UnwrappedLineParser.h revision 256281 
1//===--- UnwrappedLineParser.h - Format C++ code ----------------*- C++ -*-===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9///
10/// \file
11/// \brief This file contains the declaration of the UnwrappedLineParser,
12/// which turns a stream of tokens into UnwrappedLines.
13///
14//===----------------------------------------------------------------------===//
15
16#ifndef LLVM_CLANG_FORMAT_UNWRAPPED_LINE_PARSER_H
17#define LLVM_CLANG_FORMAT_UNWRAPPED_LINE_PARSER_H
18
19#include "clang/Basic/IdentifierTable.h"
20#include "clang/Basic/SourceManager.h"
21#include "clang/Format/Format.h"
22#include "clang/Lex/Lexer.h"
23#include <list>
24
25namespace clang {
26
27class DiagnosticsEngine;
28
29namespace format {
30
31/// \brief A wrapper around a \c Token storing information about the
32/// whitespace characters preceeding it.
33struct FormatToken {
34  FormatToken()
35      : NewlinesBefore(0), HasUnescapedNewline(false), WhiteSpaceLength(0),
36        LastNewlineOffset(0), TokenLength(0), IsFirst(false),
37        MustBreakBefore(false), TrailingWhiteSpaceLength(0) {}
38
39  /// \brief The \c Token.
40  Token Tok;
41
42  /// \brief The number of newlines immediately before the \c Token.
43  ///
44  /// This can be used to determine what the user wrote in the original code
45  /// and thereby e.g. leave an empty line between two function definitions.
46  unsigned NewlinesBefore;
47
48  /// \brief Whether there is at least one unescaped newline before the \c
49  /// Token.
50  bool HasUnescapedNewline;
51
52  /// \brief The location of the start of the whitespace immediately preceeding
53  /// the \c Token.
54  ///
55  /// Used together with \c WhiteSpaceLength to create a \c Replacement.
56  SourceLocation WhiteSpaceStart;
57
58  /// \brief The length in characters of the whitespace immediately preceeding
59  /// the \c Token.
60  unsigned WhiteSpaceLength;
61
62  /// \brief The offset just past the last '\n' in this token's leading
63  /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'.
64  unsigned LastNewlineOffset;
65
66  /// \brief The length of the non-whitespace parts of the token. This is
67  /// necessary because we need to handle escaped newlines that are stored
68  /// with the token.
69  unsigned TokenLength;
70
71  /// \brief Indicates that this is the first token.
72  bool IsFirst;
73
74  /// \brief Whether there must be a line break before this token.
75  ///
76  /// This happens for example when a preprocessor directive ended directly
77  /// before the token.
78  bool MustBreakBefore;
79
80  /// \brief Number of characters of trailing whitespace.
81  unsigned TrailingWhiteSpaceLength;
82
83  /// \brief Returns actual token start location without leading escaped
84  /// newlines and whitespace.
85  ///
86  /// This can be different to Tok.getLocation(), which includes leading escaped
87  /// newlines.
88  SourceLocation getStartOfNonWhitespace() const {
89    return WhiteSpaceStart.getLocWithOffset(WhiteSpaceLength);
90  }
91};
92
93/// \brief An unwrapped line is a sequence of \c Token, that we would like to
94/// put on a single line if there was no column limit.
95///
96/// This is used as a main interface between the \c UnwrappedLineParser and the
97/// \c UnwrappedLineFormatter. The key property is that changing the formatting
98/// within an unwrapped line does not affect any other unwrapped lines.
99struct UnwrappedLine {
100  UnwrappedLine() : Level(0), InPPDirective(false), MustBeDeclaration(false) {
101  }
102
103  // FIXME: Don't use std::list here.
104  /// \brief The \c Tokens comprising this \c UnwrappedLine.
105  std::list<FormatToken> Tokens;
106
107  /// \brief The indent level of the \c UnwrappedLine.
108  unsigned Level;
109
110  /// \brief Whether this \c UnwrappedLine is part of a preprocessor directive.
111  bool InPPDirective;
112
113  bool MustBeDeclaration;
114};
115
116class UnwrappedLineConsumer {
117public:
118  virtual ~UnwrappedLineConsumer() {
119  }
120  virtual void consumeUnwrappedLine(const UnwrappedLine &Line) = 0;
121};
122
123class FormatTokenSource {
124public:
125  virtual ~FormatTokenSource() {
126  }
127  virtual FormatToken getNextToken() = 0;
128};
129
130class UnwrappedLineParser {
131public:
132  UnwrappedLineParser(clang::DiagnosticsEngine &Diag, const FormatStyle &Style,
133                      FormatTokenSource &Tokens,
134                      UnwrappedLineConsumer &Callback);
135
136  /// Returns true in case of a structural error.
137  bool parse();
138
139private:
140  void parseFile();
141  void parseLevel(bool HasOpeningBrace);
142  void parseBlock(bool MustBeDeclaration, unsigned AddLevels = 1);
143  void parsePPDirective();
144  void parsePPDefine();
145  void parsePPUnknown();
146  void parseStructuralElement();
147  void parseBracedList();
148  void parseReturn();
149  void parseParens();
150  void parseIfThenElse();
151  void parseForOrWhileLoop();
152  void parseDoWhile();
153  void parseLabel();
154  void parseCaseLabel();
155  void parseSwitch();
156  void parseNamespace();
157  void parseAccessSpecifier();
158  void parseEnum();
159  void parseRecord();
160  void parseObjCProtocolList();
161  void parseObjCUntilAtEnd();
162  void parseObjCInterfaceOrImplementation();
163  void parseObjCProtocol();
164  void addUnwrappedLine();
165  bool eof() const;
166  void nextToken();
167  void readToken();
168  void flushComments(bool NewlineBeforeNext);
169  void pushToken(const FormatToken &Tok);
170
171  // FIXME: We are constantly running into bugs where Line.Level is incorrectly
172  // subtracted from beyond 0. Introduce a method to subtract from Line.Level
173  // and use that everywhere in the Parser.
174  OwningPtr<UnwrappedLine> Line;
175
176  // Comments are sorted into unwrapped lines by whether they are in the same
177  // line as the previous token, or not. If not, they belong to the next token.
178  // Since the next token might already be in a new unwrapped line, we need to
179  // store the comments belonging to that token.
180  SmallVector<FormatToken, 1> CommentsBeforeNextToken;
181  FormatToken FormatTok;
182  bool MustBreakBeforeNextToken;
183
184  // The parsed lines. Only added to through \c CurrentLines.
185  std::vector<UnwrappedLine> Lines;
186
187  // Preprocessor directives are parsed out-of-order from other unwrapped lines.
188  // Thus, we need to keep a list of preprocessor directives to be reported
189  // after an unwarpped line that has been started was finished.
190  std::vector<UnwrappedLine> PreprocessorDirectives;
191
192  // New unwrapped lines are added via CurrentLines.
193  // Usually points to \c &Lines. While parsing a preprocessor directive when
194  // there is an unfinished previous unwrapped line, will point to
195  // \c &PreprocessorDirectives.
196  std::vector<UnwrappedLine> *CurrentLines;
197
198  // We store for each line whether it must be a declaration depending on
199  // whether we are in a compound statement or not.
200  std::vector<bool> DeclarationScopeStack;
201
202  // Will be true if we encounter an error that leads to possibily incorrect
203  // indentation levels.
204  bool StructuralError;
205
206  clang::DiagnosticsEngine &Diag;
207  const FormatStyle &Style;
208  FormatTokenSource *Tokens;
209  UnwrappedLineConsumer &Callback;
210
211  friend class ScopedLineState;
212};
213
214} // end namespace format
215} // end namespace clang
216
217#endif // LLVM_CLANG_FORMAT_UNWRAPPED_LINE_PARSER_H
218