unknown_types_and_raw_parsing.md

Unknown Types and Delayed Parsing

Unknown JSON Data

When the data structure is unknown or partially known it can be parsed as a json_raw type. This also, allows for skipping the parsing of certain members and returning later.

[
  "a",
  2,
  "b",
  3
]

The above JSON describes an array of mixed types, string and number.

To see a working example using this code, refer to cookbook_unknown_types_and_delayed_parsing1_test.cpp.

Below is the C++ code to parse and use it.

std::string_view json_data = ...;
auto val = daw::json::json_value( json_data );
for( auto member : val ) {
  if( member.name ) {
    std::cout << *member.name << " is a ";
  } else {
    std::cout << "Array element is a ";
  }
  std::cout << to_string( member.value.type( ) ) << " with a value of "
            << member.value.get_string_view( ) << '\n';
}

Delayed Parsing

It is possible to delay the parsing of json members until a later time. The json_raw member type will store the character range and allow full parsing later. The cost of doing so is similar to that of skipping the value and is already paid for when parsing JSON data.

{
  "member0": 5,
  "member_later": {
    "a": 1,
    "b": true
  },
  "member1": "Hello world"
}

The above JSON class has 3 members, a number, a class to parse later, and a string.

To see a working example using this code, refer to cookbook_unknown_types_and_delayed_parsing2_test.cpp.

Below is the C++ code to parse and use it.

struct MyClass2 {
  int member0;
  daw::json::json_value member_later;
  std::string member1;
};


struct MyDelayedClass {
  int a;
  bool b;	
};

namespace daw::json {	
  template<>
  struct json_data_contract<MyClass2> {
    using type = json_member_list<
      json_number<"member0", int>,
      json_raw<"member_later">,
      json_string<"member1">
    >;
 
    static inline auto
    to_json_data( MyClass2 const &value ) {
      return std::forward_as_tuple( value.member0, value.member_later, value.member1 );
    }
  }; 

  // This isn't necessary to parse the values as but allows for directly constructing the object
  template<>
  struct json_data_contract<MyDelayedClass> {
    using type = json_member_list<
      json_number<"a", int>,
      json_bool<"b">
    >;
 
    static inline auto
    to_json_data( MyDelayedClass const &value ) {
      return std::forward_as_tuple( value.a, value.b );
    }
  }; 
}

std::string_view json_data = ...;
MyClass2 val = from_json<MyClass2>( json_data );

MyDelayedClass delayed_val = cls2.member_layer.template parse<MyDelayedClass>( );

RAW JSON

Parsing to a raw JSON string can be done too. the json_raw mapping type allows for an optional type argument that specifies the destination/source type. It requires that the type be constructable from a char const * and a std::size_t.

To see a working example using this code, refer to cookbook_unknown_types_and_delayed_parsing3_test.cpp.

struct Thing {
	std::string name;
	int type;
	std::string raw_json;
};

namespace daw::json {
	template<>
	struct json_data_contract<Thing> {
		using type = json_member_list<json_string<"name">, json_number<"type", int>,
		                              json_raw<"raw_json", std::string>>;

		static auto to_json_data( Thing const &value ) {
			return std::forward_as_tuple( value.name, value.type, value.raw_json );
		}
	};
} // namespace daw::json

Thing my_thing = Thing{ "skywalker", 12345, R"({ "name": "last" })" };
auto json_string = daw::json::to_json( my_thing );

json_string will have a value of

{
  "name": "skywalker",
  "type": 12345,
  "raw_json": {
    "name": "last"
  }
}