Dmitri Gribenko | aab8383 | 2012-06-20 00:34:58 +0000 | [diff] [blame] | 1 | //===--- RawCommentList.cpp - Processing raw comments -----------*- 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 | |
Chandler Carruth | 39a3e75 | 2012-06-20 09:53:52 +0000 | [diff] [blame] | 10 | #include "clang/AST/RawCommentList.h" |
Dmitri Gribenko | aab8383 | 2012-06-20 00:34:58 +0000 | [diff] [blame] | 11 | #include "llvm/ADT/STLExtras.h" |
| 12 | |
| 13 | using namespace clang; |
| 14 | |
| 15 | namespace { |
| 16 | /// Get comment kind and bool describing if it is a trailing comment. |
| 17 | std::pair<RawComment::CommentKind, bool> getCommentKind(StringRef Comment) { |
| 18 | if (Comment.size() < 3 || Comment[0] != '/') |
| 19 | return std::make_pair(RawComment::CK_Invalid, false); |
| 20 | |
| 21 | RawComment::CommentKind K; |
| 22 | if (Comment[1] == '/') { |
| 23 | if (Comment.size() < 3) |
| 24 | return std::make_pair(RawComment::CK_OrdinaryBCPL, false); |
| 25 | |
| 26 | if (Comment[2] == '/') |
| 27 | K = RawComment::CK_BCPLSlash; |
| 28 | else if (Comment[2] == '!') |
| 29 | K = RawComment::CK_BCPLExcl; |
| 30 | else |
| 31 | return std::make_pair(RawComment::CK_OrdinaryBCPL, false); |
| 32 | } else { |
| 33 | assert(Comment.size() >= 4); |
| 34 | |
| 35 | // Comment lexer does not understand escapes in comment markers, so pretend |
| 36 | // that this is not a comment. |
| 37 | if (Comment[1] != '*' || |
| 38 | Comment[Comment.size() - 2] != '*' || |
| 39 | Comment[Comment.size() - 1] != '/') |
| 40 | return std::make_pair(RawComment::CK_Invalid, false); |
| 41 | |
| 42 | if (Comment[2] == '*') |
| 43 | K = RawComment::CK_JavaDoc; |
| 44 | else if (Comment[2] == '!') |
| 45 | K = RawComment::CK_Qt; |
| 46 | else |
| 47 | return std::make_pair(RawComment::CK_OrdinaryC, false); |
| 48 | } |
| 49 | const bool TrailingComment = (Comment.size() > 3) && (Comment[3] == '<'); |
| 50 | return std::make_pair(K, TrailingComment); |
| 51 | } |
| 52 | |
| 53 | bool mergedCommentIsTrailingComment(StringRef Comment) { |
| 54 | return (Comment.size() > 3) && (Comment[3] == '<'); |
| 55 | } |
| 56 | } // unnamed namespace |
| 57 | |
| 58 | RawComment::RawComment(const SourceManager &SourceMgr, SourceRange SR, |
| 59 | bool Merged) : |
| 60 | Range(SR), RawTextValid(false), IsAlmostTrailingComment(false), |
| 61 | BeginLineValid(false), EndLineValid(false) { |
| 62 | // Extract raw comment text, if possible. |
Dmitri Gribenko | fecc2e0 | 2012-06-21 21:02:45 +0000 | [diff] [blame] | 63 | if (SR.getBegin() == SR.getEnd() || getRawText(SourceMgr).empty()) { |
Dmitri Gribenko | aab8383 | 2012-06-20 00:34:58 +0000 | [diff] [blame] | 64 | Kind = CK_Invalid; |
| 65 | return; |
| 66 | } |
| 67 | |
| 68 | if (!Merged) { |
| 69 | // Guess comment kind. |
| 70 | std::pair<CommentKind, bool> K = getCommentKind(RawText); |
| 71 | Kind = K.first; |
| 72 | IsTrailingComment = K.second; |
| 73 | |
| 74 | IsAlmostTrailingComment = RawText.startswith("//<") || |
| 75 | RawText.startswith("/*<"); |
| 76 | } else { |
| 77 | Kind = CK_Merged; |
| 78 | IsTrailingComment = mergedCommentIsTrailingComment(RawText); |
| 79 | } |
| 80 | } |
| 81 | |
| 82 | unsigned RawComment::getBeginLine(const SourceManager &SM) const { |
| 83 | if (BeginLineValid) |
| 84 | return BeginLine; |
| 85 | |
| 86 | std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Range.getBegin()); |
| 87 | BeginLine = SM.getLineNumber(LocInfo.first, LocInfo.second); |
| 88 | BeginLineValid = true; |
| 89 | return BeginLine; |
| 90 | } |
| 91 | |
| 92 | unsigned RawComment::getEndLine(const SourceManager &SM) const { |
| 93 | if (EndLineValid) |
| 94 | return EndLine; |
| 95 | |
| 96 | std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Range.getEnd()); |
| 97 | EndLine = SM.getLineNumber(LocInfo.first, LocInfo.second); |
| 98 | EndLineValid = true; |
| 99 | return EndLine; |
| 100 | } |
| 101 | |
| 102 | StringRef RawComment::getRawTextSlow(const SourceManager &SourceMgr) const { |
| 103 | FileID BeginFileID; |
| 104 | FileID EndFileID; |
| 105 | unsigned BeginOffset; |
| 106 | unsigned EndOffset; |
| 107 | |
| 108 | llvm::tie(BeginFileID, BeginOffset) = |
| 109 | SourceMgr.getDecomposedLoc(Range.getBegin()); |
| 110 | llvm::tie(EndFileID, EndOffset) = |
| 111 | SourceMgr.getDecomposedLoc(Range.getEnd()); |
| 112 | |
| 113 | const unsigned Length = EndOffset - BeginOffset; |
| 114 | if (Length < 2) |
| 115 | return StringRef(); |
| 116 | |
| 117 | // The comment can't begin in one file and end in another. |
| 118 | assert(BeginFileID == EndFileID); |
| 119 | |
| 120 | bool Invalid = false; |
| 121 | const char *BufferStart = SourceMgr.getBufferData(BeginFileID, |
| 122 | &Invalid).data(); |
| 123 | if (Invalid) |
| 124 | return StringRef(); |
| 125 | |
| 126 | return StringRef(BufferStart + BeginOffset, Length); |
| 127 | } |
| 128 | |
| 129 | namespace { |
| 130 | bool containsOnlyWhitespace(StringRef Str) { |
| 131 | return Str.find_first_not_of(" \t\f\v\r\n") == StringRef::npos; |
| 132 | } |
| 133 | |
| 134 | bool onlyWhitespaceBetweenComments(SourceManager &SM, |
| 135 | const RawComment &C1, const RawComment &C2) { |
| 136 | std::pair<FileID, unsigned> C1EndLocInfo = SM.getDecomposedLoc( |
| 137 | C1.getSourceRange().getEnd()); |
| 138 | std::pair<FileID, unsigned> C2BeginLocInfo = SM.getDecomposedLoc( |
| 139 | C2.getSourceRange().getBegin()); |
| 140 | |
| 141 | // Question does not make sense if comments are located in different files. |
| 142 | if (C1EndLocInfo.first != C2BeginLocInfo.first) |
| 143 | return false; |
| 144 | |
| 145 | bool Invalid = false; |
| 146 | const char *Buffer = SM.getBufferData(C1EndLocInfo.first, &Invalid).data(); |
| 147 | if (Invalid) |
| 148 | return false; |
| 149 | |
| 150 | StringRef TextBetweenComments(Buffer + C1EndLocInfo.second, |
| 151 | C2BeginLocInfo.second - C1EndLocInfo.second); |
| 152 | |
| 153 | return containsOnlyWhitespace(TextBetweenComments); |
| 154 | } |
| 155 | } // unnamed namespace |
| 156 | |
Dmitri Gribenko | 93b9ecb | 2012-06-20 20:39:04 +0000 | [diff] [blame] | 157 | void RawCommentList::addComment(const RawComment &RC) { |
Dmitri Gribenko | aab8383 | 2012-06-20 00:34:58 +0000 | [diff] [blame] | 158 | if (RC.isInvalid()) |
| 159 | return; |
| 160 | |
Dmitri Gribenko | 9230740 | 2012-06-21 22:04:37 +0000 | [diff] [blame] | 161 | // Check if the comments are not in source order. |
| 162 | while (!Comments.empty() && |
| 163 | !SourceMgr.isBeforeInTranslationUnit( |
| 164 | Comments.back().getSourceRange().getBegin(), |
| 165 | RC.getSourceRange().getBegin())) { |
| 166 | // If they are, just pop a few last comments that don't fit. |
| 167 | // This happens if an \#include directive contains comments. |
| 168 | Comments.pop_back(); |
| 169 | } |
Dmitri Gribenko | aab8383 | 2012-06-20 00:34:58 +0000 | [diff] [blame] | 170 | |
| 171 | if (OnlyWhitespaceSeen) { |
| 172 | if (!onlyWhitespaceBetweenComments(SourceMgr, LastComment, RC)) |
| 173 | OnlyWhitespaceSeen = false; |
| 174 | } |
| 175 | |
| 176 | LastComment = RC; |
| 177 | |
| 178 | // Ordinary comments are not interesting for us. |
| 179 | if (RC.isOrdinary()) |
| 180 | return; |
| 181 | |
| 182 | // If this is the first Doxygen comment, save it (because there isn't |
| 183 | // anything to merge it with). |
| 184 | if (Comments.empty()) { |
| 185 | Comments.push_back(RC); |
| 186 | OnlyWhitespaceSeen = true; |
| 187 | return; |
| 188 | } |
| 189 | |
| 190 | const RawComment &C1 = Comments.back(); |
| 191 | const RawComment &C2 = RC; |
| 192 | |
| 193 | // Merge comments only if there is only whitespace between them. |
| 194 | // Can't merge trailing and non-trailing comments. |
| 195 | // Merge trailing comments if they are on same or consecutive lines. |
| 196 | if (OnlyWhitespaceSeen && |
| 197 | (C1.isTrailingComment() == C2.isTrailingComment()) && |
| 198 | (!C1.isTrailingComment() || |
| 199 | C1.getEndLine(SourceMgr) + 1 >= C2.getBeginLine(SourceMgr))) { |
| 200 | SourceRange MergedRange(C1.getSourceRange().getBegin(), |
| 201 | C2.getSourceRange().getEnd()); |
| 202 | RawComment Merged(SourceMgr, MergedRange, true); |
| 203 | Comments.pop_back(); |
| 204 | Comments.push_back(Merged); |
| 205 | } else |
| 206 | Comments.push_back(RC); |
| 207 | |
| 208 | OnlyWhitespaceSeen = true; |
| 209 | } |
| 210 | |