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 "llvm/ADT/SmallString.h" |
15 | #include <cctype> |
16 | #include <optional> |
17 | |
18 | using namespace clang::ast_matchers; |
19 | |
20 | namespace clang::tidy::readability { |
21 | |
22 | namespace { |
23 | |
24 | struct IntegerLiteralCheck { |
25 | using type = clang::IntegerLiteral; |
26 | static constexpr llvm::StringLiteral Name = llvm::StringLiteral("integer" ); |
27 | // What should be skipped before looking for the Suffixes? (Nothing here.) |
28 | static constexpr llvm::StringLiteral SkipFirst = llvm::StringLiteral("" ); |
29 | // Suffix can only consist of 'u' and 'l' chars, and can be a complex number |
30 | // ('i', 'j'). In MS compatibility mode, suffixes like i32 are supported. |
31 | static constexpr llvm::StringLiteral Suffixes = |
32 | llvm::StringLiteral("uUlLiIjJ" ); |
33 | }; |
34 | constexpr llvm::StringLiteral IntegerLiteralCheck::Name; |
35 | constexpr llvm::StringLiteral IntegerLiteralCheck::SkipFirst; |
36 | constexpr llvm::StringLiteral IntegerLiteralCheck::Suffixes; |
37 | |
38 | struct FloatingLiteralCheck { |
39 | using type = clang::FloatingLiteral; |
40 | static constexpr llvm::StringLiteral Name = |
41 | llvm::StringLiteral("floating point" ); |
42 | // C++17 introduced hexadecimal floating-point literals, and 'f' is both a |
43 | // valid hexadecimal digit in a hex float literal and a valid floating-point |
44 | // literal suffix. |
45 | // So we can't just "skip to the chars that can be in the suffix". |
46 | // Since the exponent ('p'/'P') is mandatory for hexadecimal floating-point |
47 | // literals, we first skip everything before the exponent. |
48 | static constexpr llvm::StringLiteral SkipFirst = llvm::StringLiteral("pP" ); |
49 | // Suffix can only consist of 'f', 'l', "f16", 'h', 'q' chars, |
50 | // and can be a complex number ('i', 'j'). |
51 | static constexpr llvm::StringLiteral Suffixes = |
52 | llvm::StringLiteral("fFlLhHqQiIjJ" ); |
53 | }; |
54 | constexpr llvm::StringLiteral FloatingLiteralCheck::Name; |
55 | constexpr llvm::StringLiteral FloatingLiteralCheck::SkipFirst; |
56 | constexpr llvm::StringLiteral FloatingLiteralCheck::Suffixes; |
57 | |
58 | struct NewSuffix { |
59 | SourceRange LiteralLocation; |
60 | StringRef OldSuffix; |
61 | std::optional<FixItHint> FixIt; |
62 | }; |
63 | |
64 | std::optional<SourceLocation> getMacroAwareLocation(SourceLocation Loc, |
65 | const SourceManager &SM) { |
66 | // Do nothing if the provided location is invalid. |
67 | if (Loc.isInvalid()) |
68 | return std::nullopt; |
69 | // Look where the location was *actually* written. |
70 | SourceLocation SpellingLoc = SM.getSpellingLoc(Loc); |
71 | if (SpellingLoc.isInvalid()) |
72 | return std::nullopt; |
73 | return SpellingLoc; |
74 | } |
75 | |
76 | std::optional<SourceRange> getMacroAwareSourceRange(SourceRange Loc, |
77 | const SourceManager &SM) { |
78 | std::optional<SourceLocation> Begin = |
79 | getMacroAwareLocation(Loc: Loc.getBegin(), SM); |
80 | std::optional<SourceLocation> End = getMacroAwareLocation(Loc: Loc.getEnd(), SM); |
81 | if (!Begin || !End) |
82 | return std::nullopt; |
83 | return SourceRange(*Begin, *End); |
84 | } |
85 | |
86 | std::optional<std::string> |
87 | getNewSuffix(llvm::StringRef OldSuffix, |
88 | const std::vector<StringRef> &NewSuffixes) { |
89 | // If there is no config, just uppercase the entirety of the suffix. |
90 | if (NewSuffixes.empty()) |
91 | return OldSuffix.upper(); |
92 | // Else, find matching suffix, case-*insensitive*ly. |
93 | auto NewSuffix = |
94 | llvm::find_if(Range: NewSuffixes, P: [OldSuffix](StringRef PotentialNewSuffix) { |
95 | return OldSuffix.equals_insensitive(RHS: PotentialNewSuffix); |
96 | }); |
97 | // Have a match, return it. |
98 | if (NewSuffix != NewSuffixes.end()) |
99 | return NewSuffix->str(); |
100 | // Nope, I guess we have to keep it as-is. |
101 | return std::nullopt; |
102 | } |
103 | |
104 | template <typename LiteralType> |
105 | std::optional<NewSuffix> |
106 | shouldReplaceLiteralSuffix(const Expr &Literal, |
107 | const std::vector<StringRef> &NewSuffixes, |
108 | const SourceManager &SM, const LangOptions &LO) { |
109 | NewSuffix ReplacementDsc; |
110 | |
111 | const auto &L = cast<typename LiteralType::type>(Literal); |
112 | |
113 | // The naive location of the literal. Is always valid. |
114 | ReplacementDsc.LiteralLocation = L.getSourceRange(); |
115 | |
116 | // Was this literal fully spelled or is it a product of macro expansion? |
117 | bool RangeCanBeFixed = |
118 | utils::rangeCanBeFixed(Range: ReplacementDsc.LiteralLocation, SM: &SM); |
119 | |
120 | // The literal may have macro expansion, we need the final expanded src range. |
121 | std::optional<SourceRange> Range = |
122 | getMacroAwareSourceRange(Loc: ReplacementDsc.LiteralLocation, SM); |
123 | if (!Range) |
124 | return std::nullopt; |
125 | |
126 | if (RangeCanBeFixed) |
127 | ReplacementDsc.LiteralLocation = *Range; |
128 | // Else keep the naive literal location! |
129 | |
130 | // Get the whole literal from the source buffer. |
131 | bool Invalid = false; |
132 | const StringRef LiteralSourceText = Lexer::getSourceText( |
133 | Range: CharSourceRange::getTokenRange(R: *Range), SM, LangOpts: LO, Invalid: &Invalid); |
134 | assert(!Invalid && "Failed to retrieve the source text." ); |
135 | |
136 | // Make sure the first character is actually a digit, instead of |
137 | // something else, like a non-type template parameter. |
138 | if (!std::isdigit(static_cast<unsigned char>(LiteralSourceText.front()))) |
139 | return std::nullopt; |
140 | |
141 | size_t Skip = 0; |
142 | |
143 | // Do we need to ignore something before actually looking for the suffix? |
144 | if (!LiteralType::SkipFirst.empty()) { |
145 | // E.g. we can't look for 'f' suffix in hexadecimal floating-point literals |
146 | // until after we skip to the exponent (which is mandatory there), |
147 | // because hex-digit-sequence may contain 'f'. |
148 | Skip = LiteralSourceText.find_first_of(LiteralType::SkipFirst); |
149 | // We could be in non-hexadecimal floating-point literal, with no exponent. |
150 | if (Skip == StringRef::npos) |
151 | Skip = 0; |
152 | } |
153 | |
154 | // Find the beginning of the suffix by looking for the first char that is |
155 | // one of these chars that can be in the suffix, potentially starting looking |
156 | // in the exponent, if we are skipping hex-digit-sequence. |
157 | Skip = LiteralSourceText.find_first_of(LiteralType::Suffixes, /*From=*/Skip); |
158 | |
159 | // We can't check whether the *Literal has any suffix or not without actually |
160 | // looking for the suffix. So it is totally possible that there is no suffix. |
161 | if (Skip == StringRef::npos) |
162 | return std::nullopt; |
163 | |
164 | // Move the cursor in the source range to the beginning of the suffix. |
165 | Range->setBegin(Range->getBegin().getLocWithOffset(Offset: Skip)); |
166 | // And in our textual representation too. |
167 | ReplacementDsc.OldSuffix = LiteralSourceText.drop_front(N: Skip); |
168 | assert(!ReplacementDsc.OldSuffix.empty() && |
169 | "We still should have some chars left." ); |
170 | |
171 | // And get the replacement suffix. |
172 | std::optional<std::string> NewSuffix = |
173 | getNewSuffix(OldSuffix: ReplacementDsc.OldSuffix, NewSuffixes); |
174 | if (!NewSuffix || ReplacementDsc.OldSuffix == *NewSuffix) |
175 | return std::nullopt; // The suffix was already the way it should be. |
176 | |
177 | if (RangeCanBeFixed) |
178 | ReplacementDsc.FixIt = FixItHint::CreateReplacement(RemoveRange: *Range, Code: *NewSuffix); |
179 | |
180 | return ReplacementDsc; |
181 | } |
182 | |
183 | } // namespace |
184 | |
185 | UppercaseLiteralSuffixCheck::UppercaseLiteralSuffixCheck( |
186 | StringRef Name, ClangTidyContext *Context) |
187 | : ClangTidyCheck(Name, Context), |
188 | NewSuffixes( |
189 | utils::options::parseStringList(Option: Options.get(LocalName: "NewSuffixes" , Default: "" ))), |
190 | IgnoreMacros(Options.getLocalOrGlobal(LocalName: "IgnoreMacros" , Default: true)) {} |
191 | |
192 | void UppercaseLiteralSuffixCheck::storeOptions( |
193 | ClangTidyOptions::OptionMap &Opts) { |
194 | Options.store(Options&: Opts, LocalName: "NewSuffixes" , |
195 | Value: utils::options::serializeStringList(Strings: NewSuffixes)); |
196 | Options.store(Options&: Opts, LocalName: "IgnoreMacros" , Value: IgnoreMacros); |
197 | } |
198 | |
199 | void UppercaseLiteralSuffixCheck::registerMatchers(MatchFinder *Finder) { |
200 | // Sadly, we can't check whether the literal has suffix or not. |
201 | // E.g. i32 suffix still results in 'BuiltinType::Kind::Int'. |
202 | // And such an info is not stored in the *Literal itself. |
203 | Finder->addMatcher( |
204 | NodeMatch: stmt(eachOf(integerLiteral().bind(ID: IntegerLiteralCheck::Name), |
205 | floatLiteral().bind(ID: FloatingLiteralCheck::Name)), |
206 | unless(anyOf(hasParent(userDefinedLiteral()), |
207 | hasAncestor(substNonTypeTemplateParmExpr())))), |
208 | Action: this); |
209 | } |
210 | |
211 | template <typename LiteralType> |
212 | bool UppercaseLiteralSuffixCheck::checkBoundMatch( |
213 | const MatchFinder::MatchResult &Result) { |
214 | const auto *Literal = |
215 | Result.Nodes.getNodeAs<typename LiteralType::type>(LiteralType::Name); |
216 | if (!Literal) |
217 | return false; |
218 | |
219 | // We won't *always* want to diagnose. |
220 | // We might have a suffix that is already uppercase. |
221 | if (auto Details = shouldReplaceLiteralSuffix<LiteralType>( |
222 | *Literal, NewSuffixes, *Result.SourceManager, getLangOpts())) { |
223 | if (Details->LiteralLocation.getBegin().isMacroID() && IgnoreMacros) |
224 | return true; |
225 | auto Complaint = diag(Details->LiteralLocation.getBegin(), |
226 | "%0 literal has suffix '%1', which is not uppercase" ) |
227 | << LiteralType::Name << Details->OldSuffix; |
228 | if (Details->FixIt) // Similarly, a fix-it is not always possible. |
229 | Complaint << *(Details->FixIt); |
230 | } |
231 | |
232 | return true; |
233 | } |
234 | |
235 | void UppercaseLiteralSuffixCheck::check( |
236 | const MatchFinder::MatchResult &Result) { |
237 | if (checkBoundMatch<IntegerLiteralCheck>(Result)) |
238 | return; // If it *was* IntegerLiteral, don't check for FloatingLiteral. |
239 | checkBoundMatch<FloatingLiteralCheck>(Result); |
240 | } |
241 | |
242 | } // namespace clang::tidy::readability |
243 | |