| 1 | //===--- NamespaceCommentCheck.cpp - clang-tidy ---------------------------===// |
| 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 | #include "NamespaceCommentCheck.h" |
| 10 | #include "../utils/LexerUtils.h" |
| 11 | #include "clang/AST/ASTContext.h" |
| 12 | #include "clang/ASTMatchers/ASTMatchers.h" |
| 13 | #include "clang/Basic/SourceLocation.h" |
| 14 | #include "clang/Basic/TokenKinds.h" |
| 15 | #include "clang/Lex/Lexer.h" |
| 16 | #include <optional> |
| 17 | |
| 18 | using namespace clang::ast_matchers; |
| 19 | |
| 20 | namespace clang::tidy::readability { |
| 21 | |
| 22 | NamespaceCommentCheck::(StringRef Name, |
| 23 | ClangTidyContext *Context) |
| 24 | : ClangTidyCheck(Name, Context), |
| 25 | NamespaceCommentPattern( |
| 26 | "^/[/*] *(end (of )?)? *(anonymous|unnamed)? *" |
| 27 | "namespace( +(((inline )|([a-zA-Z0-9_:]))+))?\\.? *(\\*/)?$" , |
| 28 | llvm::Regex::IgnoreCase), |
| 29 | ShortNamespaceLines(Options.get(LocalName: "ShortNamespaceLines" , Default: 1U)), |
| 30 | SpacesBeforeComments(Options.get(LocalName: "SpacesBeforeComments" , Default: 1U)), |
| 31 | AllowOmittingNamespaceComments( |
| 32 | Options.get(LocalName: "AllowOmittingNamespaceComments" , Default: false)) {} |
| 33 | |
| 34 | void NamespaceCommentCheck::(ClangTidyOptions::OptionMap &Opts) { |
| 35 | Options.store(Options&: Opts, LocalName: "ShortNamespaceLines" , Value: ShortNamespaceLines); |
| 36 | Options.store(Options&: Opts, LocalName: "SpacesBeforeComments" , Value: SpacesBeforeComments); |
| 37 | Options.store(Options&: Opts, LocalName: "AllowOmittingNamespaceComments" , |
| 38 | Value: AllowOmittingNamespaceComments); |
| 39 | } |
| 40 | |
| 41 | void NamespaceCommentCheck::(MatchFinder *Finder) { |
| 42 | Finder->addMatcher(NodeMatch: namespaceDecl().bind(ID: "namespace" ), Action: this); |
| 43 | } |
| 44 | |
| 45 | static bool locationsInSameFile(const SourceManager &Sources, |
| 46 | SourceLocation Loc1, SourceLocation Loc2) { |
| 47 | return Loc1.isFileID() && Loc2.isFileID() && |
| 48 | Sources.getFileID(SpellingLoc: Loc1) == Sources.getFileID(SpellingLoc: Loc2); |
| 49 | } |
| 50 | |
| 51 | static std::optional<std::string> |
| 52 | getNamespaceNameAsWritten(SourceLocation &Loc, const SourceManager &Sources, |
| 53 | const LangOptions &LangOpts) { |
| 54 | // Loc should be at the begin of the namespace decl (usually, `namespace` |
| 55 | // token). We skip the first token right away, but in case of `inline |
| 56 | // namespace` or `namespace a::inline b` we can see both `inline` and |
| 57 | // `namespace` keywords, which we just ignore. Nested parens/squares before |
| 58 | // the opening brace can result from attributes. |
| 59 | std::string Result; |
| 60 | int Nesting = 0; |
| 61 | while (std::optional<Token> T = utils::lexer::findNextTokenSkippingComments( |
| 62 | Start: Loc, SM: Sources, LangOpts)) { |
| 63 | Loc = T->getLocation(); |
| 64 | if (T->is(K: tok::l_brace)) |
| 65 | break; |
| 66 | |
| 67 | if (T->isOneOf(K1: tok::l_square, K2: tok::l_paren)) { |
| 68 | ++Nesting; |
| 69 | } else if (T->isOneOf(K1: tok::r_square, K2: tok::r_paren)) { |
| 70 | --Nesting; |
| 71 | } else if (Nesting == 0) { |
| 72 | if (T->is(K: tok::raw_identifier)) { |
| 73 | StringRef ID = T->getRawIdentifier(); |
| 74 | if (ID != "namespace" ) |
| 75 | Result.append(str: std::string(ID)); |
| 76 | if (ID == "inline" ) |
| 77 | Result.append(s: " " ); |
| 78 | } else if (T->is(K: tok::coloncolon)) { |
| 79 | Result.append(s: "::" ); |
| 80 | } else { // Any other kind of token is unexpected here. |
| 81 | return std::nullopt; |
| 82 | } |
| 83 | } |
| 84 | } |
| 85 | return Result; |
| 86 | } |
| 87 | |
| 88 | void NamespaceCommentCheck::(const MatchFinder::MatchResult &Result) { |
| 89 | const auto *ND = Result.Nodes.getNodeAs<NamespaceDecl>(ID: "namespace" ); |
| 90 | const SourceManager &Sources = *Result.SourceManager; |
| 91 | |
| 92 | // Ignore namespaces inside macros and namespaces split across files. |
| 93 | if (ND->getBeginLoc().isMacroID() || |
| 94 | !locationsInSameFile(Sources, Loc1: ND->getBeginLoc(), Loc2: ND->getRBraceLoc())) |
| 95 | return; |
| 96 | |
| 97 | // Don't require closing comments for namespaces spanning less than certain |
| 98 | // number of lines. |
| 99 | unsigned StartLine = Sources.getSpellingLineNumber(Loc: ND->getBeginLoc()); |
| 100 | unsigned EndLine = Sources.getSpellingLineNumber(Loc: ND->getRBraceLoc()); |
| 101 | if (EndLine - StartLine + 1 <= ShortNamespaceLines) |
| 102 | return; |
| 103 | |
| 104 | // Find next token after the namespace closing brace. |
| 105 | SourceLocation AfterRBrace = Lexer::getLocForEndOfToken( |
| 106 | Loc: ND->getRBraceLoc(), /*Offset=*/0, SM: Sources, LangOpts: getLangOpts()); |
| 107 | SourceLocation Loc = AfterRBrace; |
| 108 | SourceLocation LBraceLoc = ND->getBeginLoc(); |
| 109 | |
| 110 | // Currently for nested namespace (n1::n2::...) the AST matcher will match foo |
| 111 | // then bar instead of a single match. So if we got a nested namespace we have |
| 112 | // to skip the next ones. |
| 113 | for (const SourceLocation &EndOfNameLocation : Ends) { |
| 114 | if (Sources.isBeforeInTranslationUnit(LHS: ND->getLocation(), RHS: EndOfNameLocation)) |
| 115 | return; |
| 116 | } |
| 117 | |
| 118 | std::optional<std::string> NamespaceNameAsWritten = |
| 119 | getNamespaceNameAsWritten(Loc&: LBraceLoc, Sources, LangOpts: getLangOpts()); |
| 120 | if (!NamespaceNameAsWritten) |
| 121 | return; |
| 122 | |
| 123 | if (NamespaceNameAsWritten->empty() != ND->isAnonymousNamespace()) { |
| 124 | // Apparently, we didn't find the correct namespace name. Give up. |
| 125 | return; |
| 126 | } |
| 127 | |
| 128 | Ends.push_back(Elt: LBraceLoc); |
| 129 | |
| 130 | Token Tok; |
| 131 | // Skip whitespace until we find the next token. |
| 132 | while (Lexer::getRawToken(Loc, Result&: Tok, SM: Sources, LangOpts: getLangOpts()) || |
| 133 | Tok.is(K: tok::semi)) { |
| 134 | Loc = Loc.getLocWithOffset(Offset: 1); |
| 135 | } |
| 136 | |
| 137 | if (!locationsInSameFile(Sources, Loc1: ND->getRBraceLoc(), Loc2: Loc)) |
| 138 | return; |
| 139 | |
| 140 | bool NextTokenIsOnSameLine = Sources.getSpellingLineNumber(Loc) == EndLine; |
| 141 | // If we insert a line comment before the token in the same line, we need |
| 142 | // to insert a line break. |
| 143 | bool NeedLineBreak = NextTokenIsOnSameLine && Tok.isNot(K: tok::eof); |
| 144 | |
| 145 | SourceRange (AfterRBrace, AfterRBrace); |
| 146 | std::string Message = "%0 not terminated with a closing comment" ; |
| 147 | bool = false; |
| 148 | |
| 149 | // Try to find existing namespace closing comment on the same line. |
| 150 | if (Tok.is(K: tok::comment) && NextTokenIsOnSameLine) { |
| 151 | StringRef (Sources.getCharacterData(SL: Loc), Tok.getLength()); |
| 152 | SmallVector<StringRef, 7> Groups; |
| 153 | if (NamespaceCommentPattern.match(String: Comment, Matches: &Groups)) { |
| 154 | StringRef = Groups.size() > 5 ? Groups[5] : "" ; |
| 155 | StringRef Anonymous = Groups.size() > 3 ? Groups[3] : "" ; |
| 156 | |
| 157 | if ((ND->isAnonymousNamespace() && NamespaceNameInComment.empty()) || |
| 158 | (*NamespaceNameAsWritten == NamespaceNameInComment && |
| 159 | Anonymous.empty())) { |
| 160 | // Check if the namespace in the comment is the same. |
| 161 | // FIXME: Maybe we need a strict mode, where we always fix namespace |
| 162 | // comments with different format. |
| 163 | return; |
| 164 | } |
| 165 | |
| 166 | HasComment = true; |
| 167 | |
| 168 | // Otherwise we need to fix the comment. |
| 169 | NeedLineBreak = Comment.starts_with(Prefix: "/*" ); |
| 170 | OldCommentRange = |
| 171 | SourceRange(AfterRBrace, Loc.getLocWithOffset(Offset: Tok.getLength())); |
| 172 | Message = |
| 173 | (llvm::Twine( |
| 174 | "%0 ends with a comment that refers to a wrong namespace '" ) + |
| 175 | NamespaceNameInComment + "'" ) |
| 176 | .str(); |
| 177 | } else if (Comment.starts_with(Prefix: "//" )) { |
| 178 | // Assume that this is an unrecognized form of a namespace closing line |
| 179 | // comment. Replace it. |
| 180 | NeedLineBreak = false; |
| 181 | OldCommentRange = |
| 182 | SourceRange(AfterRBrace, Loc.getLocWithOffset(Offset: Tok.getLength())); |
| 183 | Message = "%0 ends with an unrecognized comment" ; |
| 184 | } |
| 185 | // If it's a block comment, just move it to the next line, as it can be |
| 186 | // multi-line or there may be other tokens behind it. |
| 187 | } |
| 188 | |
| 189 | std::string NamespaceNameForDiag = |
| 190 | ND->isAnonymousNamespace() |
| 191 | ? "anonymous namespace" |
| 192 | : ("namespace '" + *NamespaceNameAsWritten + "'" ); |
| 193 | |
| 194 | // If no namespace comment is allowed |
| 195 | if (!HasComment && AllowOmittingNamespaceComments) |
| 196 | return; |
| 197 | |
| 198 | std::string Fix(SpacesBeforeComments, ' '); |
| 199 | Fix.append(s: "// namespace" ); |
| 200 | if (!ND->isAnonymousNamespace()) |
| 201 | Fix.append(s: " " ).append(str: *NamespaceNameAsWritten); |
| 202 | if (NeedLineBreak) |
| 203 | Fix.append(s: "\n" ); |
| 204 | |
| 205 | // Place diagnostic at an old comment, or closing brace if we did not have it. |
| 206 | SourceLocation DiagLoc = |
| 207 | OldCommentRange.getBegin() != OldCommentRange.getEnd() |
| 208 | ? OldCommentRange.getBegin() |
| 209 | : ND->getRBraceLoc(); |
| 210 | |
| 211 | diag(Loc: DiagLoc, Description: Message) << NamespaceNameForDiag |
| 212 | << FixItHint::CreateReplacement( |
| 213 | RemoveRange: CharSourceRange::getCharRange(R: OldCommentRange), |
| 214 | Code: Fix); |
| 215 | diag(ND->getLocation(), "%0 starts here" , DiagnosticIDs::Note) |
| 216 | << NamespaceNameForDiag; |
| 217 | } |
| 218 | |
| 219 | } // namespace clang::tidy::readability |
| 220 | |