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