1 //===--- SourceCode.h - Manipulating source code as strings -----*- 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 #include "SourceCode.h" 10 11 #include "Logger.h" 12 #include "clang/AST/ASTContext.h" 13 #include "clang/Basic/SourceManager.h" 14 #include "clang/Lex/Lexer.h" 15 #include "llvm/Support/Errc.h" 16 #include "llvm/Support/Error.h" 17 #include "llvm/Support/Path.h" 18 19 namespace clang { 20 namespace clangd { 21 using namespace llvm; 22 23 // Here be dragons. LSP positions use columns measured in *UTF-16 code units*! 24 // Clangd uses UTF-8 and byte-offsets internally, so conversion is nontrivial. 25 26 // Iterates over unicode codepoints in the (UTF-8) string. For each, 27 // invokes CB(UTF-8 length, UTF-16 length), and breaks if it returns true. 28 // Returns true if CB returned true, false if we hit the end of string. 29 template <typename Callback> 30 static bool iterateCodepoints(StringRef U8, const Callback &CB) { 31 for (size_t I = 0; I < U8.size();) { 32 unsigned char C = static_cast<unsigned char>(U8[I]); 33 if (LLVM_LIKELY(!(C & 0x80))) { // ASCII character. 34 if (CB(1, 1)) 35 return true; 36 ++I; 37 continue; 38 } 39 // This convenient property of UTF-8 holds for all non-ASCII characters. 40 size_t UTF8Length = countLeadingOnes(C); 41 // 0xxx is ASCII, handled above. 10xxx is a trailing byte, invalid here. 42 // 11111xxx is not valid UTF-8 at all. Assert because it's probably our bug. 43 assert((UTF8Length >= 2 && UTF8Length <= 4) && 44 "Invalid UTF-8, or transcoding bug?"); 45 I += UTF8Length; // Skip over all trailing bytes. 46 // A codepoint takes two UTF-16 code unit if it's astral (outside BMP). 47 // Astral codepoints are encoded as 4 bytes in UTF-8 (11110xxx ...) 48 if (CB(UTF8Length, UTF8Length == 4 ? 2 : 1)) 49 return true; 50 } 51 return false; 52 } 53 54 // Returns the offset into the string that matches \p Units UTF-16 code units. 55 // Conceptually, this converts to UTF-16, truncates to CodeUnits, converts back 56 // to UTF-8, and returns the length in bytes. 57 static size_t measureUTF16(StringRef U8, int U16Units, bool &Valid) { 58 size_t Result = 0; 59 Valid = U16Units == 0 || iterateCodepoints(U8, [&](int U8Len, int U16Len) { 60 Result += U8Len; 61 U16Units -= U16Len; 62 return U16Units <= 0; 63 }); 64 if (U16Units < 0) // Offset was into the middle of a surrogate pair. 65 Valid = false; 66 // Don't return an out-of-range index if we overran. 67 return std::min(Result, U8.size()); 68 } 69 70 // Counts the number of UTF-16 code units needed to represent a string. 71 // Like most strings in clangd, the input is UTF-8 encoded. 72 static size_t utf16Len(StringRef U8) { 73 // A codepoint takes two UTF-16 code unit if it's astral (outside BMP). 74 // Astral codepoints are encoded as 4 bytes in UTF-8, starting with 11110xxx. 75 size_t Count = 0; 76 iterateCodepoints(U8, [&](int U8Len, int U16Len) { 77 Count += U16Len; 78 return false; 79 }); 80 return Count; 81 } 82 83 llvm::Expected<size_t> positionToOffset(StringRef Code, Position P, 84 bool AllowColumnsBeyondLineLength) { 85 if (P.line < 0) 86 return llvm::make_error<llvm::StringError>( 87 llvm::formatv("Line value can't be negative ({0})", P.line), 88 llvm::errc::invalid_argument); 89 if (P.character < 0) 90 return llvm::make_error<llvm::StringError>( 91 llvm::formatv("Character value can't be negative ({0})", P.character), 92 llvm::errc::invalid_argument); 93 size_t StartOfLine = 0; 94 for (int I = 0; I != P.line; ++I) { 95 size_t NextNL = Code.find('\n', StartOfLine); 96 if (NextNL == StringRef::npos) 97 return llvm::make_error<llvm::StringError>( 98 llvm::formatv("Line value is out of range ({0})", P.line), 99 llvm::errc::invalid_argument); 100 StartOfLine = NextNL + 1; 101 } 102 103 size_t NextNL = Code.find('\n', StartOfLine); 104 if (NextNL == StringRef::npos) 105 NextNL = Code.size(); 106 107 bool Valid; 108 size_t ByteOffsetInLine = measureUTF16( 109 Code.substr(StartOfLine, NextNL - StartOfLine), P.character, Valid); 110 if (!Valid && !AllowColumnsBeyondLineLength) 111 return llvm::make_error<llvm::StringError>( 112 llvm::formatv("UTF-16 offset {0} is invalid for line {1}", P.character, 113 P.line), 114 llvm::errc::invalid_argument); 115 return StartOfLine + ByteOffsetInLine; 116 } 117 118 Position offsetToPosition(StringRef Code, size_t Offset) { 119 Offset = std::min(Code.size(), Offset); 120 StringRef Before = Code.substr(0, Offset); 121 int Lines = Before.count('\n'); 122 size_t PrevNL = Before.rfind('\n'); 123 size_t StartOfLine = (PrevNL == StringRef::npos) ? 0 : (PrevNL + 1); 124 Position Pos; 125 Pos.line = Lines; 126 Pos.character = utf16Len(Before.substr(StartOfLine)); 127 return Pos; 128 } 129 130 Position sourceLocToPosition(const SourceManager &SM, SourceLocation Loc) { 131 // We use the SourceManager's line tables, but its column number is in bytes. 132 FileID FID; 133 unsigned Offset; 134 std::tie(FID, Offset) = SM.getDecomposedSpellingLoc(Loc); 135 Position P; 136 P.line = static_cast<int>(SM.getLineNumber(FID, Offset)) - 1; 137 bool Invalid = false; 138 StringRef Code = SM.getBufferData(FID, &Invalid); 139 if (!Invalid) { 140 auto ColumnInBytes = SM.getColumnNumber(FID, Offset) - 1; 141 auto LineSoFar = Code.substr(Offset - ColumnInBytes, ColumnInBytes); 142 P.character = utf16Len(LineSoFar); 143 } 144 return P; 145 } 146 147 Range halfOpenToRange(const SourceManager &SM, CharSourceRange R) { 148 // Clang is 1-based, LSP uses 0-based indexes. 149 Position Begin = sourceLocToPosition(SM, R.getBegin()); 150 Position End = sourceLocToPosition(SM, R.getEnd()); 151 152 return {Begin, End}; 153 } 154 155 std::pair<size_t, size_t> offsetToClangLineColumn(StringRef Code, 156 size_t Offset) { 157 Offset = std::min(Code.size(), Offset); 158 StringRef Before = Code.substr(0, Offset); 159 int Lines = Before.count('\n'); 160 size_t PrevNL = Before.rfind('\n'); 161 size_t StartOfLine = (PrevNL == StringRef::npos) ? 0 : (PrevNL + 1); 162 return {Lines + 1, Offset - StartOfLine + 1}; 163 } 164 165 std::pair<llvm::StringRef, llvm::StringRef> 166 splitQualifiedName(llvm::StringRef QName) { 167 size_t Pos = QName.rfind("::"); 168 if (Pos == llvm::StringRef::npos) 169 return {StringRef(), QName}; 170 return {QName.substr(0, Pos + 2), QName.substr(Pos + 2)}; 171 } 172 173 TextEdit replacementToEdit(StringRef Code, const tooling::Replacement &R) { 174 Range ReplacementRange = { 175 offsetToPosition(Code, R.getOffset()), 176 offsetToPosition(Code, R.getOffset() + R.getLength())}; 177 return {ReplacementRange, R.getReplacementText()}; 178 } 179 180 std::vector<TextEdit> replacementsToEdits(StringRef Code, 181 const tooling::Replacements &Repls) { 182 std::vector<TextEdit> Edits; 183 for (const auto &R : Repls) 184 Edits.push_back(replacementToEdit(Code, R)); 185 return Edits; 186 } 187 188 llvm::Optional<std::string> getRealPath(const FileEntry *F, 189 const SourceManager &SourceMgr) { 190 // Ideally, we get the real path from the FileEntry object. 191 SmallString<128> FilePath = F->tryGetRealPathName(); 192 if (!FilePath.empty()) { 193 return FilePath.str().str(); 194 } 195 196 // Otherwise, we try to compute ourselves. 197 vlog("FileEntry for {0} did not contain the real path.", F->getName()); 198 199 llvm::SmallString<128> Path = F->getName(); 200 201 if (!llvm::sys::path::is_absolute(Path)) { 202 if (!SourceMgr.getFileManager().makeAbsolutePath(Path)) { 203 log("Could not turn relative path to absolute: {0}", Path); 204 return llvm::None; 205 } 206 } 207 208 llvm::SmallString<128> RealPath; 209 if (SourceMgr.getFileManager().getVirtualFileSystem()->getRealPath( 210 Path, RealPath)) { 211 log("Could not compute real path: {0}", Path); 212 return Path.str().str(); 213 } 214 215 return RealPath.str().str(); 216 } 217 218 TextEdit toTextEdit(const FixItHint &FixIt, const SourceManager &M, 219 const LangOptions &L) { 220 TextEdit Result; 221 Result.range = 222 halfOpenToRange(M, Lexer::makeFileCharRange(FixIt.RemoveRange, M, L)); 223 Result.newText = FixIt.CodeToInsert; 224 return Result; 225 } 226 227 bool IsRangeConsecutive(const Range &Left, const Range &Right) { 228 return Left.end.line == Right.start.line && 229 Left.end.character == Right.start.character; 230 } 231 232 } // namespace clangd 233 } // namespace clang 234