1//===--- RawCommentList.h - Classes for processing raw comments -*- 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#ifndef LLVM_CLANG_AST_RAWCOMMENTLIST_H
10#define LLVM_CLANG_AST_RAWCOMMENTLIST_H
11
12#include "clang/Basic/CommentOptions.h"
13#include "clang/Basic/SourceLocation.h"
14#include "llvm/ADT/ArrayRef.h"
15#include "llvm/ADT/DenseMap.h"
16#include "llvm/Support/Allocator.h"
17#include <map>
18
19namespace clang {
20
21class ASTContext;
22class ASTReader;
23class Decl;
24class DiagnosticsEngine;
25class Preprocessor;
26class SourceManager;
27
28namespace comments {
29 class FullComment;
30} // end namespace comments
31
32class RawComment {
33public:
34 enum CommentKind {
35 RCK_Invalid, ///< Invalid comment
36 RCK_OrdinaryBCPL, ///< Any normal BCPL comments
37 RCK_OrdinaryC, ///< Any normal C comment
38 RCK_BCPLSlash, ///< \code /// stuff \endcode
39 RCK_BCPLExcl, ///< \code //! stuff \endcode
40 RCK_JavaDoc, ///< \code /** stuff */ \endcode
41 RCK_Qt, ///< \code /*! stuff */ \endcode, also used by HeaderDoc
42 RCK_Merged ///< Two or more documentation comments merged together
43 };
44
45 RawComment() : Kind(RCK_Invalid), IsAlmostTrailingComment(false) { }
46
47 RawComment(const SourceManager &SourceMgr, SourceRange SR,
48 const CommentOptions &CommentOpts, bool Merged);
49
50 CommentKind getKind() const LLVM_READONLY {
51 return (CommentKind) Kind;
52 }
53
54 bool isInvalid() const LLVM_READONLY {
55 return Kind == RCK_Invalid;
56 }
57
58 bool isMerged() const LLVM_READONLY {
59 return Kind == RCK_Merged;
60 }
61
62 /// Is this comment attached to any declaration?
63 bool isAttached() const LLVM_READONLY {
64 return IsAttached;
65 }
66
67 void setAttached() {
68 IsAttached = true;
69 }
70
71 /// Returns true if it is a comment that should be put after a member:
72 /// \code ///< stuff \endcode
73 /// \code //!< stuff \endcode
74 /// \code /**< stuff */ \endcode
75 /// \code /*!< stuff */ \endcode
76 bool isTrailingComment() const LLVM_READONLY {
77 return IsTrailingComment;
78 }
79
80 /// Returns true if it is a probable typo:
81 /// \code //< stuff \endcode
82 /// \code /*< stuff */ \endcode
83 bool isAlmostTrailingComment() const LLVM_READONLY {
84 return IsAlmostTrailingComment;
85 }
86
87 /// Returns true if this comment is not a documentation comment.
88 bool isOrdinary() const LLVM_READONLY {
89 return ((Kind == RCK_OrdinaryBCPL) || (Kind == RCK_OrdinaryC));
90 }
91
92 /// Returns true if this comment any kind of a documentation comment.
93 bool isDocumentation() const LLVM_READONLY {
94 return !isInvalid() && !isOrdinary();
95 }
96
97 /// Returns raw comment text with comment markers.
98 StringRef getRawText(const SourceManager &SourceMgr) const {
99 if (RawTextValid)
100 return RawText;
101
102 RawText = getRawTextSlow(SourceMgr);
103 RawTextValid = true;
104 return RawText;
105 }
106
107 SourceRange getSourceRange() const LLVM_READONLY { return Range; }
108 SourceLocation getBeginLoc() const LLVM_READONLY { return Range.getBegin(); }
109 SourceLocation getEndLoc() const LLVM_READONLY { return Range.getEnd(); }
110
111 const char *getBriefText(const ASTContext &Context) const {
112 if (BriefTextValid)
113 return BriefText;
114
115 return extractBriefText(Context);
116 }
117
118 bool hasUnsupportedSplice(const SourceManager &SourceMgr) const {
119 if (!isInvalid())
120 return false;
121 StringRef Text = getRawText(SourceMgr);
122 if (Text.size() < 6 || Text[0] != '/')
123 return false;
124 if (Text[1] == '*')
125 return Text[Text.size() - 1] != '/' || Text[Text.size() - 2] != '*';
126 return Text[1] != '/';
127 }
128
129 /// Returns sanitized comment text, suitable for presentation in editor UIs.
130 /// E.g. will transform:
131 /// // This is a long multiline comment.
132 /// // Parts of it might be indented.
133 /// /* The comments styles might be mixed. */
134 /// into
135 /// "This is a long multiline comment.\n"
136 /// " Parts of it might be indented.\n"
137 /// "The comments styles might be mixed."
138 /// Also removes leading indentation and sanitizes some common cases:
139 /// /* This is a first line.
140 /// * This is a second line. It is indented.
141 /// * This is a third line. */
142 /// and
143 /// /* This is a first line.
144 /// This is a second line. It is indented.
145 /// This is a third line. */
146 /// will both turn into:
147 /// "This is a first line.\n"
148 /// " This is a second line. It is indented.\n"
149 /// "This is a third line."
150 std::string getFormattedText(const SourceManager &SourceMgr,
151 DiagnosticsEngine &Diags) const;
152
153 struct CommentLine {
154 std::string Text;
155 PresumedLoc Begin;
156 PresumedLoc End;
157
158 CommentLine(StringRef Text, PresumedLoc Begin, PresumedLoc End)
159 : Text(Text), Begin(Begin), End(End) {}
160 };
161
162 /// Returns sanitized comment text as separated lines with locations in
163 /// source, suitable for further processing and rendering requiring source
164 /// locations.
165 std::vector<CommentLine> getFormattedLines(const SourceManager &SourceMgr,
166 DiagnosticsEngine &Diags) const;
167
168 /// Parse the comment, assuming it is attached to decl \c D.
169 comments::FullComment *parse(const ASTContext &Context,
170 const Preprocessor *PP, const Decl *D) const;
171
172private:
173 SourceRange Range;
174
175 mutable StringRef RawText;
176 mutable const char *BriefText = nullptr;
177
178 LLVM_PREFERRED_TYPE(bool)
179 mutable unsigned RawTextValid : 1;
180 LLVM_PREFERRED_TYPE(bool)
181 mutable unsigned BriefTextValid : 1;
182
183 LLVM_PREFERRED_TYPE(CommentKind)
184 unsigned Kind : 3;
185
186 /// True if comment is attached to a declaration in ASTContext.
187 LLVM_PREFERRED_TYPE(bool)
188 unsigned IsAttached : 1;
189
190 LLVM_PREFERRED_TYPE(bool)
191 unsigned IsTrailingComment : 1;
192 LLVM_PREFERRED_TYPE(bool)
193 unsigned IsAlmostTrailingComment : 1;
194
195 /// Constructor for AST deserialization.
196 RawComment(SourceRange SR, CommentKind K, bool IsTrailingComment,
197 bool IsAlmostTrailingComment) :
198 Range(SR), RawTextValid(false), BriefTextValid(false), Kind(K),
199 IsAttached(false), IsTrailingComment(IsTrailingComment),
200 IsAlmostTrailingComment(IsAlmostTrailingComment)
201 { }
202
203 StringRef getRawTextSlow(const SourceManager &SourceMgr) const;
204
205 const char *extractBriefText(const ASTContext &Context) const;
206
207 friend class ASTReader;
208};
209
210/// This class represents all comments included in the translation unit,
211/// sorted in order of appearance in the translation unit.
212class RawCommentList {
213public:
214 RawCommentList(SourceManager &SourceMgr) : SourceMgr(SourceMgr) {}
215
216 void addComment(const RawComment &RC, const CommentOptions &CommentOpts,
217 llvm::BumpPtrAllocator &Allocator);
218
219 /// \returns A mapping from an offset of the start of the comment to the
220 /// comment itself, or nullptr in case there are no comments in \p File.
221 const std::map<unsigned, RawComment *> *getCommentsInFile(FileID File) const;
222
223 bool empty() const;
224
225 unsigned getCommentBeginLine(RawComment *C, FileID File,
226 unsigned Offset) const;
227 unsigned getCommentEndOffset(RawComment *C) const;
228
229private:
230 SourceManager &SourceMgr;
231 // mapping: FileId -> comment begin offset -> comment
232 llvm::DenseMap<FileID, std::map<unsigned, RawComment *>> OrderedComments;
233 mutable llvm::DenseMap<RawComment *, unsigned> CommentBeginLine;
234 mutable llvm::DenseMap<RawComment *, unsigned> CommentEndOffset;
235
236 friend class ASTReader;
237 friend class ASTWriter;
238};
239
240} // end namespace clang
241
242#endif
243