// Protocol Buffers - Google's data interchange format // Copyright 2008 Google Inc. All rights reserved. // https://developers.google.com/protocol-buffers/ // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are // met: // // * Redistributions of source code must retain the above copyright // notice, this list of conditions and the following disclaimer. // * Redistributions in binary form must reproduce the above // copyright notice, this list of conditions and the following disclaimer // in the documentation and/or other materials provided with the // distribution. // * Neither the name of Google Inc. nor the names of its // contributors may be used to endorse or promote products derived from // this software without specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. // Author: kenton@google.com (Kenton Varda) // Based on original Protocol Buffers design by // Sanjay Ghemawat, Jeff Dean, and others. #ifndef GOOGLE_PROTOBUF_COMPILER_CPP_HELPERS_H__ #define GOOGLE_PROTOBUF_COMPILER_CPP_HELPERS_H__ #include #include #include #include #include #include #include #include #include #include #include #include namespace google { namespace protobuf { namespace compiler { namespace cpp { inline std::string ProtobufNamespace(const Options& options) { return options.opensource_runtime ? "google::protobuf" : "proto2"; } inline std::string MacroPrefix(const Options& options) { return options.opensource_runtime ? "GOOGLE_PROTOBUF" : "GOOGLE_PROTOBUF"; } inline std::string DeprecatedAttribute(const Options& options, bool deprecated) { return deprecated ? "PROTOBUF_DEPRECATED " : ""; } // Commonly-used separator comments. Thick is a line of '=', thin is a line // of '-'. extern const char kThickSeparator[]; extern const char kThinSeparator[]; inline bool IsProto1(const FileDescriptor* file, const Options& options) { return false; } void SetCommonVars(const Options& options, std::map* variables); bool GetBootstrapBasename(const Options& options, const std::string& basename, std::string* bootstrap_basename); bool MaybeBootstrap(const Options& options, GeneratorContext* generator_context, bool bootstrap_flag, std::string* basename); bool IsBootstrapProto(const Options& options, const FileDescriptor* file); // Name space of the proto file. This namespace is such that the string // "::some_name" is the correct fully qualified namespace. // This means if the package is empty the namespace is "", and otherwise // the namespace is "::foo::bar::...::baz" without trailing semi-colons. std::string Namespace(const std::string& package); inline std::string Namespace(const FileDescriptor* d) { return Namespace(d->package()); } std::string Namespace(const Descriptor* d); std::string Namespace(const FieldDescriptor* d); std::string Namespace(const EnumDescriptor* d); // Returns true if it's safe to reset "field" to zero. bool CanInitializeByZeroing(const FieldDescriptor* field); std::string ClassName(const Descriptor* descriptor); std::string ClassName(const EnumDescriptor* enum_descriptor); std::string QualifiedClassName(const Descriptor* d); std::string QualifiedClassName(const EnumDescriptor* d); // DEPRECATED just use ClassName or QualifiedClassName, a boolean is very // unreadable at the callsite. // Returns the non-nested type name for the given type. If "qualified" is // true, prefix the type with the full namespace. For example, if you had: // package foo.bar; // message Baz { message Qux {} } // Then the qualified ClassName for Qux would be: // ::foo::bar::Baz_Qux // While the non-qualified version would be: // Baz_Qux inline std::string ClassName(const Descriptor* descriptor, bool qualified) { return qualified ? QualifiedClassName(descriptor) : ClassName(descriptor); } inline std::string ClassName(const EnumDescriptor* descriptor, bool qualified) { return qualified ? QualifiedClassName(descriptor) : ClassName(descriptor); } // Fully qualified name of the default_instance of this message. std::string DefaultInstanceName(const Descriptor* descriptor); // Returns the name of a no-op function that we can call to introduce a linker // dependency on the given message type. This is used to implement implicit weak // fields. std::string ReferenceFunctionName(const Descriptor* descriptor); // Name of the base class: google::protobuf::Message or google::protobuf::MessageLite. std::string SuperClassName(const Descriptor* descriptor, const Options& options); // Get the (unqualified) name that should be used for this field in C++ code. // The name is coerced to lower-case to emulate proto1 behavior. People // should be using lowercase-with-underscores style for proto field names // anyway, so normally this just returns field->name(). std::string FieldName(const FieldDescriptor* field); // Get the sanitized name that should be used for the given enum in C++ code. std::string EnumValueName(const EnumValueDescriptor* enum_value); // Returns an estimate of the compiler's alignment for the field. This // can't guarantee to be correct because the generated code could be compiled on // different systems with different alignment rules. The estimates below assume // 64-bit pointers. int EstimateAlignmentSize(const FieldDescriptor* field); // Get the unqualified name that should be used for a field's field // number constant. std::string FieldConstantName(const FieldDescriptor *field); // Returns the scope where the field was defined (for extensions, this is // different from the message type to which the field applies). inline const Descriptor* FieldScope(const FieldDescriptor* field) { return field->is_extension() ? field->extension_scope() : field->containing_type(); } // Returns the fully-qualified type name field->message_type(). Usually this // is just ClassName(field->message_type(), true); std::string FieldMessageTypeName(const FieldDescriptor* field); // Strips ".proto" or ".protodevel" from the end of a filename. PROTOC_EXPORT std::string StripProto(const std::string& filename); // Get the C++ type name for a primitive type (e.g. "double", "::google::protobuf::int32", etc.). const char* PrimitiveTypeName(FieldDescriptor::CppType type); std::string PrimitiveTypeName(const Options& options, FieldDescriptor::CppType type); // Get the declared type name in CamelCase format, as is used e.g. for the // methods of WireFormat. For example, TYPE_INT32 becomes "Int32". const char* DeclaredTypeMethodName(FieldDescriptor::Type type); // Return the code that evaluates to the number when compiled. std::string Int32ToString(int number); // Return the code that evaluates to the number when compiled. std::string Int64ToString(const Options& options, int64 number); // Get code that evaluates to the field's default value. std::string DefaultValue(const Options& options, const FieldDescriptor* field); // Compatibility function for callers outside proto2. std::string DefaultValue(const FieldDescriptor* field); // Convert a file name into a valid identifier. std::string FilenameIdentifier(const std::string& filename); // For each .proto file generates a unique name. To prevent collisions of // symbols in the global namespace std::string UniqueName(const std::string& name, const std::string& filename, const Options& options); inline std::string UniqueName(const std::string& name, const FileDescriptor* d, const Options& options) { return UniqueName(name, d->name(), options); } inline std::string UniqueName(const std::string& name, const Descriptor* d, const Options& options) { return UniqueName(name, d->file(), options); } inline std::string UniqueName(const std::string& name, const EnumDescriptor* d, const Options& options) { return UniqueName(name, d->file(), options); } inline std::string UniqueName(const std::string& name, const ServiceDescriptor* d, const Options& options) { return UniqueName(name, d->file(), options); } // Versions for call sites that only support the internal runtime (like proto1 // support). inline Options InternalRuntimeOptions() { Options options; options.opensource_runtime = false; return options; } inline std::string UniqueName(const std::string& name, const std::string& filename) { return UniqueName(name, filename, InternalRuntimeOptions()); } inline std::string UniqueName(const std::string& name, const FileDescriptor* d) { return UniqueName(name, d->name(), InternalRuntimeOptions()); } inline std::string UniqueName(const std::string& name, const Descriptor* d) { return UniqueName(name, d->file(), InternalRuntimeOptions()); } inline std::string UniqueName(const std::string& name, const EnumDescriptor* d) { return UniqueName(name, d->file(), InternalRuntimeOptions()); } inline std::string UniqueName(const std::string& name, const ServiceDescriptor* d) { return UniqueName(name, d->file(), InternalRuntimeOptions()); } // Return the qualified C++ name for a file level symbol. std::string QualifiedFileLevelSymbol(const std::string& package, const std::string& name); // Escape C++ trigraphs by escaping question marks to \? std::string EscapeTrigraphs(const std::string& to_escape); // Escaped function name to eliminate naming conflict. std::string SafeFunctionName(const Descriptor* descriptor, const FieldDescriptor* field, const std::string& prefix); // Returns true if generated messages have public unknown fields accessors inline bool PublicUnknownFieldsAccessors(const Descriptor* message) { return message->file()->syntax() != FileDescriptor::SYNTAX_PROTO3; } // Returns the optimize mode for , respecting . FileOptions_OptimizeMode GetOptimizeFor(const FileDescriptor* file, const Options& options); // Determines whether unknown fields will be stored in an UnknownFieldSet or // a string. inline bool UseUnknownFieldSet(const FileDescriptor* file, const Options& options) { return GetOptimizeFor(file, options) != FileOptions::LITE_RUNTIME; } inline bool IsWeak(const FieldDescriptor* field, const Options& options) { if (field->options().weak()) { GOOGLE_CHECK(!options.opensource_runtime); return true; } return false; } // For a string field, returns the effective ctype. If the actual ctype is // not supported, returns the default of STRING. FieldOptions::CType EffectiveStringCType(const FieldDescriptor* field, const Options& options); inline bool IsCord(const FieldDescriptor* field, const Options& options) { return field->cpp_type() == FieldDescriptor::CPPTYPE_STRING && EffectiveStringCType(field, options) == FieldOptions::CORD; } inline bool IsStringPiece(const FieldDescriptor* field, const Options& options) { return field->cpp_type() == FieldDescriptor::CPPTYPE_STRING && EffectiveStringCType(field, options) == FieldOptions::STRING_PIECE; } // Does the given FileDescriptor use lazy fields? bool HasLazyFields(const FileDescriptor* file, const Options& options); // Is the given field a supported lazy field? inline bool IsLazy(const FieldDescriptor* field, const Options& options) { return field->options().lazy() && !field->is_repeated() && field->type() == FieldDescriptor::TYPE_MESSAGE && GetOptimizeFor(field->file(), options) != FileOptions::LITE_RUNTIME && !options.opensource_runtime; } // Does the file contain any definitions that need extension_set.h? bool HasExtensionsOrExtendableMessage(const FileDescriptor* file); // Does the file have any repeated fields, necessitating the file to include // repeated_field.h? This does not include repeated extensions, since those are // all stored internally in an ExtensionSet, not a separate RepeatedField*. bool HasRepeatedFields(const FileDescriptor* file); // Does the file have any string/bytes fields with ctype=STRING_PIECE? This // does not include extensions, since ctype is ignored for extensions. bool HasStringPieceFields(const FileDescriptor* file, const Options& options); // Does the file have any string/bytes fields with ctype=CORD? This does not // include extensions, since ctype is ignored for extensions. bool HasCordFields(const FileDescriptor* file, const Options& options); // Does the file have any map fields, necessitating the file to include // map_field_inl.h and map.h. bool HasMapFields(const FileDescriptor* file); // Does this file have any enum type definitions? bool HasEnumDefinitions(const FileDescriptor* file); // Does this file have generated parsing, serialization, and other // standard methods for which reflection-based fallback implementations exist? inline bool HasGeneratedMethods(const FileDescriptor* file, const Options& options) { return GetOptimizeFor(file, options) != FileOptions::CODE_SIZE; } // Do message classes in this file have descriptor and reflection methods? inline bool HasDescriptorMethods(const FileDescriptor* file, const Options& options) { return GetOptimizeFor(file, options) != FileOptions::LITE_RUNTIME; } // Should we generate generic services for this file? inline bool HasGenericServices(const FileDescriptor* file, const Options& options) { return file->service_count() > 0 && GetOptimizeFor(file, options) != FileOptions::LITE_RUNTIME && file->options().cc_generic_services(); } // Should we generate a separate, super-optimized code path for serializing to // flat arrays? We don't do this in Lite mode because we'd rather reduce code // size. inline bool HasFastArraySerialization(const FileDescriptor* file, const Options& options) { return GetOptimizeFor(file, options) == FileOptions::SPEED; } inline bool IsProto2MessageSet(const Descriptor* descriptor, const Options& options) { return !options.opensource_runtime && !options.enforce_lite && !options.lite_implicit_weak_fields && descriptor->options().message_set_wire_format() && descriptor->full_name() == "google.protobuf.bridge.MessageSet"; } inline bool IsProto2MessageSetFile(const FileDescriptor* file, const Options& options) { return !options.opensource_runtime && !options.enforce_lite && !options.lite_implicit_weak_fields && file->name() == "net/proto2/bridge/proto/message_set.proto"; } inline bool IsMapEntryMessage(const Descriptor* descriptor) { return descriptor->options().map_entry(); } // Returns true if the field's CPPTYPE is string or message. bool IsStringOrMessage(const FieldDescriptor* field); std::string UnderscoresToCamelCase(const std::string& input, bool cap_next_letter); inline bool HasFieldPresence(const FileDescriptor* file) { return file->syntax() != FileDescriptor::SYNTAX_PROTO3; } // Returns true if 'enum' semantics are such that unknown values are preserved // in the enum field itself, rather than going to the UnknownFieldSet. inline bool HasPreservingUnknownEnumSemantics(const FileDescriptor* file) { return file->syntax() == FileDescriptor::SYNTAX_PROTO3; } inline bool SupportsArenas(const FileDescriptor* file) { return file->options().cc_enable_arenas(); } inline bool SupportsArenas(const Descriptor* desc) { return SupportsArenas(desc->file()); } inline bool SupportsArenas(const FieldDescriptor* field) { return SupportsArenas(field->file()); } inline bool IsCrossFileMessage(const FieldDescriptor* field) { return field->type() == FieldDescriptor::TYPE_MESSAGE && field->message_type()->file() != field->file(); } inline std::string MessageCreateFunction(const Descriptor* d) { return SupportsArenas(d) ? "CreateMessage" : "Create"; } inline std::string MakeDefaultName(const FieldDescriptor* field) { return "_i_give_permission_to_break_this_code_default_" + FieldName(field) + "_"; } bool IsAnyMessage(const FileDescriptor* descriptor); bool IsAnyMessage(const Descriptor* descriptor); bool IsWellKnownMessage(const FileDescriptor* descriptor); inline FileOptions_OptimizeMode GetOptimizeFor(const FileDescriptor* file, const Options& options) { return options.enforce_lite ? FileOptions::LITE_RUNTIME : file->options().optimize_for(); } // This orders the messages in a .pb.cc as it's outputted by file.cc void FlattenMessagesInFile(const FileDescriptor* file, std::vector* result); inline std::vector FlattenMessagesInFile( const FileDescriptor* file) { std::vector result; FlattenMessagesInFile(file, &result); return result; } bool HasWeakFields(const Descriptor* desc, const Options& options); bool HasWeakFields(const FileDescriptor* desc, const Options& options); // Returns true if the "required" restriction check should be ignored for the // given field. inline static bool ShouldIgnoreRequiredFieldCheck(const FieldDescriptor* field, const Options& options) { // Do not check "required" for lazy fields. return IsLazy(field, options); } struct MessageAnalysis { bool is_recursive; bool contains_cord; bool contains_extension; bool contains_required; bool constructor_requires_initialization; }; // This class is used in FileGenerator, to ensure linear instead of // quadratic performance, if we do this per message we would get O(V*(V+E)). // Logically this is just only used in message.cc, but in the header for // FileGenerator to help share it. class PROTOC_EXPORT MessageSCCAnalyzer { public: explicit MessageSCCAnalyzer(const Options& options) : options_(options) {} MessageAnalysis GetSCCAnalysis(const SCC* scc); bool HasRequiredFields(const Descriptor* descriptor) { MessageAnalysis result = GetSCCAnalysis(GetSCC(descriptor)); return result.contains_required || result.contains_extension; } const SCC* GetSCC(const Descriptor* descriptor) { return analyzer_.GetSCC(descriptor); } private: SCCAnalyzer analyzer_; Options options_; std::map analysis_cache_; }; void ListAllFields(const Descriptor* d, std::vector* fields); void ListAllFields(const FileDescriptor* d, std::vector* fields); void ListAllTypesForServices(const FileDescriptor* fd, std::vector* types); // Indicates whether we should use implicit weak fields for this file. bool UsingImplicitWeakFields(const FileDescriptor* file, const Options& options); // Indicates whether to treat this field as implicitly weak. bool IsImplicitWeakField(const FieldDescriptor* field, const Options& options, MessageSCCAnalyzer* scc_analyzer); // Formatter is a functor class which acts as a closure around printer and // the variable map. It's much like printer->Print except it supports both named // variables that are substituted using a key value map and direct arguments. In // the format string $1$, $2$, etc... are substituted for the first, second, ... // direct argument respectively in the format call, it accepts both strings and // integers. The implementation verifies all arguments are used and are "first" // used in order of appearance in the argument list. For example, // // Format("return array[$1$];", 3) -> "return array[3];" // Format("array[$2$] = $1$;", "Bla", 3) -> FATAL error (wrong order) // Format("array[$1$] = $2$;", 3, "Bla") -> "array[3] = Bla;" // // The arguments can be used more than once like // // Format("array[$1$] = $2$; // Index = $1$", 3, "Bla") -> // "array[3] = Bla; // Index = 3" // // If you use more arguments use the following style to help the reader, // // Format("int $1$() {\n" // " array[$2$] = $3$;\n" // " return $4$;" // "}\n", // funname, // 1 // idx, // 2 // varname, // 3 // retval); // 4 // // but consider using named variables. Named variables like $foo$, with some // identifier foo, are looked up in the map. One additional feature is that // spaces are accepted between the '$' delimiters, $ foo$ will // substiture to " bar" if foo stands for "bar", but in case it's empty // will substitute to "". Hence, for example, // // Format(vars, "$dllexport $void fun();") -> "void fun();" // "__declspec(export) void fun();" // // which is convenient to prevent double, leading or trailing spaces. class PROTOC_EXPORT Formatter { public: explicit Formatter(io::Printer* printer) : printer_(printer) {} Formatter(io::Printer* printer, const std::map& vars) : printer_(printer), vars_(vars) {} template void Set(const std::string& key, const T& value) { vars_[key] = ToString(value); } void AddMap(const std::map& vars) { for (const auto& keyval : vars) vars_[keyval.first] = keyval.second; } template void operator()(const char* format, const Args&... args) const { printer_->FormatInternal({ToString(args)...}, vars_, format); } void Indent() const { printer_->Indent(); } void Outdent() const { printer_->Outdent(); } io::Printer* printer() const { return printer_; } class PROTOC_EXPORT SaveState { public: explicit SaveState(Formatter* format) : format_(format), vars_(format->vars_) {} ~SaveState() { format_->vars_.swap(vars_); } private: Formatter* format_; std::map vars_; }; private: io::Printer* printer_; std::map vars_; // Convenience overloads to accept different types as arguments. static std::string ToString(const std::string& s) { return s; } template ::value>::type> static std::string ToString(I x) { return SimpleItoa(x); } static std::string ToString(strings::Hex x) { return StrCat(x); } static std::string ToString(const FieldDescriptor* d) { return Payload(d); } static std::string ToString(const Descriptor* d) { return Payload(d); } static std::string ToString(const EnumDescriptor* d) { return Payload(d); } static std::string ToString(const EnumValueDescriptor* d) { return Payload(d); } template static std::string Payload(const Descriptor* descriptor) { std::vector path; descriptor->GetLocationPath(&path); GeneratedCodeInfo::Annotation annotation; for (int i = 0; i < path.size(); ++i) { annotation.add_path(path[i]); } annotation.set_source_file(descriptor->file()->name()); return annotation.SerializeAsString(); } }; class PROTOC_EXPORT NamespaceOpener { public: explicit NamespaceOpener(const Formatter& format) : printer_(format.printer()) {} NamespaceOpener(const std::string& name, const Formatter& format) : NamespaceOpener(format) { ChangeTo(name); } ~NamespaceOpener() { ChangeTo(""); } void ChangeTo(const std::string& name) { std::vector new_stack_ = Split(name, "::", true); int len = std::min(name_stack_.size(), new_stack_.size()); int common_idx = 0; while (common_idx < len) { if (name_stack_[common_idx] != new_stack_[common_idx]) break; common_idx++; } for (int i = name_stack_.size() - 1; i >= common_idx; i--) { printer_->Print("} // namespace $ns$\n", "ns", name_stack_[i]); } name_stack_.swap(new_stack_); for (int i = common_idx; i < name_stack_.size(); i++) { printer_->Print("namespace $ns$ {\n", "ns", name_stack_[i]); } } private: io::Printer* printer_; std::vector name_stack_; }; std::string GetUtf8Suffix(const FieldDescriptor* field, const Options& options); void GenerateUtf8CheckCodeForString(const FieldDescriptor* field, const Options& options, bool for_parse, const char* parameters, const Formatter& format); void GenerateUtf8CheckCodeForCord(const FieldDescriptor* field, const Options& options, bool for_parse, const char* parameters, const Formatter& format); template struct FieldRangeImpl { struct Iterator { using iterator_category = std::forward_iterator_tag; using value_type = const FieldDescriptor*; using difference_type = int; value_type operator*() { return descriptor->field(idx); } friend bool operator==(const Iterator& a, const Iterator& b) { GOOGLE_DCHECK(a.descriptor == b.descriptor); return a.idx == b.idx; } friend bool operator!=(const Iterator& a, const Iterator& b) { return !(a == b); } Iterator& operator++() { idx++; return *this; } int idx; const T* descriptor; }; Iterator begin() const { return {0, descriptor}; } Iterator end() const { return {descriptor->field_count(), descriptor}; } const T* descriptor; }; template FieldRangeImpl FieldRange(const T* desc) { return {desc}; } struct OneOfRangeImpl { struct Iterator { using iterator_category = std::forward_iterator_tag; using value_type = const OneofDescriptor*; using difference_type = int; value_type operator*() { return descriptor->oneof_decl(idx); } friend bool operator==(const Iterator& a, const Iterator& b) { GOOGLE_DCHECK(a.descriptor == b.descriptor); return a.idx == b.idx; } friend bool operator!=(const Iterator& a, const Iterator& b) { return !(a == b); } Iterator& operator++() { idx++; return *this; } int idx; const Descriptor* descriptor; }; Iterator begin() const { return {0, descriptor}; } Iterator end() const { return {descriptor->oneof_decl_count(), descriptor}; } const Descriptor* descriptor; }; inline OneOfRangeImpl OneOfRange(const Descriptor* desc) { return {desc}; } void GenerateParserLoop(const Descriptor* descriptor, const Options& options, MessageSCCAnalyzer* scc_analyzer, io::Printer* printer); } // namespace cpp } // namespace compiler } // namespace protobuf } // namespace google #include #endif // GOOGLE_PROTOBUF_COMPILER_CPP_HELPERS_H__