1//===--- UnwrappedLineParser.h - Format C++ code ----------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8///
9/// \file
10/// This file contains the declaration of the UnwrappedLineParser,
11/// which turns a stream of tokens into UnwrappedLines.
12///
13//===----------------------------------------------------------------------===//
14
15#ifndef LLVM_CLANG_LIB_FORMAT_UNWRAPPEDLINEPARSER_H
16#define LLVM_CLANG_LIB_FORMAT_UNWRAPPEDLINEPARSER_H
17
18#include "Macros.h"
19#include <stack>
20
21namespace clang {
22namespace format {
23
24struct UnwrappedLineNode;
25
26/// An unwrapped line is a sequence of \c Token, that we would like to
27/// put on a single line if there was no column limit.
28///
29/// This is used as a main interface between the \c UnwrappedLineParser and the
30/// \c UnwrappedLineFormatter. The key property is that changing the formatting
31/// within an unwrapped line does not affect any other unwrapped lines.
32struct UnwrappedLine {
33 UnwrappedLine() = default;
34
35 /// The \c Tokens comprising this \c UnwrappedLine.
36 std::list<UnwrappedLineNode> Tokens;
37
38 /// The indent level of the \c UnwrappedLine.
39 unsigned Level = 0;
40
41 /// The \c PPBranchLevel (adjusted for header guards) if this line is a
42 /// \c InMacroBody line, and 0 otherwise.
43 unsigned PPLevel = 0;
44
45 /// Whether this \c UnwrappedLine is part of a preprocessor directive.
46 bool InPPDirective = false;
47 /// Whether this \c UnwrappedLine is part of a pramga directive.
48 bool InPragmaDirective = false;
49 /// Whether it is part of a macro body.
50 bool InMacroBody = false;
51
52 /// Nesting level of unbraced body of a control statement.
53 unsigned UnbracedBodyLevel = 0;
54
55 bool MustBeDeclaration = false;
56
57 /// Whether the parser has seen \c decltype(auto) in this line.
58 bool SeenDecltypeAuto = false;
59
60 /// \c True if this line should be indented by ContinuationIndent in
61 /// addition to the normal indention level.
62 bool IsContinuation = false;
63
64 /// If this \c UnwrappedLine closes a block in a sequence of lines,
65 /// \c MatchingOpeningBlockLineIndex stores the index of the corresponding
66 /// opening line. Otherwise, \c MatchingOpeningBlockLineIndex must be
67 /// \c kInvalidIndex.
68 size_t MatchingOpeningBlockLineIndex = kInvalidIndex;
69
70 /// If this \c UnwrappedLine opens a block, stores the index of the
71 /// line with the corresponding closing brace.
72 size_t MatchingClosingBlockLineIndex = kInvalidIndex;
73
74 static const size_t kInvalidIndex = -1;
75
76 unsigned FirstStartColumn = 0;
77};
78
79/// Interface for users of the UnwrappedLineParser to receive the parsed lines.
80/// Parsing a single snippet of code can lead to multiple runs, where each
81/// run is a coherent view of the file.
82///
83/// For example, different runs are generated:
84/// - for different combinations of #if blocks
85/// - when macros are involved, for the expanded code and the as-written code
86///
87/// Some tokens will only be visible in a subset of the runs.
88/// For each run, \c UnwrappedLineParser will call \c consumeUnwrappedLine
89/// for each parsed unwrapped line, and then \c finishRun to indicate
90/// that the set of unwrapped lines before is one coherent view of the
91/// code snippet to be formatted.
92class UnwrappedLineConsumer {
93public:
94 virtual ~UnwrappedLineConsumer() {}
95 virtual void consumeUnwrappedLine(const UnwrappedLine &Line) = 0;
96 virtual void finishRun() = 0;
97};
98
99class FormatTokenSource;
100
101class UnwrappedLineParser {
102public:
103 UnwrappedLineParser(SourceManager &SourceMgr, const FormatStyle &Style,
104 const AdditionalKeywords &Keywords,
105 unsigned FirstStartColumn, ArrayRef<FormatToken *> Tokens,
106 UnwrappedLineConsumer &Callback,
107 llvm::SpecificBumpPtrAllocator<FormatToken> &Allocator,
108 IdentifierTable &IdentTable);
109
110 void parse();
111
112private:
113 enum class IfStmtKind {
114 NotIf, // Not an if statement.
115 IfOnly, // An if statement without the else clause.
116 IfElse, // An if statement followed by else but not else if.
117 IfElseIf // An if statement followed by else if.
118 };
119
120 void reset();
121 void parseFile();
122 bool precededByCommentOrPPDirective() const;
123 bool parseLevel(const FormatToken *OpeningBrace = nullptr,
124 IfStmtKind *IfKind = nullptr,
125 FormatToken **IfLeftBrace = nullptr);
126 bool mightFitOnOneLine(UnwrappedLine &Line,
127 const FormatToken *OpeningBrace = nullptr) const;
128 FormatToken *parseBlock(bool MustBeDeclaration = false,
129 unsigned AddLevels = 1u, bool MunchSemi = true,
130 bool KeepBraces = true, IfStmtKind *IfKind = nullptr,
131 bool UnindentWhitesmithsBraces = false);
132 void parseChildBlock();
133 void parsePPDirective();
134 void parsePPDefine();
135 void parsePPIf(bool IfDef);
136 void parsePPElse();
137 void parsePPEndIf();
138 void parsePPPragma();
139 void parsePPUnknown();
140 void readTokenWithJavaScriptASI();
141 void parseStructuralElement(const FormatToken *OpeningBrace = nullptr,
142 IfStmtKind *IfKind = nullptr,
143 FormatToken **IfLeftBrace = nullptr,
144 bool *HasDoWhile = nullptr,
145 bool *HasLabel = nullptr);
146 bool tryToParseBracedList();
147 bool parseBracedList(bool IsAngleBracket = false, bool IsEnum = false);
148 bool parseParens(TokenType AmpAmpTokenType = TT_Unknown,
149 bool InMacroCall = false);
150 void parseSquare(bool LambdaIntroducer = false);
151 void keepAncestorBraces();
152 void parseUnbracedBody(bool CheckEOF = false);
153 void handleAttributes();
154 bool handleCppAttributes();
155 bool isBlockBegin(const FormatToken &Tok) const;
156 FormatToken *parseIfThenElse(IfStmtKind *IfKind, bool KeepBraces = false,
157 bool IsVerilogAssert = false);
158 void parseTryCatch();
159 void parseLoopBody(bool KeepBraces, bool WrapRightBrace);
160 void parseForOrWhileLoop(bool HasParens = true);
161 void parseDoWhile();
162 void parseLabel(bool LeftAlignLabel = false);
163 void parseCaseLabel();
164 void parseSwitch(bool IsExpr);
165 void parseNamespace();
166 bool parseModuleImport();
167 void parseNew();
168 void parseAccessSpecifier();
169 bool parseEnum();
170 bool parseStructLike();
171 bool parseRequires(bool SeenEqual);
172 void parseRequiresClause();
173 void parseRequiresExpression();
174 void parseConstraintExpression();
175 void parseCppExportBlock();
176 void parseNamespaceOrExportBlock(unsigned AddLevels);
177 void parseJavaEnumBody();
178 // Parses a record (aka class) as a top level element. If ParseAsExpr is true,
179 // parses the record as a child block, i.e. if the class declaration is an
180 // expression.
181 void parseRecord(bool ParseAsExpr = false, bool IsJavaRecord = false);
182 void parseObjCLightweightGenerics();
183 void parseObjCMethod();
184 void parseObjCProtocolList();
185 void parseObjCUntilAtEnd();
186 void parseObjCInterfaceOrImplementation();
187 bool parseObjCProtocol();
188 void parseJavaScriptEs6ImportExport();
189 void parseStatementMacro();
190 void parseCSharpAttribute();
191 // Parse a C# generic type constraint: `where T : IComparable<T>`.
192 // See:
193 // https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/keywords/where-generic-type-constraint
194 void parseCSharpGenericTypeConstraint();
195 bool tryToParseLambda();
196 bool tryToParseChildBlock();
197 bool tryToParseLambdaIntroducer();
198 bool tryToParsePropertyAccessor();
199 void tryToParseJSFunction();
200 bool tryToParseSimpleAttribute();
201 void parseVerilogHierarchyIdentifier();
202 void parseVerilogSensitivityList();
203 // Returns the number of levels of indentation in addition to the normal 1
204 // level for a block, used for indenting case labels.
205 unsigned parseVerilogHierarchyHeader();
206 void parseVerilogTable();
207 void parseVerilogCaseLabel();
208 // For import, export, and extern.
209 void parseVerilogExtern();
210 std::optional<llvm::SmallVector<llvm::SmallVector<FormatToken *, 8>, 1>>
211 parseMacroCall();
212
213 // Used by addUnwrappedLine to denote whether to keep or remove a level
214 // when resetting the line state.
215 enum class LineLevel { Remove, Keep };
216
217 void addUnwrappedLine(LineLevel AdjustLevel = LineLevel::Remove);
218 bool eof() const;
219 // LevelDifference is the difference of levels after and before the current
220 // token. For example:
221 // - if the token is '{' and opens a block, LevelDifference is 1.
222 // - if the token is '}' and closes a block, LevelDifference is -1.
223 void nextToken(int LevelDifference = 0);
224 void readToken(int LevelDifference = 0);
225
226 // Decides which comment tokens should be added to the current line and which
227 // should be added as comments before the next token.
228 //
229 // Comments specifies the sequence of comment tokens to analyze. They get
230 // either pushed to the current line or added to the comments before the next
231 // token.
232 //
233 // NextTok specifies the next token. A null pointer NextTok is supported, and
234 // signifies either the absence of a next token, or that the next token
235 // shouldn't be taken into account for the analysis.
236 void distributeComments(const ArrayRef<FormatToken *> &Comments,
237 const FormatToken *NextTok);
238
239 // Adds the comment preceding the next token to unwrapped lines.
240 void flushComments(bool NewlineBeforeNext);
241 void pushToken(FormatToken *Tok);
242 void calculateBraceTypes(bool ExpectClassBody = false);
243 void setPreviousRBraceType(TokenType Type);
244
245 // Marks a conditional compilation edge (for example, an '#if', '#ifdef',
246 // '#else' or merge conflict marker). If 'Unreachable' is true, assumes
247 // this branch either cannot be taken (for example '#if false'), or should
248 // not be taken in this round.
249 void conditionalCompilationCondition(bool Unreachable);
250 void conditionalCompilationStart(bool Unreachable);
251 void conditionalCompilationAlternative();
252 void conditionalCompilationEnd();
253
254 bool isOnNewLine(const FormatToken &FormatTok);
255
256 // Returns whether there is a macro expansion in the line, i.e. a token that
257 // was expanded from a macro call.
258 bool containsExpansion(const UnwrappedLine &Line) const;
259
260 // Compute hash of the current preprocessor branch.
261 // This is used to identify the different branches, and thus track if block
262 // open and close in the same branch.
263 size_t computePPHash() const;
264
265 bool parsingPPDirective() const { return CurrentLines != &Lines; }
266
267 // FIXME: We are constantly running into bugs where Line.Level is incorrectly
268 // subtracted from beyond 0. Introduce a method to subtract from Line.Level
269 // and use that everywhere in the Parser.
270 std::unique_ptr<UnwrappedLine> Line;
271
272 // Lines that are created by macro expansion.
273 // When formatting code containing macro calls, we first format the expanded
274 // lines to set the token types correctly. Afterwards, we format the
275 // reconstructed macro calls, re-using the token types determined in the first
276 // step.
277 // ExpandedLines will be reset every time we create a new LineAndExpansion
278 // instance once a line containing macro calls has been parsed.
279 SmallVector<UnwrappedLine, 8> CurrentExpandedLines;
280
281 // Maps from the first token of a top-level UnwrappedLine that contains
282 // a macro call to the replacement UnwrappedLines expanded from the macro
283 // call.
284 llvm::DenseMap<FormatToken *, SmallVector<UnwrappedLine, 8>> ExpandedLines;
285
286 // Map from the macro identifier to a line containing the full unexpanded
287 // macro call.
288 llvm::DenseMap<FormatToken *, std::unique_ptr<UnwrappedLine>> Unexpanded;
289
290 // For recursive macro expansions, trigger reconstruction only on the
291 // outermost expansion.
292 bool InExpansion = false;
293
294 // Set while we reconstruct a macro call.
295 // For reconstruction, we feed the expanded lines into the reconstructor
296 // until it is finished.
297 std::optional<MacroCallReconstructor> Reconstruct;
298
299 // Comments are sorted into unwrapped lines by whether they are in the same
300 // line as the previous token, or not. If not, they belong to the next token.
301 // Since the next token might already be in a new unwrapped line, we need to
302 // store the comments belonging to that token.
303 SmallVector<FormatToken *, 1> CommentsBeforeNextToken;
304
305 FormatToken *FormatTok = nullptr;
306
307 // Has just finished parsing a preprocessor line.
308 bool AtEndOfPPLine;
309
310 // The parsed lines. Only added to through \c CurrentLines.
311 SmallVector<UnwrappedLine, 8> Lines;
312
313 // Preprocessor directives are parsed out-of-order from other unwrapped lines.
314 // Thus, we need to keep a list of preprocessor directives to be reported
315 // after an unwrapped line that has been started was finished.
316 SmallVector<UnwrappedLine, 4> PreprocessorDirectives;
317
318 // New unwrapped lines are added via CurrentLines.
319 // Usually points to \c &Lines. While parsing a preprocessor directive when
320 // there is an unfinished previous unwrapped line, will point to
321 // \c &PreprocessorDirectives.
322 SmallVectorImpl<UnwrappedLine> *CurrentLines;
323
324 // We store for each line whether it must be a declaration depending on
325 // whether we are in a compound statement or not.
326 llvm::BitVector DeclarationScopeStack;
327
328 const FormatStyle &Style;
329 bool IsCpp;
330 LangOptions LangOpts;
331 const AdditionalKeywords &Keywords;
332
333 llvm::Regex CommentPragmasRegex;
334
335 FormatTokenSource *Tokens;
336 UnwrappedLineConsumer &Callback;
337
338 ArrayRef<FormatToken *> AllTokens;
339
340 // Keeps a stack of the states of nested control statements (true if the
341 // statement contains more than some predefined number of nested statements).
342 SmallVector<bool, 8> NestedTooDeep;
343
344 // Keeps a stack of the states of nested lambdas (true if the return type of
345 // the lambda is `decltype(auto)`).
346 SmallVector<bool, 4> NestedLambdas;
347
348 // Whether the parser is parsing the body of a function whose return type is
349 // `decltype(auto)`.
350 bool IsDecltypeAutoFunction = false;
351
352 // Represents preprocessor branch type, so we can find matching
353 // #if/#else/#endif directives.
354 enum PPBranchKind {
355 PP_Conditional, // Any #if, #ifdef, #ifndef, #elif, block outside #if 0
356 PP_Unreachable // #if 0 or a conditional preprocessor block inside #if 0
357 };
358
359 struct PPBranch {
360 PPBranch(PPBranchKind Kind, size_t Line) : Kind(Kind), Line(Line) {}
361 PPBranchKind Kind;
362 size_t Line;
363 };
364
365 // Keeps a stack of currently active preprocessor branching directives.
366 SmallVector<PPBranch, 16> PPStack;
367
368 // The \c UnwrappedLineParser re-parses the code for each combination
369 // of preprocessor branches that can be taken.
370 // To that end, we take the same branch (#if, #else, or one of the #elif
371 // branches) for each nesting level of preprocessor branches.
372 // \c PPBranchLevel stores the current nesting level of preprocessor
373 // branches during one pass over the code.
374 int PPBranchLevel;
375
376 // Contains the current branch (#if, #else or one of the #elif branches)
377 // for each nesting level.
378 SmallVector<int, 8> PPLevelBranchIndex;
379
380 // Contains the maximum number of branches at each nesting level.
381 SmallVector<int, 8> PPLevelBranchCount;
382
383 // Contains the number of branches per nesting level we are currently
384 // in while parsing a preprocessor branch sequence.
385 // This is used to update PPLevelBranchCount at the end of a branch
386 // sequence.
387 std::stack<int> PPChainBranchIndex;
388
389 // Include guard search state. Used to fixup preprocessor indent levels
390 // so that include guards do not participate in indentation.
391 enum IncludeGuardState {
392 IG_Inited, // Search started, looking for #ifndef.
393 IG_IfNdefed, // #ifndef found, IncludeGuardToken points to condition.
394 IG_Defined, // Matching #define found, checking other requirements.
395 IG_Found, // All requirements met, need to fix indents.
396 IG_Rejected, // Search failed or never started.
397 };
398
399 // Current state of include guard search.
400 IncludeGuardState IncludeGuard;
401
402 IncludeGuardState
403 getIncludeGuardState(FormatStyle::PPDirectiveIndentStyle Style) const {
404 return Style == FormatStyle::PPDIS_None || Style == FormatStyle::PPDIS_Leave
405 ? IG_Rejected
406 : IG_Inited;
407 }
408
409 // Points to the #ifndef condition for a potential include guard. Null unless
410 // IncludeGuardState == IG_IfNdefed.
411 FormatToken *IncludeGuardToken;
412
413 // Contains the first start column where the source begins. This is zero for
414 // normal source code and may be nonzero when formatting a code fragment that
415 // does not start at the beginning of the file.
416 unsigned FirstStartColumn;
417
418 MacroExpander Macros;
419
420 friend class ScopedLineState;
421 friend class CompoundStatementIndenter;
422};
423
424struct UnwrappedLineNode {
425 UnwrappedLineNode() : Tok(nullptr) {}
426 UnwrappedLineNode(FormatToken *Tok,
427 llvm::ArrayRef<UnwrappedLine> Children = {})
428 : Tok(Tok), Children(Children) {}
429
430 FormatToken *Tok;
431 SmallVector<UnwrappedLine, 0> Children;
432};
433
434std::ostream &operator<<(std::ostream &Stream, const UnwrappedLine &Line);
435
436} // end namespace format
437} // end namespace clang
438
439#endif
440