Alexander Kornienko | 8f7e7f7 | 2014-10-02 19:09:56 +0000 | [diff] [blame^] | 1 | //===--- BracesAroundStatementsCheck.cpp - clang-tidy ---------------------===// |
| 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 | #include "BracesAroundStatementsCheck.h" |
| 11 | #include "clang/AST/ASTContext.h" |
| 12 | #include "clang/ASTMatchers/ASTMatchers.h" |
| 13 | #include "clang/Lex/Lexer.h" |
| 14 | |
| 15 | using namespace clang::ast_matchers; |
| 16 | |
| 17 | namespace clang { |
| 18 | namespace tidy { |
| 19 | namespace { |
| 20 | |
| 21 | tok::TokenKind getTokenKind(SourceLocation Loc, const SourceManager &SM, |
| 22 | const ASTContext *Context) { |
| 23 | Token Tok; |
| 24 | SourceLocation Beginning = |
| 25 | Lexer::GetBeginningOfToken(Loc, SM, Context->getLangOpts()); |
| 26 | const bool Invalid = |
| 27 | Lexer::getRawToken(Beginning, Tok, SM, Context->getLangOpts()); |
| 28 | assert(!Invalid && "Expected a valid token."); |
| 29 | |
| 30 | if (Invalid) |
| 31 | return tok::NUM_TOKENS; |
| 32 | |
| 33 | return Tok.getKind(); |
| 34 | } |
| 35 | |
| 36 | SourceLocation forwardSkipWhitespaceAndComments(SourceLocation Loc, |
| 37 | const SourceManager &SM, |
| 38 | const ASTContext *Context) { |
| 39 | assert(Loc.isValid()); |
| 40 | for (;;) { |
| 41 | while (isWhitespace(*FullSourceLoc(Loc, SM).getCharacterData())) |
| 42 | Loc = Loc.getLocWithOffset(1); |
| 43 | |
| 44 | tok::TokenKind TokKind = getTokenKind(Loc, SM, Context); |
| 45 | if (TokKind == tok::NUM_TOKENS || TokKind != tok::comment) |
| 46 | return Loc; |
| 47 | |
| 48 | // Fast-forward current token. |
| 49 | Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, Context->getLangOpts()); |
| 50 | } |
| 51 | } |
| 52 | |
| 53 | SourceLocation findEndLocation(SourceLocation LastTokenLoc, |
| 54 | const SourceManager &SM, |
| 55 | const ASTContext *Context) { |
| 56 | SourceLocation Loc = LastTokenLoc; |
| 57 | // Loc points to the beginning of the last (non-comment non-ws) token |
| 58 | // before end or ';'. |
| 59 | assert(Loc.isValid()); |
| 60 | bool SkipEndWhitespaceAndComments = true; |
| 61 | tok::TokenKind TokKind = getTokenKind(Loc, SM, Context); |
| 62 | if (TokKind == tok::NUM_TOKENS || TokKind == tok::semi || |
| 63 | TokKind == tok::r_brace) { |
| 64 | // If we are at ";" or "}", we found the last token. We could use as well |
| 65 | // `if (isa<NullStmt>(S))`, but it wouldn't work for nested statements. |
| 66 | SkipEndWhitespaceAndComments = false; |
| 67 | } |
| 68 | |
| 69 | Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, Context->getLangOpts()); |
| 70 | // Loc points past the last token before end or after ';'. |
| 71 | |
| 72 | if (SkipEndWhitespaceAndComments) { |
| 73 | Loc = forwardSkipWhitespaceAndComments(Loc, SM, Context); |
| 74 | tok::TokenKind TokKind = getTokenKind(Loc, SM, Context); |
| 75 | if (TokKind == tok::semi) |
| 76 | Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, Context->getLangOpts()); |
| 77 | } |
| 78 | |
| 79 | for (;;) { |
| 80 | assert(Loc.isValid()); |
| 81 | while (isHorizontalWhitespace(*FullSourceLoc(Loc, SM).getCharacterData())) |
| 82 | Loc = Loc.getLocWithOffset(1); |
| 83 | |
| 84 | if (isVerticalWhitespace(*FullSourceLoc(Loc, SM).getCharacterData())) { |
| 85 | // EOL, insert brace before. |
| 86 | break; |
| 87 | } |
| 88 | tok::TokenKind TokKind = getTokenKind(Loc, SM, Context); |
| 89 | if (TokKind != tok::comment) { |
| 90 | // Non-comment token, insert brace before. |
| 91 | break; |
| 92 | } |
| 93 | |
| 94 | SourceLocation TokEndLoc = |
| 95 | Lexer::getLocForEndOfToken(Loc, 0, SM, Context->getLangOpts()); |
| 96 | SourceRange TokRange(Loc, TokEndLoc); |
| 97 | StringRef Comment = Lexer::getSourceText( |
| 98 | CharSourceRange::getTokenRange(TokRange), SM, Context->getLangOpts()); |
| 99 | if (Comment.startswith("/*") && Comment.find('\n') != StringRef::npos) { |
| 100 | // Multi-line block comment, insert brace before. |
| 101 | break; |
| 102 | } |
| 103 | // else: Trailing comment, insert brace after the newline. |
| 104 | |
| 105 | // Fast-forward current token. |
| 106 | Loc = TokEndLoc; |
| 107 | } |
| 108 | return Loc; |
| 109 | } |
| 110 | |
| 111 | } // namespace |
| 112 | |
| 113 | void BracesAroundStatementsCheck::registerMatchers(MatchFinder *Finder) { |
| 114 | Finder->addMatcher(ifStmt().bind("if"), this); |
| 115 | Finder->addMatcher(whileStmt().bind("while"), this); |
| 116 | Finder->addMatcher(doStmt().bind("do"), this); |
| 117 | Finder->addMatcher(forStmt().bind("for"), this); |
| 118 | Finder->addMatcher(forRangeStmt().bind("for-range"), this); |
| 119 | } |
| 120 | |
| 121 | void |
| 122 | BracesAroundStatementsCheck::check(const MatchFinder::MatchResult &Result) { |
| 123 | const SourceManager &SM = *Result.SourceManager; |
| 124 | const ASTContext *Context = Result.Context; |
| 125 | |
| 126 | // Get location of closing parenthesis or 'do' to insert opening brace. |
| 127 | if (auto S = Result.Nodes.getNodeAs<ForStmt>("for")) { |
| 128 | checkStmt(Result, S->getBody(), S->getRParenLoc()); |
| 129 | } else if (auto S = Result.Nodes.getNodeAs<CXXForRangeStmt>("for-range")) { |
| 130 | checkStmt(Result, S->getBody(), S->getRParenLoc()); |
| 131 | } else if (auto S = Result.Nodes.getNodeAs<DoStmt>("do")) { |
| 132 | checkStmt(Result, S->getBody(), S->getDoLoc(), S->getWhileLoc()); |
| 133 | } else if (auto S = Result.Nodes.getNodeAs<WhileStmt>("while")) { |
| 134 | SourceLocation StartLoc = findRParenLoc(S, SM, Context); |
| 135 | if (StartLoc.isInvalid()) |
| 136 | return; |
| 137 | checkStmt(Result, S->getBody(), StartLoc); |
| 138 | } else if (auto S = Result.Nodes.getNodeAs<IfStmt>("if")) { |
| 139 | SourceLocation StartLoc = findRParenLoc(S, SM, Context); |
| 140 | if (StartLoc.isInvalid()) |
| 141 | return; |
| 142 | checkStmt(Result, S->getThen(), StartLoc, S->getElseLoc()); |
| 143 | const Stmt *Else = S->getElse(); |
| 144 | if (Else && !isa<IfStmt>(Else)) { |
| 145 | // Omit 'else if' statements here, they will be handled directly. |
| 146 | checkStmt(Result, Else, S->getElseLoc()); |
| 147 | } |
| 148 | } else { |
| 149 | llvm_unreachable("Invalid match"); |
| 150 | } |
| 151 | } |
| 152 | |
| 153 | /// Find location of right parenthesis closing condition |
| 154 | template <typename IfOrWhileStmt> |
| 155 | SourceLocation |
| 156 | BracesAroundStatementsCheck::findRParenLoc(const IfOrWhileStmt *S, |
| 157 | const SourceManager &SM, |
| 158 | const ASTContext *Context) { |
| 159 | // Skip macros |
| 160 | if (S->getLocStart().isMacroID()) |
| 161 | return SourceLocation(); |
| 162 | |
| 163 | static const char *const ErrorMessage = |
| 164 | "cannot find location of closing parenthesis ')'"; |
| 165 | SourceLocation CondEndLoc = S->getCond()->getLocEnd(); |
| 166 | if (const DeclStmt *CondVar = S->getConditionVariableDeclStmt()) |
| 167 | CondEndLoc = CondVar->getLocEnd(); |
| 168 | |
| 169 | assert(CondEndLoc.isValid()); |
| 170 | SourceLocation PastCondEndLoc = |
| 171 | Lexer::getLocForEndOfToken(CondEndLoc, 0, SM, Context->getLangOpts()); |
| 172 | if (PastCondEndLoc.isInvalid()) { |
| 173 | diag(CondEndLoc, ErrorMessage); |
| 174 | return SourceLocation(); |
| 175 | } |
| 176 | SourceLocation RParenLoc = |
| 177 | forwardSkipWhitespaceAndComments(PastCondEndLoc, SM, Context); |
| 178 | if (RParenLoc.isInvalid()) { |
| 179 | diag(PastCondEndLoc, ErrorMessage); |
| 180 | return SourceLocation(); |
| 181 | } |
| 182 | tok::TokenKind TokKind = getTokenKind(RParenLoc, SM, Context); |
| 183 | if (TokKind != tok::r_paren) { |
| 184 | diag(RParenLoc, ErrorMessage); |
| 185 | return SourceLocation(); |
| 186 | } |
| 187 | return RParenLoc; |
| 188 | } |
| 189 | |
| 190 | void |
| 191 | BracesAroundStatementsCheck::checkStmt(const MatchFinder::MatchResult &Result, |
| 192 | const Stmt *S, SourceLocation InitialLoc, |
| 193 | SourceLocation EndLocHint) { |
| 194 | // 1) If there's a corresponding "else" or "while", the check inserts "} " |
| 195 | // right before that token. |
| 196 | // 2) If there's a multi-line block comment starting on the same line after |
| 197 | // the location we're inserting the closing brace at, or there's a non-comment |
| 198 | // token, the check inserts "\n}" right before that token. |
| 199 | // 3) Otherwise the check finds the end of line (possibly after some block or |
| 200 | // line comments) and inserts "\n}" right before that EOL. |
| 201 | if (!S || isa<CompoundStmt>(S)) { |
| 202 | // Already inside braces. |
| 203 | return; |
| 204 | } |
| 205 | // Skip macros. |
| 206 | if (S->getLocStart().isMacroID()) |
| 207 | return; |
| 208 | |
| 209 | // TODO: Add an option to insert braces if: |
| 210 | // * the body doesn't fit on the same line with the control statement |
| 211 | // * the body takes more than one line |
| 212 | // * always. |
| 213 | |
| 214 | const SourceManager &SM = *Result.SourceManager; |
| 215 | const ASTContext *Context = Result.Context; |
| 216 | |
| 217 | // InitialLoc points at the last token before opening brace to be inserted. |
| 218 | assert(InitialLoc.isValid()); |
| 219 | SourceLocation StartLoc = |
| 220 | Lexer::getLocForEndOfToken(InitialLoc, 0, SM, Context->getLangOpts()); |
| 221 | // StartLoc points at the location of the opening brace to be inserted. |
| 222 | SourceLocation EndLoc; |
| 223 | std::string ClosingInsertion; |
| 224 | if (EndLocHint.isValid()) { |
| 225 | EndLoc = EndLocHint; |
| 226 | ClosingInsertion = "} "; |
| 227 | } else { |
| 228 | EndLoc = findEndLocation(S->getLocEnd(), SM, Context); |
| 229 | ClosingInsertion = "\n}"; |
| 230 | } |
| 231 | |
| 232 | assert(StartLoc.isValid()); |
| 233 | auto Diag = diag(StartLoc, "statement should be inside braces"); |
| 234 | Diag << FixItHint::CreateInsertion(StartLoc, " {") |
| 235 | << FixItHint::CreateInsertion(EndLoc, ClosingInsertion); |
| 236 | } |
| 237 | |
| 238 | } // namespace tidy |
| 239 | } // namespace clang |