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 
22 // Here be dragons. LSP positions use columns measured in *UTF-16 code units*!
23 // Clangd uses UTF-8 and byte-offsets internally, so conversion is nontrivial.
24 
25 // Iterates over unicode codepoints in the (UTF-8) string. For each,
26 // invokes CB(UTF-8 length, UTF-16 length), and breaks if it returns true.
27 // Returns true if CB returned true, false if we hit the end of string.
28 template <typename Callback>
29 static bool iterateCodepoints(llvm::StringRef U8, const Callback &CB) {
30   for (size_t I = 0; I < U8.size();) {
31     unsigned char C = static_cast<unsigned char>(U8[I]);
32     if (LLVM_LIKELY(!(C & 0x80))) { // ASCII character.
33       if (CB(1, 1))
34         return true;
35       ++I;
36       continue;
37     }
38     // This convenient property of UTF-8 holds for all non-ASCII characters.
39     size_t UTF8Length = llvm::countLeadingOnes(C);
40     // 0xxx is ASCII, handled above. 10xxx is a trailing byte, invalid here.
41     // 11111xxx is not valid UTF-8 at all. Assert because it's probably our bug.
42     assert((UTF8Length >= 2 && UTF8Length <= 4) &&
43            "Invalid UTF-8, or transcoding bug?");
44     I += UTF8Length; // Skip over all trailing bytes.
45     // A codepoint takes two UTF-16 code unit if it's astral (outside BMP).
46     // Astral codepoints are encoded as 4 bytes in UTF-8 (11110xxx ...)
47     if (CB(UTF8Length, UTF8Length == 4 ? 2 : 1))
48       return true;
49   }
50   return false;
51 }
52 
53 // Returns the offset into the string that matches \p Units UTF-16 code units.
54 // Conceptually, this converts to UTF-16, truncates to CodeUnits, converts back
55 // to UTF-8, and returns the length in bytes.
56 static size_t measureUTF16(llvm::StringRef U8, int U16Units, bool &Valid) {
57   size_t Result = 0;
58   Valid = U16Units == 0 || iterateCodepoints(U8, [&](int U8Len, int U16Len) {
59             Result += U8Len;
60             U16Units -= U16Len;
61             return U16Units <= 0;
62           });
63   if (U16Units < 0) // Offset was into the middle of a surrogate pair.
64     Valid = false;
65   // Don't return an out-of-range index if we overran.
66   return std::min(Result, U8.size());
67 }
68 
69 // Like most strings in clangd, the input is UTF-8 encoded.
70 size_t lspLength(llvm::StringRef Code) {
71   // A codepoint takes two UTF-16 code unit if it's astral (outside BMP).
72   // Astral codepoints are encoded as 4 bytes in UTF-8, starting with 11110xxx.
73   size_t Count = 0;
74   iterateCodepoints(Code, [&](int U8Len, int U16Len) {
75     Count += U16Len;
76     return false;
77   });
78   return Count;
79 }
80 
81 llvm::Expected<size_t> positionToOffset(llvm::StringRef Code, Position P,
82                                         bool AllowColumnsBeyondLineLength) {
83   if (P.line < 0)
84     return llvm::make_error<llvm::StringError>(
85         llvm::formatv("Line value can't be negative ({0})", P.line),
86         llvm::errc::invalid_argument);
87   if (P.character < 0)
88     return llvm::make_error<llvm::StringError>(
89         llvm::formatv("Character value can't be negative ({0})", P.character),
90         llvm::errc::invalid_argument);
91   size_t StartOfLine = 0;
92   for (int I = 0; I != P.line; ++I) {
93     size_t NextNL = Code.find('\n', StartOfLine);
94     if (NextNL == llvm::StringRef::npos)
95       return llvm::make_error<llvm::StringError>(
96           llvm::formatv("Line value is out of range ({0})", P.line),
97           llvm::errc::invalid_argument);
98     StartOfLine = NextNL + 1;
99   }
100 
101   size_t NextNL = Code.find('\n', StartOfLine);
102   if (NextNL == llvm::StringRef::npos)
103     NextNL = Code.size();
104 
105   bool Valid;
106   size_t ByteOffsetInLine = measureUTF16(
107       Code.substr(StartOfLine, NextNL - StartOfLine), P.character, Valid);
108   if (!Valid && !AllowColumnsBeyondLineLength)
109     return llvm::make_error<llvm::StringError>(
110         llvm::formatv("UTF-16 offset {0} is invalid for line {1}", P.character,
111                       P.line),
112         llvm::errc::invalid_argument);
113   return StartOfLine + ByteOffsetInLine;
114 }
115 
116 Position offsetToPosition(llvm::StringRef Code, size_t Offset) {
117   Offset = std::min(Code.size(), Offset);
118   llvm::StringRef Before = Code.substr(0, Offset);
119   int Lines = Before.count('\n');
120   size_t PrevNL = Before.rfind('\n');
121   size_t StartOfLine = (PrevNL == llvm::StringRef::npos) ? 0 : (PrevNL + 1);
122   Position Pos;
123   Pos.line = Lines;
124   Pos.character = lspLength(Before.substr(StartOfLine));
125   return Pos;
126 }
127 
128 Position sourceLocToPosition(const SourceManager &SM, SourceLocation Loc) {
129   // We use the SourceManager's line tables, but its column number is in bytes.
130   FileID FID;
131   unsigned Offset;
132   std::tie(FID, Offset) = SM.getDecomposedSpellingLoc(Loc);
133   Position P;
134   P.line = static_cast<int>(SM.getLineNumber(FID, Offset)) - 1;
135   bool Invalid = false;
136   llvm::StringRef Code = SM.getBufferData(FID, &Invalid);
137   if (!Invalid) {
138     auto ColumnInBytes = SM.getColumnNumber(FID, Offset) - 1;
139     auto LineSoFar = Code.substr(Offset - ColumnInBytes, ColumnInBytes);
140     P.character = lspLength(LineSoFar);
141   }
142   return P;
143 }
144 
145 Range halfOpenToRange(const SourceManager &SM, CharSourceRange R) {
146   // Clang is 1-based, LSP uses 0-based indexes.
147   Position Begin = sourceLocToPosition(SM, R.getBegin());
148   Position End = sourceLocToPosition(SM, R.getEnd());
149 
150   return {Begin, End};
151 }
152 
153 std::pair<size_t, size_t> offsetToClangLineColumn(llvm::StringRef Code,
154                                                   size_t Offset) {
155   Offset = std::min(Code.size(), Offset);
156   llvm::StringRef Before = Code.substr(0, Offset);
157   int Lines = Before.count('\n');
158   size_t PrevNL = Before.rfind('\n');
159   size_t StartOfLine = (PrevNL == llvm::StringRef::npos) ? 0 : (PrevNL + 1);
160   return {Lines + 1, Offset - StartOfLine + 1};
161 }
162 
163 std::pair<llvm::StringRef, llvm::StringRef>
164 splitQualifiedName(llvm::StringRef QName) {
165   size_t Pos = QName.rfind("::");
166   if (Pos == llvm::StringRef::npos)
167     return {llvm::StringRef(), QName};
168   return {QName.substr(0, Pos + 2), QName.substr(Pos + 2)};
169 }
170 
171 TextEdit replacementToEdit(llvm::StringRef Code,
172                            const tooling::Replacement &R) {
173   Range ReplacementRange = {
174       offsetToPosition(Code, R.getOffset()),
175       offsetToPosition(Code, R.getOffset() + R.getLength())};
176   return {ReplacementRange, R.getReplacementText()};
177 }
178 
179 std::vector<TextEdit> replacementsToEdits(llvm::StringRef Code,
180                                           const tooling::Replacements &Repls) {
181   std::vector<TextEdit> Edits;
182   for (const auto &R : Repls)
183     Edits.push_back(replacementToEdit(Code, R));
184   return Edits;
185 }
186 
187 llvm::Optional<std::string> getCanonicalPath(const FileEntry *F,
188                                              const SourceManager &SourceMgr) {
189   if (!F)
190     return None;
191 
192   llvm::SmallString<128> FilePath = F->getName();
193   if (!llvm::sys::path::is_absolute(FilePath)) {
194     if (auto EC =
195             SourceMgr.getFileManager().getVirtualFileSystem()->makeAbsolute(
196                 FilePath)) {
197       elog("Could not turn relative path '{0}' to absolute: {1}", FilePath,
198            EC.message());
199       return None;
200     }
201   }
202 
203   // Handle the symbolic link path case where the current working directory
204   // (getCurrentWorkingDirectory) is a symlink./ We always want to the real
205   // file path (instead of the symlink path) for the  C++ symbols.
206   //
207   // Consider the following example:
208   //
209   //   src dir: /project/src/foo.h
210   //   current working directory (symlink): /tmp/build -> /project/src/
211   //
212   //  The file path of Symbol is "/project/src/foo.h" instead of
213   //  "/tmp/build/foo.h"
214   if (const DirectoryEntry *Dir = SourceMgr.getFileManager().getDirectory(
215           llvm::sys::path::parent_path(FilePath))) {
216     llvm::SmallString<128> RealPath;
217     llvm::StringRef DirName = SourceMgr.getFileManager().getCanonicalName(Dir);
218     llvm::sys::path::append(RealPath, DirName,
219                             llvm::sys::path::filename(FilePath));
220     return RealPath.str().str();
221   }
222 
223   return FilePath.str().str();
224 }
225 
226 TextEdit toTextEdit(const FixItHint &FixIt, const SourceManager &M,
227                     const LangOptions &L) {
228   TextEdit Result;
229   Result.range =
230       halfOpenToRange(M, Lexer::makeFileCharRange(FixIt.RemoveRange, M, L));
231   Result.newText = FixIt.CodeToInsert;
232   return Result;
233 }
234 
235 bool IsRangeConsecutive(const Range &Left, const Range &Right) {
236   return Left.end.line == Right.start.line &&
237          Left.end.character == Right.start.character;
238 }
239 
240 FileDigest digest(llvm::StringRef Content) {
241   return llvm::SHA1::hash({(const uint8_t *)Content.data(), Content.size()});
242 }
243 
244 llvm::Optional<FileDigest> digestFile(const SourceManager &SM, FileID FID) {
245   bool Invalid = false;
246   llvm::StringRef Content = SM.getBufferData(FID, &Invalid);
247   if (Invalid)
248     return None;
249   return digest(Content);
250 }
251 
252 } // namespace clangd
253 } // namespace clang
254