#include #include #include #include #include #include #include #include "libclangmm/clangmm.h" #include "libclangmm/Utility.h" #include "utils.h" #include #include #include #include //#include // While indexing, we should refer to symbols by USR. When joining into the db, we can have optimized access. struct TypeDef; struct FuncDef; struct VarDef; /* template struct Id { uint64_t file_id; uint64_t local_id; Id() : file_id(0), local_id(0) {} // Needed for containers. Do not use directly. Id(uint64_t file_id, uint64_t local_id) : file_id(file_id), local_id(local_id) {} }; */ template struct LocalId { uint64_t local_id; LocalId() : local_id(0) {} // Needed for containers. Do not use directly. explicit LocalId(uint64_t local_id) : local_id(local_id) {} }; using TypeId = LocalId; using FuncId = LocalId; using VarId = LocalId; template struct Ref { LocalId id; clang::SourceLocation loc; Ref(LocalId id, clang::SourceLocation loc) : id(id), loc(loc) {} }; using TypeRef = Ref; using FuncRef = Ref; using VarRef = Ref; // NOTE: declaration is empty if there is no forward declaration! struct TypeDef { // General metadata. TypeId id; std::string usr; std::string short_name; std::string qualified_name; std::optional declaration; // Forward decl. std::optional definition; // Immediate parent and immediate derived types. std::vector parents; std::vector derived; // Types, functions, and variables defined in this type. std::vector types; std::vector funcs; std::vector vars; // Usages. std::vector uses; TypeDef(TypeId id, const std::string& usr) : id(id), usr(usr) {} }; struct FuncDef { // General metadata. FuncId id; std::string usr; std::string short_name; std::string qualified_name; std::optional declaration; std::optional definition; // Type which declares this one (ie, it is a method) std::optional declaring_type; // Method this method overrides. std::optional base; // Methods which directly override this one. std::vector derived; // Local variables defined in this function. std::vector locals; // Functions which call this one. std::vector callers; // Functions that this function calls. std::vector callees; // Usages. std::vector uses; FuncDef(FuncId id, const std::string& usr) : id(id), usr(usr) {} }; struct VarDef { // General metadata. VarId id; std::string usr; std::string short_name; std::string qualified_name; std::optional declaration; std::vector initializations; // Type of the variable. std::optional variable_type; // Type which declares this one (ie, it is a method) std::optional declaring_type; // Usages. std::vector uses; VarDef(VarId id, const std::string& usr) : id(id), usr(usr) {} }; struct ParsingDatabase { // NOTE: Every Id is resolved to a file_id of 0. The correct file_id needs // to get fixed up when inserting into the real db. std::unordered_map usr_to_type_id; std::unordered_map usr_to_func_id; std::unordered_map usr_to_var_id; std::vector types; std::vector funcs; std::vector vars; TypeId ToTypeId(const std::string& usr); FuncId ToFuncId(const std::string& usr); VarId ToVarId(const std::string& usr); TypeDef* Resolve(TypeId id); FuncDef* Resolve(FuncId id); VarDef* Resolve(VarId id); std::string ToString(); }; TypeId ParsingDatabase::ToTypeId(const std::string& usr) { auto it = usr_to_type_id.find(usr); if (it != usr_to_type_id.end()) return it->second; TypeId id(types.size()); types.push_back(TypeDef(id, usr)); usr_to_type_id[usr] = id; return id; } FuncId ParsingDatabase::ToFuncId(const std::string& usr) { auto it = usr_to_func_id.find(usr); if (it != usr_to_func_id.end()) return it->second; FuncId id(funcs.size()); funcs.push_back(FuncDef(id, usr)); usr_to_func_id[usr] = id; return id; } VarId ParsingDatabase::ToVarId(const std::string& usr) { auto it = usr_to_var_id.find(usr); if (it != usr_to_var_id.end()) return it->second; VarId id(vars.size()); vars.push_back(VarDef(id, usr)); usr_to_var_id[usr] = id; return id; } TypeDef* ParsingDatabase::Resolve(TypeId id) { return &types[id.local_id]; } FuncDef* ParsingDatabase::Resolve(FuncId id) { return &funcs[id.local_id]; } VarDef* ParsingDatabase::Resolve(VarId id) { return &vars[id.local_id]; } using Writer = rapidjson::PrettyWriter; void Write(Writer& writer, const char* key, clang::SourceLocation location) { if (key) writer.Key(key); std::string s = location.ToString(); writer.String(s.c_str()); } void Write(Writer& writer, const char* key, std::optional location) { if (location) { Write(writer, key, location.value()); } //else { // if (key) writer.Key(key); // writer.Null(); //} } void Write(Writer& writer, const char* key, const std::vector& locs) { if (locs.size() == 0) return; if (key) writer.Key(key); writer.StartArray(); for (const clang::SourceLocation& loc : locs) Write(writer, nullptr, loc); writer.EndArray(); } template void Write(Writer& writer, const char* key, LocalId id) { if (key) writer.Key(key); writer.Uint64(id.local_id); } template void Write(Writer& writer, const char* key, std::optional> id) { if (id) { Write(writer, key, id.value()); } //else { // if (key) writer.Key(key); // writer.Null(); //} } template void Write(Writer& writer, const char* key, const std::vector>& ids) { if (ids.size() == 0) return; if (key) writer.Key(key); writer.StartArray(); for (LocalId id : ids) Write(writer, nullptr, id); writer.EndArray(); } template void Write(Writer& writer, const char* key, Ref ref) { if (key) writer.Key(key); std::string s = std::to_string(ref.id.local_id) + "@" + ref.loc.ToString(); writer.String(s.c_str()); } template void Write(Writer& writer, const char* key, const std::vector>& refs) { if (refs.size() == 0) return; if (key) writer.Key(key); writer.StartArray(); for (Ref ref : refs) Write(writer, nullptr, ref); writer.EndArray(); } void Write(Writer& writer, const char* key, const std::string& value) { if (value.size() == 0) return; if (key) writer.Key(key); writer.String(value.c_str()); } void Write(Writer& writer, const char* key, uint64_t value) { if (key) writer.Key(key); writer.Uint64(value); } std::string ParsingDatabase::ToString() { #define WRITE(name) Write(writer, #name, def.name) rapidjson::StringBuffer output; rapidjson::PrettyWriter writer(output); writer.SetFormatOptions( rapidjson::PrettyFormatOptions::kFormatSingleLineArray); writer.SetIndent(' ', 2); writer.StartObject(); // Types writer.Key("types"); writer.StartArray(); for (TypeDef& def : types) { writer.StartObject(); WRITE(id); WRITE(usr); WRITE(short_name); WRITE(qualified_name); WRITE(declaration); WRITE(definition); WRITE(parents); WRITE(derived); WRITE(types); WRITE(funcs); WRITE(vars); WRITE(uses); writer.EndObject(); } writer.EndArray(); // Functions writer.Key("functions"); writer.StartArray(); for (FuncDef& def : funcs) { writer.StartObject(); WRITE(id); WRITE(usr); WRITE(short_name); WRITE(qualified_name); WRITE(declaration); WRITE(definition); WRITE(declaring_type); WRITE(base); WRITE(derived); WRITE(locals); WRITE(callers); WRITE(callees); WRITE(uses); writer.EndObject(); } writer.EndArray(); // Variables writer.Key("variables"); writer.StartArray(); for (VarDef& def : vars) { writer.StartObject(); WRITE(id); WRITE(usr); WRITE(short_name); WRITE(qualified_name); WRITE(declaration); WRITE(initializations); WRITE(variable_type); WRITE(declaring_type); WRITE(uses); writer.EndObject(); } writer.EndArray(); writer.EndObject(); return output.GetString(); #undef WRITE } struct FileDef { uint64_t id; std::string path; std::vector types; std::vector funcs; std::vector vars; }; /* struct Database { std::unordered_map usr_to_type_id; std::unordered_map usr_to_func_id; std::unordered_map usr_to_var_id; std::vector files; TypeId ToTypeId(const std::string& usr); FuncId ToFuncId(const std::string& usr); VarId ToVarId(const std::string& usr); }; TypeId Database::ToTypeId(const std::string& usr) { auto it = usr_to_type_id.find(usr); assert(it != usr_to_type_id.end() && "Usr is not registered"); return it->second; } FuncId Database::ToFuncId(const std::string& usr) { auto it = usr_to_func_id.find(usr); assert(it != usr_to_func_id.end() && "Usr is not registered"); return it->second; } VarId Database::ToVarId(const std::string& usr) { auto it = usr_to_var_id.find(usr); assert(it != usr_to_var_id.end() && "Usr is not registered"); return it->second; } TypeDef* Resolve(FileDef* file, TypeId id) { assert(file->id == id.file_id); return &file->types[id.local_id]; } FuncDef* Resolve(FileDef* file, FuncId id) { assert(file->id == id.file_id); return &file->funcs[id.local_id]; } VarDef* Resolve(FileDef* file, VarId id) { assert(file->id == id.file_id); return &file->vars[id.local_id]; } TypeDef* Resolve(Database* db, TypeId id) { return Resolve(&db->files[id.file_id], id); } FuncDef* Resolve(Database* db, FuncId id) { return Resolve(&db->files[id.file_id], id); } VarDef* Resolve(Database* db, VarId id) { return Resolve(&db->files[id.file_id], id); } */ struct NamespaceStack { std::vector stack; void Push(const std::string& ns); void Pop(); std::string ComputeQualifiedName( ParsingDatabase* db, std::optional declaring_type, std::string short_name); static NamespaceStack kEmpty; }; NamespaceStack NamespaceStack::kEmpty; void NamespaceStack::Push(const std::string& ns) { stack.push_back(ns); } void NamespaceStack::Pop() { stack.pop_back(); } std::string NamespaceStack::ComputeQualifiedName( ParsingDatabase* db, std::optional declaring_type, std::string short_name) { if (declaring_type) { TypeDef* def = db->Resolve(declaring_type.value()); return def->qualified_name + "::" + short_name; } std::string result; for (const std::string& ns : stack) result += ns + "::"; result += short_name; return result; } std::optional ResolveDeclaringType(CXCursorKind kind, ParsingDatabase* db, const clang::Cursor& cursor, std::optional declaring_type) { // Resolve the declaring type for out-of-line method definitions. if (!declaring_type) { clang::Cursor parent = cursor.get_semantic_parent(); switch (parent.get_kind()) { case CXCursor_ClassDecl: case CXCursor_StructDecl: declaring_type = db->ToTypeId(parent.get_usr()); break; } } // FieldDecl, etc must have a declaring type. assert(cursor.get_kind() != kind || declaring_type); return declaring_type; } clang::VisiterResult DumpVisitor(clang::Cursor cursor, clang::Cursor parent, int* level) { for (int i = 0; i < *level; ++i) std::cout << " "; std::cout << clang::ToString(cursor.get_kind()) << " " << cursor.get_spelling() << std::endl; *level += 1; cursor.VisitChildren(&DumpVisitor, level); *level -= 1; return clang::VisiterResult::Continue; } void Dump(clang::Cursor cursor) { int level = 0; cursor.VisitChildren(&DumpVisitor, &level); } void HandleVarDecl(ParsingDatabase* db, NamespaceStack* ns, clang::Cursor var, std::optional declaring_type) { //Dump(var); VarId var_id = db->ToVarId(var.get_usr()); declaring_type = ResolveDeclaringType(CXCursor_FieldDecl, db, var, declaring_type); // TODO: We could use RAII to verify we don't modify db while have a *Def // instance alive. VarDef* var_def = db->Resolve(var_id); var_def->short_name = var.get_spelling(); var_def->qualified_name = ns->ComputeQualifiedName(db, declaring_type, var_def->short_name); if (declaring_type && !var_def->declaration) { db->Resolve(declaring_type.value())->vars.push_back(var_id); var_def->declaring_type = declaring_type; } // We don't do any additional processing for non-definitions. if (!var.is_definition()) { var_def->declaration = var.get_source_location(); return; } // If we're a definition and there hasn't been a forward decl, just assign // declaration location to definition location. else if (!var_def->declaration) { var_def->declaration = var.get_source_location(); } // TODO: Figure out how to scan initializations properly. We probably need // to scan for assignment statement, or definition+ctor. var_def->initializations.push_back(var.get_source_location()); var_def->variable_type = db->ToTypeId(var.get_type().get_usr()); } // |func_id| is the function definition that is currently being processed. void InsertReference(ParsingDatabase* db, FuncId func_id, clang::Cursor referencer) { clang::SourceLocation loc = referencer.get_source_location(); clang::Cursor referenced = referencer.get_referenced(); switch (referenced.get_kind()) { case CXCursor_CXXMethod: case CXCursor_FunctionDecl: { FuncId referenced_id = db->ToFuncId(referenced.get_usr()); FuncDef* referenced_def = db->Resolve(referenced_id); FuncDef* func_def = db->Resolve(func_id); func_def->callees.push_back(FuncRef(referenced_id, loc)); referenced_def->callers.push_back(FuncRef(func_id, loc)); referenced_def->uses.push_back(loc); break; } case CXCursor_VarDecl: { VarId referenced_id = db->ToVarId(referenced.get_usr()); VarDef* referenced_def = db->Resolve(referenced_id); referenced_def->uses.push_back(loc); break; } default: std::cerr << "Unhandled reference from \"" << referencer.ToString() << "\" to \"" << referenced.ToString() << "\"" << std::endl; break; } } struct FuncDefinitionParam { ParsingDatabase* db; NamespaceStack* ns; FuncId func_id; FuncDefinitionParam(ParsingDatabase* db, NamespaceStack* ns, FuncId func_id) : db(db), ns(ns), func_id(func_id) {} }; clang::VisiterResult VisitFuncDefinition(clang::Cursor cursor, clang::Cursor parent, FuncDefinitionParam* param) { //std::cout << "VistFuncDefinition got " << cursor.ToString() << std::endl; switch (cursor.get_kind()) { // TODO: Maybe we should default to recurse? /* case CXCursor_CompoundStmt: case CXCursor_DeclStmt: case CXCursor_CallExpr: case CXCursor_UnexposedExpr: case CXCursor_UnaryExpr: return clang::VisiterResult::Recurse; */ case CXCursor_CallExpr: // The called element is handled by DeclRefExpr below. //InsertReference(param->db, param->func_id, cursor); return clang::VisiterResult::Recurse; case CXCursor_MemberRefExpr: case CXCursor_DeclRefExpr: InsertReference(param->db, param->func_id, cursor); return clang::VisiterResult::Continue; case CXCursor_VarDecl: case CXCursor_ParmDecl: HandleVarDecl(param->db, param->ns, cursor, std::nullopt); return clang::VisiterResult::Recurse; case CXCursor_ReturnStmt: return clang::VisiterResult::Recurse; default: //std::cerr << "Unhandled VisitFuncDefinition kind " << clang::ToString(cursor.get_kind()) << std::endl; return clang::VisiterResult::Recurse; } } void HandleFunc(ParsingDatabase* db, NamespaceStack* ns, clang::Cursor func, std::optional declaring_type) { // What this method must process: // - function declaration // - function definition // - method declaration // - method inline definition // - method definition // Resolve id before checking for is_definition so that we insert the // function into the db even if it is only a prototype. This is needed for // various file-level operations like outlining. FuncId func_id = db->ToFuncId(func.get_usr()); // TODO: Consider skipping some of this processing if we've done it already // (ie, parsed prototype, then parse definition). declaring_type = ResolveDeclaringType(CXCursor_CXXMethod, db, func, declaring_type); FuncDef* func_def = db->Resolve(func_id); func_def->short_name = func.get_spelling(); func_def->qualified_name = ns->ComputeQualifiedName(db, declaring_type, func_def->short_name); if (declaring_type && !func_def->declaration) { db->Resolve(declaring_type.value())->funcs.push_back(func_id); func_def->declaring_type = declaring_type; } // We don't do any additional processing for non-definitions. if (!func.is_definition()) { func_def->declaration = func.get_source_location(); return; } func_def->definition = func.get_source_location(); //std::cout << "!! Types: "; //for (clang::Cursor arg : func.get_arguments()) // std::cout << arg.ToString() << ", "; //std::cout << std::endl; //std::cout << func.get_usr() << ": Set qualified name to " << db->Resolve(id)->qualified_name; //std::cout << " IsDefinition? " << func.is_definition() << std::endl; //clang::Type func_type = func.get_type(); //clang::Type return_type = func_type.get_return_type(); //std::vector argument_types = func_type.get_arguments(); //auto argument_types = func.get_arguments(); //clang::Type cursor_type = func.get_type(); //clang::Type return_type_1 = func.get_type().get_result(); //clang::Type return_type_2 = clang_getCursorResultType(func.cx_cursor); //Dump(func); FuncDefinitionParam funcDefinitionParam(db, &NamespaceStack::kEmpty, func_id); func.VisitChildren(&VisitFuncDefinition, &funcDefinitionParam); //CXType return_type = clang_getResultType(func.get_type()); //CXType_FunctionProto //std::cout << "!! HandleFunc " << func.get_type_description() << std::endl; //std::cout << " comment: " << func.get_comments() << std::endl; //std::cout << " spelling: " << func.get_spelling() << std::endl; //for (clang::Cursor argument : func.get_arguments()) // std::cout << " arg: " << clang::ToString(argument.get_kind()) << " " << argument.get_spelling() << std::endl; } struct ClassDeclParam { ParsingDatabase* db; NamespaceStack* ns; TypeId active_type; ClassDeclParam(ParsingDatabase* db, NamespaceStack* ns, TypeId active_type) : db(db), ns(ns), active_type(active_type) {} }; clang::VisiterResult VisitClassDecl(clang::Cursor cursor, clang::Cursor parent, ClassDeclParam* param) { ParsingDatabase* db = param->db; switch (cursor.get_kind()) { case CXCursor_CXXMethod: HandleFunc(param->db, param->ns, cursor, param->active_type); break; case CXCursor_FieldDecl: case CXCursor_VarDecl: HandleVarDecl(param->db, param->ns, cursor, param->active_type); break; default: std::cerr << "Unhandled VisitClassDecl kind " << clang::ToString(cursor.get_kind()) << std::endl; break; } return clang::VisiterResult::Continue; } void HandleClassDecl(clang::Cursor cursor, ParsingDatabase* db, NamespaceStack* ns) { TypeId func_id = db->ToTypeId(cursor.get_usr()); TypeDef* func_def = db->Resolve(func_id); func_def->short_name = cursor.get_spelling(); // TODO: Support nested classes (pass in declaring type insteaad of nullopt!) func_def->qualified_name = ns->ComputeQualifiedName(db, std::nullopt, func_def->short_name); if (!cursor.is_definition()) { if (!func_def->declaration) func_def->declaration = cursor.get_source_location(); return; } func_def->definition = cursor.get_source_location(); ClassDeclParam classDeclParam(db, ns, func_id); cursor.VisitChildren(&VisitClassDecl, &classDeclParam); } struct FileParam { ParsingDatabase* db; NamespaceStack* ns; FileParam(ParsingDatabase* db, NamespaceStack* ns) : db(db), ns(ns) {} }; clang::VisiterResult VisitFile(clang::Cursor cursor, clang::Cursor parent, FileParam* param) { switch (cursor.get_kind()) { case CXCursor_Namespace: // For a namespace, visit the children of the namespace, but this time with // a pushed namespace stack. param->ns->Push(cursor.get_display_name()); cursor.VisitChildren(&VisitFile, param); param->ns->Pop(); break; case CXCursor_StructDecl: case CXCursor_ClassDecl: // TODO: Cleanup Handle* param order. HandleClassDecl(cursor, param->db, param->ns); break; case CXCursor_CXXMethod: case CXCursor_FunctionDecl: HandleFunc(param->db, param->ns, cursor, std::nullopt); break; case CXCursor_VarDecl: HandleVarDecl(param->db, param->ns, cursor, std::nullopt); break; default: std::cerr << "Unhandled VisitFile kind " << clang::ToString(cursor.get_kind()) << std::endl; break; } return clang::VisiterResult::Continue; } ParsingDatabase Parse(std::string filename) { std::vector args; clang::Index index(0 /*excludeDeclarationsFromPCH*/, 0 /*displayDiagnostics*/); clang::TranslationUnit tu(index, filename, args); std::cout << "Start document dump" << std::endl; Dump(tu.document_cursor()); std::cout << "Done document dump" << std::endl << std::endl; ParsingDatabase db; NamespaceStack ns; FileParam fileParam(&db, &ns); tu.document_cursor().VisitChildren(&VisitFile, &fileParam); return db; } template bool AreEqual(const std::vector& a, const std::vector& b) { if (a.size() != b.size()) return false; for (int i = 0; i < a.size(); ++i) { if (a[i] != b[i]) return false; } return true; } void Write(const std::vector& strs) { for (const std::string& str : strs) { std::cout << str << std::endl; } } std::vector split_string(const std::string& str, const std::string& delimiter) { // http://stackoverflow.com/a/13172514 std::vector strings; std::string::size_type pos = 0; std::string::size_type prev = 0; while ((pos = str.find(delimiter, prev)) != std::string::npos) { strings.push_back(str.substr(prev, pos - prev)); prev = pos + 1; } // To get the last substring (or only, if delimiter is not found) strings.push_back(str.substr(prev)); return strings; } void DiffDocuments(rapidjson::Document& expected, rapidjson::Document& actual) { std::vector actual_output; { rapidjson::StringBuffer buffer; rapidjson::PrettyWriter writer(buffer); writer.SetFormatOptions( rapidjson::PrettyFormatOptions::kFormatSingleLineArray); writer.SetIndent(' ', 2); buffer.Clear(); actual.Accept(writer); actual_output = split_string(buffer.GetString(), "\n"); } std::vector expected_output; { rapidjson::StringBuffer buffer; rapidjson::PrettyWriter writer(buffer); writer.SetFormatOptions( rapidjson::PrettyFormatOptions::kFormatSingleLineArray); writer.SetIndent(' ', 2); buffer.Clear(); expected.Accept(writer); expected_output = split_string(buffer.GetString(), "\n"); } int len = std::min(actual_output.size(), expected_output.size()); for (int i = 0; i < len; ++i) { if (actual_output[i] != expected_output[i]) { std::cout << "Line " << i << " differs" << std::endl; std::cout << " expected: " << expected_output[i] << std::endl; std::cout << " actual: " << actual_output[i] << std::endl; } } if (actual_output.size() > len) { std::cout << "Additional output in actual:" << std::endl; for (int i = len; i < actual_output.size(); ++i) std::cout << " " << actual_output[i] << std::endl; } if (expected_output.size() > len) { std::cout << "Additional output in expected:" << std::endl; for (int i = len; i < expected_output.size(); ++i) std::cout << " " << expected_output[i] << std::endl; } } int main(int argc, char** argv) { for (std::string path : GetFilesInFolder("tests")) { // TODO: Fix all existing tests. //if (path != "tests/usage/func_usage_class_inline_var_def.cc") continue; // Parse expected output from the test, parse it into JSON document. std::string expected_output; ParseTestExpectation(path, &expected_output); rapidjson::Document expected; expected.Parse(expected_output.c_str()); // Run test. std::cout << "[START] " << path << std::endl; ParsingDatabase db = Parse(path); std::string actual_output = db.ToString(); rapidjson::Document actual; actual.Parse(actual_output.c_str()); if (actual == expected) { std::cout << "[PASSED] " << path << std::endl; } else { std::cout << "[FAILED] " << path << std::endl; std::cout << "Expected output for " << path << ":" << std::endl; std::cout << expected_output; std::cout << "Actual output for " << path << ":" << std::endl; std::cout << actual_output; std::cout << std::endl; std::cout << std::endl; DiffDocuments(expected, actual); break; } } std::cin.get(); return 0; } // TODO: ctor/dtor, copy ctor