A Universal Serialization Engine Based on compile-time Reflection

iguana is a modern, universal and easy-to-use serialization engine developed in C++20 and C++17.

OS (Compiler Version)	Status
Ubuntu 22.04 (clang 14.0.0)
Ubuntu 22.04 (gcc 11.2.0)
GCC 16.1.0 container (C++26 reflection)
macOS Monterey latest (AppleClang latest)
Windows Server 2022 (MSVC 19.33.31630.0)

线上讨论: 项目讨论

中文版

struct_pb

C++26 reflection/proto3 change notes

Motivation

Serialize an object to any other format data with compile-time reflection, such as json, xml, binary, table and so on. This library is designed to unify and simplify serialization in a portable cross-platform manner. This library is also easy to extend, and you can serialize any format of data with the library. This library provides a portable cross-platform way of:

• serialization of json
• serialization of xml
• serialization of yaml
• serialization of protobuf
• serialization of any customized format

compile time reflection

reflection lib introduction

With a C++26 static reflection compiler, iguana can read class members directly from the language reflection API. The CI job for this path uses the official gcc:16.1.0-trixie container with -std=gnu++26 -freflection and configures CMake with -DENABLE_CXX26_REFLECTION=ON.

Useful C++26 annotations:

struct [[= ylt::reflection::struct_name<"user">{}]] user_t {
  [[= ylt::reflection::field_name<"id">{}]]
  int user_id{};

  std::string name;

  [[= ylt::reflection::skip_field{}]]
  int local_cache{};
};

struct base_t {
  int internal_state{};
};

struct derived_t : [[= ylt::reflection::skip_base{}]] base_t {
  int value{};
};

struct xml_user_t {
  [[= ylt::reflection::field_name<"identifier">{}]]
  [[= iguana::xml_required{}]]
  int id{};
};

field_name changes the serialized field name for JSON/XML/YAML and generated protobuf schema. struct_name changes XML root names. skip_field excludes a member from reflection, and skip_base excludes a base class from recursive member collection. xml_required is the C++26 annotation form of the XML REQUIRED(type, fields...) macro.

Tutorial

This Tutorial is provided to give you a view of how iguana works for serialization.

Serialization of json

The first thing to do when you serialize an object is to define meta data. There is an example of defining meta data.

struct person
{
    std::string  name;
    int          age;
};
#if __cplusplus < 202002L
YLT_REFL(person, name, age) //define meta data
#endif

Defining meta data is very simple, if your compiler is C++20 compiler(gcc11+, clang13+, msvc2022), no need define YLT_REFL, other wise need to define in a YLT_REFL macro.

Now let’s serialize person to json string.

person p = { "tom", 28 };

iguana::string_stream ss; // here use std::string is also ok
iguana::to_json(p, ss);

std::cout << ss.str() << std::endl; 

This example will output:

{"name":"tom","age":28}

Serializing person to json string is also very simple, just need to call to_json method, there is nothing more.

How about deserialization of json? Look at the follow example.

std::string json = "{ \"name\" : \"tom\", \"age\" : 28}";

person p;
iguana::from_json(p, json);

It’s as simple as serialization, just need to call from_json method.

You can also use parse interface to do dom parsing:

std::string_view str = R"(false)";
iguana::jvalue val;
iguana::parse(val, str.begin(), str.end());

std::error_code ec;
auto b = val.get<bool>(ec);
CHECK(!ec);
CHECK(!b);

// or
b = val.get<bool>(); // this interface maybe throw exception
CHECK(!b);

Serialization of xml

The serialization of xml is similar to json. The first step is also defining meta data as above, and then you can call iguana::to_xml to serialization the structure, or call iguana::from_xml to deserialization the structure. The following is a complete example.

// serialization the structure to the string
person p = {"admin", 20};
iguana::string_stream ss;  // here use std::string is also ok
iguana::to_xml(p, ss);
std::cout << ss << std::endl;

// deserialization the structure from the string
std::string xml = R"(
<?xml version=\"1.0\" encoding=\"UTF-8\">  
<root> 
  <name>buke</name> 
  <age>30</age> 
</root>)";
iguana::from_xml(p, xml);

Serialization of yaml

The serialization of yaml is also as simple as the above interface. Here is a complete example:

// serialization the structure to the string
person p = {"admin", 20};
iguana::string_stream ss;  // here use std::string is also ok
iguana::to_yaml(ss, p);
std::cout << ss.str() << std::endl;
std::string yaml = R"(
name : buke
age : 30
)";
// deserialization the structure from the string
iguana::from_yaml(p, yaml);

A complicated example

json

iguana can deal with objects which contain another objects and containers. Here is the example:

At first, we define the meta data:

struct one_t
{
    int id;
};
YLT_REFL(one_t, id);

struct two
{
    std::string name;
    one_t one;
    int age;
};
YLT_REFL(two, name, one, age);

struct composit_t
{
    int a;
    std::vector<std::string> b;
    int c;
    std::map<int, int> d;
    std::unordered_map<int, int> e;
    double f;
    std::list<one_t> g;
};
YLT_REFL(composit_t, a, b, c, d, e, f, g);

Then call the simple interface:

one_t one = { 2 };
composit_t composit = { 1,{ "tom", "jack" }, 3,{ { 2,3 } },{ { 5,6 } }, 5.3,{ one } };
iguana::string_stream ss;
iguana::to_json(composit, ss);
std::cout << ss.str() << std::endl;

std::string str_comp = R"({"a":1, "b":["tom", "jack"], "c":3, "d":{"2":3,"5":6},"e":{"3":4},"f":5.3,"g":[{"id":1},{"id":2}])";
composit_t comp;
iguana::from_json(comp, str_comp);

xml

At first, define the structure and reflect the meta data.

struct book_t {
  std::string author;
  float price;
};
YLT_REFL(book_t, author, price);
struct library_t {
  std::string name;
  int id;
  std::vector<book_t> book;
};
YLT_REFL(library_t, name, id, book);

And then, simply call the interface:

// serialization the structure to the string
library_t library = {"Pro Lib", 110, {{"tom", 1.8}, {"jack", 2.1}}};
std::string ss;
iguana::to_xml(library, ss);
std::cout << ss << "\n";

// deserialization the structure from the string
std::string str = R"(
<library>
  <name>Pro Lib</name>
  <id>110</id>
  <book><author>tom</author><price>1.8</price>
  </book>
  <book><author>jack</author><price>2.1</price>
  </book>
</library>
)";
library_t lib;
iguana::from_xml(lib, str);

yaml

As always what we do, define the structure and reflect the meta data.

enum class enum_status {
  start,
  stop,
};
struct plain_type_t {
  bool isok;
  enum_status status;
  char c;
  std::optional<bool> hasprice;
  std::optional<float> num;
  std::optional<int> price;
};
YLT_REFL(plain_type_t, isok, status, c, hasprice, num, price);

And then, simply call the interface:

// deserialization the structure from the string
std::string str = R"(
isok: false
status: 1
c: a
hasprice: true
num:
price: 20
)";
plain_type_t p;
iguana::from_yaml(p, str);
// serialization the structure to the string
std::string ss;
iguana::to_yaml(ss, p);
std::cout << ss << "\n";

How to solve the problem of unicode path in a json file?

If there is an unicode string as a path in a json file, however iguana parse the file as utf-8, so maybe you can see some strange characters after parse.

It’s ok, because you see the utf-8 strings. The problem is you can’t use the string directly, such as use std::ifstream to open the file with the unicode string path.

We can slove the problem1 easily with c++17:

  //the p.path is a unicode string path
  std::ifstream in(std::filesystem::u8path(p.path)); //std::filesystem::u8path help us
  //now you can operate the file

how to handle the enum type as strings?

By default, Iguana handle enum type as number type during serialization and deserialization. To handle the enum type as strings during serialization and deserialization with Iguana, we need to define a full specialization template in the “iguana” namespace. This template is a struct that contains an array with the underlying numbers corresponding to the enum type. For example, if we have the following enum type:

enum class Status { STOP = 10, START };

And we want to handle the enum type as strings when parsing JSON:

    std::string str = R"(
{
  "a": "START",
  "b": "STOP"
}
  )";

To do this, we define the full specialization template in the “iguana” namespace:

namespace iguana {
template <> struct enum_value<Status> {
  constexpr static std::array<int, 2> value = {10, 11};
};
} // namespace iguana

Once this is done, we can continue writing the rest of the code as usual.

struct enum_t {
    Status a;
    Status b;
};
YLT_REFL(enum_t, a, b);

// deserialization
enum_t e;
iguana::from_json(e, str);
// serialization
enum_t e1;
e1.a = Status::START;
e1.b = Status::STOP;
std::string ss;
iguana::to_json(e1, ss);

Serialization of protobuf

Basic protobuf serialization uses the same object API:

struct person {
  int id;
  std::string name;
  int age;
  bool operator==(person const& rhs) const {
    return id == rhs.id && name == rhs.name && age == rhs.age;
  }
};

#if __cplusplus < 202002L
YLT_REFL(person, id, name, age) //define meta data
#endif

void test() {
  person p{1, "tom", 20};
  std::string pb;
  iguana::to_pb(p, pb);
  person p1;
  iguana::from_pb(p1, pb);
  CHECK(p == p1);
}

By default, protobuf field numbers follow member order. For stable schemas or interop with existing .proto files, specify field numbers explicitly. On the legacy/non-C++26 path, YLT_REFL_PB remains available:

struct account {
  std::string name;
  int32_t age;
  std::vector<std::string> emails;
};

YLT_REFL_PB(account, (name, 10), (age, 20), (emails, 9));

With C++26 static reflection, prefer the [[= iguana::pb_field(N)]] annotation shown below; that path reads protobuf metadata from annotations and does not depend on YLT_REFL_PB.

For advanced proto3 wire semantics on the non-C++26 path, use the descriptor helpers. The helper form keeps normal C++ field types while attaching protobuf schema metadata:

struct event_msg {
  int32_t id{};
  std::string payload;
  int32_t delta{};
  uint32_t checksum{};
  std::chrono::system_clock::time_point created_at{};
  std::chrono::nanoseconds timeout{};
  std::optional<int32_t> retry_count;
  std::variant<std::monostate, int32_t, std::string> result;
  std::string unknown;
};

inline auto get_members_impl(event_msg*) {
  return iguana::pb_members(
      iguana::pb_field<&event_msg::id, 1>("id"),
      iguana::pb_bytes_field<&event_msg::payload, 3>("payload"),
      iguana::pb_zigzag_field<&event_msg::delta, 5>("delta"),
      iguana::pb_optional_field<&event_msg::retry_count, 6>("retry_count"),
      iguana::pb_fixed_field<&event_msg::checksum, 7>("checksum"),
      iguana::as_timestamp_field<&event_msg::created_at, 8>("created_at"),
      iguana::as_duration_field<&event_msg::timeout, 9>("timeout"),
      iguana::pb_oneof_field<&event_msg::result, 10, 12>("result"),
      iguana::pb_unknown_fields_field<&event_msg::unknown>("unknown"));
}

Helper APIs:

Helper	Meaning
pb_members(...)	Returns the protobuf descriptor tuple from get_members_impl(T*).
pb_field<&T::field, N>("name")	Sets a protobuf field number and schema name.
pb_bytes_field	Emits bytes; the C++ field is std::string or std::string_view, including optional/vector forms.
pb_zigzag_field	Emits sint32 or sint64; the C++ field remains int32_t or int64_t, including optional/vector forms.
pb_fixed_field	Emits fixed32, fixed64, sfixed32, or sfixed64 for 32/64-bit integer fields, including optional/vector forms.
pb_optional_field	Emits proto3 optional; the C++ field must be std::optional<T>.
pb_timestamp_field / as_timestamp_field	Encodes std::chrono::system_clock::time_point as google.protobuf.Timestamp, including optional/vector forms.
pb_duration_field / as_duration_field	Encodes std::chrono::nanoseconds as google.protobuf.Duration, including optional/vector forms.
pb_oneof_field<&T::field, Ns...>("name")	Maps std::variant<std::monostate, ...> alternatives to oneof field numbers.
pb_unknown_fields_field<&T::field>()	Preserves unknown protobuf wire bytes in a single std::string field.

The explicit wrapper types iguana::pb_timestamp and iguana::pb_duration remain available when code wants the wire-shaped representation directly.

pb_field_ex can combine options. Supported options are pb_bytes, pb_zigzag, pb_fixed, pb_optional, pb_as_timestamp/as_timestamp, and pb_as_duration/as_duration.

iguana::pb_field_ex<&event_msg::retry_count, 6>(
    "retry_count", iguana::pb_optional, iguana::pb_zigzag);

With a C++26 reflection compiler, the same metadata can be written as annotations without YLT_REFL_PB. The current C++26 test build uses GCC 16.1 with -std=gnu++26 -freflection.

struct event_msg26 {
  [[= iguana::pb_field(1)]] int32_t id{};

  [[= iguana::pb_field(3)]]
  [[= iguana::pb_bytes]]
  std::string payload;

  [[= iguana::pb_field(5)]]
  [[= iguana::pb_zigzag]]
  int32_t delta{};

  [[= iguana::pb_field(6)]]
  [[= iguana::pb_optional]]
  std::optional<int32_t> retry_count;

  [[= iguana::pb_field(7)]]
  [[= iguana::pb_fixed]]
  uint32_t checksum{};

  [[= iguana::pb_field(8)]]
  [[= iguana::as_timestamp]]
  std::chrono::system_clock::time_point created_at{};

  [[= iguana::pb_field(9)]]
  [[= iguana::as_duration]]
  std::chrono::nanoseconds timeout{};

  [[= iguana::pb_oneof<10, 12>]]
  std::variant<std::monostate, int32_t, std::string> result;

  [[= iguana::pb_unknown_fields]]
  std::string unknown;
};

C++26 annotation equivalents:

Annotation	Meaning
[[= iguana::pb_field(N)]]	Sets the protobuf field number.
[[= iguana::pb_bytes]]	Same as pb_bytes_field.
[[= iguana::pb_zigzag]]	Same as pb_zigzag_field.
[[= iguana::pb_fixed]]	Same as pb_fixed_field.
[[= iguana::pb_optional]]	Same as pb_optional_field.
[[= iguana::pb_oneof<N...>]] / [[= iguana::oneof<N...>]]	Same as pb_oneof_field.
[[= iguana::as_timestamp]] / [[= iguana::pb_as_timestamp]]	Same as as_timestamp_field / pb_timestamp_field.
[[= iguana::as_duration]] / [[= iguana::pb_as_duration]]	Same as as_duration_field / pb_duration_field.
[[= iguana::pb_unknown_fields]]	Same as pb_unknown_fields_field.

Supported proto3 wire metadata includes custom field numbers, bytes, sint32/sint64 zigzag encoding, fixed-width integers, explicit optional presence, oneof, google.protobuf.Timestamp, google.protobuf.Duration, and unknown field preservation. Repeated primitive fields accept packed, unpacked, and mixed input; writers use proto3 default packed output where applicable.

Field numbers must be in [1, 2^29 - 1] and cannot be in protobuf’s reserved [19000, 19999] range. A message can have at most one unknown-field storage member, and it must be a std::string.

Generate a .proto view of a struct with:

std::string schema;
iguana::to_proto<event_msg>(schema, "demo");

The current conformance target covers the proto3 binary/protobuf-output wire-only subset. JSON mapping, text format, proto2, extensions, services, and custom options are outside this scope.

more detail and change notes

Full sources:

• More examples about json
• More examples about xml
• More examples about yaml
• More examples about struct_pb

Scripts

Automatically generate YLT_REFL macros based by struct.

To get a list of basic options and switches use:

python automatic_macro_generator.py -h

basic example:

The content of the test_macro_generator.cpp is as follows:

 struct person {
    std::string name;
    int age;
  };

  char *iguana = NULL;

  struct composit_t { int a; std::vector<std::string> b; int c; std::map<int, int> d; std::unordered_map<int, int> e; double f;};

  char *iguana_test = NULL;

  struct composit_t2
  {
    int a;
    std::vector<std::string> b;
    int iguana;
    std::map<int, int> example_test;
    std::unordered_map<int, int> random_name__;
    double __f__number__complex;
  };

execute script:

python automatic_macro_generator.py -i test_macro_generator.cpp

After processing by the automatic_macro_generator.py script,test_macro_generator.cpp change into:

struct person {
  std::string name;
  int age;
};
YLT_REFL(person, name, age);
char *iguana = NULL;

struct composit_t { int a; std::vector<std::string> b; int c; std::map<int, int> d; std::unordered_map<int, int> e; double f;};
YLT_REFL(composit_t, a, b, c, d, e, f);

struct composit_t2
{
  int a;
  std::vector<std::string> b;
  int iguana;
  std::map<int, int> example_test;
  std::unordered_map<int, int> random_name__;
  double __f__number__complex;
};
YLT_REFL(composit_t2, a, b, iguana, example_test, random_name__, __f__number__complex);

other example:

python automatic_macro_generator.py -i test_macro_generator.cpp -o have_macro.cpp

test_macro_generator.cpp will be unchanged, have_macro.cpp will be changed to source file with YLT_REFL macro.

scripts works out of the box with Python version 2.7 and 3.x on any platform.

Notes: In Python3,Will prompt DeprecationWarning: 'U' mode is deprecated.Ignore it.

F.A.Q

• Question: Why is the library called iguana?

  • Answer: I think serialization is like an iguana, because the only difference is the displaying format, however the meta data is never changed. With changeless meta data and YLT_REFL, you can serialize an object to any format, which is like how an iguana does.

• Question: Does iguana support raw pointer?

  • Answer: No. iguana doesn’t support raw pointer, but it will support smart pointer in the future.

• Question: Is iguana thread-safe?

  • Answer: Not yet, but it’s not a problem, you can use lock before calling from_json or to_json.

• Question: Is iguana high performance?

  • Answer: Yes, it is, because iguana is based on compile-time reflection.

• Question: I found a bug, how could I report?

  • Answer: Create an issue on GitHub with a detailed description.

deps

frozen lib

Update

1. Support C++20 and C++17
2. Refactor json reader, modification based on glaze json/read.hpp