| // 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 <algorithm> |
| #include <cstdint> |
| #include <iterator> |
| #include <map> |
| #include <string> |
| |
| #include <google/protobuf/compiler/cpp/cpp_options.h> |
| #include <google/protobuf/compiler/cpp/cpp_names.h> |
| #include <google/protobuf/compiler/scc.h> |
| #include <google/protobuf/compiler/code_generator.h> |
| #include <google/protobuf/descriptor.pb.h> |
| #include <google/protobuf/io/printer.h> |
| #include <google/protobuf/descriptor.h> |
| #include <google/protobuf/port.h> |
| #include <google/protobuf/stubs/strutil.h> |
| |
| // Must be included last. |
| #include <google/protobuf/port_def.inc> |
| |
| namespace google { |
| namespace protobuf { |
| namespace compiler { |
| namespace cpp { |
| |
| inline std::string ProtobufNamespace(const Options& /* options */) { |
| return "PROTOBUF_NAMESPACE_ID"; |
| } |
| |
| inline std::string MacroPrefix(const Options& options) { |
| return options.opensource_runtime ? "GOOGLE_PROTOBUF" : "GOOGLE_PROTOBUF"; |
| } |
| |
| inline std::string DeprecatedAttribute(const Options& /* options */, |
| const FieldDescriptor* d) { |
| return d->options().deprecated() ? "PROTOBUF_DEPRECATED " : ""; |
| } |
| |
| inline std::string DeprecatedAttribute(const Options& /* options */, |
| const EnumValueDescriptor* d) { |
| return d->options().deprecated() ? "PROTOBUF_DEPRECATED_ENUM " : ""; |
| } |
| |
| // Commonly-used separator comments. Thick is a line of '=', thin is a line |
| // of '-'. |
| extern const char kThickSeparator[]; |
| extern const char kThinSeparator[]; |
| |
| void SetCommonVars(const Options& options, |
| std::map<std::string, std::string>* variables); |
| |
| void SetUnknkownFieldsVariable(const Descriptor* descriptor, |
| const Options& options, |
| std::map<std::string, std::string>* 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 |
| // "<namespace>::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 FileDescriptor* d, const Options& options); |
| std::string Namespace(const Descriptor* d, const Options& options); |
| std::string Namespace(const FieldDescriptor* d, const Options& options); |
| std::string Namespace(const EnumDescriptor* d, const Options& options); |
| |
| // 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, const Options& options); |
| std::string QualifiedClassName(const EnumDescriptor* d, const Options& options); |
| |
| 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, Options()) |
| : ClassName(descriptor); |
| } |
| |
| inline std::string ClassName(const EnumDescriptor* descriptor, bool qualified) { |
| return qualified ? QualifiedClassName(descriptor, Options()) |
| : ClassName(descriptor); |
| } |
| |
| // Returns the extension name prefixed with the class name if nested but without |
| // the package name. |
| std::string ExtensionName(const FieldDescriptor* d); |
| |
| std::string QualifiedExtensionName(const FieldDescriptor* d, |
| const Options& options); |
| std::string QualifiedExtensionName(const FieldDescriptor* d); |
| |
| // Type name of default instance. |
| std::string DefaultInstanceType(const Descriptor* descriptor, |
| const Options& options); |
| |
| // Non-qualified name of the default_instance of this message. |
| std::string DefaultInstanceName(const Descriptor* descriptor, |
| const Options& options); |
| |
| // Non-qualified name of the default instance pointer. This is used only for |
| // implicit weak fields, where we need an extra indirection. |
| std::string DefaultInstancePtr(const Descriptor* descriptor, |
| const Options& options); |
| |
| // Fully qualified name of the default_instance of this message. |
| std::string QualifiedDefaultInstanceName(const Descriptor* descriptor, |
| const Options& options); |
| |
| // Fully qualified name of the default instance pointer. |
| std::string QualifiedDefaultInstancePtr(const Descriptor* descriptor, |
| const Options& options); |
| |
| // DescriptorTable variable name. |
| std::string DescriptorTableName(const FileDescriptor* file, |
| const Options& options); |
| |
| // When declaring symbol externs from another file, this macro will supply the |
| // dllexport needed for the target file, if any. |
| std::string FileDllExport(const FileDescriptor* file, const Options& options); |
| |
| // Name of the base class: google::protobuf::Message or google::protobuf::MessageLite. |
| std::string SuperClassName(const Descriptor* descriptor, |
| const Options& options); |
| |
| // Adds an underscore if necessary to prevent conflicting with a keyword. |
| std::string ResolveKeyword(const std::string& name); |
| |
| // 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); |
| |
| // 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, |
| const Options& options); |
| |
| // 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_t 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 FileDescriptor* file, |
| const std::string& name, |
| const Options& options); |
| |
| // 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 <file>, respecting <options.enforce_lite>. |
| 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; |
| } |
| |
| inline bool IsFieldUsed(const FieldDescriptor* /* field */, const Options& /* options */) { |
| return true; |
| } |
| |
| // Returns true if "field" is stripped. |
| inline bool IsFieldStripped(const FieldDescriptor* /*field*/, |
| const Options& /*options*/) { |
| return false; |
| } |
| |
| // 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(); |
| } |
| |
| inline bool IsProto2MessageSet(const Descriptor* descriptor, |
| const Options& options) { |
| return !options.opensource_runtime && |
| options.enforce_mode != EnforceOptimizeMode::kLiteRuntime && |
| !options.lite_implicit_weak_fields && |
| descriptor->options().message_set_wire_format() && |
| descriptor->full_name() == "google.protobuf.bridge.MessageSet"; |
| } |
| |
| 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 IsProto3(const FileDescriptor* file) { |
| return file->syntax() == FileDescriptor::SYNTAX_PROTO3; |
| } |
| |
| inline bool HasHasbit(const FieldDescriptor* field) { |
| // This predicate includes proto3 message fields only if they have "optional". |
| // Foo submsg1 = 1; // HasHasbit() == false |
| // optional Foo submsg2 = 2; // HasHasbit() == true |
| // This is slightly odd, as adding "optional" to a singular proto3 field does |
| // not change the semantics or API. However whenever any field in a message |
| // has a hasbit, it forces reflection to include hasbit offsets for *all* |
| // fields, even if almost all of them are set to -1 (no hasbit). So to avoid |
| // causing a sudden size regression for ~all proto3 messages, we give proto3 |
| // message fields a hasbit only if "optional" is present. If the user is |
| // explicitly writing "optional", it is likely they are writing it on |
| // primitive fields also. |
| return (field->has_optional_keyword() || field->is_required()) && |
| !field->options().weak(); |
| } |
| |
| // 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 FieldDescriptor* field) { |
| return field->file()->syntax() == FileDescriptor::SYNTAX_PROTO3; |
| } |
| |
| inline bool IsCrossFileMessage(const FieldDescriptor* field) { |
| return field->type() == FieldDescriptor::TYPE_MESSAGE && |
| field->message_type()->file() != field->file(); |
| } |
| |
| inline std::string MakeDefaultName(const FieldDescriptor* field) { |
| return "_i_give_permission_to_break_this_code_default_" + FieldName(field) + |
| "_"; |
| } |
| |
| bool IsAnyMessage(const FileDescriptor* descriptor, const Options& options); |
| bool IsAnyMessage(const Descriptor* descriptor, const Options& options); |
| |
| bool IsWellKnownMessage(const FileDescriptor* descriptor); |
| |
| inline std::string IncludeGuard(const FileDescriptor* file, bool pb_h, |
| const Options& options) { |
| // If we are generating a .pb.h file and the proto_h option is enabled, then |
| // the .pb.h gets an extra suffix. |
| std::string filename_identifier = FilenameIdentifier( |
| file->name() + (pb_h && options.proto_h ? ".pb.h" : "")); |
| |
| if (IsWellKnownMessage(file)) { |
| // For well-known messages we need third_party/protobuf and net/proto2 to |
| // have distinct include guards, because some source files include both and |
| // both need to be defined (the third_party copies will be in the |
| // google::protobuf_opensource namespace). |
| return MacroPrefix(options) + "_INCLUDED_" + filename_identifier; |
| } else { |
| // Ideally this case would use distinct include guards for opensource and |
| // google3 protos also. (The behavior of "first #included wins" is not |
| // ideal). But unfortunately some legacy code includes both and depends on |
| // the identical include guards to avoid compile errors. |
| // |
| // We should clean this up so that this case can be removed. |
| return "GOOGLE_PROTOBUF_INCLUDED_" + filename_identifier; |
| } |
| } |
| |
| // Returns the OptimizeMode for this file, furthermore it updates a status |
| // bool if has_opt_codesize_extension is non-null. If this status bool is true |
| // it means this file contains an extension that itself is defined as |
| // optimized_for = CODE_SIZE. |
| FileOptions_OptimizeMode GetOptimizeFor(const FileDescriptor* file, |
| const Options& options, |
| bool* has_opt_codesize_extension); |
| inline FileOptions_OptimizeMode GetOptimizeFor(const FileDescriptor* file, |
| const Options& options) { |
| return GetOptimizeFor(file, options, nullptr); |
| } |
| inline bool NeedsEagerDescriptorAssignment(const FileDescriptor* file, |
| const Options& options) { |
| bool has_opt_codesize_extension; |
| if (GetOptimizeFor(file, options, &has_opt_codesize_extension) == |
| FileOptions::CODE_SIZE && |
| has_opt_codesize_extension) { |
| // If this filedescriptor contains an extension from another file which |
| // is optimized_for = CODE_SIZE. We need to be careful in the ordering so |
| // we eagerly build the descriptors in the dependencies before building |
| // the descriptors of this file. |
| return true; |
| } else { |
| // If we have a generated code based parser we never need eager |
| // initialization of descriptors of our deps. |
| return false; |
| } |
| } |
| |
| // This orders the messages in a .pb.cc as it's outputted by file.cc |
| void FlattenMessagesInFile(const FileDescriptor* file, |
| std::vector<const Descriptor*>* result); |
| inline std::vector<const Descriptor*> FlattenMessagesInFile( |
| const FileDescriptor* file) { |
| std::vector<const Descriptor*> 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; |
| }; |
| |
| // 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: |
| struct DepsGenerator { |
| std::vector<const Descriptor*> operator()(const Descriptor* desc) const { |
| std::vector<const Descriptor*> deps; |
| for (int i = 0; i < desc->field_count(); i++) { |
| if (desc->field(i)->message_type()) { |
| deps.push_back(desc->field(i)->message_type()); |
| } |
| } |
| return deps; |
| } |
| }; |
| SCCAnalyzer<DepsGenerator> analyzer_; |
| Options options_; |
| std::map<const SCC*, MessageAnalysis> analysis_cache_; |
| }; |
| |
| void ListAllFields(const Descriptor* d, |
| std::vector<const FieldDescriptor*>* fields); |
| void ListAllFields(const FileDescriptor* d, |
| std::vector<const FieldDescriptor*>* fields); |
| |
| template <class T> |
| void ForEachField(const Descriptor* d, T&& func) { |
| for (int i = 0; i < d->nested_type_count(); i++) { |
| ForEachField(d->nested_type(i), std::forward<T&&>(func)); |
| } |
| for (int i = 0; i < d->extension_count(); i++) { |
| func(d->extension(i)); |
| } |
| for (int i = 0; i < d->field_count(); i++) { |
| func(d->field(i)); |
| } |
| } |
| |
| template <class T> |
| void ForEachField(const FileDescriptor* d, T&& func) { |
| for (int i = 0; i < d->message_type_count(); i++) { |
| ForEachField(d->message_type(i), std::forward<T&&>(func)); |
| } |
| for (int i = 0; i < d->extension_count(); i++) { |
| func(d->extension(i)); |
| } |
| } |
| |
| void ListAllTypesForServices(const FileDescriptor* fd, |
| std::vector<const Descriptor*>* 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<std::string, std::string>& vars) |
| : printer_(printer), vars_(vars) {} |
| |
| template <typename T> |
| void Set(const std::string& key, const T& value) { |
| vars_[key] = ToString(value); |
| } |
| |
| void AddMap(const std::map<std::string, std::string>& vars) { |
| for (const auto& keyval : vars) vars_[keyval.first] = keyval.second; |
| } |
| |
| template <typename... Args> |
| 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<std::string, std::string> vars_; |
| }; |
| |
| private: |
| io::Printer* printer_; |
| std::map<std::string, std::string> vars_; |
| |
| // Convenience overloads to accept different types as arguments. |
| static std::string ToString(const std::string& s) { return s; } |
| template <typename I, typename = typename std::enable_if< |
| std::is_integral<I>::value>::type> |
| static std::string ToString(I x) { |
| return StrCat(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); |
| } |
| static std::string ToString(const OneofDescriptor* d) { return Payload(d); } |
| |
| template <typename Descriptor> |
| static std::string Payload(const Descriptor* descriptor) { |
| std::vector<int> path; |
| descriptor->GetLocationPath(&path); |
| GeneratedCodeInfo::Annotation annotation; |
| for (int index: path) { |
| annotation.add_path(index); |
| } |
| annotation.set_source_file(descriptor->file()->name()); |
| return annotation.SerializeAsString(); |
| } |
| }; |
| |
| template <class T> |
| void PrintFieldComment(const Formatter& format, const T* field) { |
| // Print the field's (or oneof's) proto-syntax definition as a comment. |
| // We don't want to print group bodies so we cut off after the first |
| // line. |
| DebugStringOptions options; |
| options.elide_group_body = true; |
| options.elide_oneof_body = true; |
| std::string def = field->DebugStringWithOptions(options); |
| format("// $1$\n", def.substr(0, def.find_first_of('\n'))); |
| } |
| |
| 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<std::string> new_stack_ = |
| Split(name, "::", true); |
| size_t len = std::min(name_stack_.size(), new_stack_.size()); |
| size_t common_idx = 0; |
| while (common_idx < len) { |
| if (name_stack_[common_idx] != new_stack_[common_idx]) break; |
| common_idx++; |
| } |
| for (auto it = name_stack_.crbegin(); it != name_stack_.crend() - common_idx; ++it) { |
| if (*it == "PROTOBUF_NAMESPACE_ID") { |
| printer_->Print("PROTOBUF_NAMESPACE_CLOSE\n"); |
| } else { |
| printer_->Print("} // namespace $ns$\n", "ns", *it); |
| } |
| } |
| name_stack_.swap(new_stack_); |
| for (size_t i = common_idx; i < name_stack_.size(); ++i) { |
| if (name_stack_[i] == "PROTOBUF_NAMESPACE_ID") { |
| printer_->Print("PROTOBUF_NAMESPACE_OPEN\n"); |
| } else { |
| printer_->Print("namespace $ns$ {\n", "ns", name_stack_[i]); |
| } |
| } |
| } |
| |
| private: |
| io::Printer* printer_; |
| std::vector<std::string> name_stack_; |
| }; |
| |
| enum class Utf8CheckMode { |
| kStrict = 0, // Parsing will fail if non UTF-8 data is in string fields. |
| kVerify = 1, // Only log an error but parsing will succeed. |
| kNone = 2, // No UTF-8 check. |
| }; |
| |
| Utf8CheckMode GetUtf8CheckMode(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 <typename T> |
| 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 <typename T> |
| FieldRangeImpl<T> 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->real_oneof_decl_count(), descriptor}; |
| } |
| |
| const Descriptor* descriptor; |
| }; |
| |
| inline OneOfRangeImpl OneOfRange(const Descriptor* desc) { return {desc}; } |
| |
| PROTOC_EXPORT std::string StripProto(const std::string& filename); |
| |
| } // namespace cpp |
| } // namespace compiler |
| } // namespace protobuf |
| } // namespace google |
| |
| #include <google/protobuf/port_undef.inc> |
| |
| #endif // GOOGLE_PROTOBUF_COMPILER_CPP_HELPERS_H__ |