| 1 | //===--- UppercaseLiteralSuffixCheck.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 "UppercaseLiteralSuffixCheck.h" |
| 10 | #include "../utils/ASTUtils.h" |
| 11 | #include "clang/AST/ASTContext.h" |
| 12 | #include "clang/ASTMatchers/ASTMatchFinder.h" |
| 13 | #include "clang/Lex/Lexer.h" |
| 14 | #include <cctype> |
| 15 | #include <optional> |
| 16 | |
| 17 | using namespace clang::ast_matchers; |
| 18 | |
| 19 | namespace clang::tidy::readability { |
| 20 | |
| 21 | namespace { |
| 22 | |
| 23 | struct IntegerLiteralCheck { |
| 24 | using type = clang::IntegerLiteral; |
| 25 | static constexpr llvm::StringLiteral Name = llvm::StringLiteral("integer" ); |
| 26 | // What should be skipped before looking for the Suffixes? (Nothing here.) |
| 27 | static constexpr llvm::StringLiteral SkipFirst = llvm::StringLiteral("" ); |
| 28 | // Suffix can only consist of 'u' and 'l' chars, and can be a complex number |
| 29 | // ('i', 'j'). In MS compatibility mode, suffixes like i32 are supported. |
| 30 | static constexpr llvm::StringLiteral Suffixes = |
| 31 | llvm::StringLiteral("uUlLiIjJ" ); |
| 32 | }; |
| 33 | constexpr llvm::StringLiteral IntegerLiteralCheck::Name; |
| 34 | constexpr llvm::StringLiteral IntegerLiteralCheck::SkipFirst; |
| 35 | constexpr llvm::StringLiteral IntegerLiteralCheck::Suffixes; |
| 36 | |
| 37 | struct FloatingLiteralCheck { |
| 38 | using type = clang::FloatingLiteral; |
| 39 | static constexpr llvm::StringLiteral Name = |
| 40 | llvm::StringLiteral("floating point" ); |
| 41 | // C++17 introduced hexadecimal floating-point literals, and 'f' is both a |
| 42 | // valid hexadecimal digit in a hex float literal and a valid floating-point |
| 43 | // literal suffix. |
| 44 | // So we can't just "skip to the chars that can be in the suffix". |
| 45 | // Since the exponent ('p'/'P') is mandatory for hexadecimal floating-point |
| 46 | // literals, we first skip everything before the exponent. |
| 47 | static constexpr llvm::StringLiteral SkipFirst = llvm::StringLiteral("pP" ); |
| 48 | // Suffix can only consist of 'f', 'l', "f16", 'h', 'q' chars, |
| 49 | // and can be a complex number ('i', 'j'). |
| 50 | static constexpr llvm::StringLiteral Suffixes = |
| 51 | llvm::StringLiteral("fFlLhHqQiIjJ" ); |
| 52 | }; |
| 53 | constexpr llvm::StringLiteral FloatingLiteralCheck::Name; |
| 54 | constexpr llvm::StringLiteral FloatingLiteralCheck::SkipFirst; |
| 55 | constexpr llvm::StringLiteral FloatingLiteralCheck::Suffixes; |
| 56 | |
| 57 | struct NewSuffix { |
| 58 | SourceRange LiteralLocation; |
| 59 | StringRef OldSuffix; |
| 60 | std::optional<FixItHint> FixIt; |
| 61 | }; |
| 62 | |
| 63 | std::optional<SourceLocation> getMacroAwareLocation(SourceLocation Loc, |
| 64 | const SourceManager &SM) { |
| 65 | // Do nothing if the provided location is invalid. |
| 66 | if (Loc.isInvalid()) |
| 67 | return std::nullopt; |
| 68 | // Look where the location was *actually* written. |
| 69 | SourceLocation SpellingLoc = SM.getSpellingLoc(Loc); |
| 70 | if (SpellingLoc.isInvalid()) |
| 71 | return std::nullopt; |
| 72 | return SpellingLoc; |
| 73 | } |
| 74 | |
| 75 | std::optional<SourceRange> getMacroAwareSourceRange(SourceRange Loc, |
| 76 | const SourceManager &SM) { |
| 77 | std::optional<SourceLocation> Begin = |
| 78 | getMacroAwareLocation(Loc: Loc.getBegin(), SM); |
| 79 | std::optional<SourceLocation> End = getMacroAwareLocation(Loc: Loc.getEnd(), SM); |
| 80 | if (!Begin || !End) |
| 81 | return std::nullopt; |
| 82 | return SourceRange(*Begin, *End); |
| 83 | } |
| 84 | |
| 85 | std::optional<std::string> |
| 86 | getNewSuffix(llvm::StringRef OldSuffix, |
| 87 | const std::vector<StringRef> &NewSuffixes) { |
| 88 | // If there is no config, just uppercase the entirety of the suffix. |
| 89 | if (NewSuffixes.empty()) |
| 90 | return OldSuffix.upper(); |
| 91 | // Else, find matching suffix, case-*insensitive*ly. |
| 92 | auto NewSuffix = |
| 93 | llvm::find_if(Range: NewSuffixes, P: [OldSuffix](StringRef PotentialNewSuffix) { |
| 94 | return OldSuffix.equals_insensitive(RHS: PotentialNewSuffix); |
| 95 | }); |
| 96 | // Have a match, return it. |
| 97 | if (NewSuffix != NewSuffixes.end()) |
| 98 | return NewSuffix->str(); |
| 99 | // Nope, I guess we have to keep it as-is. |
| 100 | return std::nullopt; |
| 101 | } |
| 102 | |
| 103 | template <typename LiteralType> |
| 104 | std::optional<NewSuffix> |
| 105 | shouldReplaceLiteralSuffix(const Expr &Literal, |
| 106 | const std::vector<StringRef> &NewSuffixes, |
| 107 | const SourceManager &SM, const LangOptions &LO) { |
| 108 | NewSuffix ReplacementDsc; |
| 109 | |
| 110 | const auto &L = cast<typename LiteralType::type>(Literal); |
| 111 | |
| 112 | // The naive location of the literal. Is always valid. |
| 113 | ReplacementDsc.LiteralLocation = L.getSourceRange(); |
| 114 | |
| 115 | // Was this literal fully spelled or is it a product of macro expansion? |
| 116 | bool RangeCanBeFixed = |
| 117 | utils::rangeCanBeFixed(Range: ReplacementDsc.LiteralLocation, SM: &SM); |
| 118 | |
| 119 | // The literal may have macro expansion, we need the final expanded src range. |
| 120 | std::optional<SourceRange> Range = |
| 121 | getMacroAwareSourceRange(Loc: ReplacementDsc.LiteralLocation, SM); |
| 122 | if (!Range) |
| 123 | return std::nullopt; |
| 124 | |
| 125 | if (RangeCanBeFixed) |
| 126 | ReplacementDsc.LiteralLocation = *Range; |
| 127 | // Else keep the naive literal location! |
| 128 | |
| 129 | // Get the whole literal from the source buffer. |
| 130 | bool Invalid = false; |
| 131 | const StringRef LiteralSourceText = Lexer::getSourceText( |
| 132 | Range: CharSourceRange::getTokenRange(R: *Range), SM, LangOpts: LO, Invalid: &Invalid); |
| 133 | assert(!Invalid && "Failed to retrieve the source text." ); |
| 134 | |
| 135 | // Make sure the first character is actually a digit, instead of |
| 136 | // something else, like a non-type template parameter. |
| 137 | if (!std::isdigit(static_cast<unsigned char>(LiteralSourceText.front()))) |
| 138 | return std::nullopt; |
| 139 | |
| 140 | size_t Skip = 0; |
| 141 | |
| 142 | // Do we need to ignore something before actually looking for the suffix? |
| 143 | if (!LiteralType::SkipFirst.empty()) { |
| 144 | // E.g. we can't look for 'f' suffix in hexadecimal floating-point literals |
| 145 | // until after we skip to the exponent (which is mandatory there), |
| 146 | // because hex-digit-sequence may contain 'f'. |
| 147 | Skip = LiteralSourceText.find_first_of(LiteralType::SkipFirst); |
| 148 | // We could be in non-hexadecimal floating-point literal, with no exponent. |
| 149 | if (Skip == StringRef::npos) |
| 150 | Skip = 0; |
| 151 | } |
| 152 | |
| 153 | // Find the beginning of the suffix by looking for the first char that is |
| 154 | // one of these chars that can be in the suffix, potentially starting looking |
| 155 | // in the exponent, if we are skipping hex-digit-sequence. |
| 156 | Skip = LiteralSourceText.find_first_of(LiteralType::Suffixes, /*From=*/Skip); |
| 157 | |
| 158 | // We can't check whether the *Literal has any suffix or not without actually |
| 159 | // looking for the suffix. So it is totally possible that there is no suffix. |
| 160 | if (Skip == StringRef::npos) |
| 161 | return std::nullopt; |
| 162 | |
| 163 | // Move the cursor in the source range to the beginning of the suffix. |
| 164 | Range->setBegin(Range->getBegin().getLocWithOffset(Offset: Skip)); |
| 165 | // And in our textual representation too. |
| 166 | ReplacementDsc.OldSuffix = LiteralSourceText.drop_front(N: Skip); |
| 167 | assert(!ReplacementDsc.OldSuffix.empty() && |
| 168 | "We still should have some chars left." ); |
| 169 | |
| 170 | // And get the replacement suffix. |
| 171 | std::optional<std::string> NewSuffix = |
| 172 | getNewSuffix(OldSuffix: ReplacementDsc.OldSuffix, NewSuffixes); |
| 173 | if (!NewSuffix || ReplacementDsc.OldSuffix == *NewSuffix) |
| 174 | return std::nullopt; // The suffix was already the way it should be. |
| 175 | |
| 176 | if (RangeCanBeFixed) |
| 177 | ReplacementDsc.FixIt = FixItHint::CreateReplacement(RemoveRange: *Range, Code: *NewSuffix); |
| 178 | |
| 179 | return ReplacementDsc; |
| 180 | } |
| 181 | |
| 182 | } // namespace |
| 183 | |
| 184 | UppercaseLiteralSuffixCheck::UppercaseLiteralSuffixCheck( |
| 185 | StringRef Name, ClangTidyContext *Context) |
| 186 | : ClangTidyCheck(Name, Context), |
| 187 | NewSuffixes( |
| 188 | utils::options::parseStringList(Option: Options.get(LocalName: "NewSuffixes" , Default: "" ))), |
| 189 | IgnoreMacros(Options.getLocalOrGlobal(LocalName: "IgnoreMacros" , Default: true)) {} |
| 190 | |
| 191 | void UppercaseLiteralSuffixCheck::storeOptions( |
| 192 | ClangTidyOptions::OptionMap &Opts) { |
| 193 | Options.store(Options&: Opts, LocalName: "NewSuffixes" , |
| 194 | Value: utils::options::serializeStringList(Strings: NewSuffixes)); |
| 195 | Options.store(Options&: Opts, LocalName: "IgnoreMacros" , Value: IgnoreMacros); |
| 196 | } |
| 197 | |
| 198 | void UppercaseLiteralSuffixCheck::registerMatchers(MatchFinder *Finder) { |
| 199 | // Sadly, we can't check whether the literal has suffix or not. |
| 200 | // E.g. i32 suffix still results in 'BuiltinType::Kind::Int'. |
| 201 | // And such an info is not stored in the *Literal itself. |
| 202 | Finder->addMatcher( |
| 203 | NodeMatch: stmt(eachOf(integerLiteral().bind(ID: IntegerLiteralCheck::Name), |
| 204 | floatLiteral().bind(ID: FloatingLiteralCheck::Name)), |
| 205 | unless(anyOf(hasParent(userDefinedLiteral()), |
| 206 | hasAncestor(substNonTypeTemplateParmExpr())))), |
| 207 | Action: this); |
| 208 | } |
| 209 | |
| 210 | template <typename LiteralType> |
| 211 | bool UppercaseLiteralSuffixCheck::checkBoundMatch( |
| 212 | const MatchFinder::MatchResult &Result) { |
| 213 | const auto *Literal = |
| 214 | Result.Nodes.getNodeAs<typename LiteralType::type>(LiteralType::Name); |
| 215 | if (!Literal) |
| 216 | return false; |
| 217 | |
| 218 | // We won't *always* want to diagnose. |
| 219 | // We might have a suffix that is already uppercase. |
| 220 | if (auto Details = shouldReplaceLiteralSuffix<LiteralType>( |
| 221 | *Literal, NewSuffixes, *Result.SourceManager, getLangOpts())) { |
| 222 | if (Details->LiteralLocation.getBegin().isMacroID() && IgnoreMacros) |
| 223 | return true; |
| 224 | auto Complaint = diag(Details->LiteralLocation.getBegin(), |
| 225 | "%0 literal has suffix '%1', which is not uppercase" ) |
| 226 | << LiteralType::Name << Details->OldSuffix; |
| 227 | if (Details->FixIt) // Similarly, a fix-it is not always possible. |
| 228 | Complaint << *(Details->FixIt); |
| 229 | } |
| 230 | |
| 231 | return true; |
| 232 | } |
| 233 | |
| 234 | void UppercaseLiteralSuffixCheck::check( |
| 235 | const MatchFinder::MatchResult &Result) { |
| 236 | if (checkBoundMatch<IntegerLiteralCheck>(Result)) |
| 237 | return; // If it *was* IntegerLiteral, don't check for FloatingLiteral. |
| 238 | checkBoundMatch<FloatingLiteralCheck>(Result); |
| 239 | } |
| 240 | |
| 241 | } // namespace clang::tidy::readability |
| 242 | |