 | 0 | printLine(OS, *CI, (Prefix + " ").str()); |
59 | 0 | NewLine = true; |
60 | 0 | } |
61 | 0 | } |
62 | 0 | if (!NewLine) |
63 | 0 | OS << "\n"; |
64 | 0 | } |
65 | | |
66 | 0 | LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line) { |
67 | 0 | printLine(llvm::dbgs(), Line); |
68 | 0 | } |
69 | | |
70 | | class ScopedDeclarationState { |
71 | | public: |
72 | | ScopedDeclarationState(UnwrappedLine &Line, llvm::BitVector &Stack, |
73 | | bool MustBeDeclaration) |
74 | 271k | : Line(Line), Stack(Stack) { |
75 | 271k | Line.MustBeDeclaration = MustBeDeclaration; |
76 | 271k | Stack.push_back(MustBeDeclaration); |
77 | 271k | } |
78 | 271k | ~ScopedDeclarationState() { |
79 | 271k | Stack.pop_back(); |
80 | 271k | if (!Stack.empty()) |
81 | 270k | Line.MustBeDeclaration = Stack.back(); |
82 | 1.05k | else |
83 | 1.05k | Line.MustBeDeclaration = true; |
84 | 271k | } |
85 | | |
86 | | private: |
87 | | UnwrappedLine &Line; |
88 | | llvm::BitVector &Stack; |
89 | | }; |
90 | | |
91 | | } // end anonymous namespace |
92 | | |
93 | | class ScopedLineState { |
94 | | public: |
95 | | ScopedLineState(UnwrappedLineParser &Parser, |
96 | | bool SwitchToPreprocessorLines = false) |
97 | 275k | : Parser(Parser), OriginalLines(Parser.CurrentLines) { |
98 | 275k | if (SwitchToPreprocessorLines) |
99 | 253k | Parser.CurrentLines = &Parser.PreprocessorDirectives; |
100 | 21.7k | else if (!Parser.Line->Tokens.empty()) |
101 | 21.7k | Parser.CurrentLines = &Parser.Line->Tokens.back().Children; |
102 | 275k | PreBlockLine = std::move(Parser.Line); |
103 | 275k | Parser.Line = std::make_unique<UnwrappedLine>(); |
104 | 275k | Parser.Line->Level = PreBlockLine->Level; |
105 | 275k | Parser.Line->PPLevel = PreBlockLine->PPLevel; |
106 | 275k | Parser.Line->InPPDirective = PreBlockLine->InPPDirective; |
107 | 275k | Parser.Line->InMacroBody = PreBlockLine->InMacroBody; |
108 | 275k | } |
109 | | |
110 | 275k | ~ScopedLineState() { |
111 | 275k | if (!Parser.Line->Tokens.empty()) |
112 | 14.6k | Parser.addUnwrappedLine(); |
113 | 275k | assert(Parser.Line->Tokens.empty()); |
114 | 0 | Parser.Line = std::move(PreBlockLine); |
115 | 275k | if (Parser.CurrentLines == &Parser.PreprocessorDirectives) |
116 | 253k | Parser.MustBreakBeforeNextToken = true; |
117 | 275k | Parser.CurrentLines = OriginalLines; |
118 | 275k | } |
119 | | |
120 | | private: |
121 | | UnwrappedLineParser &Parser; |
122 | | |
123 | | std::unique_ptr<UnwrappedLine> PreBlockLine; |
124 | | SmallVectorImpl<UnwrappedLine> *OriginalLines; |
125 | | }; |
126 | | |
127 | | class CompoundStatementIndenter { |
128 | | public: |
129 | | CompoundStatementIndenter(UnwrappedLineParser *Parser, |
130 | | const FormatStyle &Style, unsigned &LineLevel) |
131 | | : CompoundStatementIndenter(Parser, LineLevel, |
132 | | Style.BraceWrapping.AfterControlStatement, |
133 | 3.31k | Style.BraceWrapping.IndentBraces) {} |
134 | | CompoundStatementIndenter(UnwrappedLineParser *Parser, unsigned &LineLevel, |
135 | | bool WrapBrace, bool IndentBrace) |
136 | 29.0k | : LineLevel(LineLevel), OldLineLevel(LineLevel) { |
137 | 29.0k | if (WrapBrace) |
138 | 0 | Parser->addUnwrappedLine(); |
139 | 29.0k | if (IndentBrace) |
140 | 0 | ++LineLevel; |
141 | 29.0k | } |
142 | 29.0k | ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } |
143 | | |
144 | | private: |
145 | | unsigned &LineLevel; |
146 | | unsigned OldLineLevel; |
147 | | }; |
148 | | |
149 | | UnwrappedLineParser::UnwrappedLineParser( |
150 | | SourceManager &SourceMgr, const FormatStyle &Style, |
151 | | const AdditionalKeywords &Keywords, unsigned FirstStartColumn, |
152 | | ArrayRef<FormatToken *> Tokens, UnwrappedLineConsumer &Callback, |
153 | | llvm::SpecificBumpPtrAllocator<FormatToken> &Allocator, |
154 | | IdentifierTable &IdentTable) |
155 | | : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), |
156 | | CurrentLines(&Lines), Style(Style), Keywords(Keywords), |
157 | | CommentPragmasRegex(Style.CommentPragmas), Tokens(nullptr), |
158 | | Callback(Callback), AllTokens(Tokens), PPBranchLevel(-1), |
159 | | IncludeGuard(Style.IndentPPDirectives == FormatStyle::PPDIS_None |
160 | | ? IG_Rejected |
161 | | : IG_Inited), |
162 | | IncludeGuardToken(nullptr), FirstStartColumn(FirstStartColumn), |
163 | 1.01k | Macros(Style.Macros, SourceMgr, Style, Allocator, IdentTable) {} |
164 | | |
165 | 1.05k | void UnwrappedLineParser::reset() { |
166 | 1.05k | PPBranchLevel = -1; |
167 | 1.05k | IncludeGuard = Style.IndentPPDirectives == FormatStyle::PPDIS_None |
168 | 1.05k | ? IG_Rejected |
169 | 1.05k | : IG_Inited; |
170 | 1.05k | IncludeGuardToken = nullptr; |
171 | 1.05k | Line.reset(new UnwrappedLine); |
172 | 1.05k | CommentsBeforeNextToken.clear(); |
173 | 1.05k | FormatTok = nullptr; |
174 | 1.05k | MustBreakBeforeNextToken = false; |
175 | 1.05k | IsDecltypeAutoFunction = false; |
176 | 1.05k | PreprocessorDirectives.clear(); |
177 | 1.05k | CurrentLines = &Lines; |
178 | 1.05k | DeclarationScopeStack.clear(); |
179 | 1.05k | NestedTooDeep.clear(); |
180 | 1.05k | NestedLambdas.clear(); |
181 | 1.05k | PPStack.clear(); |
182 | 1.05k | Line->FirstStartColumn = FirstStartColumn; |
183 | | |
184 | 1.05k | if (!Unexpanded.empty()) |
185 | 0 | for (FormatToken *Token : AllTokens) |
186 | 0 | Token->MacroCtx.reset(); |
187 | 1.05k | CurrentExpandedLines.clear(); |
188 | 1.05k | ExpandedLines.clear(); |
189 | 1.05k | Unexpanded.clear(); |
190 | 1.05k | InExpansion = false; |
191 | 1.05k | Reconstruct.reset(); |
192 | 1.05k | } |
193 | | |
194 | 1.01k | void UnwrappedLineParser::parse() { |
195 | 1.01k | IndexedTokenSource TokenSource(AllTokens); |
196 | 1.01k | Line->FirstStartColumn = FirstStartColumn; |
197 | 1.05k | do { |
198 | 1.05k | LLVM_DEBUG(llvm::dbgs() << "----\n"); |
199 | 1.05k | reset(); |
200 | 1.05k | Tokens = &TokenSource; |
201 | 1.05k | TokenSource.reset(); |
202 | | |
203 | 1.05k | readToken(); |
204 | 1.05k | parseFile(); |
205 | | |
206 | | // If we found an include guard then all preprocessor directives (other than |
207 | | // the guard) are over-indented by one. |
208 | 1.05k | if (IncludeGuard == IG_Found) { |
209 | 0 | for (auto &Line : Lines) |
210 | 0 | if (Line.InPPDirective && Line.Level > 0) |
211 | 0 | --Line.Level; |
212 | 0 | } |
213 | | |
214 | | // Create line with eof token. |
215 | 1.05k | assert(eof()); |
216 | 0 | pushToken(FormatTok); |
217 | 1.05k | addUnwrappedLine(); |
218 | | |
219 | | // In a first run, format everything with the lines containing macro calls |
220 | | // replaced by the expansion. |
221 | 1.05k | if (!ExpandedLines.empty()) { |
222 | 0 | LLVM_DEBUG(llvm::dbgs() << "Expanded lines:\n"); |
223 | 0 | for (const auto &Line : Lines) { |
224 | 0 | if (!Line.Tokens.empty()) { |
225 | 0 | auto it = ExpandedLines.find(Line.Tokens.begin()->Tok); |
226 | 0 | if (it != ExpandedLines.end()) { |
227 | 0 | for (const auto &Expanded : it->second) { |
228 | 0 | LLVM_DEBUG(printDebugInfo(Expanded)); |
229 | 0 | Callback.consumeUnwrappedLine(Expanded); |
230 | 0 | } |
231 | 0 | continue; |
232 | 0 | } |
233 | 0 | } |
234 | 0 | LLVM_DEBUG(printDebugInfo(Line)); |
235 | 0 | Callback.consumeUnwrappedLine(Line); |
236 | 0 | } |
237 | 0 | Callback.finishRun(); |
238 | 0 | } |
239 | | |
240 | 1.05k | LLVM_DEBUG(llvm::dbgs() << "Unwrapped lines:\n"); |
241 | 1.32M | for (const UnwrappedLine &Line : Lines) { |
242 | 1.32M | LLVM_DEBUG(printDebugInfo(Line)); |
243 | 1.32M | Callback.consumeUnwrappedLine(Line); |
244 | 1.32M | } |
245 | 1.05k | Callback.finishRun(); |
246 | 1.05k | Lines.clear(); |
247 | 262k | while (!PPLevelBranchIndex.empty() && |
248 | 262k | PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { |
249 | 261k | PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); |
250 | 261k | PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); |
251 | 261k | } |
252 | 1.05k | if (!PPLevelBranchIndex.empty()) { |
253 | 42 | ++PPLevelBranchIndex.back(); |
254 | 42 | assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); |
255 | 0 | assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); |
256 | 42 | } |
257 | 1.05k | } while (!PPLevelBranchIndex.empty()); |
258 | 1.01k | } |
259 | | |
260 | 3.08k | void UnwrappedLineParser::parseFile() { |
261 | | // The top-level context in a file always has declarations, except for pre- |
262 | | // processor directives and JavaScript files. |
263 | 3.08k | bool MustBeDeclaration = !Line->InPPDirective && !Style.isJavaScript(); |
264 | 3.08k | ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, |
265 | 3.08k | MustBeDeclaration); |
266 | 3.08k | if (Style.Language == FormatStyle::LK_TextProto) |
267 | 0 | parseBracedList(); |
268 | 3.08k | else |
269 | 3.08k | parseLevel(); |
270 | | // Make sure to format the remaining tokens. |
271 | | // |
272 | | // LK_TextProto is special since its top-level is parsed as the body of a |
273 | | // braced list, which does not necessarily have natural line separators such |
274 | | // as a semicolon. Comments after the last entry that have been determined to |
275 | | // not belong to that line, as in: |
276 | | // key: value |
277 | | // // endfile comment |
278 | | // do not have a chance to be put on a line of their own until this point. |
279 | | // Here we add this newline before end-of-file comments. |
280 | 3.08k | if (Style.Language == FormatStyle::LK_TextProto && |
281 | 3.08k | !CommentsBeforeNextToken.empty()) { |
282 | 0 | addUnwrappedLine(); |
283 | 0 | } |
284 | 3.08k | flushComments(true); |
285 | 3.08k | addUnwrappedLine(); |
286 | 3.08k | } |
287 | | |
288 | 0 | void UnwrappedLineParser::parseCSharpGenericTypeConstraint() { |
289 | 0 | do { |
290 | 0 | switch (FormatTok->Tok.getKind()) { |
291 | 0 | case tok::l_brace: |
292 | 0 | return; |
293 | 0 | default: |
294 | 0 | if (FormatTok->is(Keywords.kw_where)) { |
295 | 0 | addUnwrappedLine(); |
296 | 0 | nextToken(); |
297 | 0 | parseCSharpGenericTypeConstraint(); |
298 | 0 | break; |
299 | 0 | } |
300 | 0 | nextToken(); |
301 | 0 | break; |
302 | 0 | } |
303 | 0 | } while (!eof()); |
304 | 0 | } |
305 | | |
306 | 0 | void UnwrappedLineParser::parseCSharpAttribute() { |
307 | 0 | int UnpairedSquareBrackets = 1; |
308 | 0 | do { |
309 | 0 | switch (FormatTok->Tok.getKind()) { |
310 | 0 | case tok::r_square: |
311 | 0 | nextToken(); |
312 | 0 | --UnpairedSquareBrackets; |
313 | 0 | if (UnpairedSquareBrackets == 0) { |
314 | 0 | addUnwrappedLine(); |
315 | 0 | return; |
316 | 0 | } |
317 | 0 | break; |
318 | 0 | case tok::l_square: |
319 | 0 | ++UnpairedSquareBrackets; |
320 | 0 | nextToken(); |
321 | 0 | break; |
322 | 0 | default: |
323 | 0 | nextToken(); |
324 | 0 | break; |
325 | 0 | } |
326 | 0 | } while (!eof()); |
327 | 0 | } |
328 | | |
329 | 0 | bool UnwrappedLineParser::precededByCommentOrPPDirective() const { |
330 | 0 | if (!Lines.empty() && Lines.back().InPPDirective) |
331 | 0 | return true; |
332 | | |
333 | 0 | const FormatToken *Previous = Tokens->getPreviousToken(); |
334 | 0 | return Previous && Previous->is(tok::comment) && |
335 | 0 | (Previous->IsMultiline || Previous->NewlinesBefore > 0); |
336 | 0 | } |
337 | | |
338 | | /// \brief Parses a level, that is ???. |
339 | | /// \param OpeningBrace Opening brace (\p nullptr if absent) of that level. |
340 | | /// \param IfKind The \p if statement kind in the level. |
341 | | /// \param IfLeftBrace The left brace of the \p if block in the level. |
342 | | /// \returns true if a simple block of if/else/for/while, or false otherwise. |
343 | | /// (A simple block has a single statement.) |
344 | | bool UnwrappedLineParser::parseLevel(const FormatToken *OpeningBrace, |
345 | | IfStmtKind *IfKind, |
346 | 271k | FormatToken **IfLeftBrace) { |
347 | 271k | const bool InRequiresExpression = |
348 | 271k | OpeningBrace && OpeningBrace->is(TT_RequiresExpressionLBrace); |
349 | 271k | const bool IsPrecededByCommentOrPPDirective = |
350 | 271k | !Style.RemoveBracesLLVM || precededByCommentOrPPDirective(); |
351 | 271k | FormatToken *IfLBrace = nullptr; |
352 | 271k | bool HasDoWhile = false; |
353 | 271k | bool HasLabel = false; |
354 | 271k | unsigned StatementCount = 0; |
355 | 271k | bool SwitchLabelEncountered = false; |
356 | | |
357 | 1.31M | do { |
358 | 1.31M | if (FormatTok->isAttribute()) { |
359 | 0 | nextToken(); |
360 | 0 | continue; |
361 | 0 | } |
362 | 1.31M | tok::TokenKind kind = FormatTok->Tok.getKind(); |
363 | 1.31M | if (FormatTok->getType() == TT_MacroBlockBegin) |
364 | 0 | kind = tok::l_brace; |
365 | 1.31M | else if (FormatTok->getType() == TT_MacroBlockEnd) |
366 | 0 | kind = tok::r_brace; |
367 | | |
368 | 1.31M | auto ParseDefault = [this, OpeningBrace, IfKind, &IfLBrace, &HasDoWhile, |
369 | 1.31M | &HasLabel, &StatementCount] { |
370 | 722k | parseStructuralElement(OpeningBrace, IfKind, &IfLBrace, |
371 | 722k | HasDoWhile ? nullptr : &HasDoWhile, |
372 | 722k | HasLabel ? nullptr : &HasLabel); |
373 | 722k | ++StatementCount; |
374 | 722k | assert(StatementCount > 0 && "StatementCount overflow!"); |
375 | 722k | }; |
376 | | |
377 | 1.31M | switch (kind) { |
378 | 0 | case tok::comment: |
379 | 0 | nextToken(); |
380 | 0 | addUnwrappedLine(); |
381 | 0 | break; |
382 | 177k | case tok::l_brace: |
383 | 177k | if (InRequiresExpression) { |
384 | 0 | FormatTok->setFinalizedType(TT_RequiresExpressionLBrace); |
385 | 177k | } else if (FormatTok->Previous && |
386 | 177k | FormatTok->Previous->ClosesRequiresClause) { |
387 | | // We need the 'default' case here to correctly parse a function |
388 | | // l_brace. |
389 | 0 | ParseDefault(); |
390 | 0 | continue; |
391 | 0 | } |
392 | 177k | if (!InRequiresExpression && FormatTok->isNot(TT_MacroBlockBegin) && |
393 | 177k | tryToParseBracedList()) { |
394 | 1.42k | continue; |
395 | 1.42k | } |
396 | 176k | parseBlock(); |
397 | 176k | ++StatementCount; |
398 | 176k | assert(StatementCount > 0 && "StatementCount overflow!"); |
399 | 0 | addUnwrappedLine(); |
400 | 176k | break; |
401 | 417k | case tok::r_brace: |
402 | 417k | if (OpeningBrace) { |
403 | 216k | if (!Style.RemoveBracesLLVM || Line->InPPDirective || |
404 | 216k | !OpeningBrace->isOneOf(TT_ControlStatementLBrace, TT_ElseLBrace)) { |
405 | 216k | return false; |
406 | 216k | } |
407 | 0 | if (FormatTok->isNot(tok::r_brace) || StatementCount != 1 || HasLabel || |
408 | 0 | HasDoWhile || IsPrecededByCommentOrPPDirective || |
409 | 0 | precededByCommentOrPPDirective()) { |
410 | 0 | return false; |
411 | 0 | } |
412 | 0 | const FormatToken *Next = Tokens->peekNextToken(); |
413 | 0 | if (Next->is(tok::comment) && Next->NewlinesBefore == 0) |
414 | 0 | return false; |
415 | 0 | if (IfLeftBrace) |
416 | 0 | *IfLeftBrace = IfLBrace; |
417 | 0 | return true; |
418 | 0 | } |
419 | 201k | nextToken(); |
420 | 201k | addUnwrappedLine(); |
421 | 201k | break; |
422 | 0 | case tok::kw_default: { |
423 | 0 | unsigned StoredPosition = Tokens->getPosition(); |
424 | 0 | FormatToken *Next; |
425 | 0 | do { |
426 | 0 | Next = Tokens->getNextToken(); |
427 | 0 | assert(Next); |
428 | 0 | } while (Next->is(tok::comment)); |
429 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
430 | 0 | if (Next->isNot(tok::colon)) { |
431 | | // default not followed by ':' is not a case label; treat it like |
432 | | // an identifier. |
433 | 0 | parseStructuralElement(); |
434 | 0 | break; |
435 | 0 | } |
436 | | // Else, if it is 'default:', fall through to the case handling. |
437 | 0 | [[fallthrough]]; |
438 | 0 | } |
439 | 0 | case tok::kw_case: |
440 | 0 | if (Style.Language == FormatStyle::LK_Proto || Style.isVerilog() || |
441 | 0 | (Style.isJavaScript() && Line->MustBeDeclaration)) { |
442 | | // Proto: there are no switch/case statements |
443 | | // Verilog: Case labels don't have this word. We handle case |
444 | | // labels including default in TokenAnnotator. |
445 | | // JavaScript: A 'case: string' style field declaration. |
446 | 0 | ParseDefault(); |
447 | 0 | break; |
448 | 0 | } |
449 | 0 | if (!SwitchLabelEncountered && |
450 | 0 | (Style.IndentCaseLabels || |
451 | 0 | (Line->InPPDirective && Line->Level == 1))) { |
452 | 0 | ++Line->Level; |
453 | 0 | } |
454 | 0 | SwitchLabelEncountered = true; |
455 | 0 | parseStructuralElement(); |
456 | 0 | break; |
457 | 1.04k | case tok::l_square: |
458 | 1.04k | if (Style.isCSharp()) { |
459 | 0 | nextToken(); |
460 | 0 | parseCSharpAttribute(); |
461 | 0 | break; |
462 | 0 | } |
463 | 1.04k | if (handleCppAttributes()) |
464 | 0 | break; |
465 | 1.04k | [[fallthrough]]; |
466 | 722k | default: |
467 | 722k | ParseDefault(); |
468 | 722k | break; |
469 | 1.31M | } |
470 | 1.31M | } while (!eof()); |
471 | | |
472 | 55.3k | return false; |
473 | 271k | } |
474 | | |
475 | 254k | void UnwrappedLineParser::calculateBraceTypes(bool ExpectClassBody) { |
476 | | // We'll parse forward through the tokens until we hit |
477 | | // a closing brace or eof - note that getNextToken() will |
478 | | // parse macros, so this will magically work inside macro |
479 | | // definitions, too. |
480 | 254k | unsigned StoredPosition = Tokens->getPosition(); |
481 | 254k | FormatToken *Tok = FormatTok; |
482 | 254k | const FormatToken *PrevTok = Tok->Previous; |
483 | | // Keep a stack of positions of lbrace tokens. We will |
484 | | // update information about whether an lbrace starts a |
485 | | // braced init list or a different block during the loop. |
486 | 254k | struct StackEntry { |
487 | 254k | FormatToken *Tok; |
488 | 254k | const FormatToken *PrevTok; |
489 | 254k | }; |
490 | 254k | SmallVector<StackEntry, 8> LBraceStack; |
491 | 254k | assert(Tok->is(tok::l_brace)); |
492 | 54.9M | do { |
493 | | // Get next non-comment, non-preprocessor token. |
494 | 54.9M | FormatToken *NextTok; |
495 | 55.0M | do { |
496 | 55.0M | NextTok = Tokens->getNextToken(); |
497 | 55.0M | } while (NextTok->is(tok::comment)); |
498 | 55.1M | while (NextTok->is(tok::hash) && !Line->InMacroBody) { |
499 | 194k | NextTok = Tokens->getNextToken(); |
500 | 44.0M | do { |
501 | 44.0M | NextTok = Tokens->getNextToken(); |
502 | 44.0M | } while (NextTok->is(tok::comment) || |
503 | 44.0M | (NextTok->NewlinesBefore == 0 && NextTok->isNot(tok::eof))); |
504 | 194k | } |
505 | | |
506 | 54.9M | switch (Tok->Tok.getKind()) { |
507 | 5.58M | case tok::l_brace: |
508 | 5.58M | if (Style.isJavaScript() && PrevTok) { |
509 | 0 | if (PrevTok->isOneOf(tok::colon, tok::less)) { |
510 | | // A ':' indicates this code is in a type, or a braced list |
511 | | // following a label in an object literal ({a: {b: 1}}). |
512 | | // A '<' could be an object used in a comparison, but that is nonsense |
513 | | // code (can never return true), so more likely it is a generic type |
514 | | // argument (`X<{a: string; b: number}>`). |
515 | | // The code below could be confused by semicolons between the |
516 | | // individual members in a type member list, which would normally |
517 | | // trigger BK_Block. In both cases, this must be parsed as an inline |
518 | | // braced init. |
519 | 0 | Tok->setBlockKind(BK_BracedInit); |
520 | 0 | } else if (PrevTok->is(tok::r_paren)) { |
521 | | // `) { }` can only occur in function or method declarations in JS. |
522 | 0 | Tok->setBlockKind(BK_Block); |
523 | 0 | } |
524 | 5.58M | } else { |
525 | 5.58M | Tok->setBlockKind(BK_Unknown); |
526 | 5.58M | } |
527 | 5.58M | LBraceStack.push_back({Tok, PrevTok}); |
528 | 5.58M | break; |
529 | 377k | case tok::r_brace: |
530 | 377k | if (LBraceStack.empty()) |
531 | 0 | break; |
532 | 377k | if (LBraceStack.back().Tok->is(BK_Unknown)) { |
533 | 367k | bool ProbablyBracedList = false; |
534 | 367k | if (Style.Language == FormatStyle::LK_Proto) { |
535 | 0 | ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); |
536 | 367k | } else { |
537 | | // Skip NextTok over preprocessor lines, otherwise we may not |
538 | | // properly diagnose the block as a braced intializer |
539 | | // if the comma separator appears after the pp directive. |
540 | 367k | while (NextTok->is(tok::hash)) { |
541 | 0 | ScopedMacroState MacroState(*Line, Tokens, NextTok); |
542 | 0 | do { |
543 | 0 | NextTok = Tokens->getNextToken(); |
544 | 0 | } while (NextTok->isNot(tok::eof)); |
545 | 0 | } |
546 | | |
547 | | // Using OriginalColumn to distinguish between ObjC methods and |
548 | | // binary operators is a bit hacky. |
549 | 367k | bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && |
550 | 367k | NextTok->OriginalColumn == 0; |
551 | | |
552 | | // Try to detect a braced list. Note that regardless how we mark inner |
553 | | // braces here, we will overwrite the BlockKind later if we parse a |
554 | | // braced list (where all blocks inside are by default braced lists), |
555 | | // or when we explicitly detect blocks (for example while parsing |
556 | | // lambdas). |
557 | | |
558 | | // If we already marked the opening brace as braced list, the closing |
559 | | // must also be part of it. |
560 | 367k | ProbablyBracedList = LBraceStack.back().Tok->is(TT_BracedListLBrace); |
561 | | |
562 | 367k | ProbablyBracedList = ProbablyBracedList || |
563 | 367k | (Style.isJavaScript() && |
564 | 367k | NextTok->isOneOf(Keywords.kw_of, Keywords.kw_in, |
565 | 0 | Keywords.kw_as)); |
566 | 367k | ProbablyBracedList = ProbablyBracedList || |
567 | 367k | (Style.isCpp() && NextTok->is(tok::l_paren)); |
568 | | |
569 | | // If there is a comma, semicolon or right paren after the closing |
570 | | // brace, we assume this is a braced initializer list. |
571 | | // FIXME: Some of these do not apply to JS, e.g. "} {" can never be a |
572 | | // braced list in JS. |
573 | 367k | ProbablyBracedList = |
574 | 367k | ProbablyBracedList || |
575 | 367k | NextTok->isOneOf(tok::comma, tok::period, tok::colon, |
576 | 366k | tok::r_paren, tok::r_square, tok::ellipsis); |
577 | | |
578 | | // Distinguish between braced list in a constructor initializer list |
579 | | // followed by constructor body, or just adjacent blocks. |
580 | 367k | ProbablyBracedList = |
581 | 367k | ProbablyBracedList || |
582 | 367k | (NextTok->is(tok::l_brace) && LBraceStack.back().PrevTok && |
583 | 287k | LBraceStack.back().PrevTok->isOneOf(tok::identifier, |
584 | 1.33k | tok::greater)); |
585 | | |
586 | 367k | ProbablyBracedList = |
587 | 367k | ProbablyBracedList || |
588 | 367k | (NextTok->is(tok::identifier) && |
589 | 287k | !PrevTok->isOneOf(tok::semi, tok::r_brace, tok::l_brace)); |
590 | | |
591 | 367k | ProbablyBracedList = ProbablyBracedList || |
592 | 367k | (NextTok->is(tok::semi) && |
593 | 232k | (!ExpectClassBody || LBraceStack.size() != 1)); |
594 | | |
595 | 367k | ProbablyBracedList = |
596 | 367k | ProbablyBracedList || |
597 | 367k | (NextTok->isBinaryOperator() && !NextIsObjCMethod); |
598 | | |
599 | 367k | if (!Style.isCSharp() && NextTok->is(tok::l_square)) { |
600 | | // We can have an array subscript after a braced init |
601 | | // list, but C++11 attributes are expected after blocks. |
602 | 126 | NextTok = Tokens->getNextToken(); |
603 | 126 | ProbablyBracedList = NextTok->isNot(tok::l_square); |
604 | 126 | } |
605 | 367k | } |
606 | 367k | if (ProbablyBracedList) { |
607 | 158k | Tok->setBlockKind(BK_BracedInit); |
608 | 158k | LBraceStack.back().Tok->setBlockKind(BK_BracedInit); |
609 | 208k | } else { |
610 | 208k | Tok->setBlockKind(BK_Block); |
611 | 208k | LBraceStack.back().Tok->setBlockKind(BK_Block); |
612 | 208k | } |
613 | 367k | } |
614 | 377k | LBraceStack.pop_back(); |
615 | 377k | break; |
616 | 3.57M | case tok::identifier: |
617 | 3.57M | if (Tok->isNot(TT_StatementMacro)) |
618 | 3.57M | break; |
619 | 3.57M | [[fallthrough]]; |
620 | 385k | case tok::at: |
621 | 862k | case tok::semi: |
622 | 867k | case tok::kw_if: |
623 | 868k | case tok::kw_while: |
624 | 868k | case tok::kw_for: |
625 | 868k | case tok::kw_switch: |
626 | 868k | case tok::kw_try: |
627 | 868k | case tok::kw___try: |
628 | 868k | if (!LBraceStack.empty() && LBraceStack.back().Tok->is(BK_Unknown)) |
629 | 283k | LBraceStack.back().Tok->setBlockKind(BK_Block); |
630 | 868k | break; |
631 | 44.5M | default: |
632 | 44.5M | break; |
633 | 54.9M | } |
634 | 54.9M | PrevTok = Tok; |
635 | 54.9M | Tok = NextTok; |
636 | 54.9M | } while (Tok->isNot(tok::eof) && !LBraceStack.empty()); |
637 | | |
638 | | // Assume other blocks for all unclosed opening braces. |
639 | 254k | for (const auto &Entry : LBraceStack) |
640 | 5.20M | if (Entry.Tok->is(BK_Unknown)) |
641 | 4.93M | Entry.Tok->setBlockKind(BK_Block); |
642 | | |
643 | 254k | FormatTok = Tokens->setPosition(StoredPosition); |
644 | 254k | } |
645 | | |
646 | | // Sets the token type of the directly previous right brace. |
647 | 3.66k | void UnwrappedLineParser::setPreviousRBraceType(TokenType Type) { |
648 | 3.66k | if (auto Prev = FormatTok->getPreviousNonComment(); |
649 | 3.66k | Prev && Prev->is(tok::r_brace)) { |
650 | 3.54k | Prev->setFinalizedType(Type); |
651 | 3.54k | } |
652 | 3.66k | } |
653 | | |
654 | | template <class T> |
655 | 2.45G | static inline void hash_combine(std::size_t &seed, const T &v) { |
656 | 2.45G | std::hash<T> hasher; |
657 | 2.45G | seed ^= hasher(v) + 0x9e3779b9 + (seed << 6) + (seed >> 2); |
658 | 2.45G | } |
659 | | |
660 | 751k | size_t UnwrappedLineParser::computePPHash() const { |
661 | 751k | size_t h = 0; |
662 | 1.22G | for (const auto &i : PPStack) { |
663 | 1.22G | hash_combine(h, size_t(i.Kind)); |
664 | 1.22G | hash_combine(h, i.Line); |
665 | 1.22G | } |
666 | 751k | return h; |
667 | 751k | } |
668 | | |
669 | | // Checks whether \p ParsedLine might fit on a single line. If \p OpeningBrace |
670 | | // is not null, subtracts its length (plus the preceding space) when computing |
671 | | // the length of \p ParsedLine. We must clone the tokens of \p ParsedLine before |
672 | | // running the token annotator on it so that we can restore them afterward. |
673 | | bool UnwrappedLineParser::mightFitOnOneLine( |
674 | 0 | UnwrappedLine &ParsedLine, const FormatToken *OpeningBrace) const { |
675 | 0 | const auto ColumnLimit = Style.ColumnLimit; |
676 | 0 | if (ColumnLimit == 0) |
677 | 0 | return true; |
678 | | |
679 | 0 | auto &Tokens = ParsedLine.Tokens; |
680 | 0 | assert(!Tokens.empty()); |
681 | | |
682 | 0 | const auto *LastToken = Tokens.back().Tok; |
683 | 0 | assert(LastToken); |
684 | | |
685 | 0 | SmallVector<UnwrappedLineNode> SavedTokens(Tokens.size()); |
686 | |
|
687 | 0 | int Index = 0; |
688 | 0 | for (const auto &Token : Tokens) { |
689 | 0 | assert(Token.Tok); |
690 | 0 | auto &SavedToken = SavedTokens[Index++]; |
691 | 0 | SavedToken.Tok = new FormatToken; |
692 | 0 | SavedToken.Tok->copyFrom(*Token.Tok); |
693 | 0 | SavedToken.Children = std::move(Token.Children); |
694 | 0 | } |
695 | |
|
696 | 0 | AnnotatedLine Line(ParsedLine); |
697 | 0 | assert(Line.Last == LastToken); |
698 | | |
699 | 0 | TokenAnnotator Annotator(Style, Keywords); |
700 | 0 | Annotator.annotate(Line); |
701 | 0 | Annotator.calculateFormattingInformation(Line); |
702 | |
|
703 | 0 | auto Length = LastToken->TotalLength; |
704 | 0 | if (OpeningBrace) { |
705 | 0 | assert(OpeningBrace != Tokens.front().Tok); |
706 | 0 | if (auto Prev = OpeningBrace->Previous; |
707 | 0 | Prev && Prev->TotalLength + ColumnLimit == OpeningBrace->TotalLength) { |
708 | 0 | Length -= ColumnLimit; |
709 | 0 | } |
710 | 0 | Length -= OpeningBrace->TokenText.size() + 1; |
711 | 0 | } |
712 | | |
713 | 0 | if (const auto *FirstToken = Line.First; FirstToken->is(tok::r_brace)) { |
714 | 0 | assert(!OpeningBrace || OpeningBrace->is(TT_ControlStatementLBrace)); |
715 | 0 | Length -= FirstToken->TokenText.size() + 1; |
716 | 0 | } |
717 | | |
718 | 0 | Index = 0; |
719 | 0 | for (auto &Token : Tokens) { |
720 | 0 | const auto &SavedToken = SavedTokens[Index++]; |
721 | 0 | Token.Tok->copyFrom(*SavedToken.Tok); |
722 | 0 | Token.Children = std::move(SavedToken.Children); |
723 | 0 | delete SavedToken.Tok; |
724 | 0 | } |
725 | | |
726 | | // If these change PPLevel needs to be used for get correct indentation. |
727 | 0 | assert(!Line.InMacroBody); |
728 | 0 | assert(!Line.InPPDirective); |
729 | 0 | return Line.Level * Style.IndentWidth + Length <= ColumnLimit; |
730 | 0 | } |
731 | | |
732 | | FormatToken *UnwrappedLineParser::parseBlock(bool MustBeDeclaration, |
733 | | unsigned AddLevels, bool MunchSemi, |
734 | | bool KeepBraces, |
735 | | IfStmtKind *IfKind, |
736 | 540k | bool UnindentWhitesmithsBraces) { |
737 | 751k | auto HandleVerilogBlockLabel = [this]() { |
738 | | // ":" name |
739 | 751k | if (Style.isVerilog() && FormatTok->is(tok::colon)) { |
740 | 0 | nextToken(); |
741 | 0 | if (Keywords.isVerilogIdentifier(*FormatTok)) |
742 | 0 | nextToken(); |
743 | 0 | } |
744 | 751k | }; |
745 | | |
746 | | // Whether this is a Verilog-specific block that has a special header like a |
747 | | // module. |
748 | 540k | const bool VerilogHierarchy = |
749 | 540k | Style.isVerilog() && Keywords.isVerilogHierarchy(*FormatTok); |
750 | 540k | assert((FormatTok->isOneOf(tok::l_brace, TT_MacroBlockBegin) || |
751 | 540k | (Style.isVerilog() && |
752 | 540k | (Keywords.isVerilogBegin(*FormatTok) || VerilogHierarchy))) && |
753 | 540k | "'{' or macro block token expected"); |
754 | 0 | FormatToken *Tok = FormatTok; |
755 | 540k | const bool FollowedByComment = Tokens->peekNextToken()->is(tok::comment); |
756 | 540k | auto Index = CurrentLines->size(); |
757 | 540k | const bool MacroBlock = FormatTok->is(TT_MacroBlockBegin); |
758 | 540k | FormatTok->setBlockKind(BK_Block); |
759 | | |
760 | | // For Whitesmiths mode, jump to the next level prior to skipping over the |
761 | | // braces. |
762 | 540k | if (!VerilogHierarchy && AddLevels > 0 && |
763 | 540k | Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths) { |
764 | 0 | ++Line->Level; |
765 | 0 | } |
766 | | |
767 | 540k | size_t PPStartHash = computePPHash(); |
768 | | |
769 | 540k | const unsigned InitialLevel = Line->Level; |
770 | 540k | if (VerilogHierarchy) { |
771 | 0 | AddLevels += parseVerilogHierarchyHeader(); |
772 | 540k | } else { |
773 | 540k | nextToken(/*LevelDifference=*/AddLevels); |
774 | 540k | HandleVerilogBlockLabel(); |
775 | 540k | } |
776 | | |
777 | | // Bail out if there are too many levels. Otherwise, the stack might overflow. |
778 | 540k | if (Line->Level > 300) |
779 | 293k | return nullptr; |
780 | | |
781 | 246k | if (MacroBlock && FormatTok->is(tok::l_paren)) |
782 | 0 | parseParens(); |
783 | | |
784 | 246k | size_t NbPreprocessorDirectives = |
785 | 246k | !parsingPPDirective() ? PreprocessorDirectives.size() : 0; |
786 | 246k | addUnwrappedLine(); |
787 | 246k | size_t OpeningLineIndex = |
788 | 246k | CurrentLines->empty() |
789 | 246k | ? (UnwrappedLine::kInvalidIndex) |
790 | 246k | : (CurrentLines->size() - 1 - NbPreprocessorDirectives); |
791 | | |
792 | | // Whitesmiths is weird here. The brace needs to be indented for the namespace |
793 | | // block, but the block itself may not be indented depending on the style |
794 | | // settings. This allows the format to back up one level in those cases. |
795 | 246k | if (UnindentWhitesmithsBraces) |
796 | 0 | --Line->Level; |
797 | | |
798 | 246k | ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, |
799 | 246k | MustBeDeclaration); |
800 | 246k | if (AddLevels > 0u && Style.BreakBeforeBraces != FormatStyle::BS_Whitesmiths) |
801 | 246k | Line->Level += AddLevels; |
802 | | |
803 | 246k | FormatToken *IfLBrace = nullptr; |
804 | 246k | const bool SimpleBlock = parseLevel(Tok, IfKind, &IfLBrace); |
805 | | |
806 | 246k | if (eof()) |
807 | 35.5k | return IfLBrace; |
808 | | |
809 | 211k | if (MacroBlock ? FormatTok->isNot(TT_MacroBlockEnd) |
810 | 211k | : FormatTok->isNot(tok::r_brace)) { |
811 | 0 | Line->Level = InitialLevel; |
812 | 0 | FormatTok->setBlockKind(BK_Block); |
813 | 0 | return IfLBrace; |
814 | 0 | } |
815 | | |
816 | 211k | if (FormatTok->is(tok::r_brace) && Tok->is(TT_NamespaceLBrace)) |
817 | 0 | FormatTok->setFinalizedType(TT_NamespaceRBrace); |
818 | | |
819 | 211k | const bool IsFunctionRBrace = |
820 | 211k | FormatTok->is(tok::r_brace) && Tok->is(TT_FunctionLBrace); |
821 | | |
822 | 211k | auto RemoveBraces = [=]() mutable { |
823 | 211k | if (!SimpleBlock) |
824 | 211k | return false; |
825 | 0 | assert(Tok->isOneOf(TT_ControlStatementLBrace, TT_ElseLBrace)); |
826 | 0 | assert(FormatTok->is(tok::r_brace)); |
827 | 0 | const bool WrappedOpeningBrace = !Tok->Previous; |
828 | 0 | if (WrappedOpeningBrace && FollowedByComment) |
829 | 0 | return false; |
830 | 0 | const bool HasRequiredIfBraces = IfLBrace && !IfLBrace->Optional; |
831 | 0 | if (KeepBraces && !HasRequiredIfBraces) |
832 | 0 | return false; |
833 | 0 | if (Tok->isNot(TT_ElseLBrace) || !HasRequiredIfBraces) { |
834 | 0 | const FormatToken *Previous = Tokens->getPreviousToken(); |
835 | 0 | assert(Previous); |
836 | 0 | if (Previous->is(tok::r_brace) && !Previous->Optional) |
837 | 0 | return false; |
838 | 0 | } |
839 | 0 | assert(!CurrentLines->empty()); |
840 | 0 | auto &LastLine = CurrentLines->back(); |
841 | 0 | if (LastLine.Level == InitialLevel + 1 && !mightFitOnOneLine(LastLine)) |
842 | 0 | return false; |
843 | 0 | if (Tok->is(TT_ElseLBrace)) |
844 | 0 | return true; |
845 | 0 | if (WrappedOpeningBrace) { |
846 | 0 | assert(Index > 0); |
847 | 0 | --Index; // The line above the wrapped l_brace. |
848 | 0 | Tok = nullptr; |
849 | 0 | } |
850 | 0 | return mightFitOnOneLine((*CurrentLines)[Index], Tok); |
851 | 0 | }; |
852 | 211k | if (RemoveBraces()) { |
853 | 0 | Tok->MatchingParen = FormatTok; |
854 | 0 | FormatTok->MatchingParen = Tok; |
855 | 0 | } |
856 | | |
857 | 211k | size_t PPEndHash = computePPHash(); |
858 | | |
859 | | // Munch the closing brace. |
860 | 211k | nextToken(/*LevelDifference=*/-AddLevels); |
861 | | |
862 | | // When this is a function block and there is an unnecessary semicolon |
863 | | // afterwards then mark it as optional (so the RemoveSemi pass can get rid of |
864 | | // it later). |
865 | 211k | if (Style.RemoveSemicolon && IsFunctionRBrace) { |
866 | 0 | while (FormatTok->is(tok::semi)) { |
867 | 0 | FormatTok->Optional = true; |
868 | 0 | nextToken(); |
869 | 0 | } |
870 | 0 | } |
871 | | |
872 | 211k | HandleVerilogBlockLabel(); |
873 | | |
874 | 211k | if (MacroBlock && FormatTok->is(tok::l_paren)) |
875 | 0 | parseParens(); |
876 | | |
877 | 211k | Line->Level = InitialLevel; |
878 | | |
879 | 211k | if (FormatTok->is(tok::kw_noexcept)) { |
880 | | // A noexcept in a requires expression. |
881 | 0 | nextToken(); |
882 | 0 | } |
883 | | |
884 | 211k | if (FormatTok->is(tok::arrow)) { |
885 | | // Following the } or noexcept we can find a trailing return type arrow |
886 | | // as part of an implicit conversion constraint. |
887 | 0 | nextToken(); |
888 | 0 | parseStructuralElement(); |
889 | 0 | } |
890 | | |
891 | 211k | if (MunchSemi && FormatTok->is(tok::semi)) |
892 | 15 | nextToken(); |
893 | | |
894 | 211k | if (PPStartHash == PPEndHash) { |
895 | 211k | Line->MatchingOpeningBlockLineIndex = OpeningLineIndex; |
896 | 211k | if (OpeningLineIndex != UnwrappedLine::kInvalidIndex) { |
897 | | // Update the opening line to add the forward reference as well |
898 | 211k | (*CurrentLines)[OpeningLineIndex].MatchingClosingBlockLineIndex = |
899 | 211k | CurrentLines->size() - 1; |
900 | 211k | } |
901 | 211k | } |
902 | | |
903 | 211k | return IfLBrace; |
904 | 211k | } |
905 | | |
906 | 0 | static bool isGoogScope(const UnwrappedLine &Line) { |
907 | | // FIXME: Closure-library specific stuff should not be hard-coded but be |
908 | | // configurable. |
909 | 0 | if (Line.Tokens.size() < 4) |
910 | 0 | return false; |
911 | 0 | auto I = Line.Tokens.begin(); |
912 | 0 | if (I->Tok->TokenText != "goog") |
913 | 0 | return false; |
914 | 0 | ++I; |
915 | 0 | if (I->Tok->isNot(tok::period)) |
916 | 0 | return false; |
917 | 0 | ++I; |
918 | 0 | if (I->Tok->TokenText != "scope") |
919 | 0 | return false; |
920 | 0 | ++I; |
921 | 0 | return I->Tok->is(tok::l_paren); |
922 | 0 | } |
923 | | |
924 | | static bool isIIFE(const UnwrappedLine &Line, |
925 | 0 | const AdditionalKeywords &Keywords) { |
926 | | // Look for the start of an immediately invoked anonymous function. |
927 | | // https://en.wikipedia.org/wiki/Immediately-invoked_function_expression |
928 | | // This is commonly done in JavaScript to create a new, anonymous scope. |
929 | | // Example: (function() { ... })() |
930 | 0 | if (Line.Tokens.size() < 3) |
931 | 0 | return false; |
932 | 0 | auto I = Line.Tokens.begin(); |
933 | 0 | if (I->Tok->isNot(tok::l_paren)) |
934 | 0 | return false; |
935 | 0 | ++I; |
936 | 0 | if (I->Tok->isNot(Keywords.kw_function)) |
937 | 0 | return false; |
938 | 0 | ++I; |
939 | 0 | return I->Tok->is(tok::l_paren); |
940 | 0 | } |
941 | | |
942 | | static bool ShouldBreakBeforeBrace(const FormatStyle &Style, |
943 | 156 | const FormatToken &InitialToken) { |
944 | 156 | tok::TokenKind Kind = InitialToken.Tok.getKind(); |
945 | 156 | if (InitialToken.is(TT_NamespaceMacro)) |
946 | 0 | Kind = tok::kw_namespace; |
947 | | |
948 | 156 | switch (Kind) { |
949 | 0 | case tok::kw_namespace: |
950 | 0 | return Style.BraceWrapping.AfterNamespace; |
951 | 57 | case tok::kw_class: |
952 | 57 | return Style.BraceWrapping.AfterClass; |
953 | 0 | case tok::kw_union: |
954 | 0 | return Style.BraceWrapping.AfterUnion; |
955 | 99 | case tok::kw_struct: |
956 | 99 | return Style.BraceWrapping.AfterStruct; |
957 | 0 | case tok::kw_enum: |
958 | 0 | return Style.BraceWrapping.AfterEnum; |
959 | 0 | default: |
960 | 0 | return false; |
961 | 156 | } |
962 | 156 | } |
963 | | |
964 | 21.7k | void UnwrappedLineParser::parseChildBlock() { |
965 | 21.7k | assert(FormatTok->is(tok::l_brace)); |
966 | 0 | FormatTok->setBlockKind(BK_Block); |
967 | 21.7k | const FormatToken *OpeningBrace = FormatTok; |
968 | 21.7k | nextToken(); |
969 | 21.7k | { |
970 | 21.7k | bool SkipIndent = (Style.isJavaScript() && |
971 | 21.7k | (isGoogScope(*Line) || isIIFE(*Line, Keywords))); |
972 | 21.7k | ScopedLineState LineState(*this); |
973 | 21.7k | ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, |
974 | 21.7k | /*MustBeDeclaration=*/false); |
975 | 21.7k | Line->Level += SkipIndent ? 0 : 1; |
976 | 21.7k | parseLevel(OpeningBrace); |
977 | 21.7k | flushComments(isOnNewLine(*FormatTok)); |
978 | 21.7k | Line->Level -= SkipIndent ? 0 : 1; |
979 | 21.7k | } |
980 | 21.7k | nextToken(); |
981 | 21.7k | } |
982 | | |
983 | 253k | void UnwrappedLineParser::parsePPDirective() { |
984 | 253k | assert(FormatTok->is(tok::hash) && "'#' expected"); |
985 | 0 | ScopedMacroState MacroState(*Line, Tokens, FormatTok); |
986 | | |
987 | 253k | nextToken(); |
988 | | |
989 | 253k | if (!FormatTok->Tok.getIdentifierInfo()) { |
990 | 228k | parsePPUnknown(); |
991 | 228k | return; |
992 | 228k | } |
993 | | |
994 | 25.4k | switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { |
995 | 2.03k | case tok::pp_define: |
996 | 2.03k | parsePPDefine(); |
997 | 2.03k | return; |
998 | 150 | case tok::pp_if: |
999 | 150 | parsePPIf(/*IfDef=*/false); |
1000 | 150 | break; |
1001 | 201 | case tok::pp_ifdef: |
1002 | 555 | case tok::pp_ifndef: |
1003 | 555 | parsePPIf(/*IfDef=*/true); |
1004 | 555 | break; |
1005 | 168 | case tok::pp_else: |
1006 | 168 | case tok::pp_elifdef: |
1007 | 168 | case tok::pp_elifndef: |
1008 | 243 | case tok::pp_elif: |
1009 | 243 | parsePPElse(); |
1010 | 243 | break; |
1011 | 591 | case tok::pp_endif: |
1012 | 591 | parsePPEndIf(); |
1013 | 591 | break; |
1014 | 117 | case tok::pp_pragma: |
1015 | 117 | parsePPPragma(); |
1016 | 117 | break; |
1017 | 21.8k | default: |
1018 | 21.8k | parsePPUnknown(); |
1019 | 21.8k | break; |
1020 | 25.4k | } |
1021 | 25.4k | } |
1022 | | |
1023 | 262k | void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { |
1024 | 262k | size_t Line = CurrentLines->size(); |
1025 | 262k | if (CurrentLines == &PreprocessorDirectives) |
1026 | 972 | Line += Lines.size(); |
1027 | | |
1028 | 262k | if (Unreachable || |
1029 | 262k | (!PPStack.empty() && PPStack.back().Kind == PP_Unreachable)) { |
1030 | 825 | PPStack.push_back({PP_Unreachable, Line}); |
1031 | 261k | } else { |
1032 | 261k | PPStack.push_back({PP_Conditional, Line}); |
1033 | 261k | } |
1034 | 262k | } |
1035 | | |
1036 | 261k | void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { |
1037 | 261k | ++PPBranchLevel; |
1038 | 261k | assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); |
1039 | 261k | if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { |
1040 | 261k | PPLevelBranchIndex.push_back(0); |
1041 | 261k | PPLevelBranchCount.push_back(0); |
1042 | 261k | } |
1043 | 261k | PPChainBranchIndex.push(Unreachable ? -1 : 0); |
1044 | 261k | bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; |
1045 | 261k | conditionalCompilationCondition(Unreachable || Skip); |
1046 | 261k | } |
1047 | | |
1048 | 369 | void UnwrappedLineParser::conditionalCompilationAlternative() { |
1049 | 369 | if (!PPStack.empty()) |
1050 | 360 | PPStack.pop_back(); |
1051 | 369 | assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); |
1052 | 369 | if (!PPChainBranchIndex.empty()) |
1053 | 366 | ++PPChainBranchIndex.top(); |
1054 | 369 | conditionalCompilationCondition( |
1055 | 369 | PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && |
1056 | 369 | PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); |
1057 | 369 | } |
1058 | | |
1059 | 762 | void UnwrappedLineParser::conditionalCompilationEnd() { |
1060 | 762 | assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); |
1061 | 762 | if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { |
1062 | 609 | if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) |
1063 | 135 | PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; |
1064 | 609 | } |
1065 | | // Guard against #endif's without #if. |
1066 | 762 | if (PPBranchLevel > -1) |
1067 | 609 | --PPBranchLevel; |
1068 | 762 | if (!PPChainBranchIndex.empty()) |
1069 | 621 | PPChainBranchIndex.pop(); |
1070 | 762 | if (!PPStack.empty()) |
1071 | 609 | PPStack.pop_back(); |
1072 | 762 | } |
1073 | | |
1074 | 705 | void UnwrappedLineParser::parsePPIf(bool IfDef) { |
1075 | 705 | bool IfNDef = FormatTok->is(tok::pp_ifndef); |
1076 | 705 | nextToken(); |
1077 | 705 | bool Unreachable = false; |
1078 | 705 | if (!IfDef && (FormatTok->is(tok::kw_false) || FormatTok->TokenText == "0")) |
1079 | 15 | Unreachable = true; |
1080 | 705 | if (IfDef && !IfNDef && FormatTok->TokenText == "SWIG") |
1081 | 0 | Unreachable = true; |
1082 | 705 | conditionalCompilationStart(Unreachable); |
1083 | 705 | FormatToken *IfCondition = FormatTok; |
1084 | | // If there's a #ifndef on the first line, and the only lines before it are |
1085 | | // comments, it could be an include guard. |
1086 | 705 | bool MaybeIncludeGuard = IfNDef; |
1087 | 705 | if (IncludeGuard == IG_Inited && MaybeIncludeGuard) { |
1088 | 0 | for (auto &Line : Lines) { |
1089 | 0 | if (Line.Tokens.front().Tok->isNot(tok::comment)) { |
1090 | 0 | MaybeIncludeGuard = false; |
1091 | 0 | IncludeGuard = IG_Rejected; |
1092 | 0 | break; |
1093 | 0 | } |
1094 | 0 | } |
1095 | 0 | } |
1096 | 705 | --PPBranchLevel; |
1097 | 705 | parsePPUnknown(); |
1098 | 705 | ++PPBranchLevel; |
1099 | 705 | if (IncludeGuard == IG_Inited && MaybeIncludeGuard) { |
1100 | 0 | IncludeGuard = IG_IfNdefed; |
1101 | 0 | IncludeGuardToken = IfCondition; |
1102 | 0 | } |
1103 | 705 | } |
1104 | | |
1105 | 243 | void UnwrappedLineParser::parsePPElse() { |
1106 | | // If a potential include guard has an #else, it's not an include guard. |
1107 | 243 | if (IncludeGuard == IG_Defined && PPBranchLevel == 0) |
1108 | 0 | IncludeGuard = IG_Rejected; |
1109 | | // Don't crash when there is an #else without an #if. |
1110 | 243 | assert(PPBranchLevel >= -1); |
1111 | 243 | if (PPBranchLevel == -1) |
1112 | 24 | conditionalCompilationStart(/*Unreachable=*/true); |
1113 | 243 | conditionalCompilationAlternative(); |
1114 | 243 | --PPBranchLevel; |
1115 | 243 | parsePPUnknown(); |
1116 | 243 | ++PPBranchLevel; |
1117 | 243 | } |
1118 | | |
1119 | 591 | void UnwrappedLineParser::parsePPEndIf() { |
1120 | 591 | conditionalCompilationEnd(); |
1121 | 591 | parsePPUnknown(); |
1122 | | // If the #endif of a potential include guard is the last thing in the file, |
1123 | | // then we found an include guard. |
1124 | 591 | if (IncludeGuard == IG_Defined && PPBranchLevel == -1 && Tokens->isEOF() && |
1125 | 591 | Style.IndentPPDirectives != FormatStyle::PPDIS_None) { |
1126 | 0 | IncludeGuard = IG_Found; |
1127 | 0 | } |
1128 | 591 | } |
1129 | | |
1130 | 2.03k | void UnwrappedLineParser::parsePPDefine() { |
1131 | 2.03k | nextToken(); |
1132 | | |
1133 | 2.03k | if (!FormatTok->Tok.getIdentifierInfo()) { |
1134 | 6 | IncludeGuard = IG_Rejected; |
1135 | 6 | IncludeGuardToken = nullptr; |
1136 | 6 | parsePPUnknown(); |
1137 | 6 | return; |
1138 | 6 | } |
1139 | | |
1140 | 2.02k | if (IncludeGuard == IG_IfNdefed && |
1141 | 2.02k | IncludeGuardToken->TokenText == FormatTok->TokenText) { |
1142 | 0 | IncludeGuard = IG_Defined; |
1143 | 0 | IncludeGuardToken = nullptr; |
1144 | 0 | for (auto &Line : Lines) { |
1145 | 0 | if (!Line.Tokens.front().Tok->isOneOf(tok::comment, tok::hash)) { |
1146 | 0 | IncludeGuard = IG_Rejected; |
1147 | 0 | break; |
1148 | 0 | } |
1149 | 0 | } |
1150 | 0 | } |
1151 | | |
1152 | | // In the context of a define, even keywords should be treated as normal |
1153 | | // identifiers. Setting the kind to identifier is not enough, because we need |
1154 | | // to treat additional keywords like __except as well, which are already |
1155 | | // identifiers. Setting the identifier info to null interferes with include |
1156 | | // guard processing above, and changes preprocessing nesting. |
1157 | 2.02k | FormatTok->Tok.setKind(tok::identifier); |
1158 | 2.02k | FormatTok->Tok.setIdentifierInfo(Keywords.kw_internal_ident_after_define); |
1159 | 2.02k | nextToken(); |
1160 | 2.02k | if (FormatTok->Tok.getKind() == tok::l_paren && |
1161 | 2.02k | !FormatTok->hasWhitespaceBefore()) { |
1162 | 192 | parseParens(); |
1163 | 192 | } |
1164 | 2.02k | if (Style.IndentPPDirectives != FormatStyle::PPDIS_None) |
1165 | 0 | Line->Level += PPBranchLevel + 1; |
1166 | 2.02k | addUnwrappedLine(); |
1167 | 2.02k | ++Line->Level; |
1168 | | |
1169 | 2.02k | Line->PPLevel = PPBranchLevel + (IncludeGuard == IG_Defined ? 0 : 1); |
1170 | 2.02k | assert((int)Line->PPLevel >= 0); |
1171 | 0 | Line->InMacroBody = true; |
1172 | | |
1173 | 2.02k | if (FormatTok->is(tok::identifier) && |
1174 | 2.02k | Tokens->peekNextToken()->is(tok::colon)) { |
1175 | 0 | nextToken(); |
1176 | 0 | nextToken(); |
1177 | 0 | } |
1178 | | |
1179 | | // Errors during a preprocessor directive can only affect the layout of the |
1180 | | // preprocessor directive, and thus we ignore them. An alternative approach |
1181 | | // would be to use the same approach we use on the file level (no |
1182 | | // re-indentation if there was a structural error) within the macro |
1183 | | // definition. |
1184 | 2.02k | parseFile(); |
1185 | 2.02k | } |
1186 | | |
1187 | 117 | void UnwrappedLineParser::parsePPPragma() { |
1188 | 117 | Line->InPragmaDirective = true; |
1189 | 117 | parsePPUnknown(); |
1190 | 117 | } |
1191 | | |
1192 | 251k | void UnwrappedLineParser::parsePPUnknown() { |
1193 | 820k | do { |
1194 | 820k | nextToken(); |
1195 | 820k | } while (!eof()); |
1196 | 251k | if (Style.IndentPPDirectives != FormatStyle::PPDIS_None) |
1197 | 0 | Line->Level += PPBranchLevel + 1; |
1198 | 251k | addUnwrappedLine(); |
1199 | 251k | } |
1200 | | |
1201 | | // Here we exclude certain tokens that are not usually the first token in an |
1202 | | // unwrapped line. This is used in attempt to distinguish macro calls without |
1203 | | // trailing semicolons from other constructs split to several lines. |
1204 | 4.63k | static bool tokenCanStartNewLine(const FormatToken &Tok) { |
1205 | | // Semicolon can be a null-statement, l_square can be a start of a macro or |
1206 | | // a C++11 attribute, but this doesn't seem to be common. |
1207 | 4.63k | assert(Tok.isNot(TT_AttributeSquare)); |
1208 | 0 | return !Tok.isOneOf(tok::semi, tok::l_brace, |
1209 | | // Tokens that can only be used as binary operators and a |
1210 | | // part of overloaded operator names. |
1211 | 4.63k | tok::period, tok::periodstar, tok::arrow, tok::arrowstar, |
1212 | 4.63k | tok::less, tok::greater, tok::slash, tok::percent, |
1213 | 4.63k | tok::lessless, tok::greatergreater, tok::equal, |
1214 | 4.63k | tok::plusequal, tok::minusequal, tok::starequal, |
1215 | 4.63k | tok::slashequal, tok::percentequal, tok::ampequal, |
1216 | 4.63k | tok::pipeequal, tok::caretequal, tok::greatergreaterequal, |
1217 | 4.63k | tok::lesslessequal, |
1218 | | // Colon is used in labels, base class lists, initializer |
1219 | | // lists, range-based for loops, ternary operator, but |
1220 | | // should never be the first token in an unwrapped line. |
1221 | 4.63k | tok::colon, |
1222 | | // 'noexcept' is a trailing annotation. |
1223 | 4.63k | tok::kw_noexcept); |
1224 | 4.63k | } |
1225 | | |
1226 | | static bool mustBeJSIdent(const AdditionalKeywords &Keywords, |
1227 | 0 | const FormatToken *FormatTok) { |
1228 | | // FIXME: This returns true for C/C++ keywords like 'struct'. |
1229 | 0 | return FormatTok->is(tok::identifier) && |
1230 | 0 | (!FormatTok->Tok.getIdentifierInfo() || |
1231 | 0 | !FormatTok->isOneOf( |
1232 | 0 | Keywords.kw_in, Keywords.kw_of, Keywords.kw_as, Keywords.kw_async, |
1233 | 0 | Keywords.kw_await, Keywords.kw_yield, Keywords.kw_finally, |
1234 | 0 | Keywords.kw_function, Keywords.kw_import, Keywords.kw_is, |
1235 | 0 | Keywords.kw_let, Keywords.kw_var, tok::kw_const, |
1236 | 0 | Keywords.kw_abstract, Keywords.kw_extends, Keywords.kw_implements, |
1237 | 0 | Keywords.kw_instanceof, Keywords.kw_interface, |
1238 | 0 | Keywords.kw_override, Keywords.kw_throws, Keywords.kw_from)); |
1239 | 0 | } |
1240 | | |
1241 | | static bool mustBeJSIdentOrValue(const AdditionalKeywords &Keywords, |
1242 | 0 | const FormatToken *FormatTok) { |
1243 | 0 | return FormatTok->Tok.isLiteral() || |
1244 | 0 | FormatTok->isOneOf(tok::kw_true, tok::kw_false) || |
1245 | 0 | mustBeJSIdent(Keywords, FormatTok); |
1246 | 0 | } |
1247 | | |
1248 | | // isJSDeclOrStmt returns true if |FormatTok| starts a declaration or statement |
1249 | | // when encountered after a value (see mustBeJSIdentOrValue). |
1250 | | static bool isJSDeclOrStmt(const AdditionalKeywords &Keywords, |
1251 | 0 | const FormatToken *FormatTok) { |
1252 | 0 | return FormatTok->isOneOf( |
1253 | 0 | tok::kw_return, Keywords.kw_yield, |
1254 | | // conditionals |
1255 | 0 | tok::kw_if, tok::kw_else, |
1256 | | // loops |
1257 | 0 | tok::kw_for, tok::kw_while, tok::kw_do, tok::kw_continue, tok::kw_break, |
1258 | | // switch/case |
1259 | 0 | tok::kw_switch, tok::kw_case, |
1260 | | // exceptions |
1261 | 0 | tok::kw_throw, tok::kw_try, tok::kw_catch, Keywords.kw_finally, |
1262 | | // declaration |
1263 | 0 | tok::kw_const, tok::kw_class, Keywords.kw_var, Keywords.kw_let, |
1264 | 0 | Keywords.kw_async, Keywords.kw_function, |
1265 | | // import/export |
1266 | 0 | Keywords.kw_import, tok::kw_export); |
1267 | 0 | } |
1268 | | |
1269 | | // Checks whether a token is a type in K&R C (aka C78). |
1270 | 26.3k | static bool isC78Type(const FormatToken &Tok) { |
1271 | 26.3k | return Tok.isOneOf(tok::kw_char, tok::kw_short, tok::kw_int, tok::kw_long, |
1272 | 26.3k | tok::kw_unsigned, tok::kw_float, tok::kw_double, |
1273 | 26.3k | tok::identifier); |
1274 | 26.3k | } |
1275 | | |
1276 | | // This function checks whether a token starts the first parameter declaration |
1277 | | // in a K&R C (aka C78) function definition, e.g.: |
1278 | | // int f(a, b) |
1279 | | // short a, b; |
1280 | | // { |
1281 | | // return a + b; |
1282 | | // } |
1283 | | static bool isC78ParameterDecl(const FormatToken *Tok, const FormatToken *Next, |
1284 | 21.4k | const FormatToken *FuncName) { |
1285 | 21.4k | assert(Tok); |
1286 | 0 | assert(Next); |
1287 | 0 | assert(FuncName); |
1288 | | |
1289 | 21.4k | if (FuncName->isNot(tok::identifier)) |
1290 | 5.07k | return false; |
1291 | | |
1292 | 16.4k | const FormatToken *Prev = FuncName->Previous; |
1293 | 16.4k | if (!Prev || (Prev->isNot(tok::star) && !isC78Type(*Prev))) |
1294 | 6.44k | return false; |
1295 | | |
1296 | 9.96k | if (!isC78Type(*Tok) && |
1297 | 9.96k | !Tok->isOneOf(tok::kw_register, tok::kw_struct, tok::kw_union)) { |
1298 | 9.04k | return false; |
1299 | 9.04k | } |
1300 | | |
1301 | 921 | if (Next->isNot(tok::star) && !Next->Tok.getIdentifierInfo()) |
1302 | 213 | return false; |
1303 | | |
1304 | 708 | Tok = Tok->Previous; |
1305 | 708 | if (!Tok || Tok->isNot(tok::r_paren)) |
1306 | 0 | return false; |
1307 | | |
1308 | 708 | Tok = Tok->Previous; |
1309 | 708 | if (!Tok || Tok->isNot(tok::identifier)) |
1310 | 24 | return false; |
1311 | | |
1312 | 684 | return Tok->Previous && Tok->Previous->isOneOf(tok::l_paren, tok::comma); |
1313 | 708 | } |
1314 | | |
1315 | 0 | bool UnwrappedLineParser::parseModuleImport() { |
1316 | 0 | assert(FormatTok->is(Keywords.kw_import) && "'import' expected"); |
1317 | | |
1318 | 0 | if (auto Token = Tokens->peekNextToken(/*SkipComment=*/true); |
1319 | 0 | !Token->Tok.getIdentifierInfo() && |
1320 | 0 | !Token->isOneOf(tok::colon, tok::less, tok::string_literal)) { |
1321 | 0 | return false; |
1322 | 0 | } |
1323 | | |
1324 | 0 | nextToken(); |
1325 | 0 | while (!eof()) { |
1326 | 0 | if (FormatTok->is(tok::colon)) { |
1327 | 0 | FormatTok->setFinalizedType(TT_ModulePartitionColon); |
1328 | 0 | } |
1329 | | // Handle import <foo/bar.h> as we would an include statement. |
1330 | 0 | else if (FormatTok->is(tok::less)) { |
1331 | 0 | nextToken(); |
1332 | 0 | while (!FormatTok->isOneOf(tok::semi, tok::greater, tok::eof)) { |
1333 | | // Mark tokens up to the trailing line comments as implicit string |
1334 | | // literals. |
1335 | 0 | if (FormatTok->isNot(tok::comment) && |
1336 | 0 | !FormatTok->TokenText.starts_with("//")) { |
1337 | 0 | FormatTok->setFinalizedType(TT_ImplicitStringLiteral); |
1338 | 0 | } |
1339 | 0 | nextToken(); |
1340 | 0 | } |
1341 | 0 | } |
1342 | 0 | if (FormatTok->is(tok::semi)) { |
1343 | 0 | nextToken(); |
1344 | 0 | break; |
1345 | 0 | } |
1346 | 0 | nextToken(); |
1347 | 0 | } |
1348 | |
|
1349 | 0 | addUnwrappedLine(); |
1350 | 0 | return true; |
1351 | 0 | } |
1352 | | |
1353 | | // readTokenWithJavaScriptASI reads the next token and terminates the current |
1354 | | // line if JavaScript Automatic Semicolon Insertion must |
1355 | | // happen between the current token and the next token. |
1356 | | // |
1357 | | // This method is conservative - it cannot cover all edge cases of JavaScript, |
1358 | | // but only aims to correctly handle certain well known cases. It *must not* |
1359 | | // return true in speculative cases. |
1360 | 0 | void UnwrappedLineParser::readTokenWithJavaScriptASI() { |
1361 | 0 | FormatToken *Previous = FormatTok; |
1362 | 0 | readToken(); |
1363 | 0 | FormatToken *Next = FormatTok; |
1364 | |
|
1365 | 0 | bool IsOnSameLine = |
1366 | 0 | CommentsBeforeNextToken.empty() |
1367 | 0 | ? Next->NewlinesBefore == 0 |
1368 | 0 | : CommentsBeforeNextToken.front()->NewlinesBefore == 0; |
1369 | 0 | if (IsOnSameLine) |
1370 | 0 | return; |
1371 | | |
1372 | 0 | bool PreviousMustBeValue = mustBeJSIdentOrValue(Keywords, Previous); |
1373 | 0 | bool PreviousStartsTemplateExpr = |
1374 | 0 | Previous->is(TT_TemplateString) && Previous->TokenText.ends_with("${"); |
1375 | 0 | if (PreviousMustBeValue || Previous->is(tok::r_paren)) { |
1376 | | // If the line contains an '@' sign, the previous token might be an |
1377 | | // annotation, which can precede another identifier/value. |
1378 | 0 | bool HasAt = llvm::any_of(Line->Tokens, [](UnwrappedLineNode &LineNode) { |
1379 | 0 | return LineNode.Tok->is(tok::at); |
1380 | 0 | }); |
1381 | 0 | if (HasAt) |
1382 | 0 | return; |
1383 | 0 | } |
1384 | 0 | if (Next->is(tok::exclaim) && PreviousMustBeValue) |
1385 | 0 | return addUnwrappedLine(); |
1386 | 0 | bool NextMustBeValue = mustBeJSIdentOrValue(Keywords, Next); |
1387 | 0 | bool NextEndsTemplateExpr = |
1388 | 0 | Next->is(TT_TemplateString) && Next->TokenText.starts_with("}"); |
1389 | 0 | if (NextMustBeValue && !NextEndsTemplateExpr && !PreviousStartsTemplateExpr && |
1390 | 0 | (PreviousMustBeValue || |
1391 | 0 | Previous->isOneOf(tok::r_square, tok::r_paren, tok::plusplus, |
1392 | 0 | tok::minusminus))) { |
1393 | 0 | return addUnwrappedLine(); |
1394 | 0 | } |
1395 | 0 | if ((PreviousMustBeValue || Previous->is(tok::r_paren)) && |
1396 | 0 | isJSDeclOrStmt(Keywords, Next)) { |
1397 | 0 | return addUnwrappedLine(); |
1398 | 0 | } |
1399 | 0 | } |
1400 | | |
1401 | | void UnwrappedLineParser::parseStructuralElement( |
1402 | | const FormatToken *OpeningBrace, IfStmtKind *IfKind, |
1403 | 768k | FormatToken **IfLeftBrace, bool *HasDoWhile, bool *HasLabel) { |
1404 | 768k | if (Style.Language == FormatStyle::LK_TableGen && |
1405 | 768k | FormatTok->is(tok::pp_include)) { |
1406 | 0 | nextToken(); |
1407 | 0 | if (FormatTok->is(tok::string_literal)) |
1408 | 0 | nextToken(); |
1409 | 0 | addUnwrappedLine(); |
1410 | 0 | return; |
1411 | 0 | } |
1412 | | |
1413 | 768k | if (Style.isCpp()) { |
1414 | 768k | while (FormatTok->is(tok::l_square) && handleCppAttributes()) { |
1415 | 0 | } |
1416 | 768k | } else if (Style.isVerilog()) { |
1417 | 0 | if (Keywords.isVerilogStructuredProcedure(*FormatTok)) { |
1418 | 0 | parseForOrWhileLoop(/*HasParens=*/false); |
1419 | 0 | return; |
1420 | 0 | } |
1421 | 0 | if (FormatTok->isOneOf(Keywords.kw_foreach, Keywords.kw_repeat)) { |
1422 | 0 | parseForOrWhileLoop(); |
1423 | 0 | return; |
1424 | 0 | } |
1425 | 0 | if (FormatTok->isOneOf(tok::kw_restrict, Keywords.kw_assert, |
1426 | 0 | Keywords.kw_assume, Keywords.kw_cover)) { |
1427 | 0 | parseIfThenElse(IfKind, /*KeepBraces=*/false, /*IsVerilogAssert=*/true); |
1428 | 0 | return; |
1429 | 0 | } |
1430 | | |
1431 | | // Skip things that can exist before keywords like 'if' and 'case'. |
1432 | 0 | while (true) { |
1433 | 0 | if (FormatTok->isOneOf(Keywords.kw_priority, Keywords.kw_unique, |
1434 | 0 | Keywords.kw_unique0)) { |
1435 | 0 | nextToken(); |
1436 | 0 | } else if (FormatTok->is(tok::l_paren) && |
1437 | 0 | Tokens->peekNextToken()->is(tok::star)) { |
1438 | 0 | parseParens(); |
1439 | 0 | } else { |
1440 | 0 | break; |
1441 | 0 | } |
1442 | 0 | } |
1443 | 0 | } |
1444 | | |
1445 | | // Tokens that only make sense at the beginning of a line. |
1446 | 768k | switch (FormatTok->Tok.getKind()) { |
1447 | 0 | case tok::kw_asm: |
1448 | 0 | nextToken(); |
1449 | 0 | if (FormatTok->is(tok::l_brace)) { |
1450 | 0 | FormatTok->setFinalizedType(TT_InlineASMBrace); |
1451 | 0 | nextToken(); |
1452 | 0 | while (FormatTok && !eof()) { |
1453 | 0 | if (FormatTok->is(tok::r_brace)) { |
1454 | 0 | FormatTok->setFinalizedType(TT_InlineASMBrace); |
1455 | 0 | nextToken(); |
1456 | 0 | addUnwrappedLine(); |
1457 | 0 | break; |
1458 | 0 | } |
1459 | 0 | FormatTok->Finalized = true; |
1460 | 0 | nextToken(); |
1461 | 0 | } |
1462 | 0 | } |
1463 | 0 | break; |
1464 | 0 | case tok::kw_namespace: |
1465 | 0 | parseNamespace(); |
1466 | 0 | return; |
1467 | 0 | case tok::kw_public: |
1468 | 0 | case tok::kw_protected: |
1469 | 0 | case tok::kw_private: |
1470 | 0 | if (Style.Language == FormatStyle::LK_Java || Style.isJavaScript() || |
1471 | 0 | Style.isCSharp()) { |
1472 | 0 | nextToken(); |
1473 | 0 | } else { |
1474 | 0 | parseAccessSpecifier(); |
1475 | 0 | } |
1476 | 0 | return; |
1477 | 5.76k | case tok::kw_if: { |
1478 | 5.76k | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1479 | | // field/method declaration. |
1480 | 0 | break; |
1481 | 0 | } |
1482 | 5.76k | FormatToken *Tok = parseIfThenElse(IfKind); |
1483 | 5.76k | if (IfLeftBrace) |
1484 | 5.76k | *IfLeftBrace = Tok; |
1485 | 5.76k | return; |
1486 | 5.76k | } |
1487 | 294 | case tok::kw_for: |
1488 | 381 | case tok::kw_while: |
1489 | 381 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1490 | | // field/method declaration. |
1491 | 0 | break; |
1492 | 0 | } |
1493 | 381 | parseForOrWhileLoop(); |
1494 | 381 | return; |
1495 | 66 | case tok::kw_do: |
1496 | 66 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1497 | | // field/method declaration. |
1498 | 0 | break; |
1499 | 0 | } |
1500 | 66 | parseDoWhile(); |
1501 | 66 | if (HasDoWhile) |
1502 | 66 | *HasDoWhile = true; |
1503 | 66 | return; |
1504 | 0 | case tok::kw_switch: |
1505 | 0 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1506 | | // 'switch: string' field declaration. |
1507 | 0 | break; |
1508 | 0 | } |
1509 | 0 | parseSwitch(); |
1510 | 0 | return; |
1511 | 0 | case tok::kw_default: |
1512 | | // In Verilog default along with other labels are handled in the next loop. |
1513 | 0 | if (Style.isVerilog()) |
1514 | 0 | break; |
1515 | 0 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1516 | | // 'default: string' field declaration. |
1517 | 0 | break; |
1518 | 0 | } |
1519 | 0 | nextToken(); |
1520 | 0 | if (FormatTok->is(tok::colon)) { |
1521 | 0 | FormatTok->setFinalizedType(TT_CaseLabelColon); |
1522 | 0 | parseLabel(); |
1523 | 0 | return; |
1524 | 0 | } |
1525 | | // e.g. "default void f() {}" in a Java interface. |
1526 | 0 | break; |
1527 | 0 | case tok::kw_case: |
1528 | | // Proto: there are no switch/case statements. |
1529 | 0 | if (Style.Language == FormatStyle::LK_Proto) { |
1530 | 0 | nextToken(); |
1531 | 0 | return; |
1532 | 0 | } |
1533 | 0 | if (Style.isVerilog()) { |
1534 | 0 | parseBlock(); |
1535 | 0 | addUnwrappedLine(); |
1536 | 0 | return; |
1537 | 0 | } |
1538 | 0 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1539 | | // 'case: string' field declaration. |
1540 | 0 | nextToken(); |
1541 | 0 | break; |
1542 | 0 | } |
1543 | 0 | parseCaseLabel(); |
1544 | 0 | return; |
1545 | 0 | case tok::kw_try: |
1546 | 0 | case tok::kw___try: |
1547 | 0 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1548 | | // field/method declaration. |
1549 | 0 | break; |
1550 | 0 | } |
1551 | 0 | parseTryCatch(); |
1552 | 0 | return; |
1553 | 3 | case tok::kw_extern: |
1554 | 3 | nextToken(); |
1555 | 3 | if (Style.isVerilog()) { |
1556 | | // In Verilog and extern module declaration looks like a start of module. |
1557 | | // But there is no body and endmodule. So we handle it separately. |
1558 | 0 | if (Keywords.isVerilogHierarchy(*FormatTok)) { |
1559 | 0 | parseVerilogHierarchyHeader(); |
1560 | 0 | return; |
1561 | 0 | } |
1562 | 3 | } else if (FormatTok->is(tok::string_literal)) { |
1563 | 3 | nextToken(); |
1564 | 3 | if (FormatTok->is(tok::l_brace)) { |
1565 | 3 | if (Style.BraceWrapping.AfterExternBlock) |
1566 | 0 | addUnwrappedLine(); |
1567 | | // Either we indent or for backwards compatibility we follow the |
1568 | | // AfterExternBlock style. |
1569 | 3 | unsigned AddLevels = |
1570 | 3 | (Style.IndentExternBlock == FormatStyle::IEBS_Indent) || |
1571 | 3 | (Style.BraceWrapping.AfterExternBlock && |
1572 | 3 | Style.IndentExternBlock == |
1573 | 0 | FormatStyle::IEBS_AfterExternBlock) |
1574 | 3 | ? 1u |
1575 | 3 | : 0u; |
1576 | 3 | parseBlock(/*MustBeDeclaration=*/true, AddLevels); |
1577 | 3 | addUnwrappedLine(); |
1578 | 3 | return; |
1579 | 3 | } |
1580 | 3 | } |
1581 | 0 | break; |
1582 | 0 | case tok::kw_export: |
1583 | 0 | if (Style.isJavaScript()) { |
1584 | 0 | parseJavaScriptEs6ImportExport(); |
1585 | 0 | return; |
1586 | 0 | } |
1587 | 0 | if (Style.isCpp()) { |
1588 | 0 | nextToken(); |
1589 | 0 | if (FormatTok->is(tok::kw_namespace)) { |
1590 | 0 | parseNamespace(); |
1591 | 0 | return; |
1592 | 0 | } |
1593 | 0 | if (FormatTok->is(Keywords.kw_import) && parseModuleImport()) |
1594 | 0 | return; |
1595 | 0 | } |
1596 | 0 | break; |
1597 | 0 | case tok::kw_inline: |
1598 | 0 | nextToken(); |
1599 | 0 | if (FormatTok->is(tok::kw_namespace)) { |
1600 | 0 | parseNamespace(); |
1601 | 0 | return; |
1602 | 0 | } |
1603 | 0 | break; |
1604 | 390k | case tok::identifier: |
1605 | 390k | if (FormatTok->is(TT_ForEachMacro)) { |
1606 | 0 | parseForOrWhileLoop(); |
1607 | 0 | return; |
1608 | 0 | } |
1609 | 390k | if (FormatTok->is(TT_MacroBlockBegin)) { |
1610 | 0 | parseBlock(/*MustBeDeclaration=*/false, /*AddLevels=*/1u, |
1611 | 0 | /*MunchSemi=*/false); |
1612 | 0 | return; |
1613 | 0 | } |
1614 | 390k | if (FormatTok->is(Keywords.kw_import)) { |
1615 | 0 | if (Style.isJavaScript()) { |
1616 | 0 | parseJavaScriptEs6ImportExport(); |
1617 | 0 | return; |
1618 | 0 | } |
1619 | 0 | if (Style.Language == FormatStyle::LK_Proto) { |
1620 | 0 | nextToken(); |
1621 | 0 | if (FormatTok->is(tok::kw_public)) |
1622 | 0 | nextToken(); |
1623 | 0 | if (FormatTok->isNot(tok::string_literal)) |
1624 | 0 | return; |
1625 | 0 | nextToken(); |
1626 | 0 | if (FormatTok->is(tok::semi)) |
1627 | 0 | nextToken(); |
1628 | 0 | addUnwrappedLine(); |
1629 | 0 | return; |
1630 | 0 | } |
1631 | 0 | if (Style.isCpp() && parseModuleImport()) |
1632 | 0 | return; |
1633 | 0 | } |
1634 | 390k | if (Style.isCpp() && |
1635 | 390k | FormatTok->isOneOf(Keywords.kw_signals, Keywords.kw_qsignals, |
1636 | 390k | Keywords.kw_slots, Keywords.kw_qslots)) { |
1637 | 0 | nextToken(); |
1638 | 0 | if (FormatTok->is(tok::colon)) { |
1639 | 0 | nextToken(); |
1640 | 0 | addUnwrappedLine(); |
1641 | 0 | return; |
1642 | 0 | } |
1643 | 0 | } |
1644 | 390k | if (Style.isCpp() && FormatTok->is(TT_StatementMacro)) { |
1645 | 0 | parseStatementMacro(); |
1646 | 0 | return; |
1647 | 0 | } |
1648 | 390k | if (Style.isCpp() && FormatTok->is(TT_NamespaceMacro)) { |
1649 | 0 | parseNamespace(); |
1650 | 0 | return; |
1651 | 0 | } |
1652 | | // In Verilog labels can be any expression, so we don't do them here. |
1653 | | // JS doesn't have macros, and within classes colons indicate fields, not |
1654 | | // labels. |
1655 | 390k | if (!Style.isJavaScript() && !Style.isVerilog() && |
1656 | 390k | Tokens->peekNextToken()->is(tok::colon) && !Line->MustBeDeclaration) { |
1657 | 42.5k | nextToken(); |
1658 | 42.5k | Line->Tokens.begin()->Tok->MustBreakBefore = true; |
1659 | 42.5k | FormatTok->setFinalizedType(TT_GotoLabelColon); |
1660 | 42.5k | parseLabel(!Style.IndentGotoLabels); |
1661 | 42.5k | if (HasLabel) |
1662 | 33.5k | *HasLabel = true; |
1663 | 42.5k | return; |
1664 | 42.5k | } |
1665 | | // In all other cases, parse the declaration. |
1666 | 347k | break; |
1667 | 372k | default: |
1668 | 372k | break; |
1669 | 768k | } |
1670 | | |
1671 | 719k | const bool InRequiresExpression = |
1672 | 719k | OpeningBrace && OpeningBrace->is(TT_RequiresExpressionLBrace); |
1673 | 18.6M | do { |
1674 | 18.6M | const FormatToken *Previous = FormatTok->Previous; |
1675 | 18.6M | switch (FormatTok->Tok.getKind()) { |
1676 | 14.6k | case tok::at: |
1677 | 14.6k | nextToken(); |
1678 | 14.6k | if (FormatTok->is(tok::l_brace)) { |
1679 | 21 | nextToken(); |
1680 | 21 | parseBracedList(); |
1681 | 21 | break; |
1682 | 14.6k | } else if (Style.Language == FormatStyle::LK_Java && |
1683 | 14.6k | FormatTok->is(Keywords.kw_interface)) { |
1684 | 0 | nextToken(); |
1685 | 0 | break; |
1686 | 0 | } |
1687 | 14.6k | switch (FormatTok->Tok.getObjCKeywordID()) { |
1688 | 0 | case tok::objc_public: |
1689 | 0 | case tok::objc_protected: |
1690 | 0 | case tok::objc_package: |
1691 | 0 | case tok::objc_private: |
1692 | 0 | return parseAccessSpecifier(); |
1693 | 0 | case tok::objc_interface: |
1694 | 0 | case tok::objc_implementation: |
1695 | 0 | return parseObjCInterfaceOrImplementation(); |
1696 | 0 | case tok::objc_protocol: |
1697 | 0 | if (parseObjCProtocol()) |
1698 | 0 | return; |
1699 | 0 | break; |
1700 | 0 | case tok::objc_end: |
1701 | 0 | return; // Handled by the caller. |
1702 | 0 | case tok::objc_optional: |
1703 | 0 | case tok::objc_required: |
1704 | 0 | nextToken(); |
1705 | 0 | addUnwrappedLine(); |
1706 | 0 | return; |
1707 | 0 | case tok::objc_autoreleasepool: |
1708 | 0 | nextToken(); |
1709 | 0 | if (FormatTok->is(tok::l_brace)) { |
1710 | 0 | if (Style.BraceWrapping.AfterControlStatement == |
1711 | 0 | FormatStyle::BWACS_Always) { |
1712 | 0 | addUnwrappedLine(); |
1713 | 0 | } |
1714 | 0 | parseBlock(); |
1715 | 0 | } |
1716 | 0 | addUnwrappedLine(); |
1717 | 0 | return; |
1718 | 0 | case tok::objc_synchronized: |
1719 | 0 | nextToken(); |
1720 | 0 | if (FormatTok->is(tok::l_paren)) { |
1721 | | // Skip synchronization object |
1722 | 0 | parseParens(); |
1723 | 0 | } |
1724 | 0 | if (FormatTok->is(tok::l_brace)) { |
1725 | 0 | if (Style.BraceWrapping.AfterControlStatement == |
1726 | 0 | FormatStyle::BWACS_Always) { |
1727 | 0 | addUnwrappedLine(); |
1728 | 0 | } |
1729 | 0 | parseBlock(); |
1730 | 0 | } |
1731 | 0 | addUnwrappedLine(); |
1732 | 0 | return; |
1733 | 0 | case tok::objc_try: |
1734 | | // This branch isn't strictly necessary (the kw_try case below would |
1735 | | // do this too after the tok::at is parsed above). But be explicit. |
1736 | 0 | parseTryCatch(); |
1737 | 0 | return; |
1738 | 14.6k | default: |
1739 | 14.6k | break; |
1740 | 14.6k | } |
1741 | 14.6k | break; |
1742 | 14.6k | case tok::kw_requires: { |
1743 | 0 | if (Style.isCpp()) { |
1744 | 0 | bool ParsedClause = parseRequires(); |
1745 | 0 | if (ParsedClause) |
1746 | 0 | return; |
1747 | 0 | } else { |
1748 | 0 | nextToken(); |
1749 | 0 | } |
1750 | 0 | break; |
1751 | 0 | } |
1752 | 192 | case tok::kw_enum: |
1753 | | // Ignore if this is part of "template <enum ...". |
1754 | 192 | if (Previous && Previous->is(tok::less)) { |
1755 | 0 | nextToken(); |
1756 | 0 | break; |
1757 | 0 | } |
1758 | | |
1759 | | // parseEnum falls through and does not yet add an unwrapped line as an |
1760 | | // enum definition can start a structural element. |
1761 | 192 | if (!parseEnum()) |
1762 | 0 | break; |
1763 | | // This only applies to C++ and Verilog. |
1764 | 192 | if (!Style.isCpp() && !Style.isVerilog()) { |
1765 | 0 | addUnwrappedLine(); |
1766 | 0 | return; |
1767 | 0 | } |
1768 | 192 | break; |
1769 | 216 | case tok::kw_typedef: |
1770 | 216 | nextToken(); |
1771 | 216 | if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, |
1772 | 216 | Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS, |
1773 | 216 | Keywords.kw_CF_CLOSED_ENUM, |
1774 | 216 | Keywords.kw_NS_CLOSED_ENUM)) { |
1775 | 0 | parseEnum(); |
1776 | 0 | } |
1777 | 216 | break; |
1778 | 921 | case tok::kw_class: |
1779 | 921 | if (Style.isVerilog()) { |
1780 | 0 | parseBlock(); |
1781 | 0 | addUnwrappedLine(); |
1782 | 0 | return; |
1783 | 0 | } |
1784 | 921 | [[fallthrough]]; |
1785 | 1.02k | case tok::kw_struct: |
1786 | 1.02k | case tok::kw_union: |
1787 | 1.02k | if (parseStructLike()) |
1788 | 0 | return; |
1789 | 1.02k | break; |
1790 | 1.02k | case tok::kw_decltype: |
1791 | 0 | nextToken(); |
1792 | 0 | if (FormatTok->is(tok::l_paren)) { |
1793 | 0 | parseParens(); |
1794 | 0 | assert(FormatTok->Previous); |
1795 | 0 | if (FormatTok->Previous->endsSequence(tok::r_paren, tok::kw_auto, |
1796 | 0 | tok::l_paren)) { |
1797 | 0 | Line->SeenDecltypeAuto = true; |
1798 | 0 | } |
1799 | 0 | } |
1800 | 0 | break; |
1801 | 243k | case tok::period: |
1802 | 243k | nextToken(); |
1803 | | // In Java, classes have an implicit static member "class". |
1804 | 243k | if (Style.Language == FormatStyle::LK_Java && FormatTok && |
1805 | 243k | FormatTok->is(tok::kw_class)) { |
1806 | 0 | nextToken(); |
1807 | 0 | } |
1808 | 243k | if (Style.isJavaScript() && FormatTok && |
1809 | 243k | FormatTok->Tok.getIdentifierInfo()) { |
1810 | | // JavaScript only has pseudo keywords, all keywords are allowed to |
1811 | | // appear in "IdentifierName" positions. See http://es5.github.io/#x7.6 |
1812 | 0 | nextToken(); |
1813 | 0 | } |
1814 | 243k | break; |
1815 | 208k | case tok::semi: |
1816 | 208k | nextToken(); |
1817 | 208k | addUnwrappedLine(); |
1818 | 208k | return; |
1819 | 131k | case tok::r_brace: |
1820 | 131k | addUnwrappedLine(); |
1821 | 131k | return; |
1822 | 48.4k | case tok::l_paren: { |
1823 | 48.4k | parseParens(); |
1824 | | // Break the unwrapped line if a K&R C function definition has a parameter |
1825 | | // declaration. |
1826 | 48.4k | if (OpeningBrace || !Style.isCpp() || !Previous || eof()) |
1827 | 26.9k | break; |
1828 | 21.4k | if (isC78ParameterDecl(FormatTok, |
1829 | 21.4k | Tokens->peekNextToken(/*SkipComment=*/true), |
1830 | 21.4k | Previous)) { |
1831 | 0 | addUnwrappedLine(); |
1832 | 0 | return; |
1833 | 0 | } |
1834 | 21.4k | break; |
1835 | 21.4k | } |
1836 | 21.4k | case tok::kw_operator: |
1837 | 0 | nextToken(); |
1838 | 0 | if (FormatTok->isBinaryOperator()) |
1839 | 0 | nextToken(); |
1840 | 0 | break; |
1841 | 7.35k | case tok::caret: |
1842 | 7.35k | nextToken(); |
1843 | | // Block return type. |
1844 | 7.35k | if (FormatTok->Tok.isAnyIdentifier() || |
1845 | 7.35k | FormatTok->isSimpleTypeSpecifier()) { |
1846 | 936 | nextToken(); |
1847 | | // Return types: pointers are ok too. |
1848 | 936 | while (FormatTok->is(tok::star)) |
1849 | 0 | nextToken(); |
1850 | 936 | } |
1851 | | // Block argument list. |
1852 | 7.35k | if (FormatTok->is(tok::l_paren)) |
1853 | 3 | parseParens(); |
1854 | | // Block body. |
1855 | 7.35k | if (FormatTok->is(tok::l_brace)) |
1856 | 0 | parseChildBlock(); |
1857 | 7.35k | break; |
1858 | 407k | case tok::l_brace: |
1859 | 407k | if (InRequiresExpression) |
1860 | 0 | FormatTok->setFinalizedType(TT_BracedListLBrace); |
1861 | 407k | if (!tryToParsePropertyAccessor() && !tryToParseBracedList()) { |
1862 | 334k | IsDecltypeAutoFunction = Line->SeenDecltypeAuto; |
1863 | | // A block outside of parentheses must be the last part of a |
1864 | | // structural element. |
1865 | | // FIXME: Figure out cases where this is not true, and add projections |
1866 | | // for them (the one we know is missing are lambdas). |
1867 | 334k | if (Style.Language == FormatStyle::LK_Java && |
1868 | 334k | Line->Tokens.front().Tok->is(Keywords.kw_synchronized)) { |
1869 | | // If necessary, we could set the type to something different than |
1870 | | // TT_FunctionLBrace. |
1871 | 0 | if (Style.BraceWrapping.AfterControlStatement == |
1872 | 0 | FormatStyle::BWACS_Always) { |
1873 | 0 | addUnwrappedLine(); |
1874 | 0 | } |
1875 | 334k | } else if (Style.BraceWrapping.AfterFunction) { |
1876 | 0 | addUnwrappedLine(); |
1877 | 0 | } |
1878 | 334k | FormatTok->setFinalizedType(TT_FunctionLBrace); |
1879 | 334k | parseBlock(); |
1880 | 334k | IsDecltypeAutoFunction = false; |
1881 | 334k | addUnwrappedLine(); |
1882 | 334k | return; |
1883 | 334k | } |
1884 | | // Otherwise this was a braced init list, and the structural |
1885 | | // element continues. |
1886 | 72.6k | break; |
1887 | 72.6k | case tok::kw_try: |
1888 | 0 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
1889 | | // field/method declaration. |
1890 | 0 | nextToken(); |
1891 | 0 | break; |
1892 | 0 | } |
1893 | | // We arrive here when parsing function-try blocks. |
1894 | 0 | if (Style.BraceWrapping.AfterFunction) |
1895 | 0 | addUnwrappedLine(); |
1896 | 0 | parseTryCatch(); |
1897 | 0 | return; |
1898 | 2.58M | case tok::identifier: { |
1899 | 2.58M | if (Style.isCSharp() && FormatTok->is(Keywords.kw_where) && |
1900 | 2.58M | Line->MustBeDeclaration) { |
1901 | 0 | addUnwrappedLine(); |
1902 | 0 | parseCSharpGenericTypeConstraint(); |
1903 | 0 | break; |
1904 | 0 | } |
1905 | 2.58M | if (FormatTok->is(TT_MacroBlockEnd)) { |
1906 | 0 | addUnwrappedLine(); |
1907 | 0 | return; |
1908 | 0 | } |
1909 | | |
1910 | | // Function declarations (as opposed to function expressions) are parsed |
1911 | | // on their own unwrapped line by continuing this loop. Function |
1912 | | // expressions (functions that are not on their own line) must not create |
1913 | | // a new unwrapped line, so they are special cased below. |
1914 | 2.58M | size_t TokenCount = Line->Tokens.size(); |
1915 | 2.58M | if (Style.isJavaScript() && FormatTok->is(Keywords.kw_function) && |
1916 | 2.58M | (TokenCount > 1 || |
1917 | 0 | (TokenCount == 1 && |
1918 | 0 | Line->Tokens.front().Tok->isNot(Keywords.kw_async)))) { |
1919 | 0 | tryToParseJSFunction(); |
1920 | 0 | break; |
1921 | 0 | } |
1922 | 2.58M | if ((Style.isJavaScript() || Style.Language == FormatStyle::LK_Java) && |
1923 | 2.58M | FormatTok->is(Keywords.kw_interface)) { |
1924 | 0 | if (Style.isJavaScript()) { |
1925 | | // In JavaScript/TypeScript, "interface" can be used as a standalone |
1926 | | // identifier, e.g. in `var interface = 1;`. If "interface" is |
1927 | | // followed by another identifier, it is very like to be an actual |
1928 | | // interface declaration. |
1929 | 0 | unsigned StoredPosition = Tokens->getPosition(); |
1930 | 0 | FormatToken *Next = Tokens->getNextToken(); |
1931 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
1932 | 0 | if (!mustBeJSIdent(Keywords, Next)) { |
1933 | 0 | nextToken(); |
1934 | 0 | break; |
1935 | 0 | } |
1936 | 0 | } |
1937 | 0 | parseRecord(); |
1938 | 0 | addUnwrappedLine(); |
1939 | 0 | return; |
1940 | 0 | } |
1941 | | |
1942 | 2.58M | if (Style.isVerilog()) { |
1943 | 0 | if (FormatTok->is(Keywords.kw_table)) { |
1944 | 0 | parseVerilogTable(); |
1945 | 0 | return; |
1946 | 0 | } |
1947 | 0 | if (Keywords.isVerilogBegin(*FormatTok) || |
1948 | 0 | Keywords.isVerilogHierarchy(*FormatTok)) { |
1949 | 0 | parseBlock(); |
1950 | 0 | addUnwrappedLine(); |
1951 | 0 | return; |
1952 | 0 | } |
1953 | 0 | } |
1954 | | |
1955 | 2.58M | if (!Style.isCpp() && FormatTok->is(Keywords.kw_interface)) { |
1956 | 0 | if (parseStructLike()) |
1957 | 0 | return; |
1958 | 0 | break; |
1959 | 0 | } |
1960 | | |
1961 | 2.58M | if (Style.isCpp() && FormatTok->is(TT_StatementMacro)) { |
1962 | 0 | parseStatementMacro(); |
1963 | 0 | return; |
1964 | 0 | } |
1965 | | |
1966 | | // See if the following token should start a new unwrapped line. |
1967 | 2.58M | StringRef Text = FormatTok->TokenText; |
1968 | | |
1969 | 2.58M | FormatToken *PreviousToken = FormatTok; |
1970 | 2.58M | nextToken(); |
1971 | | |
1972 | | // JS doesn't have macros, and within classes colons indicate fields, not |
1973 | | // labels. |
1974 | 2.58M | if (Style.isJavaScript()) |
1975 | 0 | break; |
1976 | | |
1977 | 2.58M | auto OneTokenSoFar = [&]() { |
1978 | 2.58M | auto I = Line->Tokens.begin(), E = Line->Tokens.end(); |
1979 | 2.58M | while (I != E && I->Tok->is(tok::comment)) |
1980 | 9 | ++I; |
1981 | 2.58M | if (Style.isVerilog()) |
1982 | 0 | while (I != E && I->Tok->is(tok::hash)) |
1983 | 0 | ++I; |
1984 | 2.58M | return I != E && (++I == E); |
1985 | 2.58M | }; |
1986 | 2.58M | if (OneTokenSoFar()) { |
1987 | | // Recognize function-like macro usages without trailing semicolon as |
1988 | | // well as free-standing macros like Q_OBJECT. |
1989 | 346k | bool FunctionLike = FormatTok->is(tok::l_paren); |
1990 | 346k | if (FunctionLike) |
1991 | 4.56k | parseParens(); |
1992 | | |
1993 | 346k | bool FollowedByNewline = |
1994 | 346k | CommentsBeforeNextToken.empty() |
1995 | 346k | ? FormatTok->NewlinesBefore > 0 |
1996 | 346k | : CommentsBeforeNextToken.front()->NewlinesBefore > 0; |
1997 | | |
1998 | 346k | if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) && |
1999 | 346k | tokenCanStartNewLine(*FormatTok) && Text == Text.upper()) { |
2000 | 2.61k | if (PreviousToken->isNot(TT_UntouchableMacroFunc)) |
2001 | 2.61k | PreviousToken->setFinalizedType(TT_FunctionLikeOrFreestandingMacro); |
2002 | 2.61k | addUnwrappedLine(); |
2003 | 2.61k | return; |
2004 | 2.61k | } |
2005 | 346k | } |
2006 | 2.58M | break; |
2007 | 2.58M | } |
2008 | 2.58M | case tok::equal: |
2009 | 126k | if ((Style.isJavaScript() || Style.isCSharp()) && |
2010 | 126k | FormatTok->is(TT_FatArrow)) { |
2011 | 0 | tryToParseChildBlock(); |
2012 | 0 | break; |
2013 | 0 | } |
2014 | | |
2015 | 126k | nextToken(); |
2016 | 126k | if (FormatTok->is(tok::l_brace)) { |
2017 | | // Block kind should probably be set to BK_BracedInit for any language. |
2018 | | // C# needs this change to ensure that array initialisers and object |
2019 | | // initialisers are indented the same way. |
2020 | 192 | if (Style.isCSharp()) |
2021 | 0 | FormatTok->setBlockKind(BK_BracedInit); |
2022 | 192 | nextToken(); |
2023 | 192 | parseBracedList(); |
2024 | 125k | } else if (Style.Language == FormatStyle::LK_Proto && |
2025 | 125k | FormatTok->is(tok::less)) { |
2026 | 0 | nextToken(); |
2027 | 0 | parseBracedList(/*IsAngleBracket=*/true); |
2028 | 0 | } |
2029 | 126k | break; |
2030 | 25.1k | case tok::l_square: |
2031 | 25.1k | parseSquare(); |
2032 | 25.1k | break; |
2033 | 12 | case tok::kw_new: |
2034 | 12 | parseNew(); |
2035 | 12 | break; |
2036 | 6 | case tok::kw_case: |
2037 | | // Proto: there are no switch/case statements. |
2038 | 6 | if (Style.Language == FormatStyle::LK_Proto) { |
2039 | 0 | nextToken(); |
2040 | 0 | return; |
2041 | 0 | } |
2042 | | // In Verilog switch is called case. |
2043 | 6 | if (Style.isVerilog()) { |
2044 | 0 | parseBlock(); |
2045 | 0 | addUnwrappedLine(); |
2046 | 0 | return; |
2047 | 0 | } |
2048 | 6 | if (Style.isJavaScript() && Line->MustBeDeclaration) { |
2049 | | // 'case: string' field declaration. |
2050 | 0 | nextToken(); |
2051 | 0 | break; |
2052 | 0 | } |
2053 | 6 | parseCaseLabel(); |
2054 | 6 | break; |
2055 | 54 | case tok::kw_default: |
2056 | 54 | nextToken(); |
2057 | 54 | if (Style.isVerilog()) { |
2058 | 0 | if (FormatTok->is(tok::colon)) { |
2059 | | // The label will be handled in the next iteration. |
2060 | 0 | break; |
2061 | 0 | } |
2062 | 0 | if (FormatTok->is(Keywords.kw_clocking)) { |
2063 | | // A default clocking block. |
2064 | 0 | parseBlock(); |
2065 | 0 | addUnwrappedLine(); |
2066 | 0 | return; |
2067 | 0 | } |
2068 | 0 | parseVerilogCaseLabel(); |
2069 | 0 | return; |
2070 | 0 | } |
2071 | 54 | break; |
2072 | 235k | case tok::colon: |
2073 | 235k | nextToken(); |
2074 | 235k | if (Style.isVerilog()) { |
2075 | 0 | parseVerilogCaseLabel(); |
2076 | 0 | return; |
2077 | 0 | } |
2078 | 235k | break; |
2079 | 14.5M | default: |
2080 | 14.5M | nextToken(); |
2081 | 14.5M | break; |
2082 | 18.6M | } |
2083 | 18.6M | } while (!eof()); |
2084 | 719k | } |
2085 | | |
2086 | 407k | bool UnwrappedLineParser::tryToParsePropertyAccessor() { |
2087 | 407k | assert(FormatTok->is(tok::l_brace)); |
2088 | 407k | if (!Style.isCSharp()) |
2089 | 407k | return false; |
2090 | | // See if it's a property accessor. |
2091 | 0 | if (FormatTok->Previous->isNot(tok::identifier)) |
2092 | 0 | return false; |
2093 | | |
2094 | | // See if we are inside a property accessor. |
2095 | | // |
2096 | | // Record the current tokenPosition so that we can advance and |
2097 | | // reset the current token. `Next` is not set yet so we need |
2098 | | // another way to advance along the token stream. |
2099 | 0 | unsigned int StoredPosition = Tokens->getPosition(); |
2100 | 0 | FormatToken *Tok = Tokens->getNextToken(); |
2101 | | |
2102 | | // A trivial property accessor is of the form: |
2103 | | // { [ACCESS_SPECIFIER] [get]; [ACCESS_SPECIFIER] [set|init] } |
2104 | | // Track these as they do not require line breaks to be introduced. |
2105 | 0 | bool HasSpecialAccessor = false; |
2106 | 0 | bool IsTrivialPropertyAccessor = true; |
2107 | 0 | while (!eof()) { |
2108 | 0 | if (Tok->isOneOf(tok::semi, tok::kw_public, tok::kw_private, |
2109 | 0 | tok::kw_protected, Keywords.kw_internal, Keywords.kw_get, |
2110 | 0 | Keywords.kw_init, Keywords.kw_set)) { |
2111 | 0 | if (Tok->isOneOf(Keywords.kw_get, Keywords.kw_init, Keywords.kw_set)) |
2112 | 0 | HasSpecialAccessor = true; |
2113 | 0 | Tok = Tokens->getNextToken(); |
2114 | 0 | continue; |
2115 | 0 | } |
2116 | 0 | if (Tok->isNot(tok::r_brace)) |
2117 | 0 | IsTrivialPropertyAccessor = false; |
2118 | 0 | break; |
2119 | 0 | } |
2120 | |
|
2121 | 0 | if (!HasSpecialAccessor) { |
2122 | 0 | Tokens->setPosition(StoredPosition); |
2123 | 0 | return false; |
2124 | 0 | } |
2125 | | |
2126 | | // Try to parse the property accessor: |
2127 | | // https://docs.microsoft.com/en-us/dotnet/csharp/programming-guide/classes-and-structs/properties |
2128 | 0 | Tokens->setPosition(StoredPosition); |
2129 | 0 | if (!IsTrivialPropertyAccessor && Style.BraceWrapping.AfterFunction) |
2130 | 0 | addUnwrappedLine(); |
2131 | 0 | nextToken(); |
2132 | 0 | do { |
2133 | 0 | switch (FormatTok->Tok.getKind()) { |
2134 | 0 | case tok::r_brace: |
2135 | 0 | nextToken(); |
2136 | 0 | if (FormatTok->is(tok::equal)) { |
2137 | 0 | while (!eof() && FormatTok->isNot(tok::semi)) |
2138 | 0 | nextToken(); |
2139 | 0 | nextToken(); |
2140 | 0 | } |
2141 | 0 | addUnwrappedLine(); |
2142 | 0 | return true; |
2143 | 0 | case tok::l_brace: |
2144 | 0 | ++Line->Level; |
2145 | 0 | parseBlock(/*MustBeDeclaration=*/true); |
2146 | 0 | addUnwrappedLine(); |
2147 | 0 | --Line->Level; |
2148 | 0 | break; |
2149 | 0 | case tok::equal: |
2150 | 0 | if (FormatTok->is(TT_FatArrow)) { |
2151 | 0 | ++Line->Level; |
2152 | 0 | do { |
2153 | 0 | nextToken(); |
2154 | 0 | } while (!eof() && FormatTok->isNot(tok::semi)); |
2155 | 0 | nextToken(); |
2156 | 0 | addUnwrappedLine(); |
2157 | 0 | --Line->Level; |
2158 | 0 | break; |
2159 | 0 | } |
2160 | 0 | nextToken(); |
2161 | 0 | break; |
2162 | 0 | default: |
2163 | 0 | if (FormatTok->isOneOf(Keywords.kw_get, Keywords.kw_init, |
2164 | 0 | Keywords.kw_set) && |
2165 | 0 | !IsTrivialPropertyAccessor) { |
2166 | | // Non-trivial get/set needs to be on its own line. |
2167 | 0 | addUnwrappedLine(); |
2168 | 0 | } |
2169 | 0 | nextToken(); |
2170 | 0 | } |
2171 | 0 | } while (!eof()); |
2172 | | |
2173 | | // Unreachable for well-formed code (paired '{' and '}'). |
2174 | 0 | return true; |
2175 | 0 | } |
2176 | | |
2177 | 571k | bool UnwrappedLineParser::tryToParseLambda() { |
2178 | 571k | assert(FormatTok->is(tok::l_square)); |
2179 | 571k | if (!Style.isCpp()) { |
2180 | 0 | nextToken(); |
2181 | 0 | return false; |
2182 | 0 | } |
2183 | 571k | FormatToken &LSquare = *FormatTok; |
2184 | 571k | if (!tryToParseLambdaIntroducer()) |
2185 | 458k | return false; |
2186 | | |
2187 | 112k | bool SeenArrow = false; |
2188 | 112k | bool InTemplateParameterList = false; |
2189 | | |
2190 | 389k | while (FormatTok->isNot(tok::l_brace)) { |
2191 | 389k | if (FormatTok->isSimpleTypeSpecifier()) { |
2192 | 0 | nextToken(); |
2193 | 0 | continue; |
2194 | 0 | } |
2195 | 389k | switch (FormatTok->Tok.getKind()) { |
2196 | 0 | case tok::l_brace: |
2197 | 0 | break; |
2198 | 279 | case tok::l_paren: |
2199 | 279 | parseParens(/*AmpAmpTokenType=*/TT_PointerOrReference); |
2200 | 279 | break; |
2201 | 294 | case tok::l_square: |
2202 | 294 | parseSquare(); |
2203 | 294 | break; |
2204 | 82.4k | case tok::less: |
2205 | 82.4k | assert(FormatTok->Previous); |
2206 | 82.4k | if (FormatTok->Previous->is(tok::r_square)) |
2207 | 723 | InTemplateParameterList = true; |
2208 | 82.4k | nextToken(); |
2209 | 82.4k | break; |
2210 | 0 | case tok::kw_auto: |
2211 | 0 | case tok::kw_class: |
2212 | 0 | case tok::kw_template: |
2213 | 0 | case tok::kw_typename: |
2214 | 690 | case tok::amp: |
2215 | 768 | case tok::star: |
2216 | 768 | case tok::kw_const: |
2217 | 768 | case tok::kw_constexpr: |
2218 | 768 | case tok::kw_consteval: |
2219 | 1.52k | case tok::comma: |
2220 | 92.1k | case tok::greater: |
2221 | 188k | case tok::identifier: |
2222 | 189k | case tok::numeric_constant: |
2223 | 189k | case tok::coloncolon: |
2224 | 189k | case tok::kw_mutable: |
2225 | 189k | case tok::kw_noexcept: |
2226 | 189k | case tok::kw_static: |
2227 | 189k | nextToken(); |
2228 | 189k | break; |
2229 | | // Specialization of a template with an integer parameter can contain |
2230 | | // arithmetic, logical, comparison and ternary operators. |
2231 | | // |
2232 | | // FIXME: This also accepts sequences of operators that are not in the scope |
2233 | | // of a template argument list. |
2234 | | // |
2235 | | // In a C++ lambda a template type can only occur after an arrow. We use |
2236 | | // this as an heuristic to distinguish between Objective-C expressions |
2237 | | // followed by an `a->b` expression, such as: |
2238 | | // ([obj func:arg] + a->b) |
2239 | | // Otherwise the code below would parse as a lambda. |
2240 | 141 | case tok::plus: |
2241 | 993 | case tok::minus: |
2242 | 2.04k | case tok::exclaim: |
2243 | 2.05k | case tok::tilde: |
2244 | 2.52k | case tok::slash: |
2245 | 2.61k | case tok::percent: |
2246 | 2.62k | case tok::lessless: |
2247 | 3.66k | case tok::pipe: |
2248 | 3.66k | case tok::pipepipe: |
2249 | 3.66k | case tok::ampamp: |
2250 | 3.66k | case tok::caret: |
2251 | 3.66k | case tok::equalequal: |
2252 | 3.66k | case tok::exclaimequal: |
2253 | 3.70k | case tok::greaterequal: |
2254 | 3.70k | case tok::lessequal: |
2255 | 6.40k | case tok::question: |
2256 | 10.0k | case tok::colon: |
2257 | 10.0k | case tok::ellipsis: |
2258 | 10.0k | case tok::kw_true: |
2259 | 10.0k | case tok::kw_false: |
2260 | 10.0k | if (SeenArrow || InTemplateParameterList) { |
2261 | 3.20k | nextToken(); |
2262 | 3.20k | break; |
2263 | 3.20k | } |
2264 | 6.88k | return true; |
2265 | 969 | case tok::arrow: |
2266 | | // This might or might not actually be a lambda arrow (this could be an |
2267 | | // ObjC method invocation followed by a dereferencing arrow). We might |
2268 | | // reset this back to TT_Unknown in TokenAnnotator. |
2269 | 969 | FormatTok->setFinalizedType(TT_TrailingReturnArrow); |
2270 | 969 | SeenArrow = true; |
2271 | 969 | nextToken(); |
2272 | 969 | break; |
2273 | 0 | case tok::kw_requires: { |
2274 | 0 | auto *RequiresToken = FormatTok; |
2275 | 0 | nextToken(); |
2276 | 0 | parseRequiresClause(RequiresToken); |
2277 | 0 | break; |
2278 | 10.0k | } |
2279 | 1.02k | case tok::equal: |
2280 | 1.02k | if (!InTemplateParameterList) |
2281 | 984 | return true; |
2282 | 39 | nextToken(); |
2283 | 39 | break; |
2284 | 104k | default: |
2285 | 104k | return true; |
2286 | 389k | } |
2287 | 389k | } |
2288 | | |
2289 | 30 | FormatTok->setFinalizedType(TT_LambdaLBrace); |
2290 | 30 | LSquare.setFinalizedType(TT_LambdaLSquare); |
2291 | | |
2292 | 30 | NestedLambdas.push_back(Line->SeenDecltypeAuto); |
2293 | 30 | parseChildBlock(); |
2294 | 30 | assert(!NestedLambdas.empty()); |
2295 | 0 | NestedLambdas.pop_back(); |
2296 | | |
2297 | 30 | return true; |
2298 | 112k | } |
2299 | | |
2300 | 571k | bool UnwrappedLineParser::tryToParseLambdaIntroducer() { |
2301 | 571k | const FormatToken *Previous = FormatTok->Previous; |
2302 | 571k | const FormatToken *LeftSquare = FormatTok; |
2303 | 571k | nextToken(); |
2304 | 571k | if ((Previous && ((Previous->Tok.getIdentifierInfo() && |
2305 | 569k | !Previous->isOneOf(tok::kw_return, tok::kw_co_await, |
2306 | 50.7k | tok::kw_co_yield, tok::kw_co_return)) || |
2307 | 569k | Previous->closesScope())) || |
2308 | 571k | LeftSquare->isCppStructuredBinding(Style)) { |
2309 | 52.5k | return false; |
2310 | 52.5k | } |
2311 | 518k | if (FormatTok->is(tok::l_square) || tok::isLiteral(FormatTok->Tok.getKind())) |
2312 | 406k | return false; |
2313 | 112k | if (FormatTok->is(tok::r_square)) { |
2314 | 3.87k | const FormatToken *Next = Tokens->peekNextToken(/*SkipComment=*/true); |
2315 | 3.87k | if (Next->is(tok::greater)) |
2316 | 9 | return false; |
2317 | 3.87k | } |
2318 | 112k | parseSquare(/*LambdaIntroducer=*/true); |
2319 | 112k | return true; |
2320 | 112k | } |
2321 | | |
2322 | 0 | void UnwrappedLineParser::tryToParseJSFunction() { |
2323 | 0 | assert(FormatTok->is(Keywords.kw_function)); |
2324 | 0 | if (FormatTok->is(Keywords.kw_async)) |
2325 | 0 | nextToken(); |
2326 | | // Consume "function". |
2327 | 0 | nextToken(); |
2328 | | |
2329 | | // Consume * (generator function). Treat it like C++'s overloaded operators. |
2330 | 0 | if (FormatTok->is(tok::star)) { |
2331 | 0 | FormatTok->setFinalizedType(TT_OverloadedOperator); |
2332 | 0 | nextToken(); |
2333 | 0 | } |
2334 | | |
2335 | | // Consume function name. |
2336 | 0 | if (FormatTok->is(tok::identifier)) |
2337 | 0 | nextToken(); |
2338 | |
|
2339 | 0 | if (FormatTok->isNot(tok::l_paren)) |
2340 | 0 | return; |
2341 | | |
2342 | | // Parse formal parameter list. |
2343 | 0 | parseParens(); |
2344 | |
|
2345 | 0 | if (FormatTok->is(tok::colon)) { |
2346 | | // Parse a type definition. |
2347 | 0 | nextToken(); |
2348 | | |
2349 | | // Eat the type declaration. For braced inline object types, balance braces, |
2350 | | // otherwise just parse until finding an l_brace for the function body. |
2351 | 0 | if (FormatTok->is(tok::l_brace)) |
2352 | 0 | tryToParseBracedList(); |
2353 | 0 | else |
2354 | 0 | while (!FormatTok->isOneOf(tok::l_brace, tok::semi) && !eof()) |
2355 | 0 | nextToken(); |
2356 | 0 | } |
2357 | |
|
2358 | 0 | if (FormatTok->is(tok::semi)) |
2359 | 0 | return; |
2360 | | |
2361 | 0 | parseChildBlock(); |
2362 | 0 | } |
2363 | | |
2364 | 684k | bool UnwrappedLineParser::tryToParseBracedList() { |
2365 | 684k | if (FormatTok->is(BK_Unknown)) |
2366 | 254k | calculateBraceTypes(); |
2367 | 684k | assert(FormatTok->isNot(BK_Unknown)); |
2368 | 684k | if (FormatTok->is(BK_Block)) |
2369 | 532k | return false; |
2370 | 152k | nextToken(); |
2371 | 152k | parseBracedList(); |
2372 | 152k | return true; |
2373 | 684k | } |
2374 | | |
2375 | 0 | bool UnwrappedLineParser::tryToParseChildBlock() { |
2376 | 0 | assert(Style.isJavaScript() || Style.isCSharp()); |
2377 | 0 | assert(FormatTok->is(TT_FatArrow)); |
2378 | | // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType TT_FatArrow. |
2379 | | // They always start an expression or a child block if followed by a curly |
2380 | | // brace. |
2381 | 0 | nextToken(); |
2382 | 0 | if (FormatTok->isNot(tok::l_brace)) |
2383 | 0 | return false; |
2384 | 0 | parseChildBlock(); |
2385 | 0 | return true; |
2386 | 0 | } |
2387 | | |
2388 | 175k | bool UnwrappedLineParser::parseBracedList(bool IsAngleBracket, bool IsEnum) { |
2389 | 175k | bool HasError = false; |
2390 | | |
2391 | | // FIXME: Once we have an expression parser in the UnwrappedLineParser, |
2392 | | // replace this by using parseAssignmentExpression() inside. |
2393 | 1.00M | do { |
2394 | 1.00M | if (Style.isCSharp() && FormatTok->is(TT_FatArrow) && |
2395 | 1.00M | tryToParseChildBlock()) { |
2396 | 0 | continue; |
2397 | 0 | } |
2398 | 1.00M | if (Style.isJavaScript()) { |
2399 | 0 | if (FormatTok->is(Keywords.kw_function)) { |
2400 | 0 | tryToParseJSFunction(); |
2401 | 0 | continue; |
2402 | 0 | } |
2403 | 0 | if (FormatTok->is(tok::l_brace)) { |
2404 | | // Could be a method inside of a braced list `{a() { return 1; }}`. |
2405 | 0 | if (tryToParseBracedList()) |
2406 | 0 | continue; |
2407 | 0 | parseChildBlock(); |
2408 | 0 | } |
2409 | 0 | } |
2410 | 1.00M | if (FormatTok->is(IsAngleBracket ? tok::greater : tok::r_brace)) { |
2411 | 160k | if (IsEnum && !Style.AllowShortEnumsOnASingleLine) |
2412 | 0 | addUnwrappedLine(); |
2413 | 160k | nextToken(); |
2414 | 160k | return !HasError; |
2415 | 160k | } |
2416 | 842k | switch (FormatTok->Tok.getKind()) { |
2417 | 3.30k | case tok::l_square: |
2418 | 3.30k | if (Style.isCSharp()) |
2419 | 0 | parseSquare(); |
2420 | 3.30k | else |
2421 | 3.30k | tryToParseLambda(); |
2422 | 3.30k | break; |
2423 | 1.74k | case tok::l_paren: |
2424 | 1.74k | parseParens(); |
2425 | | // JavaScript can just have free standing methods and getters/setters in |
2426 | | // object literals. Detect them by a "{" following ")". |
2427 | 1.74k | if (Style.isJavaScript()) { |
2428 | 0 | if (FormatTok->is(tok::l_brace)) |
2429 | 0 | parseChildBlock(); |
2430 | 0 | break; |
2431 | 0 | } |
2432 | 1.74k | break; |
2433 | 18.8k | case tok::l_brace: |
2434 | | // Assume there are no blocks inside a braced init list apart |
2435 | | // from the ones we explicitly parse out (like lambdas). |
2436 | 18.8k | FormatTok->setBlockKind(BK_BracedInit); |
2437 | 18.8k | nextToken(); |
2438 | 18.8k | parseBracedList(); |
2439 | 18.8k | break; |
2440 | 13.1k | case tok::less: |
2441 | 13.1k | nextToken(); |
2442 | 13.1k | if (IsAngleBracket) |
2443 | 0 | parseBracedList(/*IsAngleBracket=*/true); |
2444 | 13.1k | break; |
2445 | 14.8k | case tok::semi: |
2446 | | // JavaScript (or more precisely TypeScript) can have semicolons in braced |
2447 | | // lists (in so-called TypeMemberLists). Thus, the semicolon cannot be |
2448 | | // used for error recovery if we have otherwise determined that this is |
2449 | | // a braced list. |
2450 | 14.8k | if (Style.isJavaScript()) { |
2451 | 0 | nextToken(); |
2452 | 0 | break; |
2453 | 0 | } |
2454 | 14.8k | HasError = true; |
2455 | 14.8k | if (!IsEnum) |
2456 | 14.6k | return false; |
2457 | 225 | nextToken(); |
2458 | 225 | break; |
2459 | 83.3k | case tok::comma: |
2460 | 83.3k | nextToken(); |
2461 | 83.3k | if (IsEnum && !Style.AllowShortEnumsOnASingleLine) |
2462 | 0 | addUnwrappedLine(); |
2463 | 83.3k | break; |
2464 | 707k | default: |
2465 | 707k | nextToken(); |
2466 | 707k | break; |
2467 | 842k | } |
2468 | 842k | } while (!eof()); |
2469 | 117 | return false; |
2470 | 175k | } |
2471 | | |
2472 | | /// \brief Parses a pair of parentheses (and everything between them). |
2473 | | /// \param AmpAmpTokenType If different than TT_Unknown sets this type for all |
2474 | | /// double ampersands. This applies for all nested scopes as well. |
2475 | | /// |
2476 | | /// Returns whether there is a `=` token between the parentheses. |
2477 | 444k | bool UnwrappedLineParser::parseParens(TokenType AmpAmpTokenType) { |
2478 | 444k | assert(FormatTok->is(tok::l_paren) && "'(' expected."); |
2479 | 0 | auto *LeftParen = FormatTok; |
2480 | 444k | bool SeenEqual = false; |
2481 | 444k | const bool MightBeStmtExpr = Tokens->peekNextToken()->is(tok::l_brace); |
2482 | 444k | nextToken(); |
2483 | 16.2M | do { |
2484 | 16.2M | switch (FormatTok->Tok.getKind()) { |
2485 | 357k | case tok::l_paren: |
2486 | 357k | if (parseParens(AmpAmpTokenType)) |
2487 | 190k | SeenEqual = true; |
2488 | 357k | if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) |
2489 | 0 | parseChildBlock(); |
2490 | 357k | break; |
2491 | 165k | case tok::r_paren: |
2492 | 165k | if (!MightBeStmtExpr && |
2493 | 165k | Style.RemoveParentheses > FormatStyle::RPS_Leave) { |
2494 | 0 | const auto *Prev = LeftParen->Previous; |
2495 | 0 | const auto *Next = Tokens->peekNextToken(); |
2496 | 0 | const bool DoubleParens = |
2497 | 0 | Prev && Prev->is(tok::l_paren) && Next && Next->is(tok::r_paren); |
2498 | 0 | const auto *PrevPrev = Prev ? Prev->getPreviousNonComment() : nullptr; |
2499 | 0 | const bool Blacklisted = |
2500 | 0 | PrevPrev && |
2501 | 0 | (PrevPrev->isOneOf(tok::kw___attribute, tok::kw_decltype) || |
2502 | 0 | (SeenEqual && |
2503 | 0 | (PrevPrev->isOneOf(tok::kw_if, tok::kw_while) || |
2504 | 0 | PrevPrev->endsSequence(tok::kw_constexpr, tok::kw_if)))); |
2505 | 0 | const bool ReturnParens = |
2506 | 0 | Style.RemoveParentheses == FormatStyle::RPS_ReturnStatement && |
2507 | 0 | ((NestedLambdas.empty() && !IsDecltypeAutoFunction) || |
2508 | 0 | (!NestedLambdas.empty() && !NestedLambdas.back())) && |
2509 | 0 | Prev && Prev->isOneOf(tok::kw_return, tok::kw_co_return) && Next && |
2510 | 0 | Next->is(tok::semi); |
2511 | 0 | if ((DoubleParens && !Blacklisted) || ReturnParens) { |
2512 | 0 | LeftParen->Optional = true; |
2513 | 0 | FormatTok->Optional = true; |
2514 | 0 | } |
2515 | 0 | } |
2516 | 165k | nextToken(); |
2517 | 165k | return SeenEqual; |
2518 | 125k | case tok::r_brace: |
2519 | | // A "}" inside parenthesis is an error if there wasn't a matching "{". |
2520 | 125k | return SeenEqual; |
2521 | 75.9k | case tok::l_square: |
2522 | 75.9k | tryToParseLambda(); |
2523 | 75.9k | break; |
2524 | 15.5k | case tok::l_brace: |
2525 | 15.5k | if (!tryToParseBracedList()) |
2526 | 14.1k | parseChildBlock(); |
2527 | 15.5k | break; |
2528 | 29.8k | case tok::at: |
2529 | 29.8k | nextToken(); |
2530 | 29.8k | if (FormatTok->is(tok::l_brace)) { |
2531 | 1.76k | nextToken(); |
2532 | 1.76k | parseBracedList(); |
2533 | 1.76k | } |
2534 | 29.8k | break; |
2535 | 99.1k | case tok::equal: |
2536 | 99.1k | SeenEqual = true; |
2537 | 99.1k | if (Style.isCSharp() && FormatTok->is(TT_FatArrow)) |
2538 | 0 | tryToParseChildBlock(); |
2539 | 99.1k | else |
2540 | 99.1k | nextToken(); |
2541 | 99.1k | break; |
2542 | 24 | case tok::kw_class: |
2543 | 24 | if (Style.isJavaScript()) |
2544 | 0 | parseRecord(/*ParseAsExpr=*/true); |
2545 | 24 | else |
2546 | 24 | nextToken(); |
2547 | 24 | break; |
2548 | 1.95M | case tok::identifier: |
2549 | 1.95M | if (Style.isJavaScript() && (FormatTok->is(Keywords.kw_function))) |
2550 | 0 | tryToParseJSFunction(); |
2551 | 1.95M | else |
2552 | 1.95M | nextToken(); |
2553 | 1.95M | break; |
2554 | 0 | case tok::kw_requires: { |
2555 | 0 | auto RequiresToken = FormatTok; |
2556 | 0 | nextToken(); |
2557 | 0 | parseRequiresExpression(RequiresToken); |
2558 | 0 | break; |
2559 | 0 | } |
2560 | 3.49k | case tok::ampamp: |
2561 | 3.49k | if (AmpAmpTokenType != TT_Unknown) |
2562 | 0 | FormatTok->setFinalizedType(AmpAmpTokenType); |
2563 | 3.49k | [[fallthrough]]; |
2564 | 13.4M | default: |
2565 | 13.4M | nextToken(); |
2566 | 13.4M | break; |
2567 | 16.2M | } |
2568 | 16.2M | } while (!eof()); |
2569 | 153k | return SeenEqual; |
2570 | 444k | } |
2571 | | |
2572 | 604k | void UnwrappedLineParser::parseSquare(bool LambdaIntroducer) { |
2573 | 604k | if (!LambdaIntroducer) { |
2574 | 492k | assert(FormatTok->is(tok::l_square) && "'[' expected."); |
2575 | 492k | if (tryToParseLambda()) |
2576 | 99.4k | return; |
2577 | 492k | } |
2578 | 14.7M | do { |
2579 | 14.7M | switch (FormatTok->Tok.getKind()) { |
2580 | 26.1k | case tok::l_paren: |
2581 | 26.1k | parseParens(); |
2582 | 26.1k | break; |
2583 | 47.9k | case tok::r_square: |
2584 | 47.9k | nextToken(); |
2585 | 47.9k | return; |
2586 | 26.5k | case tok::r_brace: |
2587 | | // A "}" inside parenthesis is an error if there wasn't a matching "{". |
2588 | 26.5k | return; |
2589 | 466k | case tok::l_square: |
2590 | 466k | parseSquare(); |
2591 | 466k | break; |
2592 | 84.2k | case tok::l_brace: { |
2593 | 84.2k | if (!tryToParseBracedList()) |
2594 | 7.58k | parseChildBlock(); |
2595 | 84.2k | break; |
2596 | 0 | } |
2597 | 31.5k | case tok::at: |
2598 | 31.5k | nextToken(); |
2599 | 31.5k | if (FormatTok->is(tok::l_brace)) { |
2600 | 1.88k | nextToken(); |
2601 | 1.88k | parseBracedList(); |
2602 | 1.88k | } |
2603 | 31.5k | break; |
2604 | 14.0M | default: |
2605 | 14.0M | nextToken(); |
2606 | 14.0M | break; |
2607 | 14.7M | } |
2608 | 14.7M | } while (!eof()); |
2609 | 505k | } |
2610 | | |
2611 | 6.39k | void UnwrappedLineParser::keepAncestorBraces() { |
2612 | 6.39k | if (!Style.RemoveBracesLLVM) |
2613 | 6.39k | return; |
2614 | | |
2615 | 0 | const int MaxNestingLevels = 2; |
2616 | 0 | const int Size = NestedTooDeep.size(); |
2617 | 0 | if (Size >= MaxNestingLevels) |
2618 | 0 | NestedTooDeep[Size - MaxNestingLevels] = true; |
2619 | 0 | NestedTooDeep.push_back(false); |
2620 | 0 | } |
2621 | | |
2622 | 0 | static FormatToken *getLastNonComment(const UnwrappedLine &Line) { |
2623 | 0 | for (const auto &Token : llvm::reverse(Line.Tokens)) |
2624 | 0 | if (Token.Tok->isNot(tok::comment)) |
2625 | 0 | return Token.Tok; |
2626 | | |
2627 | 0 | return nullptr; |
2628 | 0 | } |
2629 | | |
2630 | 3.46k | void UnwrappedLineParser::parseUnbracedBody(bool CheckEOF) { |
2631 | 3.46k | FormatToken *Tok = nullptr; |
2632 | | |
2633 | 3.46k | if (Style.InsertBraces && !Line->InPPDirective && !Line->Tokens.empty() && |
2634 | 3.46k | PreprocessorDirectives.empty() && FormatTok->isNot(tok::semi)) { |
2635 | 0 | Tok = Style.BraceWrapping.AfterControlStatement == FormatStyle::BWACS_Never |
2636 | 0 | ? getLastNonComment(*Line) |
2637 | 0 | : Line->Tokens.back().Tok; |
2638 | 0 | assert(Tok); |
2639 | 0 | if (Tok->BraceCount < 0) { |
2640 | 0 | assert(Tok->BraceCount == -1); |
2641 | 0 | Tok = nullptr; |
2642 | 0 | } else { |
2643 | 0 | Tok->BraceCount = -1; |
2644 | 0 | } |
2645 | 0 | } |
2646 | | |
2647 | 0 | addUnwrappedLine(); |
2648 | 3.46k | ++Line->Level; |
2649 | 3.46k | parseStructuralElement(); |
2650 | | |
2651 | 3.46k | if (Tok) { |
2652 | 0 | assert(!Line->InPPDirective); |
2653 | 0 | Tok = nullptr; |
2654 | 0 | for (const auto &L : llvm::reverse(*CurrentLines)) { |
2655 | 0 | if (!L.InPPDirective && getLastNonComment(L)) { |
2656 | 0 | Tok = L.Tokens.back().Tok; |
2657 | 0 | break; |
2658 | 0 | } |
2659 | 0 | } |
2660 | 0 | assert(Tok); |
2661 | 0 | ++Tok->BraceCount; |
2662 | 0 | } |
2663 | | |
2664 | 3.46k | if (CheckEOF && eof()) |
2665 | 0 | addUnwrappedLine(); |
2666 | | |
2667 | 3.46k | --Line->Level; |
2668 | 3.46k | } |
2669 | | |
2670 | 0 | static void markOptionalBraces(FormatToken *LeftBrace) { |
2671 | 0 | if (!LeftBrace) |
2672 | 0 | return; |
2673 | | |
2674 | 0 | assert(LeftBrace->is(tok::l_brace)); |
2675 | | |
2676 | 0 | FormatToken *RightBrace = LeftBrace->MatchingParen; |
2677 | 0 | if (!RightBrace) { |
2678 | 0 | assert(!LeftBrace->Optional); |
2679 | 0 | return; |
2680 | 0 | } |
2681 | | |
2682 | 0 | assert(RightBrace->is(tok::r_brace)); |
2683 | 0 | assert(RightBrace->MatchingParen == LeftBrace); |
2684 | 0 | assert(LeftBrace->Optional == RightBrace->Optional); |
2685 | | |
2686 | 0 | LeftBrace->Optional = true; |
2687 | 0 | RightBrace->Optional = true; |
2688 | 0 | } |
2689 | | |
2690 | 6.90k | void UnwrappedLineParser::handleAttributes() { |
2691 | | // Handle AttributeMacro, e.g. `if (x) UNLIKELY`. |
2692 | 6.90k | if (FormatTok->isAttribute()) |
2693 | 0 | nextToken(); |
2694 | 6.90k | else if (FormatTok->is(tok::l_square)) |
2695 | 0 | handleCppAttributes(); |
2696 | 6.90k | } |
2697 | | |
2698 | 2.77k | bool UnwrappedLineParser::handleCppAttributes() { |
2699 | | // Handle [[likely]] / [[unlikely]] attributes. |
2700 | 2.77k | assert(FormatTok->is(tok::l_square)); |
2701 | 2.77k | if (!tryToParseSimpleAttribute()) |
2702 | 2.77k | return false; |
2703 | 0 | parseSquare(); |
2704 | 0 | return true; |
2705 | 2.77k | } |
2706 | | |
2707 | | /// Returns whether \c Tok begins a block. |
2708 | 6.96k | bool UnwrappedLineParser::isBlockBegin(const FormatToken &Tok) const { |
2709 | | // FIXME: rename the function or make |
2710 | | // Tok.isOneOf(tok::l_brace, TT_MacroBlockBegin) work. |
2711 | 6.96k | return Style.isVerilog() ? Keywords.isVerilogBegin(Tok) |
2712 | 6.96k | : Tok.is(tok::l_brace); |
2713 | 6.96k | } |
2714 | | |
2715 | | FormatToken *UnwrappedLineParser::parseIfThenElse(IfStmtKind *IfKind, |
2716 | | bool KeepBraces, |
2717 | 5.94k | bool IsVerilogAssert) { |
2718 | 5.94k | assert((FormatTok->is(tok::kw_if) || |
2719 | 5.94k | (Style.isVerilog() && |
2720 | 5.94k | FormatTok->isOneOf(tok::kw_restrict, Keywords.kw_assert, |
2721 | 5.94k | Keywords.kw_assume, Keywords.kw_cover))) && |
2722 | 5.94k | "'if' expected"); |
2723 | 0 | nextToken(); |
2724 | | |
2725 | 5.94k | if (IsVerilogAssert) { |
2726 | | // Handle `assert #0` and `assert final`. |
2727 | 0 | if (FormatTok->is(Keywords.kw_verilogHash)) { |
2728 | 0 | nextToken(); |
2729 | 0 | if (FormatTok->is(tok::numeric_constant)) |
2730 | 0 | nextToken(); |
2731 | 0 | } else if (FormatTok->isOneOf(Keywords.kw_final, Keywords.kw_property, |
2732 | 0 | Keywords.kw_sequence)) { |
2733 | 0 | nextToken(); |
2734 | 0 | } |
2735 | 0 | } |
2736 | | |
2737 | | // Handle `if !consteval`. |
2738 | 5.94k | if (FormatTok->is(tok::exclaim)) |
2739 | 0 | nextToken(); |
2740 | | |
2741 | 5.94k | bool KeepIfBraces = true; |
2742 | 5.94k | if (FormatTok->is(tok::kw_consteval)) { |
2743 | 0 | nextToken(); |
2744 | 5.94k | } else { |
2745 | 5.94k | KeepIfBraces = !Style.RemoveBracesLLVM || KeepBraces; |
2746 | 5.94k | if (FormatTok->isOneOf(tok::kw_constexpr, tok::identifier)) |
2747 | 348 | nextToken(); |
2748 | 5.94k | if (FormatTok->is(tok::l_paren)) { |
2749 | 5.51k | FormatTok->setFinalizedType(TT_ConditionLParen); |
2750 | 5.51k | parseParens(); |
2751 | 5.51k | } |
2752 | 5.94k | } |
2753 | 5.94k | handleAttributes(); |
2754 | | // The then action is optional in Verilog assert statements. |
2755 | 5.94k | if (IsVerilogAssert && FormatTok->is(tok::semi)) { |
2756 | 0 | nextToken(); |
2757 | 0 | addUnwrappedLine(); |
2758 | 0 | return nullptr; |
2759 | 0 | } |
2760 | | |
2761 | 5.94k | bool NeedsUnwrappedLine = false; |
2762 | 5.94k | keepAncestorBraces(); |
2763 | | |
2764 | 5.94k | FormatToken *IfLeftBrace = nullptr; |
2765 | 5.94k | IfStmtKind IfBlockKind = IfStmtKind::NotIf; |
2766 | | |
2767 | 5.94k | if (isBlockBegin(*FormatTok)) { |
2768 | 2.63k | FormatTok->setFinalizedType(TT_ControlStatementLBrace); |
2769 | 2.63k | IfLeftBrace = FormatTok; |
2770 | 2.63k | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
2771 | 2.63k | parseBlock(/*MustBeDeclaration=*/false, /*AddLevels=*/1u, |
2772 | 2.63k | /*MunchSemi=*/true, KeepIfBraces, &IfBlockKind); |
2773 | 2.63k | setPreviousRBraceType(TT_ControlStatementRBrace); |
2774 | 2.63k | if (Style.BraceWrapping.BeforeElse) |
2775 | 0 | addUnwrappedLine(); |
2776 | 2.63k | else |
2777 | 2.63k | NeedsUnwrappedLine = true; |
2778 | 3.31k | } else if (IsVerilogAssert && FormatTok->is(tok::kw_else)) { |
2779 | 0 | addUnwrappedLine(); |
2780 | 3.31k | } else { |
2781 | 3.31k | parseUnbracedBody(); |
2782 | 3.31k | } |
2783 | | |
2784 | 5.94k | if (Style.RemoveBracesLLVM) { |
2785 | 0 | assert(!NestedTooDeep.empty()); |
2786 | 0 | KeepIfBraces = KeepIfBraces || |
2787 | 0 | (IfLeftBrace && !IfLeftBrace->MatchingParen) || |
2788 | 0 | NestedTooDeep.back() || IfBlockKind == IfStmtKind::IfOnly || |
2789 | 0 | IfBlockKind == IfStmtKind::IfElseIf; |
2790 | 0 | } |
2791 | | |
2792 | 0 | bool KeepElseBraces = KeepIfBraces; |
2793 | 5.94k | FormatToken *ElseLeftBrace = nullptr; |
2794 | 5.94k | IfStmtKind Kind = IfStmtKind::IfOnly; |
2795 | | |
2796 | 5.94k | if (FormatTok->is(tok::kw_else)) { |
2797 | 573 | if (Style.RemoveBracesLLVM) { |
2798 | 0 | NestedTooDeep.back() = false; |
2799 | 0 | Kind = IfStmtKind::IfElse; |
2800 | 0 | } |
2801 | 573 | nextToken(); |
2802 | 573 | handleAttributes(); |
2803 | 573 | if (isBlockBegin(*FormatTok)) { |
2804 | 366 | const bool FollowedByIf = Tokens->peekNextToken()->is(tok::kw_if); |
2805 | 366 | FormatTok->setFinalizedType(TT_ElseLBrace); |
2806 | 366 | ElseLeftBrace = FormatTok; |
2807 | 366 | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
2808 | 366 | IfStmtKind ElseBlockKind = IfStmtKind::NotIf; |
2809 | 366 | FormatToken *IfLBrace = |
2810 | 366 | parseBlock(/*MustBeDeclaration=*/false, /*AddLevels=*/1u, |
2811 | 366 | /*MunchSemi=*/true, KeepElseBraces, &ElseBlockKind); |
2812 | 366 | setPreviousRBraceType(TT_ElseRBrace); |
2813 | 366 | if (FormatTok->is(tok::kw_else)) { |
2814 | 0 | KeepElseBraces = KeepElseBraces || |
2815 | 0 | ElseBlockKind == IfStmtKind::IfOnly || |
2816 | 0 | ElseBlockKind == IfStmtKind::IfElseIf; |
2817 | 366 | } else if (FollowedByIf && IfLBrace && !IfLBrace->Optional) { |
2818 | 0 | KeepElseBraces = true; |
2819 | 0 | assert(ElseLeftBrace->MatchingParen); |
2820 | 0 | markOptionalBraces(ElseLeftBrace); |
2821 | 0 | } |
2822 | 0 | addUnwrappedLine(); |
2823 | 366 | } else if (!IsVerilogAssert && FormatTok->is(tok::kw_if)) { |
2824 | 183 | const FormatToken *Previous = Tokens->getPreviousToken(); |
2825 | 183 | assert(Previous); |
2826 | 0 | const bool IsPrecededByComment = Previous->is(tok::comment); |
2827 | 183 | if (IsPrecededByComment) { |
2828 | 0 | addUnwrappedLine(); |
2829 | 0 | ++Line->Level; |
2830 | 0 | } |
2831 | 183 | bool TooDeep = true; |
2832 | 183 | if (Style.RemoveBracesLLVM) { |
2833 | 0 | Kind = IfStmtKind::IfElseIf; |
2834 | 0 | TooDeep = NestedTooDeep.pop_back_val(); |
2835 | 0 | } |
2836 | 183 | ElseLeftBrace = parseIfThenElse(/*IfKind=*/nullptr, KeepIfBraces); |
2837 | 183 | if (Style.RemoveBracesLLVM) |
2838 | 0 | NestedTooDeep.push_back(TooDeep); |
2839 | 183 | if (IsPrecededByComment) |
2840 | 0 | --Line->Level; |
2841 | 183 | } else { |
2842 | 24 | parseUnbracedBody(/*CheckEOF=*/true); |
2843 | 24 | } |
2844 | 5.37k | } else { |
2845 | 5.37k | KeepIfBraces = KeepIfBraces || IfBlockKind == IfStmtKind::IfElse; |
2846 | 5.37k | if (NeedsUnwrappedLine) |
2847 | 2.08k | addUnwrappedLine(); |
2848 | 5.37k | } |
2849 | | |
2850 | 5.94k | if (!Style.RemoveBracesLLVM) |
2851 | 5.94k | return nullptr; |
2852 | | |
2853 | 0 | assert(!NestedTooDeep.empty()); |
2854 | 0 | KeepElseBraces = KeepElseBraces || |
2855 | 0 | (ElseLeftBrace && !ElseLeftBrace->MatchingParen) || |
2856 | 0 | NestedTooDeep.back(); |
2857 | |
|
2858 | 0 | NestedTooDeep.pop_back(); |
2859 | |
|
2860 | 0 | if (!KeepIfBraces && !KeepElseBraces) { |
2861 | 0 | markOptionalBraces(IfLeftBrace); |
2862 | 0 | markOptionalBraces(ElseLeftBrace); |
2863 | 0 | } else if (IfLeftBrace) { |
2864 | 0 | FormatToken *IfRightBrace = IfLeftBrace->MatchingParen; |
2865 | 0 | if (IfRightBrace) { |
2866 | 0 | assert(IfRightBrace->MatchingParen == IfLeftBrace); |
2867 | 0 | assert(!IfLeftBrace->Optional); |
2868 | 0 | assert(!IfRightBrace->Optional); |
2869 | 0 | IfLeftBrace->MatchingParen = nullptr; |
2870 | 0 | IfRightBrace->MatchingParen = nullptr; |
2871 | 0 | } |
2872 | 0 | } |
2873 | | |
2874 | 0 | if (IfKind) |
2875 | 0 | *IfKind = Kind; |
2876 | |
|
2877 | 0 | return IfLeftBrace; |
2878 | 5.94k | } |
2879 | | |
2880 | 0 | void UnwrappedLineParser::parseTryCatch() { |
2881 | 0 | assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); |
2882 | 0 | nextToken(); |
2883 | 0 | bool NeedsUnwrappedLine = false; |
2884 | 0 | if (FormatTok->is(tok::colon)) { |
2885 | | // We are in a function try block, what comes is an initializer list. |
2886 | 0 | nextToken(); |
2887 | | |
2888 | | // In case identifiers were removed by clang-tidy, what might follow is |
2889 | | // multiple commas in sequence - before the first identifier. |
2890 | 0 | while (FormatTok->is(tok::comma)) |
2891 | 0 | nextToken(); |
2892 | |
|
2893 | 0 | while (FormatTok->is(tok::identifier)) { |
2894 | 0 | nextToken(); |
2895 | 0 | if (FormatTok->is(tok::l_paren)) |
2896 | 0 | parseParens(); |
2897 | 0 | if (FormatTok->Previous && FormatTok->Previous->is(tok::identifier) && |
2898 | 0 | FormatTok->is(tok::l_brace)) { |
2899 | 0 | do { |
2900 | 0 | nextToken(); |
2901 | 0 | } while (FormatTok->isNot(tok::r_brace)); |
2902 | 0 | nextToken(); |
2903 | 0 | } |
2904 | | |
2905 | | // In case identifiers were removed by clang-tidy, what might follow is |
2906 | | // multiple commas in sequence - after the first identifier. |
2907 | 0 | while (FormatTok->is(tok::comma)) |
2908 | 0 | nextToken(); |
2909 | 0 | } |
2910 | 0 | } |
2911 | | // Parse try with resource. |
2912 | 0 | if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) |
2913 | 0 | parseParens(); |
2914 | |
|
2915 | 0 | keepAncestorBraces(); |
2916 | |
|
2917 | 0 | if (FormatTok->is(tok::l_brace)) { |
2918 | 0 | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
2919 | 0 | parseBlock(); |
2920 | 0 | if (Style.BraceWrapping.BeforeCatch) |
2921 | 0 | addUnwrappedLine(); |
2922 | 0 | else |
2923 | 0 | NeedsUnwrappedLine = true; |
2924 | 0 | } else if (FormatTok->isNot(tok::kw_catch)) { |
2925 | | // The C++ standard requires a compound-statement after a try. |
2926 | | // If there's none, we try to assume there's a structuralElement |
2927 | | // and try to continue. |
2928 | 0 | addUnwrappedLine(); |
2929 | 0 | ++Line->Level; |
2930 | 0 | parseStructuralElement(); |
2931 | 0 | --Line->Level; |
2932 | 0 | } |
2933 | 0 | while (true) { |
2934 | 0 | if (FormatTok->is(tok::at)) |
2935 | 0 | nextToken(); |
2936 | 0 | if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, |
2937 | 0 | tok::kw___finally) || |
2938 | 0 | ((Style.Language == FormatStyle::LK_Java || Style.isJavaScript()) && |
2939 | 0 | FormatTok->is(Keywords.kw_finally)) || |
2940 | 0 | (FormatTok->isObjCAtKeyword(tok::objc_catch) || |
2941 | 0 | FormatTok->isObjCAtKeyword(tok::objc_finally)))) { |
2942 | 0 | break; |
2943 | 0 | } |
2944 | 0 | nextToken(); |
2945 | 0 | while (FormatTok->isNot(tok::l_brace)) { |
2946 | 0 | if (FormatTok->is(tok::l_paren)) { |
2947 | 0 | parseParens(); |
2948 | 0 | continue; |
2949 | 0 | } |
2950 | 0 | if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) { |
2951 | 0 | if (Style.RemoveBracesLLVM) |
2952 | 0 | NestedTooDeep.pop_back(); |
2953 | 0 | return; |
2954 | 0 | } |
2955 | 0 | nextToken(); |
2956 | 0 | } |
2957 | 0 | NeedsUnwrappedLine = false; |
2958 | 0 | Line->MustBeDeclaration = false; |
2959 | 0 | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
2960 | 0 | parseBlock(); |
2961 | 0 | if (Style.BraceWrapping.BeforeCatch) |
2962 | 0 | addUnwrappedLine(); |
2963 | 0 | else |
2964 | 0 | NeedsUnwrappedLine = true; |
2965 | 0 | } |
2966 | | |
2967 | 0 | if (Style.RemoveBracesLLVM) |
2968 | 0 | NestedTooDeep.pop_back(); |
2969 | |
|
2970 | 0 | if (NeedsUnwrappedLine) |
2971 | 0 | addUnwrappedLine(); |
2972 | 0 | } |
2973 | | |
2974 | 0 | void UnwrappedLineParser::parseNamespace() { |
2975 | 0 | assert(FormatTok->isOneOf(tok::kw_namespace, TT_NamespaceMacro) && |
2976 | 0 | "'namespace' expected"); |
2977 | | |
2978 | 0 | const FormatToken &InitialToken = *FormatTok; |
2979 | 0 | nextToken(); |
2980 | 0 | if (InitialToken.is(TT_NamespaceMacro)) { |
2981 | 0 | parseParens(); |
2982 | 0 | } else { |
2983 | 0 | while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::kw_inline, |
2984 | 0 | tok::l_square, tok::period, tok::l_paren) || |
2985 | 0 | (Style.isCSharp() && FormatTok->is(tok::kw_union))) { |
2986 | 0 | if (FormatTok->is(tok::l_square)) |
2987 | 0 | parseSquare(); |
2988 | 0 | else if (FormatTok->is(tok::l_paren)) |
2989 | 0 | parseParens(); |
2990 | 0 | else |
2991 | 0 | nextToken(); |
2992 | 0 | } |
2993 | 0 | } |
2994 | 0 | if (FormatTok->is(tok::l_brace)) { |
2995 | 0 | FormatTok->setFinalizedType(TT_NamespaceLBrace); |
2996 | |
|
2997 | 0 | if (ShouldBreakBeforeBrace(Style, InitialToken)) |
2998 | 0 | addUnwrappedLine(); |
2999 | |
|
3000 | 0 | unsigned AddLevels = |
3001 | 0 | Style.NamespaceIndentation == FormatStyle::NI_All || |
3002 | 0 | (Style.NamespaceIndentation == FormatStyle::NI_Inner && |
3003 | 0 | DeclarationScopeStack.size() > 1) |
3004 | 0 | ? 1u |
3005 | 0 | : 0u; |
3006 | 0 | bool ManageWhitesmithsBraces = |
3007 | 0 | AddLevels == 0u && |
3008 | 0 | Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths; |
3009 | | |
3010 | | // If we're in Whitesmiths mode, indent the brace if we're not indenting |
3011 | | // the whole block. |
3012 | 0 | if (ManageWhitesmithsBraces) |
3013 | 0 | ++Line->Level; |
3014 | | |
3015 | | // Munch the semicolon after a namespace. This is more common than one would |
3016 | | // think. Putting the semicolon into its own line is very ugly. |
3017 | 0 | parseBlock(/*MustBeDeclaration=*/true, AddLevels, /*MunchSemi=*/true, |
3018 | 0 | /*KeepBraces=*/true, /*IfKind=*/nullptr, |
3019 | 0 | ManageWhitesmithsBraces); |
3020 | |
|
3021 | 0 | addUnwrappedLine(AddLevels > 0 ? LineLevel::Remove : LineLevel::Keep); |
3022 | |
|
3023 | 0 | if (ManageWhitesmithsBraces) |
3024 | 0 | --Line->Level; |
3025 | 0 | } |
3026 | | // FIXME: Add error handling. |
3027 | 0 | } |
3028 | | |
3029 | 12 | void UnwrappedLineParser::parseNew() { |
3030 | 12 | assert(FormatTok->is(tok::kw_new) && "'new' expected"); |
3031 | 0 | nextToken(); |
3032 | | |
3033 | 12 | if (Style.isCSharp()) { |
3034 | 0 | do { |
3035 | | // Handle constructor invocation, e.g. `new(field: value)`. |
3036 | 0 | if (FormatTok->is(tok::l_paren)) |
3037 | 0 | parseParens(); |
3038 | | |
3039 | | // Handle array initialization syntax, e.g. `new[] {10, 20, 30}`. |
3040 | 0 | if (FormatTok->is(tok::l_brace)) |
3041 | 0 | parseBracedList(); |
3042 | |
|
3043 | 0 | if (FormatTok->isOneOf(tok::semi, tok::comma)) |
3044 | 0 | return; |
3045 | | |
3046 | 0 | nextToken(); |
3047 | 0 | } while (!eof()); |
3048 | 0 | } |
3049 | | |
3050 | 12 | if (Style.Language != FormatStyle::LK_Java) |
3051 | 12 | return; |
3052 | | |
3053 | | // In Java, we can parse everything up to the parens, which aren't optional. |
3054 | 0 | do { |
3055 | | // There should not be a ;, { or } before the new's open paren. |
3056 | 0 | if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) |
3057 | 0 | return; |
3058 | | |
3059 | | // Consume the parens. |
3060 | 0 | if (FormatTok->is(tok::l_paren)) { |
3061 | 0 | parseParens(); |
3062 | | |
3063 | | // If there is a class body of an anonymous class, consume that as child. |
3064 | 0 | if (FormatTok->is(tok::l_brace)) |
3065 | 0 | parseChildBlock(); |
3066 | 0 | return; |
3067 | 0 | } |
3068 | 0 | nextToken(); |
3069 | 0 | } while (!eof()); |
3070 | 0 | } |
3071 | | |
3072 | 444 | void UnwrappedLineParser::parseLoopBody(bool KeepBraces, bool WrapRightBrace) { |
3073 | 444 | keepAncestorBraces(); |
3074 | | |
3075 | 444 | if (isBlockBegin(*FormatTok)) { |
3076 | 318 | FormatTok->setFinalizedType(TT_ControlStatementLBrace); |
3077 | 318 | FormatToken *LeftBrace = FormatTok; |
3078 | 318 | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
3079 | 318 | parseBlock(/*MustBeDeclaration=*/false, /*AddLevels=*/1u, |
3080 | 318 | /*MunchSemi=*/true, KeepBraces); |
3081 | 318 | setPreviousRBraceType(TT_ControlStatementRBrace); |
3082 | 318 | if (!KeepBraces) { |
3083 | 0 | assert(!NestedTooDeep.empty()); |
3084 | 0 | if (!NestedTooDeep.back()) |
3085 | 0 | markOptionalBraces(LeftBrace); |
3086 | 0 | } |
3087 | 318 | if (WrapRightBrace) |
3088 | 252 | addUnwrappedLine(); |
3089 | 318 | } else { |
3090 | 126 | parseUnbracedBody(); |
3091 | 126 | } |
3092 | | |
3093 | 444 | if (!KeepBraces) |
3094 | 0 | NestedTooDeep.pop_back(); |
3095 | 444 | } |
3096 | | |
3097 | 381 | void UnwrappedLineParser::parseForOrWhileLoop(bool HasParens) { |
3098 | 381 | assert((FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) || |
3099 | 381 | (Style.isVerilog() && |
3100 | 381 | FormatTok->isOneOf(Keywords.kw_always, Keywords.kw_always_comb, |
3101 | 381 | Keywords.kw_always_ff, Keywords.kw_always_latch, |
3102 | 381 | Keywords.kw_final, Keywords.kw_initial, |
3103 | 381 | Keywords.kw_foreach, Keywords.kw_forever, |
3104 | 381 | Keywords.kw_repeat))) && |
3105 | 381 | "'for', 'while' or foreach macro expected"); |
3106 | 381 | const bool KeepBraces = !Style.RemoveBracesLLVM || |
3107 | 381 | !FormatTok->isOneOf(tok::kw_for, tok::kw_while); |
3108 | | |
3109 | 381 | nextToken(); |
3110 | | // JS' for await ( ... |
3111 | 381 | if (Style.isJavaScript() && FormatTok->is(Keywords.kw_await)) |
3112 | 0 | nextToken(); |
3113 | 381 | if (Style.isCpp() && FormatTok->is(tok::kw_co_await)) |
3114 | 0 | nextToken(); |
3115 | 381 | if (HasParens && FormatTok->is(tok::l_paren)) { |
3116 | | // The type is only set for Verilog basically because we were afraid to |
3117 | | // change the existing behavior for loops. See the discussion on D121756 for |
3118 | | // details. |
3119 | 381 | if (Style.isVerilog()) |
3120 | 0 | FormatTok->setFinalizedType(TT_ConditionLParen); |
3121 | 381 | parseParens(); |
3122 | 381 | } |
3123 | | |
3124 | 381 | if (Style.isVerilog()) { |
3125 | | // Event control. |
3126 | 0 | parseVerilogSensitivityList(); |
3127 | 381 | } else if (Style.AllowShortLoopsOnASingleLine && FormatTok->is(tok::semi) && |
3128 | 381 | Tokens->getPreviousToken()->is(tok::r_paren)) { |
3129 | 3 | nextToken(); |
3130 | 3 | addUnwrappedLine(); |
3131 | 3 | return; |
3132 | 3 | } |
3133 | | |
3134 | 378 | handleAttributes(); |
3135 | 378 | parseLoopBody(KeepBraces, /*WrapRightBrace=*/true); |
3136 | 378 | } |
3137 | | |
3138 | 66 | void UnwrappedLineParser::parseDoWhile() { |
3139 | 66 | assert(FormatTok->is(tok::kw_do) && "'do' expected"); |
3140 | 0 | nextToken(); |
3141 | | |
3142 | 66 | parseLoopBody(/*KeepBraces=*/true, Style.BraceWrapping.BeforeWhile); |
3143 | | |
3144 | | // FIXME: Add error handling. |
3145 | 66 | if (FormatTok->isNot(tok::kw_while)) { |
3146 | 24 | addUnwrappedLine(); |
3147 | 24 | return; |
3148 | 24 | } |
3149 | | |
3150 | 42 | FormatTok->setFinalizedType(TT_DoWhile); |
3151 | | |
3152 | | // If in Whitesmiths mode, the line with the while() needs to be indented |
3153 | | // to the same level as the block. |
3154 | 42 | if (Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths) |
3155 | 0 | ++Line->Level; |
3156 | | |
3157 | 42 | nextToken(); |
3158 | 42 | parseStructuralElement(); |
3159 | 42 | } |
3160 | | |
3161 | 42.6k | void UnwrappedLineParser::parseLabel(bool LeftAlignLabel) { |
3162 | 42.6k | nextToken(); |
3163 | 42.6k | unsigned OldLineLevel = Line->Level; |
3164 | 42.6k | if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) |
3165 | 42.6k | --Line->Level; |
3166 | 42.6k | if (LeftAlignLabel) |
3167 | 0 | Line->Level = 0; |
3168 | | |
3169 | 42.6k | if (!Style.IndentCaseBlocks && CommentsBeforeNextToken.empty() && |
3170 | 42.6k | FormatTok->is(tok::l_brace)) { |
3171 | | |
3172 | 25.7k | CompoundStatementIndenter Indenter(this, Line->Level, |
3173 | 25.7k | Style.BraceWrapping.AfterCaseLabel, |
3174 | 25.7k | Style.BraceWrapping.IndentBraces); |
3175 | 25.7k | parseBlock(); |
3176 | 25.7k | if (FormatTok->is(tok::kw_break)) { |
3177 | 0 | if (Style.BraceWrapping.AfterControlStatement == |
3178 | 0 | FormatStyle::BWACS_Always) { |
3179 | 0 | addUnwrappedLine(); |
3180 | 0 | if (!Style.IndentCaseBlocks && |
3181 | 0 | Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths) { |
3182 | 0 | ++Line->Level; |
3183 | 0 | } |
3184 | 0 | } |
3185 | 0 | parseStructuralElement(); |
3186 | 0 | } |
3187 | 25.7k | addUnwrappedLine(); |
3188 | 25.7k | } else { |
3189 | 16.8k | if (FormatTok->is(tok::semi)) |
3190 | 0 | nextToken(); |
3191 | 16.8k | addUnwrappedLine(); |
3192 | 16.8k | } |
3193 | 42.6k | Line->Level = OldLineLevel; |
3194 | 42.6k | if (FormatTok->isNot(tok::l_brace)) { |
3195 | 42.6k | parseStructuralElement(); |
3196 | 42.6k | addUnwrappedLine(); |
3197 | 42.6k | } |
3198 | 42.6k | } |
3199 | | |
3200 | 6 | void UnwrappedLineParser::parseCaseLabel() { |
3201 | 6 | assert(FormatTok->is(tok::kw_case) && "'case' expected"); |
3202 | | |
3203 | | // FIXME: fix handling of complex expressions here. |
3204 | 1.20k | do { |
3205 | 1.20k | nextToken(); |
3206 | 1.20k | if (FormatTok->is(tok::colon)) { |
3207 | 6 | FormatTok->setFinalizedType(TT_CaseLabelColon); |
3208 | 6 | break; |
3209 | 6 | } |
3210 | 1.20k | } while (!eof()); |
3211 | 0 | parseLabel(); |
3212 | 6 | } |
3213 | | |
3214 | 0 | void UnwrappedLineParser::parseSwitch() { |
3215 | 0 | assert(FormatTok->is(tok::kw_switch) && "'switch' expected"); |
3216 | 0 | nextToken(); |
3217 | 0 | if (FormatTok->is(tok::l_paren)) |
3218 | 0 | parseParens(); |
3219 | |
|
3220 | 0 | keepAncestorBraces(); |
3221 | |
|
3222 | 0 | if (FormatTok->is(tok::l_brace)) { |
3223 | 0 | CompoundStatementIndenter Indenter(this, Style, Line->Level); |
3224 | 0 | FormatTok->setFinalizedType(TT_ControlStatementLBrace); |
3225 | 0 | parseBlock(); |
3226 | 0 | setPreviousRBraceType(TT_ControlStatementRBrace); |
3227 | 0 | addUnwrappedLine(); |
3228 | 0 | } else { |
3229 | 0 | addUnwrappedLine(); |
3230 | 0 | ++Line->Level; |
3231 | 0 | parseStructuralElement(); |
3232 | 0 | --Line->Level; |
3233 | 0 | } |
3234 | |
|
3235 | 0 | if (Style.RemoveBracesLLVM) |
3236 | 0 | NestedTooDeep.pop_back(); |
3237 | 0 | } |
3238 | | |
3239 | | // Operators that can follow a C variable. |
3240 | 0 | static bool isCOperatorFollowingVar(tok::TokenKind kind) { |
3241 | 0 | switch (kind) { |
3242 | 0 | case tok::ampamp: |
3243 | 0 | case tok::ampequal: |
3244 | 0 | case tok::arrow: |
3245 | 0 | case tok::caret: |
3246 | 0 | case tok::caretequal: |
3247 | 0 | case tok::comma: |
3248 | 0 | case tok::ellipsis: |
3249 | 0 | case tok::equal: |
3250 | 0 | case tok::equalequal: |
3251 | 0 | case tok::exclaim: |
3252 | 0 | case tok::exclaimequal: |
3253 | 0 | case tok::greater: |
3254 | 0 | case tok::greaterequal: |
3255 | 0 | case tok::greatergreater: |
3256 | 0 | case tok::greatergreaterequal: |
3257 | 0 | case tok::l_paren: |
3258 | 0 | case tok::l_square: |
3259 | 0 | case tok::less: |
3260 | 0 | case tok::lessequal: |
3261 | 0 | case tok::lessless: |
3262 | 0 | case tok::lesslessequal: |
3263 | 0 | case tok::minus: |
3264 | 0 | case tok::minusequal: |
3265 | 0 | case tok::minusminus: |
3266 | 0 | case tok::percent: |
3267 | 0 | case tok::percentequal: |
3268 | 0 | case tok::period: |
3269 | 0 | case tok::pipe: |
3270 | 0 | case tok::pipeequal: |
3271 | 0 | case tok::pipepipe: |
3272 | 0 | case tok::plus: |
3273 | 0 | case tok::plusequal: |
3274 | 0 | case tok::plusplus: |
3275 | 0 | case tok::question: |
3276 | 0 | case tok::r_brace: |
3277 | 0 | case tok::r_paren: |
3278 | 0 | case tok::r_square: |
3279 | 0 | case tok::semi: |
3280 | 0 | case tok::slash: |
3281 | 0 | case tok::slashequal: |
3282 | 0 | case tok::star: |
3283 | 0 | case tok::starequal: |
3284 | 0 | return true; |
3285 | 0 | default: |
3286 | 0 | return false; |
3287 | 0 | } |
3288 | 0 | } |
3289 | | |
3290 | 0 | void UnwrappedLineParser::parseAccessSpecifier() { |
3291 | 0 | FormatToken *AccessSpecifierCandidate = FormatTok; |
3292 | 0 | nextToken(); |
3293 | | // Understand Qt's slots. |
3294 | 0 | if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) |
3295 | 0 | nextToken(); |
3296 | | // Otherwise, we don't know what it is, and we'd better keep the next token. |
3297 | 0 | if (FormatTok->is(tok::colon)) { |
3298 | 0 | nextToken(); |
3299 | 0 | addUnwrappedLine(); |
3300 | 0 | } else if (FormatTok->isNot(tok::coloncolon) && |
3301 | 0 | !isCOperatorFollowingVar(FormatTok->Tok.getKind())) { |
3302 | | // Not a variable name nor namespace name. |
3303 | 0 | addUnwrappedLine(); |
3304 | 0 | } else if (AccessSpecifierCandidate) { |
3305 | | // Consider the access specifier to be a C identifier. |
3306 | 0 | AccessSpecifierCandidate->Tok.setKind(tok::identifier); |
3307 | 0 | } |
3308 | 0 | } |
3309 | | |
3310 | | /// \brief Parses a requires, decides if it is a clause or an expression. |
3311 | | /// \pre The current token has to be the requires keyword. |
3312 | | /// \returns true if it parsed a clause. |
3313 | 0 | bool clang::format::UnwrappedLineParser::parseRequires() { |
3314 | 0 | assert(FormatTok->is(tok::kw_requires) && "'requires' expected"); |
3315 | 0 | auto RequiresToken = FormatTok; |
3316 | | |
3317 | | // We try to guess if it is a requires clause, or a requires expression. For |
3318 | | // that we first consume the keyword and check the next token. |
3319 | 0 | nextToken(); |
3320 | |
|
3321 | 0 | switch (FormatTok->Tok.getKind()) { |
3322 | 0 | case tok::l_brace: |
3323 | | // This can only be an expression, never a clause. |
3324 | 0 | parseRequiresExpression(RequiresToken); |
3325 | 0 | return false; |
3326 | 0 | case tok::l_paren: |
3327 | | // Clauses and expression can start with a paren, it's unclear what we have. |
3328 | 0 | break; |
3329 | 0 | default: |
3330 | | // All other tokens can only be a clause. |
3331 | 0 | parseRequiresClause(RequiresToken); |
3332 | 0 | return true; |
3333 | 0 | } |
3334 | | |
3335 | | // Looking forward we would have to decide if there are function declaration |
3336 | | // like arguments to the requires expression: |
3337 | | // requires (T t) { |
3338 | | // Or there is a constraint expression for the requires clause: |
3339 | | // requires (C<T> && ... |
3340 | | |
3341 | | // But first let's look behind. |
3342 | 0 | auto *PreviousNonComment = RequiresToken->getPreviousNonComment(); |
3343 | |
|
3344 | 0 | if (!PreviousNonComment || |
3345 | 0 | PreviousNonComment->is(TT_RequiresExpressionLBrace)) { |
3346 | | // If there is no token, or an expression left brace, we are a requires |
3347 | | // clause within a requires expression. |
3348 | 0 | parseRequiresClause(RequiresToken); |
3349 | 0 | return true; |
3350 | 0 | } |
3351 | | |
3352 | 0 | switch (PreviousNonComment->Tok.getKind()) { |
3353 | 0 | case tok::greater: |
3354 | 0 | case tok::r_paren: |
3355 | 0 | case tok::kw_noexcept: |
3356 | 0 | case tok::kw_const: |
3357 | | // This is a requires clause. |
3358 | 0 | parseRequiresClause(RequiresToken); |
3359 | 0 | return true; |
3360 | 0 | case tok::amp: |
3361 | 0 | case tok::ampamp: { |
3362 | | // This can be either: |
3363 | | // if (... && requires (T t) ...) |
3364 | | // Or |
3365 | | // void member(...) && requires (C<T> ... |
3366 | | // We check the one token before that for a const: |
3367 | | // void member(...) const && requires (C<T> ... |
3368 | 0 | auto PrevPrev = PreviousNonComment->getPreviousNonComment(); |
3369 | 0 | if (PrevPrev && PrevPrev->is(tok::kw_const)) { |
3370 | 0 | parseRequiresClause(RequiresToken); |
3371 | 0 | return true; |
3372 | 0 | } |
3373 | 0 | break; |
3374 | 0 | } |
3375 | 0 | default: |
3376 | 0 | if (PreviousNonComment->isTypeOrIdentifier()) { |
3377 | | // This is a requires clause. |
3378 | 0 | parseRequiresClause(RequiresToken); |
3379 | 0 | return true; |
3380 | 0 | } |
3381 | | // It's an expression. |
3382 | 0 | parseRequiresExpression(RequiresToken); |
3383 | 0 | return false; |
3384 | 0 | } |
3385 | | |
3386 | | // Now we look forward and try to check if the paren content is a parameter |
3387 | | // list. The parameters can be cv-qualified and contain references or |
3388 | | // pointers. |
3389 | | // So we want basically to check for TYPE NAME, but TYPE can contain all kinds |
3390 | | // of stuff: typename, const, *, &, &&, ::, identifiers. |
3391 | | |
3392 | 0 | unsigned StoredPosition = Tokens->getPosition(); |
3393 | 0 | FormatToken *NextToken = Tokens->getNextToken(); |
3394 | 0 | int Lookahead = 0; |
3395 | 0 | auto PeekNext = [&Lookahead, &NextToken, this] { |
3396 | 0 | ++Lookahead; |
3397 | 0 | NextToken = Tokens->getNextToken(); |
3398 | 0 | }; |
3399 | |
|
3400 | 0 | bool FoundType = false; |
3401 | 0 | bool LastWasColonColon = false; |
3402 | 0 | int OpenAngles = 0; |
3403 | |
|
3404 | 0 | for (; Lookahead < 50; PeekNext()) { |
3405 | 0 | switch (NextToken->Tok.getKind()) { |
3406 | 0 | case tok::kw_volatile: |
3407 | 0 | case tok::kw_const: |
3408 | 0 | case tok::comma: |
3409 | 0 | if (OpenAngles == 0) { |
3410 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3411 | 0 | parseRequiresExpression(RequiresToken); |
3412 | 0 | return false; |
3413 | 0 | } |
3414 | 0 | break; |
3415 | 0 | case tok::r_paren: |
3416 | 0 | case tok::pipepipe: |
3417 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3418 | 0 | parseRequiresClause(RequiresToken); |
3419 | 0 | return true; |
3420 | 0 | case tok::eof: |
3421 | | // Break out of the loop. |
3422 | 0 | Lookahead = 50; |
3423 | 0 | break; |
3424 | 0 | case tok::coloncolon: |
3425 | 0 | LastWasColonColon = true; |
3426 | 0 | break; |
3427 | 0 | case tok::identifier: |
3428 | 0 | if (FoundType && !LastWasColonColon && OpenAngles == 0) { |
3429 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3430 | 0 | parseRequiresExpression(RequiresToken); |
3431 | 0 | return false; |
3432 | 0 | } |
3433 | 0 | FoundType = true; |
3434 | 0 | LastWasColonColon = false; |
3435 | 0 | break; |
3436 | 0 | case tok::less: |
3437 | 0 | ++OpenAngles; |
3438 | 0 | break; |
3439 | 0 | case tok::greater: |
3440 | 0 | --OpenAngles; |
3441 | 0 | break; |
3442 | 0 | default: |
3443 | 0 | if (NextToken->isSimpleTypeSpecifier()) { |
3444 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3445 | 0 | parseRequiresExpression(RequiresToken); |
3446 | 0 | return false; |
3447 | 0 | } |
3448 | 0 | break; |
3449 | 0 | } |
3450 | 0 | } |
3451 | | // This seems to be a complicated expression, just assume it's a clause. |
3452 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3453 | 0 | parseRequiresClause(RequiresToken); |
3454 | 0 | return true; |
3455 | 0 | } |
3456 | | |
3457 | | /// \brief Parses a requires clause. |
3458 | | /// \param RequiresToken The requires keyword token, which starts this clause. |
3459 | | /// \pre We need to be on the next token after the requires keyword. |
3460 | | /// \sa parseRequiresExpression |
3461 | | /// |
3462 | | /// Returns if it either has finished parsing the clause, or it detects, that |
3463 | | /// the clause is incorrect. |
3464 | 0 | void UnwrappedLineParser::parseRequiresClause(FormatToken *RequiresToken) { |
3465 | 0 | assert(FormatTok->getPreviousNonComment() == RequiresToken); |
3466 | 0 | assert(RequiresToken->is(tok::kw_requires) && "'requires' expected"); |
3467 | | |
3468 | | // If there is no previous token, we are within a requires expression, |
3469 | | // otherwise we will always have the template or function declaration in front |
3470 | | // of it. |
3471 | 0 | bool InRequiresExpression = |
3472 | 0 | !RequiresToken->Previous || |
3473 | 0 | RequiresToken->Previous->is(TT_RequiresExpressionLBrace); |
3474 | |
|
3475 | 0 | RequiresToken->setFinalizedType(InRequiresExpression |
3476 | 0 | ? TT_RequiresClauseInARequiresExpression |
3477 | 0 | : TT_RequiresClause); |
3478 | | |
3479 | | // NOTE: parseConstraintExpression is only ever called from this function. |
3480 | | // It could be inlined into here. |
3481 | 0 | parseConstraintExpression(); |
3482 | |
|
3483 | 0 | if (!InRequiresExpression) |
3484 | 0 | FormatTok->Previous->ClosesRequiresClause = true; |
3485 | 0 | } |
3486 | | |
3487 | | /// \brief Parses a requires expression. |
3488 | | /// \param RequiresToken The requires keyword token, which starts this clause. |
3489 | | /// \pre We need to be on the next token after the requires keyword. |
3490 | | /// \sa parseRequiresClause |
3491 | | /// |
3492 | | /// Returns if it either has finished parsing the expression, or it detects, |
3493 | | /// that the expression is incorrect. |
3494 | 0 | void UnwrappedLineParser::parseRequiresExpression(FormatToken *RequiresToken) { |
3495 | 0 | assert(FormatTok->getPreviousNonComment() == RequiresToken); |
3496 | 0 | assert(RequiresToken->is(tok::kw_requires) && "'requires' expected"); |
3497 | | |
3498 | 0 | RequiresToken->setFinalizedType(TT_RequiresExpression); |
3499 | |
|
3500 | 0 | if (FormatTok->is(tok::l_paren)) { |
3501 | 0 | FormatTok->setFinalizedType(TT_RequiresExpressionLParen); |
3502 | 0 | parseParens(); |
3503 | 0 | } |
3504 | |
|
3505 | 0 | if (FormatTok->is(tok::l_brace)) { |
3506 | 0 | FormatTok->setFinalizedType(TT_RequiresExpressionLBrace); |
3507 | 0 | parseChildBlock(); |
3508 | 0 | } |
3509 | 0 | } |
3510 | | |
3511 | | /// \brief Parses a constraint expression. |
3512 | | /// |
3513 | | /// This is the body of a requires clause. It returns, when the parsing is |
3514 | | /// complete, or the expression is incorrect. |
3515 | 0 | void UnwrappedLineParser::parseConstraintExpression() { |
3516 | | // The special handling for lambdas is needed since tryToParseLambda() eats a |
3517 | | // token and if a requires expression is the last part of a requires clause |
3518 | | // and followed by an attribute like [[nodiscard]] the ClosesRequiresClause is |
3519 | | // not set on the correct token. Thus we need to be aware if we even expect a |
3520 | | // lambda to be possible. |
3521 | | // template <typename T> requires requires { ... } [[nodiscard]] ...; |
3522 | 0 | bool LambdaNextTimeAllowed = true; |
3523 | | |
3524 | | // Within lambda declarations, it is permitted to put a requires clause after |
3525 | | // its template parameter list, which would place the requires clause right |
3526 | | // before the parentheses of the parameters of the lambda declaration. Thus, |
3527 | | // we track if we expect to see grouping parentheses at all. |
3528 | | // Without this check, `requires foo<T> (T t)` in the below example would be |
3529 | | // seen as the whole requires clause, accidentally eating the parameters of |
3530 | | // the lambda. |
3531 | | // [&]<typename T> requires foo<T> (T t) { ... }; |
3532 | 0 | bool TopLevelParensAllowed = true; |
3533 | |
|
3534 | 0 | do { |
3535 | 0 | bool LambdaThisTimeAllowed = std::exchange(LambdaNextTimeAllowed, false); |
3536 | |
|
3537 | 0 | switch (FormatTok->Tok.getKind()) { |
3538 | 0 | case tok::kw_requires: { |
3539 | 0 | auto RequiresToken = FormatTok; |
3540 | 0 | nextToken(); |
3541 | 0 | parseRequiresExpression(RequiresToken); |
3542 | 0 | break; |
3543 | 0 | } |
3544 | | |
3545 | 0 | case tok::l_paren: |
3546 | 0 | if (!TopLevelParensAllowed) |
3547 | 0 | return; |
3548 | 0 | parseParens(/*AmpAmpTokenType=*/TT_BinaryOperator); |
3549 | 0 | TopLevelParensAllowed = false; |
3550 | 0 | break; |
3551 | | |
3552 | 0 | case tok::l_square: |
3553 | 0 | if (!LambdaThisTimeAllowed || !tryToParseLambda()) |
3554 | 0 | return; |
3555 | 0 | break; |
3556 | | |
3557 | 0 | case tok::kw_const: |
3558 | 0 | case tok::semi: |
3559 | 0 | case tok::kw_class: |
3560 | 0 | case tok::kw_struct: |
3561 | 0 | case tok::kw_union: |
3562 | 0 | return; |
3563 | | |
3564 | 0 | case tok::l_brace: |
3565 | | // Potential function body. |
3566 | 0 | return; |
3567 | | |
3568 | 0 | case tok::ampamp: |
3569 | 0 | case tok::pipepipe: |
3570 | 0 | FormatTok->setFinalizedType(TT_BinaryOperator); |
3571 | 0 | nextToken(); |
3572 | 0 | LambdaNextTimeAllowed = true; |
3573 | 0 | TopLevelParensAllowed = true; |
3574 | 0 | break; |
3575 | | |
3576 | 0 | case tok::comma: |
3577 | 0 | case tok::comment: |
3578 | 0 | LambdaNextTimeAllowed = LambdaThisTimeAllowed; |
3579 | 0 | nextToken(); |
3580 | 0 | break; |
3581 | | |
3582 | 0 | case tok::kw_sizeof: |
3583 | 0 | case tok::greater: |
3584 | 0 | case tok::greaterequal: |
3585 | 0 | case tok::greatergreater: |
3586 | 0 | case tok::less: |
3587 | 0 | case tok::lessequal: |
3588 | 0 | case tok::lessless: |
3589 | 0 | case tok::equalequal: |
3590 | 0 | case tok::exclaim: |
3591 | 0 | case tok::exclaimequal: |
3592 | 0 | case tok::plus: |
3593 | 0 | case tok::minus: |
3594 | 0 | case tok::star: |
3595 | 0 | case tok::slash: |
3596 | 0 | LambdaNextTimeAllowed = true; |
3597 | 0 | TopLevelParensAllowed = true; |
3598 | | // Just eat them. |
3599 | 0 | nextToken(); |
3600 | 0 | break; |
3601 | | |
3602 | 0 | case tok::numeric_constant: |
3603 | 0 | case tok::coloncolon: |
3604 | 0 | case tok::kw_true: |
3605 | 0 | case tok::kw_false: |
3606 | 0 | TopLevelParensAllowed = false; |
3607 | | // Just eat them. |
3608 | 0 | nextToken(); |
3609 | 0 | break; |
3610 | | |
3611 | 0 | case tok::kw_static_cast: |
3612 | 0 | case tok::kw_const_cast: |
3613 | 0 | case tok::kw_reinterpret_cast: |
3614 | 0 | case tok::kw_dynamic_cast: |
3615 | 0 | nextToken(); |
3616 | 0 | if (FormatTok->isNot(tok::less)) |
3617 | 0 | return; |
3618 | | |
3619 | 0 | nextToken(); |
3620 | 0 | parseBracedList(/*IsAngleBracket=*/true); |
3621 | 0 | break; |
3622 | | |
3623 | 0 | default: |
3624 | 0 | if (!FormatTok->Tok.getIdentifierInfo()) { |
3625 | | // Identifiers are part of the default case, we check for more then |
3626 | | // tok::identifier to handle builtin type traits. |
3627 | 0 | return; |
3628 | 0 | } |
3629 | | |
3630 | | // We need to differentiate identifiers for a template deduction guide, |
3631 | | // variables, or function return types (the constraint expression has |
3632 | | // ended before that), and basically all other cases. But it's easier to |
3633 | | // check the other way around. |
3634 | 0 | assert(FormatTok->Previous); |
3635 | 0 | switch (FormatTok->Previous->Tok.getKind()) { |
3636 | 0 | case tok::coloncolon: // Nested identifier. |
3637 | 0 | case tok::ampamp: // Start of a function or variable for the |
3638 | 0 | case tok::pipepipe: // constraint expression. (binary) |
3639 | 0 | case tok::exclaim: // The same as above, but unary. |
3640 | 0 | case tok::kw_requires: // Initial identifier of a requires clause. |
3641 | 0 | case tok::equal: // Initial identifier of a concept declaration. |
3642 | 0 | break; |
3643 | 0 | default: |
3644 | 0 | return; |
3645 | 0 | } |
3646 | | |
3647 | | // Read identifier with optional template declaration. |
3648 | 0 | nextToken(); |
3649 | 0 | if (FormatTok->is(tok::less)) { |
3650 | 0 | nextToken(); |
3651 | 0 | parseBracedList(/*IsAngleBracket=*/true); |
3652 | 0 | } |
3653 | 0 | TopLevelParensAllowed = false; |
3654 | 0 | break; |
3655 | 0 | } |
3656 | 0 | } while (!eof()); |
3657 | 0 | } |
3658 | | |
3659 | 192 | bool UnwrappedLineParser::parseEnum() { |
3660 | 192 | const FormatToken &InitialToken = *FormatTok; |
3661 | | |
3662 | | // Won't be 'enum' for NS_ENUMs. |
3663 | 192 | if (FormatTok->is(tok::kw_enum)) |
3664 | 192 | nextToken(); |
3665 | | |
3666 | | // In TypeScript, "enum" can also be used as property name, e.g. in interface |
3667 | | // declarations. An "enum" keyword followed by a colon would be a syntax |
3668 | | // error and thus assume it is just an identifier. |
3669 | 192 | if (Style.isJavaScript() && FormatTok->isOneOf(tok::colon, tok::question)) |
3670 | 0 | return false; |
3671 | | |
3672 | | // In protobuf, "enum" can be used as a field name. |
3673 | 192 | if (Style.Language == FormatStyle::LK_Proto && FormatTok->is(tok::equal)) |
3674 | 0 | return false; |
3675 | | |
3676 | | // Eat up enum class ... |
3677 | 192 | if (FormatTok->isOneOf(tok::kw_class, tok::kw_struct)) |
3678 | 0 | nextToken(); |
3679 | | |
3680 | 192 | while (FormatTok->Tok.getIdentifierInfo() || |
3681 | 192 | FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, |
3682 | 192 | tok::greater, tok::comma, tok::question, |
3683 | 192 | tok::l_square, tok::r_square)) { |
3684 | 0 | if (Style.isVerilog()) { |
3685 | 0 | FormatTok->setFinalizedType(TT_VerilogDimensionedTypeName); |
3686 | 0 | nextToken(); |
3687 | | // In Verilog the base type can have dimensions. |
3688 | 0 | while (FormatTok->is(tok::l_square)) |
3689 | 0 | parseSquare(); |
3690 | 0 | } else { |
3691 | 0 | nextToken(); |
3692 | 0 | } |
3693 | | // We can have macros or attributes in between 'enum' and the enum name. |
3694 | 0 | if (FormatTok->is(tok::l_paren)) |
3695 | 0 | parseParens(); |
3696 | 0 | assert(FormatTok->isNot(TT_AttributeSquare)); |
3697 | 0 | if (FormatTok->is(tok::identifier)) { |
3698 | 0 | nextToken(); |
3699 | | // If there are two identifiers in a row, this is likely an elaborate |
3700 | | // return type. In Java, this can be "implements", etc. |
3701 | 0 | if (Style.isCpp() && FormatTok->is(tok::identifier)) |
3702 | 0 | return false; |
3703 | 0 | } |
3704 | 0 | } |
3705 | | |
3706 | | // Just a declaration or something is wrong. |
3707 | 192 | if (FormatTok->isNot(tok::l_brace)) |
3708 | 0 | return true; |
3709 | 192 | FormatTok->setFinalizedType(TT_EnumLBrace); |
3710 | 192 | FormatTok->setBlockKind(BK_Block); |
3711 | | |
3712 | 192 | if (Style.Language == FormatStyle::LK_Java) { |
3713 | | // Java enums are different. |
3714 | 0 | parseJavaEnumBody(); |
3715 | 0 | return true; |
3716 | 0 | } |
3717 | 192 | if (Style.Language == FormatStyle::LK_Proto) { |
3718 | 0 | parseBlock(/*MustBeDeclaration=*/true); |
3719 | 0 | return true; |
3720 | 0 | } |
3721 | | |
3722 | 192 | if (!Style.AllowShortEnumsOnASingleLine && |
3723 | 192 | ShouldBreakBeforeBrace(Style, InitialToken)) { |
3724 | 0 | addUnwrappedLine(); |
3725 | 0 | } |
3726 | | // Parse enum body. |
3727 | 192 | nextToken(); |
3728 | 192 | if (!Style.AllowShortEnumsOnASingleLine) { |
3729 | 0 | addUnwrappedLine(); |
3730 | 0 | Line->Level += 1; |
3731 | 0 | } |
3732 | 192 | bool HasError = !parseBracedList(/*IsAngleBracket=*/false, /*IsEnum=*/true); |
3733 | 192 | if (!Style.AllowShortEnumsOnASingleLine) |
3734 | 0 | Line->Level -= 1; |
3735 | 192 | if (HasError) { |
3736 | 27 | if (FormatTok->is(tok::semi)) |
3737 | 0 | nextToken(); |
3738 | 27 | addUnwrappedLine(); |
3739 | 27 | } |
3740 | 192 | setPreviousRBraceType(TT_EnumRBrace); |
3741 | 192 | return true; |
3742 | | |
3743 | | // There is no addUnwrappedLine() here so that we fall through to parsing a |
3744 | | // structural element afterwards. Thus, in "enum A {} n, m;", |
3745 | | // "} n, m;" will end up in one unwrapped line. |
3746 | 192 | } |
3747 | | |
3748 | 1.02k | bool UnwrappedLineParser::parseStructLike() { |
3749 | | // parseRecord falls through and does not yet add an unwrapped line as a |
3750 | | // record declaration or definition can start a structural element. |
3751 | 1.02k | parseRecord(); |
3752 | | // This does not apply to Java, JavaScript and C#. |
3753 | 1.02k | if (Style.Language == FormatStyle::LK_Java || Style.isJavaScript() || |
3754 | 1.02k | Style.isCSharp()) { |
3755 | 0 | if (FormatTok->is(tok::semi)) |
3756 | 0 | nextToken(); |
3757 | 0 | addUnwrappedLine(); |
3758 | 0 | return true; |
3759 | 0 | } |
3760 | 1.02k | return false; |
3761 | 1.02k | } |
3762 | | |
3763 | | namespace { |
3764 | | // A class used to set and restore the Token position when peeking |
3765 | | // ahead in the token source. |
3766 | | class ScopedTokenPosition { |
3767 | | unsigned StoredPosition; |
3768 | | FormatTokenSource *Tokens; |
3769 | | |
3770 | | public: |
3771 | 2.77k | ScopedTokenPosition(FormatTokenSource *Tokens) : Tokens(Tokens) { |
3772 | 2.77k | assert(Tokens && "Tokens expected to not be null"); |
3773 | 0 | StoredPosition = Tokens->getPosition(); |
3774 | 2.77k | } |
3775 | | |
3776 | 2.77k | ~ScopedTokenPosition() { Tokens->setPosition(StoredPosition); } |
3777 | | }; |
3778 | | } // namespace |
3779 | | |
3780 | | // Look to see if we have [[ by looking ahead, if |
3781 | | // its not then rewind to the original position. |
3782 | 2.77k | bool UnwrappedLineParser::tryToParseSimpleAttribute() { |
3783 | 2.77k | ScopedTokenPosition AutoPosition(Tokens); |
3784 | 2.77k | FormatToken *Tok = Tokens->getNextToken(); |
3785 | | // We already read the first [ check for the second. |
3786 | 2.77k | if (Tok->isNot(tok::l_square)) |
3787 | 2.75k | return false; |
3788 | | // Double check that the attribute is just something |
3789 | | // fairly simple. |
3790 | 270 | while (Tok->isNot(tok::eof)) { |
3791 | 264 | if (Tok->is(tok::r_square)) |
3792 | 12 | break; |
3793 | 252 | Tok = Tokens->getNextToken(); |
3794 | 252 | } |
3795 | 18 | if (Tok->is(tok::eof)) |
3796 | 6 | return false; |
3797 | 12 | Tok = Tokens->getNextToken(); |
3798 | 12 | if (Tok->isNot(tok::r_square)) |
3799 | 12 | return false; |
3800 | 0 | Tok = Tokens->getNextToken(); |
3801 | 0 | if (Tok->is(tok::semi)) |
3802 | 0 | return false; |
3803 | 0 | return true; |
3804 | 0 | } |
3805 | | |
3806 | 0 | void UnwrappedLineParser::parseJavaEnumBody() { |
3807 | 0 | assert(FormatTok->is(tok::l_brace)); |
3808 | 0 | const FormatToken *OpeningBrace = FormatTok; |
3809 | | |
3810 | | // Determine whether the enum is simple, i.e. does not have a semicolon or |
3811 | | // constants with class bodies. Simple enums can be formatted like braced |
3812 | | // lists, contracted to a single line, etc. |
3813 | 0 | unsigned StoredPosition = Tokens->getPosition(); |
3814 | 0 | bool IsSimple = true; |
3815 | 0 | FormatToken *Tok = Tokens->getNextToken(); |
3816 | 0 | while (Tok->isNot(tok::eof)) { |
3817 | 0 | if (Tok->is(tok::r_brace)) |
3818 | 0 | break; |
3819 | 0 | if (Tok->isOneOf(tok::l_brace, tok::semi)) { |
3820 | 0 | IsSimple = false; |
3821 | 0 | break; |
3822 | 0 | } |
3823 | | // FIXME: This will also mark enums with braces in the arguments to enum |
3824 | | // constants as "not simple". This is probably fine in practice, though. |
3825 | 0 | Tok = Tokens->getNextToken(); |
3826 | 0 | } |
3827 | 0 | FormatTok = Tokens->setPosition(StoredPosition); |
3828 | |
|
3829 | 0 | if (IsSimple) { |
3830 | 0 | nextToken(); |
3831 | 0 | parseBracedList(); |
3832 | 0 | addUnwrappedLine(); |
3833 | 0 | return; |
3834 | 0 | } |
3835 | | |
3836 | | // Parse the body of a more complex enum. |
3837 | | // First add a line for everything up to the "{". |
3838 | 0 | nextToken(); |
3839 | 0 | addUnwrappedLine(); |
3840 | 0 | ++Line->Level; |
3841 | | |
3842 | | // Parse the enum constants. |
3843 | 0 | while (!eof()) { |
3844 | 0 | if (FormatTok->is(tok::l_brace)) { |
3845 | | // Parse the constant's class body. |
3846 | 0 | parseBlock(/*MustBeDeclaration=*/true, /*AddLevels=*/1u, |
3847 | 0 | /*MunchSemi=*/false); |
3848 | 0 | } else if (FormatTok->is(tok::l_paren)) { |
3849 | 0 | parseParens(); |
3850 | 0 | } else if (FormatTok->is(tok::comma)) { |
3851 | 0 | nextToken(); |
3852 | 0 | addUnwrappedLine(); |
3853 | 0 | } else if (FormatTok->is(tok::semi)) { |
3854 | 0 | nextToken(); |
3855 | 0 | addUnwrappedLine(); |
3856 | 0 | break; |
3857 | 0 | } else if (FormatTok->is(tok::r_brace)) { |
3858 | 0 | addUnwrappedLine(); |
3859 | 0 | break; |
3860 | 0 | } else { |
3861 | 0 | nextToken(); |
3862 | 0 | } |
3863 | 0 | } |
3864 | | |
3865 | | // Parse the class body after the enum's ";" if any. |
3866 | 0 | parseLevel(OpeningBrace); |
3867 | 0 | nextToken(); |
3868 | 0 | --Line->Level; |
3869 | 0 | addUnwrappedLine(); |
3870 | 0 | } |
3871 | | |
3872 | 1.02k | void UnwrappedLineParser::parseRecord(bool ParseAsExpr) { |
3873 | 1.02k | const FormatToken &InitialToken = *FormatTok; |
3874 | 1.02k | nextToken(); |
3875 | | |
3876 | | // The actual identifier can be a nested name specifier, and in macros |
3877 | | // it is often token-pasted. |
3878 | | // An [[attribute]] can be before the identifier. |
3879 | 1.11k | while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, |
3880 | 1.11k | tok::kw_alignas, tok::l_square) || |
3881 | 1.11k | FormatTok->isAttribute() || |
3882 | 1.11k | ((Style.Language == FormatStyle::LK_Java || Style.isJavaScript()) && |
3883 | 1.02k | FormatTok->isOneOf(tok::period, tok::comma))) { |
3884 | 81 | if (Style.isJavaScript() && |
3885 | 81 | FormatTok->isOneOf(Keywords.kw_extends, Keywords.kw_implements)) { |
3886 | | // JavaScript/TypeScript supports inline object types in |
3887 | | // extends/implements positions: |
3888 | | // class Foo implements {bar: number} { } |
3889 | 0 | nextToken(); |
3890 | 0 | if (FormatTok->is(tok::l_brace)) { |
3891 | 0 | tryToParseBracedList(); |
3892 | 0 | continue; |
3893 | 0 | } |
3894 | 0 | } |
3895 | 81 | if (FormatTok->is(tok::l_square) && handleCppAttributes()) |
3896 | 0 | continue; |
3897 | 81 | bool IsNonMacroIdentifier = |
3898 | 81 | FormatTok->is(tok::identifier) && |
3899 | 81 | FormatTok->TokenText != FormatTok->TokenText.upper(); |
3900 | 81 | nextToken(); |
3901 | | // We can have macros in between 'class' and the class name. |
3902 | 81 | if (!IsNonMacroIdentifier && FormatTok->is(tok::l_paren)) |
3903 | 0 | parseParens(); |
3904 | 81 | } |
3905 | | |
3906 | | // Note that parsing away template declarations here leads to incorrectly |
3907 | | // accepting function declarations as record declarations. |
3908 | | // In general, we cannot solve this problem. Consider: |
3909 | | // class A<int> B() {} |
3910 | | // which can be a function definition or a class definition when B() is a |
3911 | | // macro. If we find enough real-world cases where this is a problem, we |
3912 | | // can parse for the 'template' keyword in the beginning of the statement, |
3913 | | // and thus rule out the record production in case there is no template |
3914 | | // (this would still leave us with an ambiguity between template function |
3915 | | // and class declarations). |
3916 | 1.02k | if (FormatTok->isOneOf(tok::colon, tok::less)) { |
3917 | 3.06k | do { |
3918 | 3.06k | if (FormatTok->is(tok::l_brace)) { |
3919 | 75 | calculateBraceTypes(/*ExpectClassBody=*/true); |
3920 | 75 | if (!tryToParseBracedList()) |
3921 | 57 | break; |
3922 | 75 | } |
3923 | 3.01k | if (FormatTok->is(tok::l_square)) { |
3924 | 0 | FormatToken *Previous = FormatTok->Previous; |
3925 | 0 | if (!Previous || |
3926 | 0 | !(Previous->is(tok::r_paren) || Previous->isTypeOrIdentifier())) { |
3927 | | // Don't try parsing a lambda if we had a closing parenthesis before, |
3928 | | // it was probably a pointer to an array: int (*)[]. |
3929 | 0 | if (!tryToParseLambda()) |
3930 | 0 | continue; |
3931 | 0 | } else { |
3932 | 0 | parseSquare(); |
3933 | 0 | continue; |
3934 | 0 | } |
3935 | 0 | } |
3936 | 3.01k | if (FormatTok->is(tok::semi)) |
3937 | 0 | return; |
3938 | 3.01k | if (Style.isCSharp() && FormatTok->is(Keywords.kw_where)) { |
3939 | 0 | addUnwrappedLine(); |
3940 | 0 | nextToken(); |
3941 | 0 | parseCSharpGenericTypeConstraint(); |
3942 | 0 | break; |
3943 | 0 | } |
3944 | 3.01k | nextToken(); |
3945 | 3.01k | } while (!eof()); |
3946 | 57 | } |
3947 | | |
3948 | 1.02k | auto GetBraceTypes = |
3949 | 1.02k | [](const FormatToken &RecordTok) -> std::pair<TokenType, TokenType> { |
3950 | 156 | switch (RecordTok.Tok.getKind()) { |
3951 | 57 | case tok::kw_class: |
3952 | 57 | return {TT_ClassLBrace, TT_ClassRBrace}; |
3953 | 99 | case tok::kw_struct: |
3954 | 99 | return {TT_StructLBrace, TT_StructRBrace}; |
3955 | 0 | case tok::kw_union: |
3956 | 0 | return {TT_UnionLBrace, TT_UnionRBrace}; |
3957 | 0 | default: |
3958 | | // Useful for e.g. interface. |
3959 | 0 | return {TT_RecordLBrace, TT_RecordRBrace}; |
3960 | 156 | } |
3961 | 156 | }; |
3962 | 1.02k | if (FormatTok->is(tok::l_brace)) { |
3963 | 156 | auto [OpenBraceType, ClosingBraceType] = GetBraceTypes(InitialToken); |
3964 | 156 | FormatTok->setFinalizedType(OpenBraceType); |
3965 | 156 | if (ParseAsExpr) { |
3966 | 0 | parseChildBlock(); |
3967 | 156 | } else { |
3968 | 156 | if (ShouldBreakBeforeBrace(Style, InitialToken)) |
3969 | 0 | addUnwrappedLine(); |
3970 | | |
3971 | 156 | unsigned AddLevels = Style.IndentAccessModifiers ? 2u : 1u; |
3972 | 156 | parseBlock(/*MustBeDeclaration=*/true, AddLevels, /*MunchSemi=*/false); |
3973 | 156 | } |
3974 | 156 | setPreviousRBraceType(ClosingBraceType); |
3975 | 156 | } |
3976 | | // There is no addUnwrappedLine() here so that we fall through to parsing a |
3977 | | // structural element afterwards. Thus, in "class A {} n, m;", |
3978 | | // "} n, m;" will end up in one unwrapped line. |
3979 | 1.02k | } |
3980 | | |
3981 | 0 | void UnwrappedLineParser::parseObjCMethod() { |
3982 | 0 | assert(FormatTok->isOneOf(tok::l_paren, tok::identifier) && |
3983 | 0 | "'(' or identifier expected."); |
3984 | 0 | do { |
3985 | 0 | if (FormatTok->is(tok::semi)) { |
3986 | 0 | nextToken(); |
3987 | 0 | addUnwrappedLine(); |
3988 | 0 | return; |
3989 | 0 | } else if (FormatTok->is(tok::l_brace)) { |
3990 | 0 | if (Style.BraceWrapping.AfterFunction) |
3991 | 0 | addUnwrappedLine(); |
3992 | 0 | parseBlock(); |
3993 | 0 | addUnwrappedLine(); |
3994 | 0 | return; |
3995 | 0 | } else { |
3996 | 0 | nextToken(); |
3997 | 0 | } |
3998 | 0 | } while (!eof()); |
3999 | 0 | } |
4000 | | |
4001 | 0 | void UnwrappedLineParser::parseObjCProtocolList() { |
4002 | 0 | assert(FormatTok->is(tok::less) && "'<' expected."); |
4003 | 0 | do { |
4004 | 0 | nextToken(); |
4005 | | // Early exit in case someone forgot a close angle. |
4006 | 0 | if (FormatTok->isOneOf(tok::semi, tok::l_brace) || |
4007 | 0 | FormatTok->isObjCAtKeyword(tok::objc_end)) { |
4008 | 0 | return; |
4009 | 0 | } |
4010 | 0 | } while (!eof() && FormatTok->isNot(tok::greater)); |
4011 | 0 | nextToken(); // Skip '>'. |
4012 | 0 | } |
4013 | | |
4014 | 0 | void UnwrappedLineParser::parseObjCUntilAtEnd() { |
4015 | 0 | do { |
4016 | 0 | if (FormatTok->isObjCAtKeyword(tok::objc_end)) { |
4017 | 0 | nextToken(); |
4018 | 0 | addUnwrappedLine(); |
4019 | 0 | break; |
4020 | 0 | } |
4021 | 0 | if (FormatTok->is(tok::l_brace)) { |
4022 | 0 | parseBlock(); |
4023 | | // In ObjC interfaces, nothing should be following the "}". |
4024 | 0 | addUnwrappedLine(); |
4025 | 0 | } else if (FormatTok->is(tok::r_brace)) { |
4026 | | // Ignore stray "}". parseStructuralElement doesn't consume them. |
4027 | 0 | nextToken(); |
4028 | 0 | addUnwrappedLine(); |
4029 | 0 | } else if (FormatTok->isOneOf(tok::minus, tok::plus)) { |
4030 | 0 | nextToken(); |
4031 | 0 | parseObjCMethod(); |
4032 | 0 | } else { |
4033 | 0 | parseStructuralElement(); |
4034 | 0 | } |
4035 | 0 | } while (!eof()); |
4036 | 0 | } |
4037 | | |
4038 | 0 | void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { |
4039 | 0 | assert(FormatTok->Tok.getObjCKeywordID() == tok::objc_interface || |
4040 | 0 | FormatTok->Tok.getObjCKeywordID() == tok::objc_implementation); |
4041 | 0 | nextToken(); |
4042 | 0 | nextToken(); // interface name |
4043 | | |
4044 | | // @interface can be followed by a lightweight generic |
4045 | | // specialization list, then either a base class or a category. |
4046 | 0 | if (FormatTok->is(tok::less)) |
4047 | 0 | parseObjCLightweightGenerics(); |
4048 | 0 | if (FormatTok->is(tok::colon)) { |
4049 | 0 | nextToken(); |
4050 | 0 | nextToken(); // base class name |
4051 | | // The base class can also have lightweight generics applied to it. |
4052 | 0 | if (FormatTok->is(tok::less)) |
4053 | 0 | parseObjCLightweightGenerics(); |
4054 | 0 | } else if (FormatTok->is(tok::l_paren)) { |
4055 | | // Skip category, if present. |
4056 | 0 | parseParens(); |
4057 | 0 | } |
4058 | |
|
4059 | 0 | if (FormatTok->is(tok::less)) |
4060 | 0 | parseObjCProtocolList(); |
4061 | |
|
4062 | 0 | if (FormatTok->is(tok::l_brace)) { |
4063 | 0 | if (Style.BraceWrapping.AfterObjCDeclaration) |
4064 | 0 | addUnwrappedLine(); |
4065 | 0 | parseBlock(/*MustBeDeclaration=*/true); |
4066 | 0 | } |
4067 | | |
4068 | | // With instance variables, this puts '}' on its own line. Without instance |
4069 | | // variables, this ends the @interface line. |
4070 | 0 | addUnwrappedLine(); |
4071 | |
|
4072 | 0 | parseObjCUntilAtEnd(); |
4073 | 0 | } |
4074 | | |
4075 | 0 | void UnwrappedLineParser::parseObjCLightweightGenerics() { |
4076 | 0 | assert(FormatTok->is(tok::less)); |
4077 | | // Unlike protocol lists, generic parameterizations support |
4078 | | // nested angles: |
4079 | | // |
4080 | | // @interface Foo<ValueType : id <NSCopying, NSSecureCoding>> : |
4081 | | // NSObject <NSCopying, NSSecureCoding> |
4082 | | // |
4083 | | // so we need to count how many open angles we have left. |
4084 | 0 | unsigned NumOpenAngles = 1; |
4085 | 0 | do { |
4086 | 0 | nextToken(); |
4087 | | // Early exit in case someone forgot a close angle. |
4088 | 0 | if (FormatTok->isOneOf(tok::semi, tok::l_brace) || |
4089 | 0 | FormatTok->isObjCAtKeyword(tok::objc_end)) { |
4090 | 0 | break; |
4091 | 0 | } |
4092 | 0 | if (FormatTok->is(tok::less)) { |
4093 | 0 | ++NumOpenAngles; |
4094 | 0 | } else if (FormatTok->is(tok::greater)) { |
4095 | 0 | assert(NumOpenAngles > 0 && "'>' makes NumOpenAngles negative"); |
4096 | 0 | --NumOpenAngles; |
4097 | 0 | } |
4098 | 0 | } while (!eof() && NumOpenAngles != 0); |
4099 | 0 | nextToken(); // Skip '>'. |
4100 | 0 | } |
4101 | | |
4102 | | // Returns true for the declaration/definition form of @protocol, |
4103 | | // false for the expression form. |
4104 | 0 | bool UnwrappedLineParser::parseObjCProtocol() { |
4105 | 0 | assert(FormatTok->Tok.getObjCKeywordID() == tok::objc_protocol); |
4106 | 0 | nextToken(); |
4107 | |
|
4108 | 0 | if (FormatTok->is(tok::l_paren)) { |
4109 | | // The expression form of @protocol, e.g. "Protocol* p = @protocol(foo);". |
4110 | 0 | return false; |
4111 | 0 | } |
4112 | | |
4113 | | // The definition/declaration form, |
4114 | | // @protocol Foo |
4115 | | // - (int)someMethod; |
4116 | | // @end |
4117 | | |
4118 | 0 | nextToken(); // protocol name |
4119 | |
|
4120 | 0 | if (FormatTok->is(tok::less)) |
4121 | 0 | parseObjCProtocolList(); |
4122 | | |
4123 | | // Check for protocol declaration. |
4124 | 0 | if (FormatTok->is(tok::semi)) { |
4125 | 0 | nextToken(); |
4126 | 0 | addUnwrappedLine(); |
4127 | 0 | return true; |
4128 | 0 | } |
4129 | | |
4130 | 0 | addUnwrappedLine(); |
4131 | 0 | parseObjCUntilAtEnd(); |
4132 | 0 | return true; |
4133 | 0 | } |
4134 | | |
4135 | 0 | void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { |
4136 | 0 | bool IsImport = FormatTok->is(Keywords.kw_import); |
4137 | 0 | assert(IsImport || FormatTok->is(tok::kw_export)); |
4138 | 0 | nextToken(); |
4139 | | |
4140 | | // Consume the "default" in "export default class/function". |
4141 | 0 | if (FormatTok->is(tok::kw_default)) |
4142 | 0 | nextToken(); |
4143 | | |
4144 | | // Consume "async function", "function" and "default function", so that these |
4145 | | // get parsed as free-standing JS functions, i.e. do not require a trailing |
4146 | | // semicolon. |
4147 | 0 | if (FormatTok->is(Keywords.kw_async)) |
4148 | 0 | nextToken(); |
4149 | 0 | if (FormatTok->is(Keywords.kw_function)) { |
4150 | 0 | nextToken(); |
4151 | 0 | return; |
4152 | 0 | } |
4153 | | |
4154 | | // For imports, `export *`, `export {...}`, consume the rest of the line up |
4155 | | // to the terminating `;`. For everything else, just return and continue |
4156 | | // parsing the structural element, i.e. the declaration or expression for |
4157 | | // `export default`. |
4158 | 0 | if (!IsImport && !FormatTok->isOneOf(tok::l_brace, tok::star) && |
4159 | 0 | !FormatTok->isStringLiteral() && |
4160 | 0 | !(FormatTok->is(Keywords.kw_type) && |
4161 | 0 | Tokens->peekNextToken()->isOneOf(tok::l_brace, tok::star))) { |
4162 | 0 | return; |
4163 | 0 | } |
4164 | | |
4165 | 0 | while (!eof()) { |
4166 | 0 | if (FormatTok->is(tok::semi)) |
4167 | 0 | return; |
4168 | 0 | if (Line->Tokens.empty()) { |
4169 | | // Common issue: Automatic Semicolon Insertion wrapped the line, so the |
4170 | | // import statement should terminate. |
4171 | 0 | return; |
4172 | 0 | } |
4173 | 0 | if (FormatTok->is(tok::l_brace)) { |
4174 | 0 | FormatTok->setBlockKind(BK_Block); |
4175 | 0 | nextToken(); |
4176 | 0 | parseBracedList(); |
4177 | 0 | } else { |
4178 | 0 | nextToken(); |
4179 | 0 | } |
4180 | 0 | } |
4181 | 0 | } |
4182 | | |
4183 | 0 | void UnwrappedLineParser::parseStatementMacro() { |
4184 | 0 | nextToken(); |
4185 | 0 | if (FormatTok->is(tok::l_paren)) |
4186 | 0 | parseParens(); |
4187 | 0 | if (FormatTok->is(tok::semi)) |
4188 | 0 | nextToken(); |
4189 | 0 | addUnwrappedLine(); |
4190 | 0 | } |
4191 | | |
4192 | 0 | void UnwrappedLineParser::parseVerilogHierarchyIdentifier() { |
4193 | | // consume things like a::`b.c[d:e] or a::* |
4194 | 0 | while (true) { |
4195 | 0 | if (FormatTok->isOneOf(tok::star, tok::period, tok::periodstar, |
4196 | 0 | tok::coloncolon, tok::hash) || |
4197 | 0 | Keywords.isVerilogIdentifier(*FormatTok)) { |
4198 | 0 | nextToken(); |
4199 | 0 | } else if (FormatTok->is(tok::l_square)) { |
4200 | 0 | parseSquare(); |
4201 | 0 | } else { |
4202 | 0 | break; |
4203 | 0 | } |
4204 | 0 | } |
4205 | 0 | } |
4206 | | |
4207 | 0 | void UnwrappedLineParser::parseVerilogSensitivityList() { |
4208 | 0 | if (FormatTok->isNot(tok::at)) |
4209 | 0 | return; |
4210 | 0 | nextToken(); |
4211 | | // A block event expression has 2 at signs. |
4212 | 0 | if (FormatTok->is(tok::at)) |
4213 | 0 | nextToken(); |
4214 | 0 | switch (FormatTok->Tok.getKind()) { |
4215 | 0 | case tok::star: |
4216 | 0 | nextToken(); |
4217 | 0 | break; |
4218 | 0 | case tok::l_paren: |
4219 | 0 | parseParens(); |
4220 | 0 | break; |
4221 | 0 | default: |
4222 | 0 | parseVerilogHierarchyIdentifier(); |
4223 | 0 | break; |
4224 | 0 | } |
4225 | 0 | } |
4226 | | |
4227 | 0 | unsigned UnwrappedLineParser::parseVerilogHierarchyHeader() { |
4228 | 0 | unsigned AddLevels = 0; |
4229 | |
|
4230 | 0 | if (FormatTok->is(Keywords.kw_clocking)) { |
4231 | 0 | nextToken(); |
4232 | 0 | if (Keywords.isVerilogIdentifier(*FormatTok)) |
4233 | 0 | nextToken(); |
4234 | 0 | parseVerilogSensitivityList(); |
4235 | 0 | if (FormatTok->is(tok::semi)) |
4236 | 0 | nextToken(); |
4237 | 0 | } else if (FormatTok->isOneOf(tok::kw_case, Keywords.kw_casex, |
4238 | 0 | Keywords.kw_casez, Keywords.kw_randcase, |
4239 | 0 | Keywords.kw_randsequence)) { |
4240 | 0 | if (Style.IndentCaseLabels) |
4241 | 0 | AddLevels++; |
4242 | 0 | nextToken(); |
4243 | 0 | if (FormatTok->is(tok::l_paren)) { |
4244 | 0 | FormatTok->setFinalizedType(TT_ConditionLParen); |
4245 | 0 | parseParens(); |
4246 | 0 | } |
4247 | 0 | if (FormatTok->isOneOf(Keywords.kw_inside, Keywords.kw_matches)) |
4248 | 0 | nextToken(); |
4249 | | // The case header has no semicolon. |
4250 | 0 | } else { |
4251 | | // "module" etc. |
4252 | 0 | nextToken(); |
4253 | | // all the words like the name of the module and specifiers like |
4254 | | // "automatic" and the width of function return type |
4255 | 0 | while (true) { |
4256 | 0 | if (FormatTok->is(tok::l_square)) { |
4257 | 0 | auto Prev = FormatTok->getPreviousNonComment(); |
4258 | 0 | if (Prev && Keywords.isVerilogIdentifier(*Prev)) |
4259 | 0 | Prev->setFinalizedType(TT_VerilogDimensionedTypeName); |
4260 | 0 | parseSquare(); |
4261 | 0 | } else if (Keywords.isVerilogIdentifier(*FormatTok) || |
4262 | 0 | FormatTok->isOneOf(Keywords.kw_automatic, tok::kw_static)) { |
4263 | 0 | nextToken(); |
4264 | 0 | } else { |
4265 | 0 | break; |
4266 | 0 | } |
4267 | 0 | } |
4268 | |
|
4269 | 0 | auto NewLine = [this]() { |
4270 | 0 | addUnwrappedLine(); |
4271 | 0 | Line->IsContinuation = true; |
4272 | 0 | }; |
4273 | | |
4274 | | // package imports |
4275 | 0 | while (FormatTok->is(Keywords.kw_import)) { |
4276 | 0 | NewLine(); |
4277 | 0 | nextToken(); |
4278 | 0 | parseVerilogHierarchyIdentifier(); |
4279 | 0 | if (FormatTok->is(tok::semi)) |
4280 | 0 | nextToken(); |
4281 | 0 | } |
4282 | | |
4283 | | // parameters and ports |
4284 | 0 | if (FormatTok->is(Keywords.kw_verilogHash)) { |
4285 | 0 | NewLine(); |
4286 | 0 | nextToken(); |
4287 | 0 | if (FormatTok->is(tok::l_paren)) { |
4288 | 0 | FormatTok->setFinalizedType(TT_VerilogMultiLineListLParen); |
4289 | 0 | parseParens(); |
4290 | 0 | } |
4291 | 0 | } |
4292 | 0 | if (FormatTok->is(tok::l_paren)) { |
4293 | 0 | NewLine(); |
4294 | 0 | FormatTok->setFinalizedType(TT_VerilogMultiLineListLParen); |
4295 | 0 | parseParens(); |
4296 | 0 | } |
4297 | | |
4298 | | // extends and implements |
4299 | 0 | if (FormatTok->is(Keywords.kw_extends)) { |
4300 | 0 | NewLine(); |
4301 | 0 | nextToken(); |
4302 | 0 | parseVerilogHierarchyIdentifier(); |
4303 | 0 | if (FormatTok->is(tok::l_paren)) |
4304 | 0 | parseParens(); |
4305 | 0 | } |
4306 | 0 | if (FormatTok->is(Keywords.kw_implements)) { |
4307 | 0 | NewLine(); |
4308 | 0 | do { |
4309 | 0 | nextToken(); |
4310 | 0 | parseVerilogHierarchyIdentifier(); |
4311 | 0 | } while (FormatTok->is(tok::comma)); |
4312 | 0 | } |
4313 | | |
4314 | | // Coverage event for cover groups. |
4315 | 0 | if (FormatTok->is(tok::at)) { |
4316 | 0 | NewLine(); |
4317 | 0 | parseVerilogSensitivityList(); |
4318 | 0 | } |
4319 | |
|
4320 | 0 | if (FormatTok->is(tok::semi)) |
4321 | 0 | nextToken(/*LevelDifference=*/1); |
4322 | 0 | addUnwrappedLine(); |
4323 | 0 | } |
4324 | |
|
4325 | 0 | return AddLevels; |
4326 | 0 | } |
4327 | | |
4328 | 0 | void UnwrappedLineParser::parseVerilogTable() { |
4329 | 0 | assert(FormatTok->is(Keywords.kw_table)); |
4330 | 0 | nextToken(/*LevelDifference=*/1); |
4331 | 0 | addUnwrappedLine(); |
4332 | |
|
4333 | 0 | auto InitialLevel = Line->Level++; |
4334 | 0 | while (!eof() && !Keywords.isVerilogEnd(*FormatTok)) { |
4335 | 0 | FormatToken *Tok = FormatTok; |
4336 | 0 | nextToken(); |
4337 | 0 | if (Tok->is(tok::semi)) |
4338 | 0 | addUnwrappedLine(); |
4339 | 0 | else if (Tok->isOneOf(tok::star, tok::colon, tok::question, tok::minus)) |
4340 | 0 | Tok->setFinalizedType(TT_VerilogTableItem); |
4341 | 0 | } |
4342 | 0 | Line->Level = InitialLevel; |
4343 | 0 | nextToken(/*LevelDifference=*/-1); |
4344 | 0 | addUnwrappedLine(); |
4345 | 0 | } |
4346 | | |
4347 | 0 | void UnwrappedLineParser::parseVerilogCaseLabel() { |
4348 | | // The label will get unindented in AnnotatingParser. If there are no leading |
4349 | | // spaces, indent the rest here so that things inside the block will be |
4350 | | // indented relative to things outside. We don't use parseLabel because we |
4351 | | // don't know whether this colon is a label or a ternary expression at this |
4352 | | // point. |
4353 | 0 | auto OrigLevel = Line->Level; |
4354 | 0 | auto FirstLine = CurrentLines->size(); |
4355 | 0 | if (Line->Level == 0 || (Line->InPPDirective && Line->Level <= 1)) |
4356 | 0 | ++Line->Level; |
4357 | 0 | else if (!Style.IndentCaseBlocks && Keywords.isVerilogBegin(*FormatTok)) |
4358 | 0 | --Line->Level; |
4359 | 0 | parseStructuralElement(); |
4360 | | // Restore the indentation in both the new line and the line that has the |
4361 | | // label. |
4362 | 0 | if (CurrentLines->size() > FirstLine) |
4363 | 0 | (*CurrentLines)[FirstLine].Level = OrigLevel; |
4364 | 0 | Line->Level = OrigLevel; |
4365 | 0 | } |
4366 | | |
4367 | 1.33M | bool UnwrappedLineParser::containsExpansion(const UnwrappedLine &Line) const { |
4368 | 51.3M | for (const auto &N : Line.Tokens) { |
4369 | 51.3M | if (N.Tok->MacroCtx) |
4370 | 0 | return true; |
4371 | 51.3M | for (const UnwrappedLine &Child : N.Children) |
4372 | 270k | if (containsExpansion(Child)) |
4373 | 0 | return true; |
4374 | 51.3M | } |
4375 | 1.33M | return false; |
4376 | 1.33M | } |
4377 | | |
4378 | 2.44M | void UnwrappedLineParser::addUnwrappedLine(LineLevel AdjustLevel) { |
4379 | 2.44M | if (Line->Tokens.empty()) |
4380 | 850k | return; |
4381 | 1.59M | LLVM_DEBUG({ |
4382 | 1.59M | if (!parsingPPDirective()) { |
4383 | 1.59M | llvm::dbgs() << "Adding unwrapped line:\n"; |
4384 | 1.59M | printDebugInfo(*Line); |
4385 | 1.59M | } |
4386 | 1.59M | }); |
4387 | | |
4388 | | // If this line closes a block when in Whitesmiths mode, remember that |
4389 | | // information so that the level can be decreased after the line is added. |
4390 | | // This has to happen after the addition of the line since the line itself |
4391 | | // needs to be indented. |
4392 | 1.59M | bool ClosesWhitesmithsBlock = |
4393 | 1.59M | Line->MatchingOpeningBlockLineIndex != UnwrappedLine::kInvalidIndex && |
4394 | 1.59M | Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths; |
4395 | | |
4396 | | // If the current line was expanded from a macro call, we use it to |
4397 | | // reconstruct an unwrapped line from the structure of the expanded unwrapped |
4398 | | // line and the unexpanded token stream. |
4399 | 1.59M | if (!parsingPPDirective() && !InExpansion && containsExpansion(*Line)) { |
4400 | 0 | if (!Reconstruct) |
4401 | 0 | Reconstruct.emplace(Line->Level, Unexpanded); |
4402 | 0 | Reconstruct->addLine(*Line); |
4403 | | |
4404 | | // While the reconstructed unexpanded lines are stored in the normal |
4405 | | // flow of lines, the expanded lines are stored on the side to be analyzed |
4406 | | // in an extra step. |
4407 | 0 | CurrentExpandedLines.push_back(std::move(*Line)); |
4408 | |
|
4409 | 0 | if (Reconstruct->finished()) { |
4410 | 0 | UnwrappedLine Reconstructed = std::move(*Reconstruct).takeResult(); |
4411 | 0 | assert(!Reconstructed.Tokens.empty() && |
4412 | 0 | "Reconstructed must at least contain the macro identifier."); |
4413 | 0 | assert(!parsingPPDirective()); |
4414 | 0 | LLVM_DEBUG({ |
4415 | 0 | llvm::dbgs() << "Adding unexpanded line:\n"; |
4416 | 0 | printDebugInfo(Reconstructed); |
4417 | 0 | }); |
4418 | 0 | ExpandedLines[Reconstructed.Tokens.begin()->Tok] = CurrentExpandedLines; |
4419 | 0 | Lines.push_back(std::move(Reconstructed)); |
4420 | 0 | CurrentExpandedLines.clear(); |
4421 | 0 | Reconstruct.reset(); |
4422 | 0 | } |
4423 | 1.59M | } else { |
4424 | | // At the top level we only get here when no unexpansion is going on, or |
4425 | | // when conditional formatting led to unfinished macro reconstructions. |
4426 | 1.59M | assert(!Reconstruct || (CurrentLines != &Lines) || PPStack.size() > 0); |
4427 | 0 | CurrentLines->push_back(std::move(*Line)); |
4428 | 1.59M | } |
4429 | 0 | Line->Tokens.clear(); |
4430 | 1.59M | Line->MatchingOpeningBlockLineIndex = UnwrappedLine::kInvalidIndex; |
4431 | 1.59M | Line->FirstStartColumn = 0; |
4432 | 1.59M | Line->IsContinuation = false; |
4433 | 1.59M | Line->SeenDecltypeAuto = false; |
4434 | | |
4435 | 1.59M | if (ClosesWhitesmithsBlock && AdjustLevel == LineLevel::Remove) |
4436 | 0 | --Line->Level; |
4437 | 1.59M | if (!parsingPPDirective() && !PreprocessorDirectives.empty()) { |
4438 | 1.12k | CurrentLines->append( |
4439 | 1.12k | std::make_move_iterator(PreprocessorDirectives.begin()), |
4440 | 1.12k | std::make_move_iterator(PreprocessorDirectives.end())); |
4441 | 1.12k | PreprocessorDirectives.clear(); |
4442 | 1.12k | } |
4443 | | // Disconnect the current token from the last token on the previous line. |
4444 | 1.59M | FormatTok->Previous = nullptr; |
4445 | 1.59M | } |
4446 | | |
4447 | 106M | bool UnwrappedLineParser::eof() const { return FormatTok->is(tok::eof); } |
4448 | | |
4449 | 52.5M | bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { |
4450 | 52.5M | return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && |
4451 | 52.5M | FormatTok.NewlinesBefore > 0; |
4452 | 52.5M | } |
4453 | | |
4454 | | // Checks if \p FormatTok is a line comment that continues the line comment |
4455 | | // section on \p Line. |
4456 | | static bool |
4457 | | continuesLineCommentSection(const FormatToken &FormatTok, |
4458 | | const UnwrappedLine &Line, |
4459 | 55.5k | const llvm::Regex &CommentPragmasRegex) { |
4460 | 55.5k | if (Line.Tokens.empty()) |
4461 | 2.73k | return false; |
4462 | | |
4463 | 52.8k | StringRef IndentContent = FormatTok.TokenText; |
4464 | 52.8k | if (FormatTok.TokenText.starts_with("//") || |
4465 | 52.8k | FormatTok.TokenText.starts_with("/*")) { |
4466 | 52.8k | IndentContent = FormatTok.TokenText.substr(2); |
4467 | 52.8k | } |
4468 | 52.8k | if (CommentPragmasRegex.match(IndentContent)) |
4469 | 0 | return false; |
4470 | | |
4471 | | // If Line starts with a line comment, then FormatTok continues the comment |
4472 | | // section if its original column is greater or equal to the original start |
4473 | | // column of the line. |
4474 | | // |
4475 | | // Define the min column token of a line as follows: if a line ends in '{' or |
4476 | | // contains a '{' followed by a line comment, then the min column token is |
4477 | | // that '{'. Otherwise, the min column token of the line is the first token of |
4478 | | // the line. |
4479 | | // |
4480 | | // If Line starts with a token other than a line comment, then FormatTok |
4481 | | // continues the comment section if its original column is greater than the |
4482 | | // original start column of the min column token of the line. |
4483 | | // |
4484 | | // For example, the second line comment continues the first in these cases: |
4485 | | // |
4486 | | // // first line |
4487 | | // // second line |
4488 | | // |
4489 | | // and: |
4490 | | // |
4491 | | // // first line |
4492 | | // // second line |
4493 | | // |
4494 | | // and: |
4495 | | // |
4496 | | // int i; // first line |
4497 | | // // second line |
4498 | | // |
4499 | | // and: |
4500 | | // |
4501 | | // do { // first line |
4502 | | // // second line |
4503 | | // int i; |
4504 | | // } while (true); |
4505 | | // |
4506 | | // and: |
4507 | | // |
4508 | | // enum { |
4509 | | // a, // first line |
4510 | | // // second line |
4511 | | // b |
4512 | | // }; |
4513 | | // |
4514 | | // The second line comment doesn't continue the first in these cases: |
4515 | | // |
4516 | | // // first line |
4517 | | // // second line |
4518 | | // |
4519 | | // and: |
4520 | | // |
4521 | | // int i; // first line |
4522 | | // // second line |
4523 | | // |
4524 | | // and: |
4525 | | // |
4526 | | // do { // first line |
4527 | | // // second line |
4528 | | // int i; |
4529 | | // } while (true); |
4530 | | // |
4531 | | // and: |
4532 | | // |
4533 | | // enum { |
4534 | | // a, // first line |
4535 | | // // second line |
4536 | | // }; |
4537 | 52.8k | const FormatToken *MinColumnToken = Line.Tokens.front().Tok; |
4538 | | |
4539 | | // Scan for '{//'. If found, use the column of '{' as a min column for line |
4540 | | // comment section continuation. |
4541 | 52.8k | const FormatToken *PreviousToken = nullptr; |
4542 | 74.0M | for (const UnwrappedLineNode &Node : Line.Tokens) { |
4543 | 74.0M | if (PreviousToken && PreviousToken->is(tok::l_brace) && |
4544 | 74.0M | isLineComment(*Node.Tok)) { |
4545 | 0 | MinColumnToken = PreviousToken; |
4546 | 0 | break; |
4547 | 0 | } |
4548 | 74.0M | PreviousToken = Node.Tok; |
4549 | | |
4550 | | // Grab the last newline preceding a token in this unwrapped line. |
4551 | 74.0M | if (Node.Tok->NewlinesBefore > 0) |
4552 | 4.70M | MinColumnToken = Node.Tok; |
4553 | 74.0M | } |
4554 | 52.8k | if (PreviousToken && PreviousToken->is(tok::l_brace)) |
4555 | 882 | MinColumnToken = PreviousToken; |
4556 | | |
4557 | 52.8k | return continuesLineComment(FormatTok, /*Previous=*/Line.Tokens.back().Tok, |
4558 | 52.8k | MinColumnToken); |
4559 | 52.8k | } |
4560 | | |
4561 | 52.4M | void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { |
4562 | 52.4M | bool JustComments = Line->Tokens.empty(); |
4563 | 52.4M | for (FormatToken *Tok : CommentsBeforeNextToken) { |
4564 | | // Line comments that belong to the same line comment section are put on the |
4565 | | // same line since later we might want to reflow content between them. |
4566 | | // Additional fine-grained breaking of line comment sections is controlled |
4567 | | // by the class BreakableLineCommentSection in case it is desirable to keep |
4568 | | // several line comment sections in the same unwrapped line. |
4569 | | // |
4570 | | // FIXME: Consider putting separate line comment sections as children to the |
4571 | | // unwrapped line instead. |
4572 | 12.1k | Tok->ContinuesLineCommentSection = |
4573 | 12.1k | continuesLineCommentSection(*Tok, *Line, CommentPragmasRegex); |
4574 | 12.1k | if (isOnNewLine(*Tok) && JustComments && !Tok->ContinuesLineCommentSection) |
4575 | 4.32k | addUnwrappedLine(); |
4576 | 12.1k | pushToken(Tok); |
4577 | 12.1k | } |
4578 | 52.4M | if (NewlineBeforeNext && JustComments) |
4579 | 774k | addUnwrappedLine(); |
4580 | 52.4M | CommentsBeforeNextToken.clear(); |
4581 | 52.4M | } |
4582 | | |
4583 | 52.4M | void UnwrappedLineParser::nextToken(int LevelDifference) { |
4584 | 52.4M | if (eof()) |
4585 | 258k | return; |
4586 | 52.1M | flushComments(isOnNewLine(*FormatTok)); |
4587 | 52.1M | pushToken(FormatTok); |
4588 | 52.1M | FormatToken *Previous = FormatTok; |
4589 | 52.1M | if (!Style.isJavaScript()) |
4590 | 52.1M | readToken(LevelDifference); |
4591 | 0 | else |
4592 | 0 | readTokenWithJavaScriptASI(); |
4593 | 52.1M | FormatTok->Previous = Previous; |
4594 | 52.1M | if (Style.isVerilog()) { |
4595 | | // Blocks in Verilog can have `begin` and `end` instead of braces. For |
4596 | | // keywords like `begin`, we can't treat them the same as left braces |
4597 | | // because some contexts require one of them. For example structs use |
4598 | | // braces and if blocks use keywords, and a left brace can occur in an if |
4599 | | // statement, but it is not a block. For keywords like `end`, we simply |
4600 | | // treat them the same as right braces. |
4601 | 0 | if (Keywords.isVerilogEnd(*FormatTok)) |
4602 | 0 | FormatTok->Tok.setKind(tok::r_brace); |
4603 | 0 | } |
4604 | 52.1M | } |
4605 | | |
4606 | | void UnwrappedLineParser::distributeComments( |
4607 | | const SmallVectorImpl<FormatToken *> &Comments, |
4608 | 52.4M | const FormatToken *NextTok) { |
4609 | | // Whether or not a line comment token continues a line is controlled by |
4610 | | // the method continuesLineCommentSection, with the following caveat: |
4611 | | // |
4612 | | // Define a trail of Comments to be a nonempty proper postfix of Comments such |
4613 | | // that each comment line from the trail is aligned with the next token, if |
4614 | | // the next token exists. If a trail exists, the beginning of the maximal |
4615 | | // trail is marked as a start of a new comment section. |
4616 | | // |
4617 | | // For example in this code: |
4618 | | // |
4619 | | // int a; // line about a |
4620 | | // // line 1 about b |
4621 | | // // line 2 about b |
4622 | | // int b; |
4623 | | // |
4624 | | // the two lines about b form a maximal trail, so there are two sections, the |
4625 | | // first one consisting of the single comment "// line about a" and the |
4626 | | // second one consisting of the next two comments. |
4627 | 52.4M | if (Comments.empty()) |
4628 | 52.4M | return; |
4629 | 38.8k | bool ShouldPushCommentsInCurrentLine = true; |
4630 | 38.8k | bool HasTrailAlignedWithNextToken = false; |
4631 | 38.8k | unsigned StartOfTrailAlignedWithNextToken = 0; |
4632 | 38.8k | if (NextTok) { |
4633 | | // We are skipping the first element intentionally. |
4634 | 45.7k | for (unsigned i = Comments.size() - 1; i > 0; --i) { |
4635 | 6.90k | if (Comments[i]->OriginalColumn == NextTok->OriginalColumn) { |
4636 | 6.21k | HasTrailAlignedWithNextToken = true; |
4637 | 6.21k | StartOfTrailAlignedWithNextToken = i; |
4638 | 6.21k | } |
4639 | 6.90k | } |
4640 | 38.8k | } |
4641 | 84.6k | for (unsigned i = 0, e = Comments.size(); i < e; ++i) { |
4642 | 45.7k | FormatToken *FormatTok = Comments[i]; |
4643 | 45.7k | if (HasTrailAlignedWithNextToken && i == StartOfTrailAlignedWithNextToken) { |
4644 | 2.32k | FormatTok->ContinuesLineCommentSection = false; |
4645 | 43.4k | } else { |
4646 | 43.4k | FormatTok->ContinuesLineCommentSection = |
4647 | 43.4k | continuesLineCommentSection(*FormatTok, *Line, CommentPragmasRegex); |
4648 | 43.4k | } |
4649 | 45.7k | if (!FormatTok->ContinuesLineCommentSection && |
4650 | 45.7k | (isOnNewLine(*FormatTok) || FormatTok->IsFirst)) { |
4651 | 12.0k | ShouldPushCommentsInCurrentLine = false; |
4652 | 12.0k | } |
4653 | 45.7k | if (ShouldPushCommentsInCurrentLine) |
4654 | 33.6k | pushToken(FormatTok); |
4655 | 12.1k | else |
4656 | 12.1k | CommentsBeforeNextToken.push_back(FormatTok); |
4657 | 45.7k | } |
4658 | 38.8k | } |
4659 | | |
4660 | 52.1M | void UnwrappedLineParser::readToken(int LevelDifference) { |
4661 | 52.1M | SmallVector<FormatToken *, 1> Comments; |
4662 | 52.1M | bool PreviousWasComment = false; |
4663 | 52.1M | bool FirstNonCommentOnLine = false; |
4664 | 54.7M | do { |
4665 | 54.7M | FormatTok = Tokens->getNextToken(); |
4666 | 54.7M | assert(FormatTok); |
4667 | 55.0M | while (FormatTok->getType() == TT_ConflictStart || |
4668 | 55.0M | FormatTok->getType() == TT_ConflictEnd || |
4669 | 55.0M | FormatTok->getType() == TT_ConflictAlternative) { |
4670 | 261k | if (FormatTok->getType() == TT_ConflictStart) |
4671 | 261k | conditionalCompilationStart(/*Unreachable=*/false); |
4672 | 297 | else if (FormatTok->getType() == TT_ConflictAlternative) |
4673 | 126 | conditionalCompilationAlternative(); |
4674 | 171 | else if (FormatTok->getType() == TT_ConflictEnd) |
4675 | 171 | conditionalCompilationEnd(); |
4676 | 261k | FormatTok = Tokens->getNextToken(); |
4677 | 261k | FormatTok->MustBreakBefore = true; |
4678 | 261k | FormatTok->MustBreakBeforeFinalized = true; |
4679 | 261k | } |
4680 | | |
4681 | 54.7M | auto IsFirstNonCommentOnLine = [](bool FirstNonCommentOnLine, |
4682 | 54.7M | const FormatToken &Tok, |
4683 | 55.0M | bool PreviousWasComment) { |
4684 | 55.0M | auto IsFirstOnLine = [](const FormatToken &Tok) { |
4685 | 55.0M | return Tok.HasUnescapedNewline || Tok.IsFirst; |
4686 | 55.0M | }; |
4687 | | |
4688 | | // Consider preprocessor directives preceded by block comments as first |
4689 | | // on line. |
4690 | 55.0M | if (PreviousWasComment) |
4691 | 50.8k | return FirstNonCommentOnLine || IsFirstOnLine(Tok); |
4692 | 54.9M | return IsFirstOnLine(Tok); |
4693 | 55.0M | }; |
4694 | | |
4695 | 54.7M | FirstNonCommentOnLine = IsFirstNonCommentOnLine( |
4696 | 54.7M | FirstNonCommentOnLine, *FormatTok, PreviousWasComment); |
4697 | 54.7M | PreviousWasComment = FormatTok->is(tok::comment); |
4698 | | |
4699 | 55.0M | while (!Line->InPPDirective && FormatTok->is(tok::hash) && |
4700 | 55.0M | (!Style.isVerilog() || |
4701 | 412k | Keywords.isVerilogPPDirective(*Tokens->peekNextToken())) && |
4702 | 55.0M | FirstNonCommentOnLine) { |
4703 | 253k | distributeComments(Comments, FormatTok); |
4704 | 253k | Comments.clear(); |
4705 | | // If there is an unfinished unwrapped line, we flush the preprocessor |
4706 | | // directives only after that unwrapped line was finished later. |
4707 | 253k | bool SwitchToPreprocessorLines = !Line->Tokens.empty(); |
4708 | 253k | ScopedLineState BlockState(*this, SwitchToPreprocessorLines); |
4709 | 253k | assert((LevelDifference >= 0 || |
4710 | 253k | static_cast<unsigned>(-LevelDifference) <= Line->Level) && |
4711 | 253k | "LevelDifference makes Line->Level negative"); |
4712 | 0 | Line->Level += LevelDifference; |
4713 | | // Comments stored before the preprocessor directive need to be output |
4714 | | // before the preprocessor directive, at the same level as the |
4715 | | // preprocessor directive, as we consider them to apply to the directive. |
4716 | 253k | if (Style.IndentPPDirectives == FormatStyle::PPDIS_BeforeHash && |
4717 | 253k | PPBranchLevel > 0) { |
4718 | 0 | Line->Level += PPBranchLevel; |
4719 | 0 | } |
4720 | 253k | flushComments(isOnNewLine(*FormatTok)); |
4721 | 253k | parsePPDirective(); |
4722 | 253k | PreviousWasComment = FormatTok->is(tok::comment); |
4723 | 253k | FirstNonCommentOnLine = IsFirstNonCommentOnLine( |
4724 | 253k | FirstNonCommentOnLine, *FormatTok, PreviousWasComment); |
4725 | 253k | } |
4726 | | |
4727 | 54.7M | if (!PPStack.empty() && (PPStack.back().Kind == PP_Unreachable) && |
4728 | 54.7M | !Line->InPPDirective) { |
4729 | 2.56M | continue; |
4730 | 2.56M | } |
4731 | | |
4732 | 52.2M | if (FormatTok->is(tok::identifier) && |
4733 | 52.2M | Macros.defined(FormatTok->TokenText) && |
4734 | | // FIXME: Allow expanding macros in preprocessor directives. |
4735 | 52.2M | !Line->InPPDirective) { |
4736 | 0 | FormatToken *ID = FormatTok; |
4737 | 0 | unsigned Position = Tokens->getPosition(); |
4738 | | |
4739 | | // To correctly parse the code, we need to replace the tokens of the macro |
4740 | | // call with its expansion. |
4741 | 0 | auto PreCall = std::move(Line); |
4742 | 0 | Line.reset(new UnwrappedLine); |
4743 | 0 | bool OldInExpansion = InExpansion; |
4744 | 0 | InExpansion = true; |
4745 | | // We parse the macro call into a new line. |
4746 | 0 | auto Args = parseMacroCall(); |
4747 | 0 | InExpansion = OldInExpansion; |
4748 | 0 | assert(Line->Tokens.front().Tok == ID); |
4749 | | // And remember the unexpanded macro call tokens. |
4750 | 0 | auto UnexpandedLine = std::move(Line); |
4751 | | // Reset to the old line. |
4752 | 0 | Line = std::move(PreCall); |
4753 | |
|
4754 | 0 | LLVM_DEBUG({ |
4755 | 0 | llvm::dbgs() << "Macro call: " << ID->TokenText << "("; |
4756 | 0 | if (Args) { |
4757 | 0 | llvm::dbgs() << "("; |
4758 | 0 | for (const auto &Arg : Args.value()) |
4759 | 0 | for (const auto &T : Arg) |
4760 | 0 | llvm::dbgs() << T->TokenText << " "; |
4761 | 0 | llvm::dbgs() << ")"; |
4762 | 0 | } |
4763 | 0 | llvm::dbgs() << "\n"; |
4764 | 0 | }); |
4765 | 0 | if (Macros.objectLike(ID->TokenText) && Args && |
4766 | 0 | !Macros.hasArity(ID->TokenText, Args->size())) { |
4767 | | // The macro is either |
4768 | | // - object-like, but we got argumnets, or |
4769 | | // - overloaded to be both object-like and function-like, but none of |
4770 | | // the function-like arities match the number of arguments. |
4771 | | // Thus, expand as object-like macro. |
4772 | 0 | LLVM_DEBUG(llvm::dbgs() |
4773 | 0 | << "Macro \"" << ID->TokenText |
4774 | 0 | << "\" not overloaded for arity " << Args->size() |
4775 | 0 | << "or not function-like, using object-like overload."); |
4776 | 0 | Args.reset(); |
4777 | 0 | UnexpandedLine->Tokens.resize(1); |
4778 | 0 | Tokens->setPosition(Position); |
4779 | 0 | nextToken(); |
4780 | 0 | assert(!Args && Macros.objectLike(ID->TokenText)); |
4781 | 0 | } |
4782 | 0 | if ((!Args && Macros.objectLike(ID->TokenText)) || |
4783 | 0 | (Args && Macros.hasArity(ID->TokenText, Args->size()))) { |
4784 | | // Next, we insert the expanded tokens in the token stream at the |
4785 | | // current position, and continue parsing. |
4786 | 0 | Unexpanded[ID] = std::move(UnexpandedLine); |
4787 | 0 | SmallVector<FormatToken *, 8> Expansion = |
4788 | 0 | Macros.expand(ID, std::move(Args)); |
4789 | 0 | if (!Expansion.empty()) |
4790 | 0 | FormatTok = Tokens->insertTokens(Expansion); |
4791 | |
|
4792 | 0 | LLVM_DEBUG({ |
4793 | 0 | llvm::dbgs() << "Expanded: "; |
4794 | 0 | for (const auto &T : Expansion) |
4795 | 0 | llvm::dbgs() << T->TokenText << " "; |
4796 | 0 | llvm::dbgs() << "\n"; |
4797 | 0 | }); |
4798 | 0 | } else { |
4799 | 0 | LLVM_DEBUG({ |
4800 | 0 | llvm::dbgs() << "Did not expand macro \"" << ID->TokenText |
4801 | 0 | << "\", because it was used "; |
4802 | 0 | if (Args) |
4803 | 0 | llvm::dbgs() << "with " << Args->size(); |
4804 | 0 | else |
4805 | 0 | llvm::dbgs() << "without"; |
4806 | 0 | llvm::dbgs() << " arguments, which doesn't match any definition.\n"; |
4807 | 0 | }); |
4808 | 0 | Tokens->setPosition(Position); |
4809 | 0 | FormatTok = ID; |
4810 | 0 | } |
4811 | 0 | } |
4812 | | |
4813 | 52.2M | if (FormatTok->isNot(tok::comment)) { |
4814 | 52.1M | distributeComments(Comments, FormatTok); |
4815 | 52.1M | Comments.clear(); |
4816 | 52.1M | return; |
4817 | 52.1M | } |
4818 | | |
4819 | 45.7k | Comments.push_back(FormatTok); |
4820 | 2.60M | } while (!eof()); |
4821 | | |
4822 | 39 | distributeComments(Comments, nullptr); |
4823 | 39 | Comments.clear(); |
4824 | 39 | } |
4825 | | |
4826 | | namespace { |
4827 | | template <typename Iterator> |
4828 | | void pushTokens(Iterator Begin, Iterator End, |
4829 | 0 | llvm::SmallVectorImpl<FormatToken *> &Into) { |
4830 | 0 | for (auto I = Begin; I != End; ++I) { |
4831 | 0 | Into.push_back(I->Tok); |
4832 | 0 | for (const auto &Child : I->Children) |
4833 | 0 | pushTokens(Child.Tokens.begin(), Child.Tokens.end(), Into); |
4834 | 0 | } |
4835 | 0 | } Unexecuted instantiation: UnwrappedLineParser.cpp:void clang::format::(anonymous namespace)::pushTokens<std::__1::__list_iterator<clang::format::UnwrappedLineNode, void*> >(std::__1::__list_iterator<clang::format::UnwrappedLineNode, void*>, std::__1::__list_iterator<clang::format::UnwrappedLineNode, void*>, llvm::SmallVectorImpl<clang::format::FormatToken*>&) Unexecuted instantiation: UnwrappedLineParser.cpp:void clang::format::(anonymous namespace)::pushTokens<std::__1::__list_const_iterator<clang::format::UnwrappedLineNode, void*> >(std::__1::__list_const_iterator<clang::format::UnwrappedLineNode, void*>, std::__1::__list_const_iterator<clang::format::UnwrappedLineNode, void*>, llvm::SmallVectorImpl<clang::format::FormatToken*>&) |
4836 | | } // namespace |
4837 | | |
4838 | | std::optional<llvm::SmallVector<llvm::SmallVector<FormatToken *, 8>, 1>> |
4839 | 0 | UnwrappedLineParser::parseMacroCall() { |
4840 | 0 | std::optional<llvm::SmallVector<llvm::SmallVector<FormatToken *, 8>, 1>> Args; |
4841 | 0 | assert(Line->Tokens.empty()); |
4842 | 0 | nextToken(); |
4843 | 0 | if (FormatTok->isNot(tok::l_paren)) |
4844 | 0 | return Args; |
4845 | 0 | unsigned Position = Tokens->getPosition(); |
4846 | 0 | FormatToken *Tok = FormatTok; |
4847 | 0 | nextToken(); |
4848 | 0 | Args.emplace(); |
4849 | 0 | auto ArgStart = std::prev(Line->Tokens.end()); |
4850 | |
|
4851 | 0 | int Parens = 0; |
4852 | 0 | do { |
4853 | 0 | switch (FormatTok->Tok.getKind()) { |
4854 | 0 | case tok::l_paren: |
4855 | 0 | ++Parens; |
4856 | 0 | nextToken(); |
4857 | 0 | break; |
4858 | 0 | case tok::r_paren: { |
4859 | 0 | if (Parens > 0) { |
4860 | 0 | --Parens; |
4861 | 0 | nextToken(); |
4862 | 0 | break; |
4863 | 0 | } |
4864 | 0 | Args->push_back({}); |
4865 | 0 | pushTokens(std::next(ArgStart), Line->Tokens.end(), Args->back()); |
4866 | 0 | nextToken(); |
4867 | 0 | return Args; |
4868 | 0 | } |
4869 | 0 | case tok::comma: { |
4870 | 0 | if (Parens > 0) { |
4871 | 0 | nextToken(); |
4872 | 0 | break; |
4873 | 0 | } |
4874 | 0 | Args->push_back({}); |
4875 | 0 | pushTokens(std::next(ArgStart), Line->Tokens.end(), Args->back()); |
4876 | 0 | nextToken(); |
4877 | 0 | ArgStart = std::prev(Line->Tokens.end()); |
4878 | 0 | break; |
4879 | 0 | } |
4880 | 0 | default: |
4881 | 0 | nextToken(); |
4882 | 0 | break; |
4883 | 0 | } |
4884 | 0 | } while (!eof()); |
4885 | 0 | Line->Tokens.resize(1); |
4886 | 0 | Tokens->setPosition(Position); |
4887 | 0 | FormatTok = Tok; |
4888 | 0 | return {}; |
4889 | 0 | } |
4890 | | |
4891 | 52.2M | void UnwrappedLineParser::pushToken(FormatToken *Tok) { |
4892 | 52.2M | Line->Tokens.push_back(UnwrappedLineNode(Tok)); |
4893 | 52.2M | if (MustBreakBeforeNextToken) { |
4894 | 253k | Line->Tokens.back().Tok->MustBreakBefore = true; |
4895 | 253k | Line->Tokens.back().Tok->MustBreakBeforeFinalized = true; |
4896 | 253k | MustBreakBeforeNextToken = false; |
4897 | 253k | } |
4898 | 52.2M | } |
4899 | | |
4900 | | } // end namespace format |
4901 | | } // end namespace clang |