1 //===- ExtractAPI/Serialization/SymbolGraphSerializer.cpp -------*- C++ -*-===// 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 /// \file 10 /// This file implements the SymbolGraphSerializer. 11 /// 12 //===----------------------------------------------------------------------===// 13 14 #include "clang/ExtractAPI/Serialization/SymbolGraphSerializer.h" 15 #include "clang/Basic/SourceLocation.h" 16 #include "clang/Basic/Version.h" 17 #include "clang/ExtractAPI/API.h" 18 #include "clang/ExtractAPI/DeclarationFragments.h" 19 #include "llvm/ADT/STLExtras.h" 20 #include "llvm/ADT/STLFunctionalExtras.h" 21 #include "llvm/ADT/SmallVector.h" 22 #include "llvm/Support/Casting.h" 23 #include "llvm/Support/Compiler.h" 24 #include "llvm/Support/Path.h" 25 #include "llvm/Support/VersionTuple.h" 26 #include "llvm/Support/raw_ostream.h" 27 #include <iterator> 28 #include <optional> 29 #include <type_traits> 30 31 using namespace clang; 32 using namespace clang::extractapi; 33 using namespace llvm; 34 35 namespace { 36 37 /// Helper function to inject a JSON object \p Obj into another object \p Paren 38 /// at position \p Key. 39 void serializeObject(Object &Paren, StringRef Key, 40 std::optional<Object> &&Obj) { 41 if (Obj) 42 Paren[Key] = std::move(*Obj); 43 } 44 45 /// Helper function to inject a JSON array \p Array into object \p Paren at 46 /// position \p Key. 47 void serializeArray(Object &Paren, StringRef Key, 48 std::optional<Array> &&Array) { 49 if (Array) 50 Paren[Key] = std::move(*Array); 51 } 52 53 /// Helper function to inject a JSON array composed of the values in \p C into 54 /// object \p Paren at position \p Key. 55 template <typename ContainerTy> 56 void serializeArray(Object &Paren, StringRef Key, ContainerTy &&C) { 57 Paren[Key] = Array(C); 58 } 59 60 /// Serialize a \c VersionTuple \p V with the Symbol Graph semantic version 61 /// format. 62 /// 63 /// A semantic version object contains three numeric fields, representing the 64 /// \c major, \c minor, and \c patch parts of the version tuple. 65 /// For example version tuple 1.0.3 is serialized as: 66 /// \code 67 /// { 68 /// "major" : 1, 69 /// "minor" : 0, 70 /// "patch" : 3 71 /// } 72 /// \endcode 73 /// 74 /// \returns \c std::nullopt if the version \p V is empty, or an \c Object 75 /// containing the semantic version representation of \p V. 76 std::optional<Object> serializeSemanticVersion(const VersionTuple &V) { 77 if (V.empty()) 78 return std::nullopt; 79 80 Object Version; 81 Version["major"] = V.getMajor(); 82 Version["minor"] = V.getMinor().value_or(0); 83 Version["patch"] = V.getSubminor().value_or(0); 84 return Version; 85 } 86 87 /// Serialize the OS information in the Symbol Graph platform property. 88 /// 89 /// The OS information in Symbol Graph contains the \c name of the OS, and an 90 /// optional \c minimumVersion semantic version field. 91 Object serializeOperatingSystem(const Triple &T) { 92 Object OS; 93 OS["name"] = T.getOSTypeName(T.getOS()); 94 serializeObject(OS, "minimumVersion", 95 serializeSemanticVersion(T.getMinimumSupportedOSVersion())); 96 return OS; 97 } 98 99 /// Serialize the platform information in the Symbol Graph module section. 100 /// 101 /// The platform object describes a target platform triple in corresponding 102 /// three fields: \c architecture, \c vendor, and \c operatingSystem. 103 Object serializePlatform(const Triple &T) { 104 Object Platform; 105 Platform["architecture"] = T.getArchName(); 106 Platform["vendor"] = T.getVendorName(); 107 108 if (!T.getEnvironmentName().empty()) 109 Platform["environment"] = T.getEnvironmentName(); 110 111 Platform["operatingSystem"] = serializeOperatingSystem(T); 112 return Platform; 113 } 114 115 /// Serialize a source position. 116 Object serializeSourcePosition(const PresumedLoc &Loc) { 117 assert(Loc.isValid() && "invalid source position"); 118 119 Object SourcePosition; 120 SourcePosition["line"] = Loc.getLine() - 1; 121 SourcePosition["character"] = Loc.getColumn() - 1; 122 123 return SourcePosition; 124 } 125 126 /// Serialize a source location in file. 127 /// 128 /// \param Loc The presumed location to serialize. 129 /// \param IncludeFileURI If true, include the file path of \p Loc as a URI. 130 /// Defaults to false. 131 Object serializeSourceLocation(const PresumedLoc &Loc, 132 bool IncludeFileURI = false) { 133 Object SourceLocation; 134 serializeObject(SourceLocation, "position", serializeSourcePosition(Loc)); 135 136 if (IncludeFileURI) { 137 std::string FileURI = "file://"; 138 // Normalize file path to use forward slashes for the URI. 139 FileURI += sys::path::convert_to_slash(Loc.getFilename()); 140 SourceLocation["uri"] = FileURI; 141 } 142 143 return SourceLocation; 144 } 145 146 /// Serialize a source range with begin and end locations. 147 Object serializeSourceRange(const PresumedLoc &BeginLoc, 148 const PresumedLoc &EndLoc) { 149 Object SourceRange; 150 serializeObject(SourceRange, "start", serializeSourcePosition(BeginLoc)); 151 serializeObject(SourceRange, "end", serializeSourcePosition(EndLoc)); 152 return SourceRange; 153 } 154 155 /// Serialize the availability attributes of a symbol. 156 /// 157 /// Availability information contains the introduced, deprecated, and obsoleted 158 /// versions of the symbol as semantic versions, if not default. 159 /// Availability information also contains flags to indicate if the symbol is 160 /// unconditionally unavailable or deprecated, 161 /// i.e. \c __attribute__((unavailable)) and \c __attribute__((deprecated)). 162 /// 163 /// \returns \c std::nullopt if the symbol has default availability attributes, 164 /// or an \c Array containing an object with the formatted availability 165 /// information. 166 std::optional<Array> serializeAvailability(const AvailabilityInfo &Avail) { 167 if (Avail.isDefault()) 168 return std::nullopt; 169 170 Array AvailabilityArray; 171 172 if (Avail.isUnconditionallyDeprecated()) { 173 Object UnconditionallyDeprecated; 174 UnconditionallyDeprecated["domain"] = "*"; 175 UnconditionallyDeprecated["isUnconditionallyDeprecated"] = true; 176 AvailabilityArray.emplace_back(std::move(UnconditionallyDeprecated)); 177 } 178 179 if (Avail.Domain.str() != "") { 180 Object Availability; 181 Availability["domain"] = Avail.Domain; 182 183 if (Avail.isUnavailable()) { 184 Availability["isUnconditionallyUnavailable"] = true; 185 } else { 186 serializeObject(Availability, "introduced", 187 serializeSemanticVersion(Avail.Introduced)); 188 serializeObject(Availability, "deprecated", 189 serializeSemanticVersion(Avail.Deprecated)); 190 serializeObject(Availability, "obsoleted", 191 serializeSemanticVersion(Avail.Obsoleted)); 192 } 193 194 AvailabilityArray.emplace_back(std::move(Availability)); 195 } 196 197 return AvailabilityArray; 198 } 199 200 /// Get the language name string for interface language references. 201 StringRef getLanguageName(Language Lang) { 202 switch (Lang) { 203 case Language::C: 204 return "c"; 205 case Language::ObjC: 206 return "objective-c"; 207 case Language::CXX: 208 return "c++"; 209 case Language::ObjCXX: 210 return "objective-c++"; 211 212 // Unsupported language currently 213 case Language::OpenCL: 214 case Language::OpenCLCXX: 215 case Language::CUDA: 216 case Language::HIP: 217 case Language::HLSL: 218 219 // Languages that the frontend cannot parse and compile 220 case Language::Unknown: 221 case Language::Asm: 222 case Language::LLVM_IR: 223 case Language::CIR: 224 llvm_unreachable("Unsupported language kind"); 225 } 226 227 llvm_unreachable("Unhandled language kind"); 228 } 229 230 /// Serialize the identifier object as specified by the Symbol Graph format. 231 /// 232 /// The identifier property of a symbol contains the USR for precise and unique 233 /// references, and the interface language name. 234 Object serializeIdentifier(const APIRecord &Record, Language Lang) { 235 Object Identifier; 236 Identifier["precise"] = Record.USR; 237 Identifier["interfaceLanguage"] = getLanguageName(Lang); 238 239 return Identifier; 240 } 241 242 /// Serialize the documentation comments attached to a symbol, as specified by 243 /// the Symbol Graph format. 244 /// 245 /// The Symbol Graph \c docComment object contains an array of lines. Each line 246 /// represents one line of striped documentation comment, with source range 247 /// information. 248 /// e.g. 249 /// \code 250 /// /// This is a documentation comment 251 /// ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~' First line. 252 /// /// with multiple lines. 253 /// ^~~~~~~~~~~~~~~~~~~~~~~' Second line. 254 /// \endcode 255 /// 256 /// \returns \c std::nullopt if \p Comment is empty, or an \c Object containing 257 /// the formatted lines. 258 std::optional<Object> serializeDocComment(const DocComment &Comment) { 259 if (Comment.empty()) 260 return std::nullopt; 261 262 Object DocComment; 263 264 Array LinesArray; 265 for (const auto &CommentLine : Comment) { 266 Object Line; 267 Line["text"] = CommentLine.Text; 268 serializeObject(Line, "range", 269 serializeSourceRange(CommentLine.Begin, CommentLine.End)); 270 LinesArray.emplace_back(std::move(Line)); 271 } 272 273 serializeArray(DocComment, "lines", std::move(LinesArray)); 274 275 return DocComment; 276 } 277 278 /// Serialize the declaration fragments of a symbol. 279 /// 280 /// The Symbol Graph declaration fragments is an array of tagged important 281 /// parts of a symbol's declaration. The fragments sequence can be joined to 282 /// form spans of declaration text, with attached information useful for 283 /// purposes like syntax-highlighting etc. For example: 284 /// \code 285 /// const int pi; -> "declarationFragments" : [ 286 /// { 287 /// "kind" : "keyword", 288 /// "spelling" : "const" 289 /// }, 290 /// { 291 /// "kind" : "text", 292 /// "spelling" : " " 293 /// }, 294 /// { 295 /// "kind" : "typeIdentifier", 296 /// "preciseIdentifier" : "c:I", 297 /// "spelling" : "int" 298 /// }, 299 /// { 300 /// "kind" : "text", 301 /// "spelling" : " " 302 /// }, 303 /// { 304 /// "kind" : "identifier", 305 /// "spelling" : "pi" 306 /// } 307 /// ] 308 /// \endcode 309 /// 310 /// \returns \c std::nullopt if \p DF is empty, or an \c Array containing the 311 /// formatted declaration fragments array. 312 std::optional<Array> 313 serializeDeclarationFragments(const DeclarationFragments &DF) { 314 if (DF.getFragments().empty()) 315 return std::nullopt; 316 317 Array Fragments; 318 for (const auto &F : DF.getFragments()) { 319 Object Fragment; 320 Fragment["spelling"] = F.Spelling; 321 Fragment["kind"] = DeclarationFragments::getFragmentKindString(F.Kind); 322 if (!F.PreciseIdentifier.empty()) 323 Fragment["preciseIdentifier"] = F.PreciseIdentifier; 324 Fragments.emplace_back(std::move(Fragment)); 325 } 326 327 return Fragments; 328 } 329 330 /// Serialize the \c names field of a symbol as specified by the Symbol Graph 331 /// format. 332 /// 333 /// The Symbol Graph names field contains multiple representations of a symbol 334 /// that can be used for different applications: 335 /// - \c title : The simple declared name of the symbol; 336 /// - \c subHeading : An array of declaration fragments that provides tags, 337 /// and potentially more tokens (for example the \c +/- symbol for 338 /// Objective-C methods). Can be used as sub-headings for documentation. 339 Object serializeNames(const APIRecord *Record) { 340 Object Names; 341 Names["title"] = Record->Name; 342 343 serializeArray(Names, "subHeading", 344 serializeDeclarationFragments(Record->SubHeading)); 345 DeclarationFragments NavigatorFragments; 346 NavigatorFragments.append(Record->Name, 347 DeclarationFragments::FragmentKind::Identifier, 348 /*PreciseIdentifier*/ ""); 349 serializeArray(Names, "navigator", 350 serializeDeclarationFragments(NavigatorFragments)); 351 352 return Names; 353 } 354 355 Object serializeSymbolKind(APIRecord::RecordKind RK, Language Lang) { 356 auto AddLangPrefix = [&Lang](StringRef S) -> std::string { 357 return (getLanguageName(Lang) + "." + S).str(); 358 }; 359 360 Object Kind; 361 switch (RK) { 362 case APIRecord::RK_Unknown: 363 Kind["identifier"] = AddLangPrefix("unknown"); 364 Kind["displayName"] = "Unknown"; 365 break; 366 case APIRecord::RK_Namespace: 367 Kind["identifier"] = AddLangPrefix("namespace"); 368 Kind["displayName"] = "Namespace"; 369 break; 370 case APIRecord::RK_GlobalFunction: 371 Kind["identifier"] = AddLangPrefix("func"); 372 Kind["displayName"] = "Function"; 373 break; 374 case APIRecord::RK_GlobalFunctionTemplate: 375 Kind["identifier"] = AddLangPrefix("func"); 376 Kind["displayName"] = "Function Template"; 377 break; 378 case APIRecord::RK_GlobalFunctionTemplateSpecialization: 379 Kind["identifier"] = AddLangPrefix("func"); 380 Kind["displayName"] = "Function Template Specialization"; 381 break; 382 case APIRecord::RK_GlobalVariableTemplate: 383 Kind["identifier"] = AddLangPrefix("var"); 384 Kind["displayName"] = "Global Variable Template"; 385 break; 386 case APIRecord::RK_GlobalVariableTemplateSpecialization: 387 Kind["identifier"] = AddLangPrefix("var"); 388 Kind["displayName"] = "Global Variable Template Specialization"; 389 break; 390 case APIRecord::RK_GlobalVariableTemplatePartialSpecialization: 391 Kind["identifier"] = AddLangPrefix("var"); 392 Kind["displayName"] = "Global Variable Template Partial Specialization"; 393 break; 394 case APIRecord::RK_GlobalVariable: 395 Kind["identifier"] = AddLangPrefix("var"); 396 Kind["displayName"] = "Global Variable"; 397 break; 398 case APIRecord::RK_EnumConstant: 399 Kind["identifier"] = AddLangPrefix("enum.case"); 400 Kind["displayName"] = "Enumeration Case"; 401 break; 402 case APIRecord::RK_Enum: 403 Kind["identifier"] = AddLangPrefix("enum"); 404 Kind["displayName"] = "Enumeration"; 405 break; 406 case APIRecord::RK_StructField: 407 Kind["identifier"] = AddLangPrefix("property"); 408 Kind["displayName"] = "Instance Property"; 409 break; 410 case APIRecord::RK_Struct: 411 Kind["identifier"] = AddLangPrefix("struct"); 412 Kind["displayName"] = "Structure"; 413 break; 414 case APIRecord::RK_UnionField: 415 Kind["identifier"] = AddLangPrefix("property"); 416 Kind["displayName"] = "Instance Property"; 417 break; 418 case APIRecord::RK_Union: 419 Kind["identifier"] = AddLangPrefix("union"); 420 Kind["displayName"] = "Union"; 421 break; 422 case APIRecord::RK_CXXField: 423 Kind["identifier"] = AddLangPrefix("property"); 424 Kind["displayName"] = "Instance Property"; 425 break; 426 case APIRecord::RK_StaticField: 427 Kind["identifier"] = AddLangPrefix("type.property"); 428 Kind["displayName"] = "Type Property"; 429 break; 430 case APIRecord::RK_ClassTemplate: 431 case APIRecord::RK_ClassTemplateSpecialization: 432 case APIRecord::RK_ClassTemplatePartialSpecialization: 433 case APIRecord::RK_CXXClass: 434 Kind["identifier"] = AddLangPrefix("class"); 435 Kind["displayName"] = "Class"; 436 break; 437 case APIRecord::RK_CXXMethodTemplate: 438 Kind["identifier"] = AddLangPrefix("method"); 439 Kind["displayName"] = "Method Template"; 440 break; 441 case APIRecord::RK_CXXMethodTemplateSpecialization: 442 Kind["identifier"] = AddLangPrefix("method"); 443 Kind["displayName"] = "Method Template Specialization"; 444 break; 445 case APIRecord::RK_CXXFieldTemplate: 446 Kind["identifier"] = AddLangPrefix("property"); 447 Kind["displayName"] = "Template Property"; 448 break; 449 case APIRecord::RK_Concept: 450 Kind["identifier"] = AddLangPrefix("concept"); 451 Kind["displayName"] = "Concept"; 452 break; 453 case APIRecord::RK_CXXStaticMethod: 454 Kind["identifier"] = AddLangPrefix("type.method"); 455 Kind["displayName"] = "Static Method"; 456 break; 457 case APIRecord::RK_CXXInstanceMethod: 458 Kind["identifier"] = AddLangPrefix("method"); 459 Kind["displayName"] = "Instance Method"; 460 break; 461 case APIRecord::RK_CXXConstructorMethod: 462 Kind["identifier"] = AddLangPrefix("method"); 463 Kind["displayName"] = "Constructor"; 464 break; 465 case APIRecord::RK_CXXDestructorMethod: 466 Kind["identifier"] = AddLangPrefix("method"); 467 Kind["displayName"] = "Destructor"; 468 break; 469 case APIRecord::RK_ObjCIvar: 470 Kind["identifier"] = AddLangPrefix("ivar"); 471 Kind["displayName"] = "Instance Variable"; 472 break; 473 case APIRecord::RK_ObjCInstanceMethod: 474 Kind["identifier"] = AddLangPrefix("method"); 475 Kind["displayName"] = "Instance Method"; 476 break; 477 case APIRecord::RK_ObjCClassMethod: 478 Kind["identifier"] = AddLangPrefix("type.method"); 479 Kind["displayName"] = "Type Method"; 480 break; 481 case APIRecord::RK_ObjCInstanceProperty: 482 Kind["identifier"] = AddLangPrefix("property"); 483 Kind["displayName"] = "Instance Property"; 484 break; 485 case APIRecord::RK_ObjCClassProperty: 486 Kind["identifier"] = AddLangPrefix("type.property"); 487 Kind["displayName"] = "Type Property"; 488 break; 489 case APIRecord::RK_ObjCInterface: 490 Kind["identifier"] = AddLangPrefix("class"); 491 Kind["displayName"] = "Class"; 492 break; 493 case APIRecord::RK_ObjCCategory: 494 Kind["identifier"] = AddLangPrefix("class.extension"); 495 Kind["displayName"] = "Class Extension"; 496 break; 497 case APIRecord::RK_ObjCProtocol: 498 Kind["identifier"] = AddLangPrefix("protocol"); 499 Kind["displayName"] = "Protocol"; 500 break; 501 case APIRecord::RK_MacroDefinition: 502 Kind["identifier"] = AddLangPrefix("macro"); 503 Kind["displayName"] = "Macro"; 504 break; 505 case APIRecord::RK_Typedef: 506 Kind["identifier"] = AddLangPrefix("typealias"); 507 Kind["displayName"] = "Type Alias"; 508 break; 509 default: 510 llvm_unreachable("API Record with uninstantiable kind"); 511 } 512 513 return Kind; 514 } 515 516 /// Serialize the symbol kind information. 517 /// 518 /// The Symbol Graph symbol kind property contains a shorthand \c identifier 519 /// which is prefixed by the source language name, useful for tooling to parse 520 /// the kind, and a \c displayName for rendering human-readable names. 521 Object serializeSymbolKind(const APIRecord &Record, Language Lang) { 522 return serializeSymbolKind(Record.KindForDisplay, Lang); 523 } 524 525 /// Serialize the function signature field, as specified by the 526 /// Symbol Graph format. 527 /// 528 /// The Symbol Graph function signature property contains two arrays. 529 /// - The \c returns array is the declaration fragments of the return type; 530 /// - The \c parameters array contains names and declaration fragments of the 531 /// parameters. 532 template <typename RecordTy> 533 void serializeFunctionSignatureMixin(Object &Paren, const RecordTy &Record) { 534 const auto &FS = Record.Signature; 535 if (FS.empty()) 536 return; 537 538 Object Signature; 539 serializeArray(Signature, "returns", 540 serializeDeclarationFragments(FS.getReturnType())); 541 542 Array Parameters; 543 for (const auto &P : FS.getParameters()) { 544 Object Parameter; 545 Parameter["name"] = P.Name; 546 serializeArray(Parameter, "declarationFragments", 547 serializeDeclarationFragments(P.Fragments)); 548 Parameters.emplace_back(std::move(Parameter)); 549 } 550 551 if (!Parameters.empty()) 552 Signature["parameters"] = std::move(Parameters); 553 554 serializeObject(Paren, "functionSignature", std::move(Signature)); 555 } 556 557 template <typename RecordTy> 558 void serializeTemplateMixin(Object &Paren, const RecordTy &Record) { 559 const auto &Template = Record.Templ; 560 if (Template.empty()) 561 return; 562 563 Object Generics; 564 Array GenericParameters; 565 for (const auto &Param : Template.getParameters()) { 566 Object Parameter; 567 Parameter["name"] = Param.Name; 568 Parameter["index"] = Param.Index; 569 Parameter["depth"] = Param.Depth; 570 GenericParameters.emplace_back(std::move(Parameter)); 571 } 572 if (!GenericParameters.empty()) 573 Generics["parameters"] = std::move(GenericParameters); 574 575 Array GenericConstraints; 576 for (const auto &Constr : Template.getConstraints()) { 577 Object Constraint; 578 Constraint["kind"] = Constr.Kind; 579 Constraint["lhs"] = Constr.LHS; 580 Constraint["rhs"] = Constr.RHS; 581 GenericConstraints.emplace_back(std::move(Constraint)); 582 } 583 584 if (!GenericConstraints.empty()) 585 Generics["constraints"] = std::move(GenericConstraints); 586 587 serializeObject(Paren, "swiftGenerics", Generics); 588 } 589 590 Array generateParentContexts(const SmallVectorImpl<SymbolReference> &Parents, 591 Language Lang) { 592 Array ParentContexts; 593 594 for (const auto &Parent : Parents) { 595 Object Elem; 596 Elem["usr"] = Parent.USR; 597 Elem["name"] = Parent.Name; 598 if (Parent.Record) 599 Elem["kind"] = serializeSymbolKind(Parent.Record->KindForDisplay, 600 Lang)["identifier"]; 601 else 602 Elem["kind"] = 603 serializeSymbolKind(APIRecord::RK_Unknown, Lang)["identifier"]; 604 ParentContexts.emplace_back(std::move(Elem)); 605 } 606 607 return ParentContexts; 608 } 609 610 /// Walk the records parent information in reverse to generate a hierarchy 611 /// suitable for serialization. 612 SmallVector<SymbolReference, 8> 613 generateHierarchyFromRecord(const APIRecord *Record) { 614 SmallVector<SymbolReference, 8> ReverseHierarchy; 615 for (const auto *Current = Record; Current != nullptr; 616 Current = Current->Parent.Record) 617 ReverseHierarchy.emplace_back(Current); 618 619 return SmallVector<SymbolReference, 8>( 620 std::make_move_iterator(ReverseHierarchy.rbegin()), 621 std::make_move_iterator(ReverseHierarchy.rend())); 622 } 623 624 SymbolReference getHierarchyReference(const APIRecord *Record, 625 const APISet &API) { 626 // If the parent is a category extended from internal module then we need to 627 // pretend this belongs to the associated interface. 628 if (auto *CategoryRecord = dyn_cast_or_null<ObjCCategoryRecord>(Record)) { 629 return CategoryRecord->Interface; 630 // FIXME: TODO generate path components correctly for categories extending 631 // an external module. 632 } 633 634 return SymbolReference(Record); 635 } 636 637 } // namespace 638 639 Object *ExtendedModule::addSymbol(Object &&Symbol) { 640 Symbols.emplace_back(std::move(Symbol)); 641 return Symbols.back().getAsObject(); 642 } 643 644 void ExtendedModule::addRelationship(Object &&Relationship) { 645 Relationships.emplace_back(std::move(Relationship)); 646 } 647 648 /// Defines the format version emitted by SymbolGraphSerializer. 649 const VersionTuple SymbolGraphSerializer::FormatVersion{0, 5, 3}; 650 651 Object SymbolGraphSerializer::serializeMetadata() const { 652 Object Metadata; 653 serializeObject(Metadata, "formatVersion", 654 serializeSemanticVersion(FormatVersion)); 655 Metadata["generator"] = clang::getClangFullVersion(); 656 return Metadata; 657 } 658 659 Object 660 SymbolGraphSerializer::serializeModuleObject(StringRef ModuleName) const { 661 Object Module; 662 Module["name"] = ModuleName; 663 serializeObject(Module, "platform", serializePlatform(API.getTarget())); 664 return Module; 665 } 666 667 bool SymbolGraphSerializer::shouldSkip(const APIRecord *Record) const { 668 if (!Record) 669 return true; 670 671 // Skip unconditionally unavailable symbols 672 if (Record->Availability.isUnconditionallyUnavailable()) 673 return true; 674 675 // Filter out symbols prefixed with an underscored as they are understood to 676 // be symbols clients should not use. 677 if (Record->Name.starts_with("_")) 678 return true; 679 680 // Skip explicitly ignored symbols. 681 if (IgnoresList.shouldIgnore(Record->Name)) 682 return true; 683 684 return false; 685 } 686 687 ExtendedModule &SymbolGraphSerializer::getModuleForCurrentSymbol() { 688 if (!ForceEmitToMainModule && ModuleForCurrentSymbol) 689 return *ModuleForCurrentSymbol; 690 691 return MainModule; 692 } 693 694 Array SymbolGraphSerializer::serializePathComponents( 695 const APIRecord *Record) const { 696 return Array(map_range(Hierarchy, [](auto Elt) { return Elt.Name; })); 697 } 698 699 StringRef SymbolGraphSerializer::getRelationshipString(RelationshipKind Kind) { 700 switch (Kind) { 701 case RelationshipKind::MemberOf: 702 return "memberOf"; 703 case RelationshipKind::InheritsFrom: 704 return "inheritsFrom"; 705 case RelationshipKind::ConformsTo: 706 return "conformsTo"; 707 case RelationshipKind::ExtensionTo: 708 return "extensionTo"; 709 } 710 llvm_unreachable("Unhandled relationship kind"); 711 } 712 713 void SymbolGraphSerializer::serializeRelationship(RelationshipKind Kind, 714 const SymbolReference &Source, 715 const SymbolReference &Target, 716 ExtendedModule &Into) { 717 Object Relationship; 718 SmallString<64> TestRelLabel; 719 if (EmitSymbolLabelsForTesting) { 720 llvm::raw_svector_ostream OS(TestRelLabel); 721 OS << SymbolGraphSerializer::getRelationshipString(Kind) << " $ " 722 << Source.USR << " $ "; 723 if (Target.USR.empty()) 724 OS << Target.Name; 725 else 726 OS << Target.USR; 727 Relationship["!testRelLabel"] = TestRelLabel; 728 } 729 Relationship["source"] = Source.USR; 730 Relationship["target"] = Target.USR; 731 Relationship["targetFallback"] = Target.Name; 732 Relationship["kind"] = SymbolGraphSerializer::getRelationshipString(Kind); 733 734 if (ForceEmitToMainModule) 735 MainModule.addRelationship(std::move(Relationship)); 736 else 737 Into.addRelationship(std::move(Relationship)); 738 } 739 740 StringRef SymbolGraphSerializer::getConstraintString(ConstraintKind Kind) { 741 switch (Kind) { 742 case ConstraintKind::Conformance: 743 return "conformance"; 744 case ConstraintKind::ConditionalConformance: 745 return "conditionalConformance"; 746 } 747 llvm_unreachable("Unhandled constraint kind"); 748 } 749 750 void SymbolGraphSerializer::serializeAPIRecord(const APIRecord *Record) { 751 Object Obj; 752 753 // If we need symbol labels for testing emit the USR as the value and the key 754 // starts with '!'' to ensure it ends up at the top of the object. 755 if (EmitSymbolLabelsForTesting) 756 Obj["!testLabel"] = Record->USR; 757 758 serializeObject(Obj, "identifier", 759 serializeIdentifier(*Record, API.getLanguage())); 760 serializeObject(Obj, "kind", serializeSymbolKind(*Record, API.getLanguage())); 761 serializeObject(Obj, "names", serializeNames(Record)); 762 serializeObject( 763 Obj, "location", 764 serializeSourceLocation(Record->Location, /*IncludeFileURI=*/true)); 765 serializeArray(Obj, "availability", 766 serializeAvailability(Record->Availability)); 767 serializeObject(Obj, "docComment", serializeDocComment(Record->Comment)); 768 serializeArray(Obj, "declarationFragments", 769 serializeDeclarationFragments(Record->Declaration)); 770 771 Obj["pathComponents"] = serializePathComponents(Record); 772 Obj["accessLevel"] = Record->Access.getAccess(); 773 774 ExtendedModule &Module = getModuleForCurrentSymbol(); 775 // If the hierarchy has at least one parent and child. 776 if (Hierarchy.size() >= 2) 777 serializeRelationship(MemberOf, Hierarchy.back(), 778 Hierarchy[Hierarchy.size() - 2], Module); 779 780 CurrentSymbol = Module.addSymbol(std::move(Obj)); 781 } 782 783 bool SymbolGraphSerializer::traverseAPIRecord(const APIRecord *Record) { 784 if (!Record) 785 return true; 786 if (shouldSkip(Record)) 787 return true; 788 Hierarchy.push_back(getHierarchyReference(Record, API)); 789 // Defer traversal mechanics to APISetVisitor base implementation 790 auto RetVal = Base::traverseAPIRecord(Record); 791 Hierarchy.pop_back(); 792 return RetVal; 793 } 794 795 bool SymbolGraphSerializer::visitAPIRecord(const APIRecord *Record) { 796 serializeAPIRecord(Record); 797 return true; 798 } 799 800 bool SymbolGraphSerializer::visitGlobalFunctionRecord( 801 const GlobalFunctionRecord *Record) { 802 if (!CurrentSymbol) 803 return true; 804 805 serializeFunctionSignatureMixin(*CurrentSymbol, *Record); 806 return true; 807 } 808 809 bool SymbolGraphSerializer::visitCXXClassRecord(const CXXClassRecord *Record) { 810 if (!CurrentSymbol) 811 return true; 812 813 for (const auto &Base : Record->Bases) 814 serializeRelationship(RelationshipKind::InheritsFrom, Record, Base, 815 getModuleForCurrentSymbol()); 816 return true; 817 } 818 819 bool SymbolGraphSerializer::visitClassTemplateRecord( 820 const ClassTemplateRecord *Record) { 821 if (!CurrentSymbol) 822 return true; 823 824 serializeTemplateMixin(*CurrentSymbol, *Record); 825 return true; 826 } 827 828 bool SymbolGraphSerializer::visitClassTemplatePartialSpecializationRecord( 829 const ClassTemplatePartialSpecializationRecord *Record) { 830 if (!CurrentSymbol) 831 return true; 832 833 serializeTemplateMixin(*CurrentSymbol, *Record); 834 return true; 835 } 836 837 bool SymbolGraphSerializer::visitCXXMethodRecord( 838 const CXXMethodRecord *Record) { 839 if (!CurrentSymbol) 840 return true; 841 842 serializeFunctionSignatureMixin(*CurrentSymbol, *Record); 843 return true; 844 } 845 846 bool SymbolGraphSerializer::visitCXXMethodTemplateRecord( 847 const CXXMethodTemplateRecord *Record) { 848 if (!CurrentSymbol) 849 return true; 850 851 serializeTemplateMixin(*CurrentSymbol, *Record); 852 return true; 853 } 854 855 bool SymbolGraphSerializer::visitCXXFieldTemplateRecord( 856 const CXXFieldTemplateRecord *Record) { 857 if (!CurrentSymbol) 858 return true; 859 860 serializeTemplateMixin(*CurrentSymbol, *Record); 861 return true; 862 } 863 864 bool SymbolGraphSerializer::visitConceptRecord(const ConceptRecord *Record) { 865 if (!CurrentSymbol) 866 return true; 867 868 serializeTemplateMixin(*CurrentSymbol, *Record); 869 return true; 870 } 871 872 bool SymbolGraphSerializer::visitGlobalVariableTemplateRecord( 873 const GlobalVariableTemplateRecord *Record) { 874 if (!CurrentSymbol) 875 return true; 876 877 serializeTemplateMixin(*CurrentSymbol, *Record); 878 return true; 879 } 880 881 bool SymbolGraphSerializer:: 882 visitGlobalVariableTemplatePartialSpecializationRecord( 883 const GlobalVariableTemplatePartialSpecializationRecord *Record) { 884 if (!CurrentSymbol) 885 return true; 886 887 serializeTemplateMixin(*CurrentSymbol, *Record); 888 return true; 889 } 890 891 bool SymbolGraphSerializer::visitGlobalFunctionTemplateRecord( 892 const GlobalFunctionTemplateRecord *Record) { 893 if (!CurrentSymbol) 894 return true; 895 896 serializeTemplateMixin(*CurrentSymbol, *Record); 897 return true; 898 } 899 900 bool SymbolGraphSerializer::visitObjCContainerRecord( 901 const ObjCContainerRecord *Record) { 902 if (!CurrentSymbol) 903 return true; 904 905 for (const auto &Protocol : Record->Protocols) 906 serializeRelationship(ConformsTo, Record, Protocol, 907 getModuleForCurrentSymbol()); 908 909 return true; 910 } 911 912 bool SymbolGraphSerializer::visitObjCInterfaceRecord( 913 const ObjCInterfaceRecord *Record) { 914 if (!CurrentSymbol) 915 return true; 916 917 if (!Record->SuperClass.empty()) 918 serializeRelationship(InheritsFrom, Record, Record->SuperClass, 919 getModuleForCurrentSymbol()); 920 return true; 921 } 922 923 bool SymbolGraphSerializer::traverseObjCCategoryRecord( 924 const ObjCCategoryRecord *Record) { 925 if (SkipSymbolsInCategoriesToExternalTypes && 926 !API.findRecordForUSR(Record->Interface.USR)) 927 return true; 928 929 auto *CurrentModule = ModuleForCurrentSymbol; 930 if (auto ModuleExtendedByRecord = Record->getExtendedExternalModule()) 931 ModuleForCurrentSymbol = &ExtendedModules[*ModuleExtendedByRecord]; 932 933 if (!walkUpFromObjCCategoryRecord(Record)) 934 return false; 935 936 bool RetVal = traverseRecordContext(Record); 937 ModuleForCurrentSymbol = CurrentModule; 938 return RetVal; 939 } 940 941 bool SymbolGraphSerializer::walkUpFromObjCCategoryRecord( 942 const ObjCCategoryRecord *Record) { 943 return visitObjCCategoryRecord(Record); 944 } 945 946 bool SymbolGraphSerializer::visitObjCCategoryRecord( 947 const ObjCCategoryRecord *Record) { 948 // If we need to create a record for the category in the future do so here, 949 // otherwise everything is set up to pretend that the category is in fact the 950 // interface it extends. 951 for (const auto &Protocol : Record->Protocols) 952 serializeRelationship(ConformsTo, Record->Interface, Protocol, 953 getModuleForCurrentSymbol()); 954 955 return true; 956 } 957 958 bool SymbolGraphSerializer::visitObjCMethodRecord( 959 const ObjCMethodRecord *Record) { 960 if (!CurrentSymbol) 961 return true; 962 963 serializeFunctionSignatureMixin(*CurrentSymbol, *Record); 964 return true; 965 } 966 967 bool SymbolGraphSerializer::visitObjCInstanceVariableRecord( 968 const ObjCInstanceVariableRecord *Record) { 969 // FIXME: serialize ivar access control here. 970 return true; 971 } 972 973 bool SymbolGraphSerializer::walkUpFromTypedefRecord( 974 const TypedefRecord *Record) { 975 // Short-circuit walking up the class hierarchy and handle creating typedef 976 // symbol objects manually as there are additional symbol dropping rules to 977 // respect. 978 return visitTypedefRecord(Record); 979 } 980 981 bool SymbolGraphSerializer::visitTypedefRecord(const TypedefRecord *Record) { 982 // Typedefs of anonymous types have their entries unified with the underlying 983 // type. 984 bool ShouldDrop = Record->UnderlyingType.Name.empty(); 985 // enums declared with `NS_OPTION` have a named enum and a named typedef, with 986 // the same name 987 ShouldDrop |= (Record->UnderlyingType.Name == Record->Name); 988 if (ShouldDrop) 989 return true; 990 991 // Create the symbol record if the other symbol droppping rules permit it. 992 serializeAPIRecord(Record); 993 if (!CurrentSymbol) 994 return true; 995 996 (*CurrentSymbol)["type"] = Record->UnderlyingType.USR; 997 998 return true; 999 } 1000 1001 void SymbolGraphSerializer::serializeSingleRecord(const APIRecord *Record) { 1002 switch (Record->getKind()) { 1003 // dispatch to the relevant walkUpFromMethod 1004 #define CONCRETE_RECORD(CLASS, BASE, KIND) \ 1005 case APIRecord::KIND: { \ 1006 walkUpFrom##CLASS(static_cast<const CLASS *>(Record)); \ 1007 break; \ 1008 } 1009 #include "clang/ExtractAPI/APIRecords.inc" 1010 // otherwise fallback on the only behavior we can implement safely. 1011 case APIRecord::RK_Unknown: 1012 visitAPIRecord(Record); 1013 break; 1014 default: 1015 llvm_unreachable("API Record with uninstantiable kind"); 1016 } 1017 } 1018 1019 Object SymbolGraphSerializer::serializeGraph(StringRef ModuleName, 1020 ExtendedModule &&EM) { 1021 Object Root; 1022 serializeObject(Root, "metadata", serializeMetadata()); 1023 serializeObject(Root, "module", serializeModuleObject(ModuleName)); 1024 1025 Root["symbols"] = std::move(EM.Symbols); 1026 Root["relationships"] = std::move(EM.Relationships); 1027 1028 return Root; 1029 } 1030 1031 void SymbolGraphSerializer::serializeGraphToStream( 1032 raw_ostream &OS, SymbolGraphSerializerOption Options, StringRef ModuleName, 1033 ExtendedModule &&EM) { 1034 Object Root = serializeGraph(ModuleName, std::move(EM)); 1035 if (Options.Compact) 1036 OS << formatv("{0}", json::Value(std::move(Root))) << "\n"; 1037 else 1038 OS << formatv("{0:2}", json::Value(std::move(Root))) << "\n"; 1039 } 1040 1041 void SymbolGraphSerializer::serializeMainSymbolGraph( 1042 raw_ostream &OS, const APISet &API, const APIIgnoresList &IgnoresList, 1043 SymbolGraphSerializerOption Options) { 1044 SymbolGraphSerializer Serializer( 1045 API, IgnoresList, Options.EmitSymbolLabelsForTesting, 1046 /*ForceEmitToMainModule=*/true, 1047 /*SkipSymbolsInCategoriesToExternalTypes=*/true); 1048 1049 Serializer.traverseAPISet(); 1050 Serializer.serializeGraphToStream(OS, Options, API.ProductName, 1051 std::move(Serializer.MainModule)); 1052 // FIXME: TODO handle extended modules here 1053 } 1054 1055 void SymbolGraphSerializer::serializeWithExtensionGraphs( 1056 raw_ostream &MainOutput, const APISet &API, 1057 const APIIgnoresList &IgnoresList, 1058 llvm::function_ref<std::unique_ptr<llvm::raw_pwrite_stream>(Twine BaseName)> 1059 CreateOutputStream, 1060 SymbolGraphSerializerOption Options) { 1061 SymbolGraphSerializer Serializer(API, IgnoresList, 1062 Options.EmitSymbolLabelsForTesting); 1063 Serializer.traverseAPISet(); 1064 1065 Serializer.serializeGraphToStream(MainOutput, Options, API.ProductName, 1066 std::move(Serializer.MainModule)); 1067 1068 for (auto &ExtensionSGF : Serializer.ExtendedModules) { 1069 if (auto ExtensionOS = 1070 CreateOutputStream(API.ProductName + "@" + ExtensionSGF.getKey())) 1071 Serializer.serializeGraphToStream(*ExtensionOS, Options, API.ProductName, 1072 std::move(ExtensionSGF.getValue())); 1073 } 1074 } 1075 1076 std::optional<Object> 1077 SymbolGraphSerializer::serializeSingleSymbolSGF(StringRef USR, 1078 const APISet &API) { 1079 APIRecord *Record = API.findRecordForUSR(USR); 1080 if (!Record) 1081 return {}; 1082 1083 Object Root; 1084 APIIgnoresList EmptyIgnores; 1085 SymbolGraphSerializer Serializer(API, EmptyIgnores, 1086 /*EmitSymbolLabelsForTesting*/ false, 1087 /*ForceEmitToMainModule*/ true); 1088 1089 // Set up serializer parent chain 1090 Serializer.Hierarchy = generateHierarchyFromRecord(Record); 1091 1092 Serializer.serializeSingleRecord(Record); 1093 serializeObject(Root, "symbolGraph", 1094 Serializer.serializeGraph(API.ProductName, 1095 std::move(Serializer.MainModule))); 1096 1097 Language Lang = API.getLanguage(); 1098 serializeArray(Root, "parentContexts", 1099 generateParentContexts(Serializer.Hierarchy, Lang)); 1100 1101 Array RelatedSymbols; 1102 1103 for (const auto &Fragment : Record->Declaration.getFragments()) { 1104 // If we don't have a USR there isn't much we can do. 1105 if (Fragment.PreciseIdentifier.empty()) 1106 continue; 1107 1108 APIRecord *RelatedRecord = API.findRecordForUSR(Fragment.PreciseIdentifier); 1109 1110 // If we can't find the record let's skip. 1111 if (!RelatedRecord) 1112 continue; 1113 1114 Object RelatedSymbol; 1115 RelatedSymbol["usr"] = RelatedRecord->USR; 1116 RelatedSymbol["declarationLanguage"] = getLanguageName(Lang); 1117 RelatedSymbol["accessLevel"] = RelatedRecord->Access.getAccess(); 1118 RelatedSymbol["filePath"] = RelatedRecord->Location.getFilename(); 1119 RelatedSymbol["moduleName"] = API.ProductName; 1120 RelatedSymbol["isSystem"] = RelatedRecord->IsFromSystemHeader; 1121 1122 serializeArray(RelatedSymbol, "parentContexts", 1123 generateParentContexts( 1124 generateHierarchyFromRecord(RelatedRecord), Lang)); 1125 1126 RelatedSymbols.push_back(std::move(RelatedSymbol)); 1127 } 1128 1129 serializeArray(Root, "relatedSymbols", RelatedSymbols); 1130 return Root; 1131 } 1132