Description
This is a high-performance JSON library written in C++ 11 with the following features:
- header only: no third-party dependencies;
- fast: parsing speed is half of simdjson, much faster than rapidjson;
- simple api: the common interface is similar to nlohmann / json;
- modularity: You can customize various sub-modules to meet customized needs;
- json value type extension: You can extend the new JSON basic type by yourself;
- support types serialization: You can use JSON as an any class;
Supported compilers
- GCC 4.8.5 or later
- Clang 3.4 or later
- Microsoft Visual C++ 2015 or later
Performance
- Run the test program in the "./tests" directory to see the performance comparison between this library and rapidjson;
- Direct manipulation of raw data can also improve performance
Examples
Read the test code in the tests directory for more usage. Basic usage:
#include "json.hpp"
#include <iostream>
int main(int argc, char** argv) {
amo::json json_empty; // json_value_empty
amo::json json_string = "ttt"; // json_value_string
amo::json json_number = 12.33; // json_value_float
json_number = 3; // json_value_signed
json_number = 3u; // json_value_unsigned
amo::json json_boolean = true; // json_value_boolean
amo::json json_null = nullptr; // json_value_null;
amo::json json_object_{ amo::json_value_object }; // json_value_object;
amo::json json_array_{ amo::json_value_array }; // json_value_array;
amo::json json; // empty json
json["string"] = "amo::json"; // {"string": "amo::json"}
json["string"] = "amo.json"; // {"string": "amo.json"}
json["array"] = { 1, 0, 0, 0 }; // {"string": "amo.json", "array": [1,0,0,0]};
json["sub"] = { {"a", 1}, {"b", 1.1} }; // {"string": "amo.json", "array": [1,0,0,0], "sub":{"a": 1, "b": 1.1} };
std::string str = json.to_string(); // to string
std::cout << str << std::endl;
std::string s = json["string"]; // get string from json;
amo::json json2 = amo::string_reader(str); // parse json from string
bool has_string = json2.has("string"); // has key
if (json2["sub"].is_object()) { // is json object
amo::json sub = json2["sub"]; // ref from json["sub"];
amo::json sub2 = json2["sub"].get<amo::json>(); // copy from json["sub"];
sub["a"] = 10.23;
sub["c"] = true;
std::cout << "[sub][a]=" << json2["sub"]["a"].get<int>() << std::endl; // get<T>()
std::cout << "[sub][a]=" << json2["sub"]["a"].get<double>() << std::endl; // get<T>()
std::cout << "[sub][b]=" << json2["sub"]["b"] << std::endl; // operator T()
std::cout << "[sub][c]=" << json2["sub"]["c"] << std::endl; // operator T()
if (sub2.has("c")) {
// never execute
std::cout << "[sub][c]=" << json2["sub"]["c"] << std::endl; // operator <<
}
}
if (json2["array"].is(amo::json_value_array)) { // is json array
amo::json array = json2["array"]; // ref from json2["array"]
array[-1] = 3; // push back
array.push_back(1); // push back
array[0] = 2; // modify value at index 0
}
std::cout << json2.to_string(4) << std::endl; // prettify
json2.to_file("./xxx.json"); // write to file
amo::json json3 = amo::file_reader("./xxx.json"); // parse json from file
amo::json arr; // empty json
// arr.set_array(); // set to array
arr.push_back(0);
arr.push_back({1, 2, 3, 4, 5, false, true, "string txt" }); // push back with initializer_list
std::vector<int> vec{9, 10, 11, 12 };
arr.push_back(vec.begin(), vec.end()); // push back with iterator
// transfer json array
for (const amo::json& p : arr) {
std::cout << p << std::endl;
}
// transfer json array with array index
arr.transfer<int>([](const int& index, amo::json & p) {
std::cout << index << ": " << p << std::endl;
});
// transfer json object
for (auto& p : json2) {
std::cout << p << std::endl; // without key , only value
}
// transfer json object with key
json2.transfer<amo::json::key_type>([](const amo::json::key_type & key, amo::json & p) {
std::cout << key << ": " << p << std::endl;
});
return 0;
}
Custom json value type:
#include "json.hpp"
class custom_json_value {
public:
const static int __json_type_value__ = 136; // require custom json_value_type must > 128
public:
custom_json_value() {}
custom_json_value(const std::string& str) : m_str(str) {}
std::string to_string() const { return m_str; } // require
static custom_json_value from_string(const std::string& str) { return custom_json_value(str); } // require
std::string m_str;
};
int main(int argc, char** argv){
amo::json json = custom_json_value("123"); // construct json with custom json value
bool b = json.is(custom_json_value::__json_type_value__); //
std::cout << "json value type: " << b << std::endl;
custom_json_value value = json; // get custom_json_value from json
std::cout << value.m_str << std::endl;
custom_json_value* value2 = json; // get custom_json_value* from json
std::cout << value2->m_str << std::endl;
delete value2; // delete pointer
json = "456";
std::shared_ptr<custom_json_value> value3 = json; // smart ptr
std::cout << value3->m_str << std::endl;
// use custom_json_value as a normal json value type
json.set_object();
json["custom_json_value"] = value;
json["custom_json_value3"] = value3;
std::cout << json.to_string() << std::endl;
return 0;
}Serialization / Deserialization:
#include "json.hpp"
#include <iostream>
// simplest class
class custom_json_object {
public:
custom_json_object() {}
custom_json_object(const std::string& str) : m_str(str) {}
amo::json to_json() const {
amo::json json;
json["m_str"] = m_str;
return (json);
}
static custom_json_object* from_json(const amo::json& json, custom_json_object* ptr) {
if (ptr == nullptr) {
ptr = new custom_json_object();
}
ptr->m_str = json["m_str"].get<std::string>();
return ptr;
}
std::string m_str;
};
// use macro
class custom_json_object2 {
public:
typedef amo::json JsonType; // require, json type
public:
custom_json_object2() { m_int = 3; }
AMO_ENTITY_ARGS_GET_BEGIN_WITH_JSON_CONSTRUCTOR(custom_json_object2)
AMO_ENTITY_ARGS_GET(m_int) // get value from json["m_int"] if json["m_int"] is not empty
AMO_ENTITY_ARGS_GET(m_int2)
AMO_ENTITY_ARGS_GET(m_vec)
AMO_ENTITY_ARGS_GET(m_custom_json_object)
AMO_ENTITY_ARGS_GET_END()
AMO_ENTITY_ARGS_SET_BEGIN(custom_json_object2)
AMO_ENTITY_ARGS_SET(m_int) // set value to json["m_int"]
AMO_ENTITY_ARGS_SET(m_int2)
AMO_ENTITY_ARGS_SET(m_vec)
AMO_ENTITY_ARGS_SET(m_custom_json_object)
AMO_ENTITY_ARGS_SET_END()
int m_int; // pod
std::shared_ptr<int> m_int2; // smart ptr
std::vector<int> m_vec; // stl container
std::shared_ptr<custom_json_object> m_custom_json_object; // smart ptr , class
};
int main(int argc, char** argv){
custom_json_object json_class;
json_class.m_str = "3";
amo::json simplejson = json_class;
std::cout << simplejson << std::endl; // {"m_str":"3"}
custom_json_object2 c1;
c1.m_vec.push_back(1);
c1.m_vec.push_back(2);
amo::json json = c1;
std::cout << c1.to_string() << std::endl; // {"m_int":3,"m_vec":[1,2]}
std::cout << json.to_string() << std::endl; // {"m_int":3,"m_vec":[1,2]}
json["m_int2"] = 3;
custom_json_object2 c2 = json;
std::cout << *c2.m_int2 << std::endl; // 3
json["m_custom_json_object"] = custom_json_object("657");
std::shared_ptr<custom_json_object2> c3 = json;
std::cout << c3->m_custom_json_object->m_str << std::endl; // 567
std::cout << c3->to_string() << std::endl;//{"m_int":3,"m_int2":3,"m_vec":[1,2],"m_custom_json_object":{"m_str":"657"}}
return 0;
}
Use STL containers:
// construct from stl container
std::vector<int> vec{1,2,3,4,5,6,7};
amo::json json_vec = vec; // json_value_array;
amo::json json_list = std::list<int>{1,2,3,4}; // json_value_array;
amo::json json_set = std::set<int>{1,2,3,4}; // json_value_array;
amo::json json_deque = std::deque<int>{1,2,3,4}; // json_value_array;
amo::json json_object_map = std::map<std::string, int>{{"a",1}, {"b", 2}}; // json_value_object
amo::json json_arr_map = std::map<int, int>{{1,1},{1,1}}; // json_value_array
amo::json json_vec2 = std::vector<amo::json::object_type>{"1",2,false, 33}
std::vector<int> my_vec = json_vec; // ok
You can directly use native data to improve program performance:
amo::json json;
for (int i =0; i < 10000; ++i) {
json.push_back(i);
}
// Use the find function to avoid temporary JSON objects
for(int i=0; i < 10000; ++i){
json.find(i).set_value(i+1, json.get_allocator()); //
}
// Use the get_data_object function to use native data directly.
for(int i=0; i < 10000; ++i){
json.get_data_object().d.a.elements.at(i) = i+2;
}Defects
- The program uses pre-allocated memory to improve Performance, it will waste some memory;
- JSON will not release any memory during the life cycle, and it will also waste some memory;
- When parsing a JSON string, the keys in key-value pairs are not processed by default;
std::string sb = R"({"\u4E2D\u6587": "\u4e2d\u6587\u5b57\u7b26\u4e32"})"; // u8 {"中文": "中文字符串"}
amo::json json = amo::string_reader(sb);
std::cout << json.to_string() << std::endl; // u8 {"\u4E2D\u6587":"中文字符串"}, key unchanged- When you call the function operator ["key"], if the key-value does not exist, a empty value will be created. This feature will cause the return value of the size function to be inaccurate. You can use has ( "key") to determine whether the corresponding key exists;
amo::json json;
json["a"];
json["b"] = 1;
std::cout << json.size() << std::endl; // 2
std::cout << json.to_string() << std::endl; // {"b": 1}
std::cout << json.has("a") << std::endl; // true
std::cout << json.has("c") << std::endl; // false
- The program uses the fast insert mode by default when parsing a JSON string. It may behave inconsistently when there are duplicate key values in the string. 16 key-values are inserted using vector.push_back, and later key-values are used map.insert; you can modify the configuration to filter duplicate key-value pairs;
std::string str = R"({"a": 1, "a": 2, "a": 3, "a":4, "a": 5, "a": 6, "a": 7, "a": 8, "a": 9, "a": 10, "a": 11, "a": 12, "a": 13, "a": 14, "a": 15, "a": 16, "a": 17, "a": 18, "a": 19, "a": 20 , "b": 1, "b":2 })";
amo::json json = amo::string_reader(str);
std::cout << json["a"] << std::endl; // 1 index < 16 use vector.push_back
std::cout << json["b"] << std::endl; // 2 index >= 16 use map.insert()
std::cout << json.to_string() << std::endl; // {"a": 1, "a": 2, "a": 3, "a":4, "a": 5, "a": 6, "a": 7, "a": 8, "a": 9, "a": 10, "a": 11, "a": 12, "a": 13, "a": 14, "a": 15, "a": 16, "a": 20, "b":2 }