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 using namespace llvm;
20 namespace clang {
21 namespace clangd {
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 // Like most strings in clangd, the input is UTF-8 encoded.
71 size_t lspLength(StringRef Code) {
72   // A codepoint takes two UTF-16 code unit if it's astral (outside BMP).
73   // Astral codepoints are encoded as 4 bytes in UTF-8, starting with 11110xxx.
74   size_t Count = 0;
75   iterateCodepoints(Code, [&](int U8Len, int U16Len) {
76     Count += U16Len;
77     return false;
78   });
79   return Count;
80 }
81 
82 Expected<size_t> positionToOffset(StringRef Code, Position P,
83                                   bool AllowColumnsBeyondLineLength) {
84   if (P.line < 0)
85     return make_error<StringError>(
86         formatv("Line value can't be negative ({0})", P.line),
87         errc::invalid_argument);
88   if (P.character < 0)
89     return make_error<StringError>(
90         formatv("Character value can't be negative ({0})", P.character),
91         errc::invalid_argument);
92   size_t StartOfLine = 0;
93   for (int I = 0; I != P.line; ++I) {
94     size_t NextNL = Code.find('\n', StartOfLine);
95     if (NextNL == StringRef::npos)
96       return make_error<StringError>(
97           formatv("Line value is out of range ({0})", P.line),
98           errc::invalid_argument);
99     StartOfLine = NextNL + 1;
100   }
101 
102   size_t NextNL = Code.find('\n', StartOfLine);
103   if (NextNL == StringRef::npos)
104     NextNL = Code.size();
105 
106   bool Valid;
107   size_t ByteOffsetInLine = measureUTF16(
108       Code.substr(StartOfLine, NextNL - StartOfLine), P.character, Valid);
109   if (!Valid && !AllowColumnsBeyondLineLength)
110     return make_error<StringError>(
111         formatv("UTF-16 offset {0} is invalid for line {1}", P.character,
112                 P.line),
113         errc::invalid_argument);
114   return StartOfLine + ByteOffsetInLine;
115 }
116 
117 Position offsetToPosition(StringRef Code, size_t Offset) {
118   Offset = std::min(Code.size(), Offset);
119   StringRef Before = Code.substr(0, Offset);
120   int Lines = Before.count('\n');
121   size_t PrevNL = Before.rfind('\n');
122   size_t StartOfLine = (PrevNL == StringRef::npos) ? 0 : (PrevNL + 1);
123   Position Pos;
124   Pos.line = Lines;
125   Pos.character = lspLength(Before.substr(StartOfLine));
126   return Pos;
127 }
128 
129 Position sourceLocToPosition(const SourceManager &SM, SourceLocation Loc) {
130   // We use the SourceManager's line tables, but its column number is in bytes.
131   FileID FID;
132   unsigned Offset;
133   std::tie(FID, Offset) = SM.getDecomposedSpellingLoc(Loc);
134   Position P;
135   P.line = static_cast<int>(SM.getLineNumber(FID, Offset)) - 1;
136   bool Invalid = false;
137   StringRef Code = SM.getBufferData(FID, &Invalid);
138   if (!Invalid) {
139     auto ColumnInBytes = SM.getColumnNumber(FID, Offset) - 1;
140     auto LineSoFar = Code.substr(Offset - ColumnInBytes, ColumnInBytes);
141     P.character = lspLength(LineSoFar);
142   }
143   return P;
144 }
145 
146 Range halfOpenToRange(const SourceManager &SM, CharSourceRange R) {
147   // Clang is 1-based, LSP uses 0-based indexes.
148   Position Begin = sourceLocToPosition(SM, R.getBegin());
149   Position End = sourceLocToPosition(SM, R.getEnd());
150 
151   return {Begin, End};
152 }
153 
154 std::pair<size_t, size_t> offsetToClangLineColumn(StringRef Code,
155                                                   size_t Offset) {
156   Offset = std::min(Code.size(), Offset);
157   StringRef Before = Code.substr(0, Offset);
158   int Lines = Before.count('\n');
159   size_t PrevNL = Before.rfind('\n');
160   size_t StartOfLine = (PrevNL == StringRef::npos) ? 0 : (PrevNL + 1);
161   return {Lines + 1, Offset - StartOfLine + 1};
162 }
163 
164 std::pair<StringRef, StringRef> splitQualifiedName(StringRef QName) {
165   size_t Pos = QName.rfind("::");
166   if (Pos == StringRef::npos)
167     return {StringRef(), QName};
168   return {QName.substr(0, Pos + 2), QName.substr(Pos + 2)};
169 }
170 
171 TextEdit replacementToEdit(StringRef Code, const tooling::Replacement &R) {
172   Range ReplacementRange = {
173       offsetToPosition(Code, R.getOffset()),
174       offsetToPosition(Code, R.getOffset() + R.getLength())};
175   return {ReplacementRange, R.getReplacementText()};
176 }
177 
178 std::vector<TextEdit> replacementsToEdits(StringRef Code,
179                                           const tooling::Replacements &Repls) {
180   std::vector<TextEdit> Edits;
181   for (const auto &R : Repls)
182     Edits.push_back(replacementToEdit(Code, R));
183   return Edits;
184 }
185 
186 Optional<std::string> getCanonicalPath(const FileEntry *F,
187                                        const SourceManager &SourceMgr) {
188   if (!F)
189     return None;
190 
191   SmallString<128> FilePath = F->getName();
192   if (!sys::path::is_absolute(FilePath)) {
193     if (auto EC =
194             SourceMgr.getFileManager().getVirtualFileSystem()->makeAbsolute(
195                 FilePath)) {
196       elog("Could not turn relative path '{0}' to absolute: {1}", FilePath,
197            EC.message());
198       return None;
199     }
200   }
201 
202   // Handle the symbolic link path case where the current working directory
203   // (getCurrentWorkingDirectory) is a symlink./ We always want to the real
204   // file path (instead of the symlink path) for the  C++ symbols.
205   //
206   // Consider the following example:
207   //
208   //   src dir: /project/src/foo.h
209   //   current working directory (symlink): /tmp/build -> /project/src/
210   //
211   //  The file path of Symbol is "/project/src/foo.h" instead of
212   //  "/tmp/build/foo.h"
213   if (const DirectoryEntry *Dir = SourceMgr.getFileManager().getDirectory(
214           sys::path::parent_path(FilePath))) {
215     SmallString<128> RealPath;
216     StringRef DirName = SourceMgr.getFileManager().getCanonicalName(Dir);
217     sys::path::append(RealPath, DirName, sys::path::filename(FilePath));
218     return RealPath.str().str();
219   }
220 
221   return FilePath.str().str();
222 }
223 
224 TextEdit toTextEdit(const FixItHint &FixIt, const SourceManager &M,
225                     const LangOptions &L) {
226   TextEdit Result;
227   Result.range =
228       halfOpenToRange(M, Lexer::makeFileCharRange(FixIt.RemoveRange, M, L));
229   Result.newText = FixIt.CodeToInsert;
230   return Result;
231 }
232 
233 bool IsRangeConsecutive(const Range &Left, const Range &Right) {
234   return Left.end.line == Right.start.line &&
235          Left.end.character == Right.start.character;
236 }
237 
238 FileDigest digest(StringRef Content) {
239   return llvm::SHA1::hash({(const uint8_t *)Content.data(), Content.size()});
240 }
241 
242 Optional<FileDigest> digestFile(const SourceManager &SM, FileID FID) {
243   bool Invalid = false;
244   StringRef Content = SM.getBufferData(FID, &Invalid);
245   if (Invalid)
246     return None;
247   return digest(Content);
248 }
249 
250 } // namespace clangd
251 } // namespace clang
252