clang-tools  10.0.0svn
Serialization.cpp
Go to the documentation of this file.
1 //===-- Serialization.cpp - Binary serialization of index data ------------===//
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 "Serialization.h"
10 #include "Headers.h"
11 #include "Logger.h"
12 #include "RIFF.h"
13 #include "SymbolLocation.h"
14 #include "SymbolOrigin.h"
15 #include "Trace.h"
16 #include "dex/Dex.h"
17 #include "clang/Tooling/CompilationDatabase.h"
18 #include "llvm/ADT/StringRef.h"
19 #include "llvm/Support/Compression.h"
20 #include "llvm/Support/Endian.h"
21 #include "llvm/Support/Error.h"
22 #include "llvm/Support/raw_ostream.h"
23 #include <vector>
24 
25 namespace clang {
26 namespace clangd {
27 namespace {
28 llvm::Error makeError(const llvm::Twine &Msg) {
29  return llvm::make_error<llvm::StringError>(Msg,
30  llvm::inconvertibleErrorCode());
31 }
32 
33 // IO PRIMITIVES
34 // We use little-endian 32 bit ints, sometimes with variable-length encoding.
35 //
36 // Variable-length int encoding (varint) uses the bottom 7 bits of each byte
37 // to encode the number, and the top bit to indicate whether more bytes follow.
38 // e.g. 9a 2f means [0x1a and keep reading, 0x2f and stop].
39 // This represents 0x1a | 0x2f<<7 = 6042.
40 // A 32-bit integer takes 1-5 bytes to encode; small numbers are more compact.
41 
42 // Reads binary data from a StringRef, and keeps track of position.
43 class Reader {
44  const char *Begin, *End;
45  bool Err = false;
46 
47 public:
48  Reader(llvm::StringRef Data) : Begin(Data.begin()), End(Data.end()) {}
49  // The "error" bit is set by reading past EOF or reading invalid data.
50  // When in an error state, reads may return zero values: callers should check.
51  bool err() const { return Err; }
52  // Did we read all the data, or encounter an error?
53  bool eof() const { return Begin == End || Err; }
54  // All the data we didn't read yet.
55  llvm::StringRef rest() const { return llvm::StringRef(Begin, End - Begin); }
56 
57  uint8_t consume8() {
58  if (LLVM_UNLIKELY(Begin == End)) {
59  Err = true;
60  return 0;
61  }
62  return *Begin++;
63  }
64 
65  uint32_t consume32() {
66  if (LLVM_UNLIKELY(Begin + 4 > End)) {
67  Err = true;
68  return 0;
69  }
70  auto Ret = llvm::support::endian::read32le(Begin);
71  Begin += 4;
72  return Ret;
73  }
74 
75  llvm::StringRef consume(int N) {
76  if (LLVM_UNLIKELY(Begin + N > End)) {
77  Err = true;
78  return llvm::StringRef();
79  }
80  llvm::StringRef Ret(Begin, N);
81  Begin += N;
82  return Ret;
83  }
84 
85  uint32_t consumeVar() {
86  constexpr static uint8_t More = 1 << 7;
87  uint8_t B = consume8();
88  if (LLVM_LIKELY(!(B & More)))
89  return B;
90  uint32_t Val = B & ~More;
91  for (int Shift = 7; B & More && Shift < 32; Shift += 7) {
92  B = consume8();
93  Val |= (B & ~More) << Shift;
94  }
95  return Val;
96  }
97 
98  llvm::StringRef consumeString(llvm::ArrayRef<llvm::StringRef> Strings) {
99  auto StringIndex = consumeVar();
100  if (LLVM_UNLIKELY(StringIndex >= Strings.size())) {
101  Err = true;
102  return llvm::StringRef();
103  }
104  return Strings[StringIndex];
105  }
106 
107  SymbolID consumeID() {
108  llvm::StringRef Raw = consume(SymbolID::RawSize); // short if truncated.
109  return LLVM_UNLIKELY(err()) ? SymbolID() : SymbolID::fromRaw(Raw);
110  }
111 };
112 
113 void write32(uint32_t I, llvm::raw_ostream &OS) {
114  char Buf[4];
115  llvm::support::endian::write32le(Buf, I);
116  OS.write(Buf, sizeof(Buf));
117 }
118 
119 void writeVar(uint32_t I, llvm::raw_ostream &OS) {
120  constexpr static uint8_t More = 1 << 7;
121  if (LLVM_LIKELY(I < 1 << 7)) {
122  OS.write(I);
123  return;
124  }
125  for (;;) {
126  OS.write(I | More);
127  I >>= 7;
128  if (I < 1 << 7) {
129  OS.write(I);
130  return;
131  }
132  }
133 }
134 
135 // STRING TABLE ENCODING
136 // Index data has many string fields, and many strings are identical.
137 // We store each string once, and refer to them by index.
138 //
139 // The string table's format is:
140 // - UncompressedSize : uint32 (or 0 for no compression)
141 // - CompressedData : byte[CompressedSize]
142 //
143 // CompressedData is a zlib-compressed byte[UncompressedSize].
144 // It contains a sequence of null-terminated strings, e.g. "foo\0bar\0".
145 // These are sorted to improve compression.
146 
147 // Maps each string to a canonical representation.
148 // Strings remain owned externally (e.g. by SymbolSlab).
149 class StringTableOut {
150  llvm::DenseSet<llvm::StringRef> Unique;
151  std::vector<llvm::StringRef> Sorted;
152  // Since strings are interned, look up can be by pointer.
153  llvm::DenseMap<std::pair<const char *, size_t>, unsigned> Index;
154 
155 public:
156  StringTableOut() {
157  // Ensure there's at least one string in the table.
158  // Table size zero is reserved to indicate no compression.
159  Unique.insert("");
160  }
161  // Add a string to the table. Overwrites S if an identical string exists.
162  void intern(llvm::StringRef &S) { S = *Unique.insert(S).first; };
163  // Finalize the table and write it to OS. No more strings may be added.
164  void finalize(llvm::raw_ostream &OS) {
165  Sorted = {Unique.begin(), Unique.end()};
166  llvm::sort(Sorted);
167  for (unsigned I = 0; I < Sorted.size(); ++I)
168  Index.try_emplace({Sorted[I].data(), Sorted[I].size()}, I);
169 
170  std::string RawTable;
171  for (llvm::StringRef S : Sorted) {
172  RawTable.append(S);
173  RawTable.push_back(0);
174  }
175  if (llvm::zlib::isAvailable()) {
176  llvm::SmallString<1> Compressed;
177  llvm::cantFail(llvm::zlib::compress(RawTable, Compressed));
178  write32(RawTable.size(), OS);
179  OS << Compressed;
180  } else {
181  write32(0, OS); // No compression.
182  OS << RawTable;
183  }
184  }
185  // Get the ID of an string, which must be interned. Table must be finalized.
186  unsigned index(llvm::StringRef S) const {
187  assert(!Sorted.empty() && "table not finalized");
188  assert(Index.count({S.data(), S.size()}) && "string not interned");
189  return Index.find({S.data(), S.size()})->second;
190  }
191 };
192 
193 struct StringTableIn {
194  llvm::BumpPtrAllocator Arena;
195  std::vector<llvm::StringRef> Strings;
196 };
197 
198 llvm::Expected<StringTableIn> readStringTable(llvm::StringRef Data) {
199  Reader R(Data);
200  size_t UncompressedSize = R.consume32();
201  if (R.err())
202  return makeError("Truncated string table");
203 
204  llvm::StringRef Uncompressed;
205  llvm::SmallString<1> UncompressedStorage;
206  if (UncompressedSize == 0) // No compression
207  Uncompressed = R.rest();
208  else {
209  if (llvm::Error E = llvm::zlib::uncompress(R.rest(), UncompressedStorage,
210  UncompressedSize))
211  return std::move(E);
212  Uncompressed = UncompressedStorage;
213  }
214 
215  StringTableIn Table;
216  llvm::StringSaver Saver(Table.Arena);
217  R = Reader(Uncompressed);
218  for (Reader R(Uncompressed); !R.eof();) {
219  auto Len = R.rest().find(0);
220  if (Len == llvm::StringRef::npos)
221  return makeError("Bad string table: not null terminated");
222  Table.Strings.push_back(Saver.save(R.consume(Len)));
223  R.consume8();
224  }
225  if (R.err())
226  return makeError("Truncated string table");
227  return std::move(Table);
228 }
229 
230 // SYMBOL ENCODING
231 // Each field of clangd::Symbol is encoded in turn (see implementation).
232 // - StringRef fields encode as varint (index into the string table)
233 // - enums encode as the underlying type
234 // - most numbers encode as varint
235 
236 void writeLocation(const SymbolLocation &Loc, const StringTableOut &Strings,
237  llvm::raw_ostream &OS) {
238  writeVar(Strings.index(Loc.FileURI), OS);
239  for (const auto &Endpoint : {Loc.Start, Loc.End}) {
240  writeVar(Endpoint.line(), OS);
241  writeVar(Endpoint.column(), OS);
242  }
243 }
244 
245 SymbolLocation readLocation(Reader &Data,
246  llvm::ArrayRef<llvm::StringRef> Strings) {
247  SymbolLocation Loc;
248  Loc.FileURI = Data.consumeString(Strings).data();
249  for (auto *Endpoint : {&Loc.Start, &Loc.End}) {
250  Endpoint->setLine(Data.consumeVar());
251  Endpoint->setColumn(Data.consumeVar());
252  }
253  return Loc;
254 }
255 
256 IncludeGraphNode readIncludeGraphNode(Reader &Data,
257  llvm::ArrayRef<llvm::StringRef> Strings) {
258  IncludeGraphNode IGN;
259  IGN.Flags = static_cast<IncludeGraphNode::SourceFlag>(Data.consume8());
260  IGN.URI = Data.consumeString(Strings);
261  llvm::StringRef Digest = Data.consume(IGN.Digest.size());
262  std::copy(Digest.bytes_begin(), Digest.bytes_end(), IGN.Digest.begin());
263  IGN.DirectIncludes.resize(Data.consumeVar());
264  for (llvm::StringRef &Include : IGN.DirectIncludes)
265  Include = Data.consumeString(Strings);
266  return IGN;
267 }
268 
269 void writeIncludeGraphNode(const IncludeGraphNode &IGN,
270  const StringTableOut &Strings,
271  llvm::raw_ostream &OS) {
272  OS.write(static_cast<uint8_t>(IGN.Flags));
273  writeVar(Strings.index(IGN.URI), OS);
274  llvm::StringRef Hash(reinterpret_cast<const char *>(IGN.Digest.data()),
275  IGN.Digest.size());
276  OS << Hash;
277  writeVar(IGN.DirectIncludes.size(), OS);
278  for (llvm::StringRef Include : IGN.DirectIncludes)
279  writeVar(Strings.index(Include), OS);
280 }
281 
282 void writeSymbol(const Symbol &Sym, const StringTableOut &Strings,
283  llvm::raw_ostream &OS) {
284  OS << Sym.ID.raw(); // TODO: once we start writing xrefs and posting lists,
285  // symbol IDs should probably be in a string table.
286  OS.write(static_cast<uint8_t>(Sym.SymInfo.Kind));
287  OS.write(static_cast<uint8_t>(Sym.SymInfo.Lang));
288  writeVar(Strings.index(Sym.Name), OS);
289  writeVar(Strings.index(Sym.Scope), OS);
290  writeVar(Strings.index(Sym.TemplateSpecializationArgs), OS);
291  writeLocation(Sym.Definition, Strings, OS);
292  writeLocation(Sym.CanonicalDeclaration, Strings, OS);
293  writeVar(Sym.References, OS);
294  OS.write(static_cast<uint8_t>(Sym.Flags));
295  OS.write(static_cast<uint8_t>(Sym.Origin));
296  writeVar(Strings.index(Sym.Signature), OS);
297  writeVar(Strings.index(Sym.CompletionSnippetSuffix), OS);
298  writeVar(Strings.index(Sym.Documentation), OS);
299  writeVar(Strings.index(Sym.ReturnType), OS);
300  writeVar(Strings.index(Sym.Type), OS);
301 
302  auto WriteInclude = [&](const Symbol::IncludeHeaderWithReferences &Include) {
303  writeVar(Strings.index(Include.IncludeHeader), OS);
304  writeVar(Include.References, OS);
305  };
306  writeVar(Sym.IncludeHeaders.size(), OS);
307  for (const auto &Include : Sym.IncludeHeaders)
308  WriteInclude(Include);
309 }
310 
311 Symbol readSymbol(Reader &Data, llvm::ArrayRef<llvm::StringRef> Strings) {
312  Symbol Sym;
313  Sym.ID = Data.consumeID();
314  Sym.SymInfo.Kind = static_cast<index::SymbolKind>(Data.consume8());
315  Sym.SymInfo.Lang = static_cast<index::SymbolLanguage>(Data.consume8());
316  Sym.Name = Data.consumeString(Strings);
317  Sym.Scope = Data.consumeString(Strings);
318  Sym.TemplateSpecializationArgs = Data.consumeString(Strings);
319  Sym.Definition = readLocation(Data, Strings);
320  Sym.CanonicalDeclaration = readLocation(Data, Strings);
321  Sym.References = Data.consumeVar();
322  Sym.Flags = static_cast<Symbol::SymbolFlag>(Data.consume8());
323  Sym.Origin = static_cast<SymbolOrigin>(Data.consume8());
324  Sym.Signature = Data.consumeString(Strings);
325  Sym.CompletionSnippetSuffix = Data.consumeString(Strings);
326  Sym.Documentation = Data.consumeString(Strings);
327  Sym.ReturnType = Data.consumeString(Strings);
328  Sym.Type = Data.consumeString(Strings);
329  Sym.IncludeHeaders.resize(Data.consumeVar());
330  for (auto &I : Sym.IncludeHeaders) {
331  I.IncludeHeader = Data.consumeString(Strings);
332  I.References = Data.consumeVar();
333  }
334  return Sym;
335 }
336 
337 // REFS ENCODING
338 // A refs section has data grouped by Symbol. Each symbol has:
339 // - SymbolID: 8 bytes
340 // - NumRefs: varint
341 // - Ref[NumRefs]
342 // Fields of Ref are encoded in turn, see implementation.
343 
344 void writeRefs(const SymbolID &ID, llvm::ArrayRef<Ref> Refs,
345  const StringTableOut &Strings, llvm::raw_ostream &OS) {
346  OS << ID.raw();
347  writeVar(Refs.size(), OS);
348  for (const auto &Ref : Refs) {
349  OS.write(static_cast<unsigned char>(Ref.Kind));
350  writeLocation(Ref.Location, Strings, OS);
351  }
352 }
353 
354 std::pair<SymbolID, std::vector<Ref>>
355 readRefs(Reader &Data, llvm::ArrayRef<llvm::StringRef> Strings) {
356  std::pair<SymbolID, std::vector<Ref>> Result;
357  Result.first = Data.consumeID();
358  Result.second.resize(Data.consumeVar());
359  for (auto &Ref : Result.second) {
360  Ref.Kind = static_cast<RefKind>(Data.consume8());
361  Ref.Location = readLocation(Data, Strings);
362  }
363  return Result;
364 }
365 
366 // RELATIONS ENCODING
367 // A relations section is a flat list of relations. Each relation has:
368 // - SymbolID (subject): 8 bytes
369 // - relation kind (predicate): 1 byte
370 // - SymbolID (object): 8 bytes
371 // In the future, we might prefer a packed representation if the need arises.
372 
373 void writeRelation(const Relation &R, llvm::raw_ostream &OS) {
374  OS << R.Subject.raw();
375  OS.write(static_cast<uint8_t>(R.Predicate));
376  OS << R.Object.raw();
377 }
378 
379 Relation readRelation(Reader &Data) {
380  SymbolID Subject = Data.consumeID();
381  RelationKind Predicate = static_cast<RelationKind>(Data.consume8());
382  SymbolID Object = Data.consumeID();
383  return {Subject, Predicate, Object};
384 }
385 
386 struct InternedCompileCommand {
387  llvm::StringRef Directory;
388  std::vector<llvm::StringRef> CommandLine;
389 };
390 
391 void writeCompileCommand(const InternedCompileCommand &Cmd,
392  const StringTableOut &Strings,
393  llvm::raw_ostream &CmdOS) {
394  writeVar(Strings.index(Cmd.Directory), CmdOS);
395  writeVar(Cmd.CommandLine.size(), CmdOS);
396  for (llvm::StringRef C : Cmd.CommandLine)
397  writeVar(Strings.index(C), CmdOS);
398 }
399 
400 InternedCompileCommand
401 readCompileCommand(Reader CmdReader, llvm::ArrayRef<llvm::StringRef> Strings) {
402  InternedCompileCommand Cmd;
403  Cmd.Directory = CmdReader.consumeString(Strings);
404  Cmd.CommandLine.resize(CmdReader.consumeVar());
405  for (llvm::StringRef &C : Cmd.CommandLine)
406  C = CmdReader.consumeString(Strings);
407  return Cmd;
408 }
409 
410 // FILE ENCODING
411 // A file is a RIFF chunk with type 'CdIx'.
412 // It contains the sections:
413 // - meta: version number
414 // - srcs: information related to include graph
415 // - stri: string table
416 // - symb: symbols
417 // - refs: references to symbols
418 
419 // The current versioning scheme is simple - non-current versions are rejected.
420 // If you make a breaking change, bump this version number to invalidate stored
421 // data. Later we may want to support some backward compatibility.
422 constexpr static uint32_t Version = 12;
423 
424 llvm::Expected<IndexFileIn> readRIFF(llvm::StringRef Data) {
425  auto RIFF = riff::readFile(Data);
426  if (!RIFF)
427  return RIFF.takeError();
428  if (RIFF->Type != riff::fourCC("CdIx"))
429  return makeError("wrong RIFF type");
430  llvm::StringMap<llvm::StringRef> Chunks;
431  for (const auto &Chunk : RIFF->Chunks)
432  Chunks.try_emplace(llvm::StringRef(Chunk.ID.data(), Chunk.ID.size()),
433  Chunk.Data);
434 
435  for (llvm::StringRef RequiredChunk : {"meta", "stri"})
436  if (!Chunks.count(RequiredChunk))
437  return makeError("missing required chunk " + RequiredChunk);
438 
439  Reader Meta(Chunks.lookup("meta"));
440  if (Meta.consume32() != Version)
441  return makeError("wrong version");
442 
443  auto Strings = readStringTable(Chunks.lookup("stri"));
444  if (!Strings)
445  return Strings.takeError();
446 
447  IndexFileIn Result;
448  if (Chunks.count("srcs")) {
449  Reader SrcsReader(Chunks.lookup("srcs"));
450  Result.Sources.emplace();
451  while (!SrcsReader.eof()) {
452  auto IGN = readIncludeGraphNode(SrcsReader, Strings->Strings);
453  auto Entry = Result.Sources->try_emplace(IGN.URI).first;
454  Entry->getValue() = std::move(IGN);
455  // We change all the strings inside the structure to point at the keys in
456  // the map, since it is the only copy of the string that's going to live.
457  Entry->getValue().URI = Entry->getKey();
458  for (auto &Include : Entry->getValue().DirectIncludes)
459  Include = Result.Sources->try_emplace(Include).first->getKey();
460  }
461  if (SrcsReader.err())
462  return makeError("malformed or truncated include uri");
463  }
464 
465  if (Chunks.count("symb")) {
466  Reader SymbolReader(Chunks.lookup("symb"));
468  while (!SymbolReader.eof())
469  Symbols.insert(readSymbol(SymbolReader, Strings->Strings));
470  if (SymbolReader.err())
471  return makeError("malformed or truncated symbol");
472  Result.Symbols = std::move(Symbols).build();
473  }
474  if (Chunks.count("refs")) {
475  Reader RefsReader(Chunks.lookup("refs"));
477  while (!RefsReader.eof()) {
478  auto RefsBundle = readRefs(RefsReader, Strings->Strings);
479  for (const auto &Ref : RefsBundle.second) // FIXME: bulk insert?
480  Refs.insert(RefsBundle.first, Ref);
481  }
482  if (RefsReader.err())
483  return makeError("malformed or truncated refs");
484  Result.Refs = std::move(Refs).build();
485  }
486  if (Chunks.count("rela")) {
487  Reader RelationsReader(Chunks.lookup("rela"));
488  RelationSlab::Builder Relations;
489  while (!RelationsReader.eof()) {
490  auto Relation = readRelation(RelationsReader);
491  Relations.insert(Relation);
492  }
493  if (RelationsReader.err())
494  return makeError("malformed or truncated relations");
495  Result.Relations = std::move(Relations).build();
496  }
497  if (Chunks.count("cmdl")) {
498  Reader CmdReader(Chunks.lookup("cmdl"));
499  if (CmdReader.err())
500  return makeError("malformed or truncated commandline section");
501  InternedCompileCommand Cmd =
502  readCompileCommand(CmdReader, Strings->Strings);
503  Result.Cmd.emplace();
504  Result.Cmd->Directory = Cmd.Directory;
505  Result.Cmd->CommandLine.reserve(Cmd.CommandLine.size());
506  for (llvm::StringRef C : Cmd.CommandLine)
507  Result.Cmd->CommandLine.emplace_back(C);
508  }
509  return std::move(Result);
510 }
511 
512 template <class Callback>
513 void visitStrings(IncludeGraphNode &IGN, const Callback &CB) {
514  CB(IGN.URI);
515  for (llvm::StringRef &Include : IGN.DirectIncludes)
516  CB(Include);
517 }
518 
519 void writeRIFF(const IndexFileOut &Data, llvm::raw_ostream &OS) {
520  assert(Data.Symbols && "An index file without symbols makes no sense!");
521  riff::File RIFF;
522  RIFF.Type = riff::fourCC("CdIx");
523 
524  llvm::SmallString<4> Meta;
525  {
526  llvm::raw_svector_ostream MetaOS(Meta);
527  write32(Version, MetaOS);
528  }
529  RIFF.Chunks.push_back({riff::fourCC("meta"), Meta});
530 
531  StringTableOut Strings;
532  std::vector<Symbol> Symbols;
533  for (const auto &Sym : *Data.Symbols) {
534  Symbols.emplace_back(Sym);
535  visitStrings(Symbols.back(),
536  [&](llvm::StringRef &S) { Strings.intern(S); });
537  }
538  std::vector<IncludeGraphNode> Sources;
539  if (Data.Sources)
540  for (const auto &Source : *Data.Sources) {
541  Sources.push_back(Source.getValue());
542  visitStrings(Sources.back(),
543  [&](llvm::StringRef &S) { Strings.intern(S); });
544  }
545 
546  std::vector<std::pair<SymbolID, std::vector<Ref>>> Refs;
547  if (Data.Refs) {
548  for (const auto &Sym : *Data.Refs) {
549  Refs.emplace_back(Sym);
550  for (auto &Ref : Refs.back().second) {
551  llvm::StringRef File = Ref.Location.FileURI;
552  Strings.intern(File);
553  Ref.Location.FileURI = File.data();
554  }
555  }
556  }
557 
558  std::vector<Relation> Relations;
559  if (Data.Relations) {
560  for (const auto &Relation : *Data.Relations) {
561  Relations.emplace_back(Relation);
562  // No strings to be interned in relations.
563  }
564  }
565 
566  InternedCompileCommand InternedCmd;
567  if (Data.Cmd) {
568  InternedCmd.CommandLine.reserve(Data.Cmd->CommandLine.size());
569  InternedCmd.Directory = Data.Cmd->Directory;
570  Strings.intern(InternedCmd.Directory);
571  for (llvm::StringRef C : Data.Cmd->CommandLine) {
572  InternedCmd.CommandLine.emplace_back(C);
573  Strings.intern(InternedCmd.CommandLine.back());
574  }
575  }
576 
577  std::string StringSection;
578  {
579  llvm::raw_string_ostream StringOS(StringSection);
580  Strings.finalize(StringOS);
581  }
582  RIFF.Chunks.push_back({riff::fourCC("stri"), StringSection});
583 
584  std::string SymbolSection;
585  {
586  llvm::raw_string_ostream SymbolOS(SymbolSection);
587  for (const auto &Sym : Symbols)
588  writeSymbol(Sym, Strings, SymbolOS);
589  }
590  RIFF.Chunks.push_back({riff::fourCC("symb"), SymbolSection});
591 
592  std::string RefsSection;
593  if (Data.Refs) {
594  {
595  llvm::raw_string_ostream RefsOS(RefsSection);
596  for (const auto &Sym : Refs)
597  writeRefs(Sym.first, Sym.second, Strings, RefsOS);
598  }
599  RIFF.Chunks.push_back({riff::fourCC("refs"), RefsSection});
600  }
601 
602  std::string RelationSection;
603  if (Data.Relations) {
604  {
605  llvm::raw_string_ostream RelationOS{RelationSection};
606  for (const auto &Relation : Relations)
607  writeRelation(Relation, RelationOS);
608  }
609  RIFF.Chunks.push_back({riff::fourCC("rela"), RelationSection});
610  }
611 
612  std::string SrcsSection;
613  {
614  {
615  llvm::raw_string_ostream SrcsOS(SrcsSection);
616  for (const auto &SF : Sources)
617  writeIncludeGraphNode(SF, Strings, SrcsOS);
618  }
619  RIFF.Chunks.push_back({riff::fourCC("srcs"), SrcsSection});
620  }
621 
622  std::string CmdlSection;
623  if (Data.Cmd) {
624  {
625  llvm::raw_string_ostream CmdOS(CmdlSection);
626  writeCompileCommand(InternedCmd, Strings, CmdOS);
627  }
628  RIFF.Chunks.push_back({riff::fourCC("cmdl"), CmdlSection});
629  }
630 
631  OS << RIFF;
632 }
633 
634 } // namespace
635 
636 // Defined in YAMLSerialization.cpp.
637 void writeYAML(const IndexFileOut &, llvm::raw_ostream &);
638 llvm::Expected<IndexFileIn> readYAML(llvm::StringRef);
639 
640 llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, const IndexFileOut &O) {
641  switch (O.Format) {
643  writeRIFF(O, OS);
644  break;
646  writeYAML(O, OS);
647  break;
648  }
649  return OS;
650 }
651 
652 llvm::Expected<IndexFileIn> readIndexFile(llvm::StringRef Data) {
653  if (Data.startswith("RIFF")) {
654  return readRIFF(Data);
655  } else if (auto YAMLContents = readYAML(Data)) {
656  return std::move(*YAMLContents);
657  } else {
658  return makeError("Not a RIFF file and failed to parse as YAML: " +
659  llvm::toString(YAMLContents.takeError()));
660  }
661 }
662 
663 std::unique_ptr<SymbolIndex> loadIndex(llvm::StringRef SymbolFilename,
664  bool UseDex) {
665  trace::Span OverallTracer("LoadIndex");
666  auto Buffer = llvm::MemoryBuffer::getFile(SymbolFilename);
667  if (!Buffer) {
668  elog("Can't open {0}", SymbolFilename);
669  return nullptr;
670  }
671 
673  RefSlab Refs;
674  RelationSlab Relations;
675  {
676  trace::Span Tracer("ParseIndex");
677  if (auto I = readIndexFile(Buffer->get()->getBuffer())) {
678  if (I->Symbols)
679  Symbols = std::move(*I->Symbols);
680  if (I->Refs)
681  Refs = std::move(*I->Refs);
682  if (I->Relations)
683  Relations = std::move(*I->Relations);
684  } else {
685  elog("Bad Index: {0}", I.takeError());
686  return nullptr;
687  }
688  }
689 
690  size_t NumSym = Symbols.size();
691  size_t NumRefs = Refs.numRefs();
692  size_t NumRelations = Relations.size();
693 
694  trace::Span Tracer("BuildIndex");
695  auto Index = UseDex ? dex::Dex::build(std::move(Symbols), std::move(Refs),
696  std::move(Relations))
697  : MemIndex::build(std::move(Symbols), std::move(Refs),
698  std::move(Relations));
699  vlog("Loaded {0} from {1} with estimated memory usage {2} bytes\n"
700  " - number of symbols: {3}\n"
701  " - number of refs: {4}\n"
702  " - numnber of relations: {5}",
703  UseDex ? "Dex" : "MemIndex", SymbolFilename,
704  Index->estimateMemoryUsage(), NumSym, NumRefs, NumRelations);
705  return Index;
706 }
707 
708 } // namespace clangd
709 } // namespace clang
SourceLocation Loc
&#39;#&#39; location in the include directive
llvm::Expected< IndexFileIn > readYAML(llvm::StringRef)
An immutable symbol container that stores a set of symbols.
Definition: Symbol.h:177
std::unique_ptr< SymbolIndex > loadIndex(llvm::StringRef SymbolFilename, bool UseDex)
An efficient structure of storing large set of symbol references in memory.
Definition: Ref.h:69
static llvm::Error makeError(const char *Msg)
Definition: RIFF.cpp:16
This defines Dex - a symbol index implementation based on query iterators over symbol tokens...
void writeYAML(const IndexFileOut &, llvm::raw_ostream &)
clang::find_all_symbols::SymbolInfo::SymbolKind SymbolKind
Definition: SymbolInfo.cpp:21
constexpr FourCC fourCC(const char(&Literal)[5])
Definition: RIFF.h:44
static llvm::StringRef toString(SpecialMemberFunctionsCheck::SpecialMemberFunctionKind K)
llvm::unique_function< void(llvm::Expected< T >)> Callback
A Callback<T> is a void function that accepts Expected<T>.
Definition: Function.h:28
llvm::Expected< File > readFile(llvm::StringRef Stream)
Definition: RIFF.cpp:51
size_t numRefs() const
Definition: Ref.h:84
llvm::Expected< IndexFileIn > readIndexFile(llvm::StringRef Data)
void vlog(const char *Fmt, Ts &&... Vals)
Definition: Logger.h:67
void elog(const char *Fmt, Ts &&... Vals)
Definition: Logger.h:56
std::vector< llvm::StringRef > CommandLine
static constexpr size_t RawSize
Definition: SymbolID.h:45
llvm::BumpPtrAllocator Arena
std::vector< std::pair< DocID, float > > consume(Iterator &It)
Advances the iterator until it is exhausted.
Definition: Iterator.cpp:350
static std::unique_ptr< SymbolIndex > build(SymbolSlab Symbols, RefSlab Refs, RelationSlab Relations)
Builds an index from slabs. The index takes ownership of the data.
Definition: MemIndex.cpp:19
static std::unique_ptr< SymbolIndex > build(SymbolSlab, RefSlab, RelationSlab)
Builds an index from slabs. The index takes ownership of the slab.
Definition: Dex.cpp:26
RefKind
Describes the kind of a cross-reference.
Definition: Ref.h:28
SymbolSlab Symbols
llvm::StringRef Directory
std::vector< llvm::StringRef > Strings
CodeCompletionBuilder Builder
===– Representation.cpp - ClangDoc Representation --------—*- C++ -*-===//
size_t size() const
Definition: Relation.h:57
RefSlab Refs
llvm::raw_ostream & operator<<(llvm::raw_ostream &OS, const CodeCompletion &C)
Records an event whose duration is the lifetime of the Span object.
Definition: Trace.h:81
size_t size() const
Definition: Symbol.h:189
std::array< uint8_t, 20 > SymbolID
void visitStrings(Symbol &S, const Callback &CB)
Invokes Callback with each StringRef& contained in the Symbol.
Definition: Symbol.h:147
const SymbolIndex * Index
Definition: Dexp.cpp:84