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 void parseSquare(bool LambdaIntroducer = false);
150 void keepAncestorBraces();
151 void parseUnbracedBody(bool CheckEOF = false);
152 void handleAttributes();
153 bool handleCppAttributes();
154 bool isBlockBegin(const FormatToken &Tok) const;
155 FormatToken *parseIfThenElse(IfStmtKind *IfKind, bool KeepBraces = false,
156 bool IsVerilogAssert = false);
157 void parseTryCatch();
158 void parseLoopBody(bool KeepBraces, bool WrapRightBrace);
159 void parseForOrWhileLoop(bool HasParens = true);
160 void parseDoWhile();
161 void parseLabel(bool LeftAlignLabel = false);
162 void parseCaseLabel();
163 void parseSwitch(bool IsExpr);
164 void parseNamespace();
165 bool parseModuleImport();
166 void parseNew();
167 void parseAccessSpecifier();
168 bool parseEnum();
169 bool parseStructLike();
170 bool parseRequires(bool SeenEqual);
171 void parseRequiresClause(FormatToken *RequiresToken);
172 void parseRequiresExpression(FormatToken *RequiresToken);
173 void parseConstraintExpression();
174 void parseCppExportBlock();
175 void parseNamespaceOrExportBlock(unsigned AddLevels);
176 void parseJavaEnumBody();
177 // Parses a record (aka class) as a top level element. If ParseAsExpr is true,
178 // parses the record as a child block, i.e. if the class declaration is an
179 // expression.
180 void parseRecord(bool ParseAsExpr = false, bool IsJavaRecord = false);
181 void parseObjCLightweightGenerics();
182 void parseObjCMethod();
183 void parseObjCProtocolList();
184 void parseObjCUntilAtEnd();
185 void parseObjCInterfaceOrImplementation();
186 bool parseObjCProtocol();
187 void parseJavaScriptEs6ImportExport();
188 void parseStatementMacro();
189 void parseCSharpAttribute();
190 // Parse a C# generic type constraint: `where T : IComparable<T>`.
191 // See:
192 // https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/keywords/where-generic-type-constraint
193 void parseCSharpGenericTypeConstraint();
194 bool tryToParseLambda();
195 bool tryToParseChildBlock();
196 bool tryToParseLambdaIntroducer();
197 bool tryToParsePropertyAccessor();
198 void tryToParseJSFunction();
199 bool tryToParseSimpleAttribute();
200 void parseVerilogHierarchyIdentifier();
201 void parseVerilogSensitivityList();
202 // Returns the number of levels of indentation in addition to the normal 1
203 // level for a block, used for indenting case labels.
204 unsigned parseVerilogHierarchyHeader();
205 void parseVerilogTable();
206 void parseVerilogCaseLabel();
207 std::optional<llvm::SmallVector<llvm::SmallVector<FormatToken *, 8>, 1>>
208 parseMacroCall();
209
210 // Used by addUnwrappedLine to denote whether to keep or remove a level
211 // when resetting the line state.
212 enum class LineLevel { Remove, Keep };
213
214 void addUnwrappedLine(LineLevel AdjustLevel = LineLevel::Remove);
215 bool eof() const;
216 // LevelDifference is the difference of levels after and before the current
217 // token. For example:
218 // - if the token is '{' and opens a block, LevelDifference is 1.
219 // - if the token is '}' and closes a block, LevelDifference is -1.
220 void nextToken(int LevelDifference = 0);
221 void readToken(int LevelDifference = 0);
222
223 // Decides which comment tokens should be added to the current line and which
224 // should be added as comments before the next token.
225 //
226 // Comments specifies the sequence of comment tokens to analyze. They get
227 // either pushed to the current line or added to the comments before the next
228 // token.
229 //
230 // NextTok specifies the next token. A null pointer NextTok is supported, and
231 // signifies either the absence of a next token, or that the next token
232 // shouldn't be taken into account for the analysis.
233 void distributeComments(const ArrayRef<FormatToken *> &Comments,
234 const FormatToken *NextTok);
235
236 // Adds the comment preceding the next token to unwrapped lines.
237 void flushComments(bool NewlineBeforeNext);
238 void pushToken(FormatToken *Tok);
239 void calculateBraceTypes(bool ExpectClassBody = false);
240 void setPreviousRBraceType(TokenType Type);
241
242 // Marks a conditional compilation edge (for example, an '#if', '#ifdef',
243 // '#else' or merge conflict marker). If 'Unreachable' is true, assumes
244 // this branch either cannot be taken (for example '#if false'), or should
245 // not be taken in this round.
246 void conditionalCompilationCondition(bool Unreachable);
247 void conditionalCompilationStart(bool Unreachable);
248 void conditionalCompilationAlternative();
249 void conditionalCompilationEnd();
250
251 bool isOnNewLine(const FormatToken &FormatTok);
252
253 // Returns whether there is a macro expansion in the line, i.e. a token that
254 // was expanded from a macro call.
255 bool containsExpansion(const UnwrappedLine &Line) const;
256
257 // Compute hash of the current preprocessor branch.
258 // This is used to identify the different branches, and thus track if block
259 // open and close in the same branch.
260 size_t computePPHash() const;
261
262 bool parsingPPDirective() const { return CurrentLines != &Lines; }
263
264 // FIXME: We are constantly running into bugs where Line.Level is incorrectly
265 // subtracted from beyond 0. Introduce a method to subtract from Line.Level
266 // and use that everywhere in the Parser.
267 std::unique_ptr<UnwrappedLine> Line;
268
269 // Lines that are created by macro expansion.
270 // When formatting code containing macro calls, we first format the expanded
271 // lines to set the token types correctly. Afterwards, we format the
272 // reconstructed macro calls, re-using the token types determined in the first
273 // step.
274 // ExpandedLines will be reset every time we create a new LineAndExpansion
275 // instance once a line containing macro calls has been parsed.
276 SmallVector<UnwrappedLine, 8> CurrentExpandedLines;
277
278 // Maps from the first token of a top-level UnwrappedLine that contains
279 // a macro call to the replacement UnwrappedLines expanded from the macro
280 // call.
281 llvm::DenseMap<FormatToken *, SmallVector<UnwrappedLine, 8>> ExpandedLines;
282
283 // Map from the macro identifier to a line containing the full unexpanded
284 // macro call.
285 llvm::DenseMap<FormatToken *, std::unique_ptr<UnwrappedLine>> Unexpanded;
286
287 // For recursive macro expansions, trigger reconstruction only on the
288 // outermost expansion.
289 bool InExpansion = false;
290
291 // Set while we reconstruct a macro call.
292 // For reconstruction, we feed the expanded lines into the reconstructor
293 // until it is finished.
294 std::optional<MacroCallReconstructor> Reconstruct;
295
296 // Comments are sorted into unwrapped lines by whether they are in the same
297 // line as the previous token, or not. If not, they belong to the next token.
298 // Since the next token might already be in a new unwrapped line, we need to
299 // store the comments belonging to that token.
300 SmallVector<FormatToken *, 1> CommentsBeforeNextToken;
301
302 FormatToken *FormatTok = nullptr;
303
304 // Has just finished parsing a preprocessor line.
305 bool AtEndOfPPLine;
306
307 // The parsed lines. Only added to through \c CurrentLines.
308 SmallVector<UnwrappedLine, 8> Lines;
309
310 // Preprocessor directives are parsed out-of-order from other unwrapped lines.
311 // Thus, we need to keep a list of preprocessor directives to be reported
312 // after an unwrapped line that has been started was finished.
313 SmallVector<UnwrappedLine, 4> PreprocessorDirectives;
314
315 // New unwrapped lines are added via CurrentLines.
316 // Usually points to \c &Lines. While parsing a preprocessor directive when
317 // there is an unfinished previous unwrapped line, will point to
318 // \c &PreprocessorDirectives.
319 SmallVectorImpl<UnwrappedLine> *CurrentLines;
320
321 // We store for each line whether it must be a declaration depending on
322 // whether we are in a compound statement or not.
323 llvm::BitVector DeclarationScopeStack;
324
325 const FormatStyle &Style;
326 bool IsCpp;
327 LangOptions LangOpts;
328 const AdditionalKeywords &Keywords;
329
330 llvm::Regex CommentPragmasRegex;
331
332 FormatTokenSource *Tokens;
333 UnwrappedLineConsumer &Callback;
334
335 ArrayRef<FormatToken *> AllTokens;
336
337 // Keeps a stack of the states of nested control statements (true if the
338 // statement contains more than some predefined number of nested statements).
339 SmallVector<bool, 8> NestedTooDeep;
340
341 // Keeps a stack of the states of nested lambdas (true if the return type of
342 // the lambda is `decltype(auto)`).
343 SmallVector<bool, 4> NestedLambdas;
344
345 // Whether the parser is parsing the body of a function whose return type is
346 // `decltype(auto)`.
347 bool IsDecltypeAutoFunction = false;
348
349 // Represents preprocessor branch type, so we can find matching
350 // #if/#else/#endif directives.
351 enum PPBranchKind {
352 PP_Conditional, // Any #if, #ifdef, #ifndef, #elif, block outside #if 0
353 PP_Unreachable // #if 0 or a conditional preprocessor block inside #if 0
354 };
355
356 struct PPBranch {
357 PPBranch(PPBranchKind Kind, size_t Line) : Kind(Kind), Line(Line) {}
358 PPBranchKind Kind;
359 size_t Line;
360 };
361
362 // Keeps a stack of currently active preprocessor branching directives.
363 SmallVector<PPBranch, 16> PPStack;
364
365 // The \c UnwrappedLineParser re-parses the code for each combination
366 // of preprocessor branches that can be taken.
367 // To that end, we take the same branch (#if, #else, or one of the #elif
368 // branches) for each nesting level of preprocessor branches.
369 // \c PPBranchLevel stores the current nesting level of preprocessor
370 // branches during one pass over the code.
371 int PPBranchLevel;
372
373 // Contains the current branch (#if, #else or one of the #elif branches)
374 // for each nesting level.
375 SmallVector<int, 8> PPLevelBranchIndex;
376
377 // Contains the maximum number of branches at each nesting level.
378 SmallVector<int, 8> PPLevelBranchCount;
379
380 // Contains the number of branches per nesting level we are currently
381 // in while parsing a preprocessor branch sequence.
382 // This is used to update PPLevelBranchCount at the end of a branch
383 // sequence.
384 std::stack<int> PPChainBranchIndex;
385
386 // Include guard search state. Used to fixup preprocessor indent levels
387 // so that include guards do not participate in indentation.
388 enum IncludeGuardState {
389 IG_Inited, // Search started, looking for #ifndef.
390 IG_IfNdefed, // #ifndef found, IncludeGuardToken points to condition.
391 IG_Defined, // Matching #define found, checking other requirements.
392 IG_Found, // All requirements met, need to fix indents.
393 IG_Rejected, // Search failed or never started.
394 };
395
396 // Current state of include guard search.
397 IncludeGuardState IncludeGuard;
398
399 // Points to the #ifndef condition for a potential include guard. Null unless
400 // IncludeGuardState == IG_IfNdefed.
401 FormatToken *IncludeGuardToken;
402
403 // Contains the first start column where the source begins. This is zero for
404 // normal source code and may be nonzero when formatting a code fragment that
405 // does not start at the beginning of the file.
406 unsigned FirstStartColumn;
407
408 MacroExpander Macros;
409
410 friend class ScopedLineState;
411 friend class CompoundStatementIndenter;
412};
413
414struct UnwrappedLineNode {
415 UnwrappedLineNode() : Tok(nullptr) {}
416 UnwrappedLineNode(FormatToken *Tok,
417 llvm::ArrayRef<UnwrappedLine> Children = {})
418 : Tok(Tok), Children(Children) {}
419
420 FormatToken *Tok;
421 SmallVector<UnwrappedLine, 0> Children;
422};
423
424std::ostream &operator<<(std::ostream &Stream, const UnwrappedLine &Line);
425
426} // end namespace format
427} // end namespace clang
428
429#endif
430