GitHub - nlohmann/json: JSON for Modern C++
(
,
,
,
,
)
Creating json objects from JSON literals
JSON as a first-class data type
Serialization / Deserialization
Conversion from STL containers
Conversions to/from arbitrary types
Binary formats (BSON, CBOR, MessagePack, UBJSON, and BJData)
Design goals
There are myriads of
libraries out there, and each may even have its reason to exist. Our class had these design goals:
Intuitive syntax. In languages such as Python, JSON feels like a first-class data type. We used all the operator magic of modern C++ to achieve the same feeling in your code. Check out the
and you'll know what I mean.
Trivial integration. Our whole code consists of a single header file
. That's it. No library, no subproject, no dependencies, no complex build system. The class is written in vanilla C++11. All in all, everything should require no adjustment of your compiler flags or project settings. The library is also included in all popular
.
Serious testing. Our code is heavily
and covers
of the code, including all exceptional behavior. Furthermore, we checked with
and the
that there are no memory leaks.
additionally runs fuzz tests against all parsers 24/7, effectively executing billions of tests so far. To maintain high quality, the project is following the
Core Infrastructure Initiative (CII) best practices
. See the
overview documentation.
Other aspects were not so important to us:
Memory efficiency. Each JSON object has an overhead of one pointer (the maximal size of a union) and one enumeration element (1 byte). The default generalization uses the following C++ data types: std::string for strings, int64_t, uint64_t or double for numbers, std::map for objects, std::vector for arrays, and bool for Booleans. However, you can template the generalized class basic_json to your needs.
Speed. There are certainly
out there. However, if your goal is to speed up your development by adding JSON support with a single header, then this library is the way to go. If you know how to use a std::vector or std::map, you are already set.
See the
for more information.
Sponsors
You can sponsor this library at
.
🙋 Priority Sponsor
🏷️ Named Sponsors
Further support
The development of the library is further supported by JetBrains by providing free access to their IDE tools.
Thanks everyone!
Support
❓ If you have a question, please check if it is already answered in the
or the
section. If not, please
there.
📚 If you want to learn more about how to use the library, check out the rest of the
, have a look at
, or browse through the
.
🚧 If you want to understand the API better, check out the
or have a look at the
below.
🐛 If you found a bug, please check the
if it is a known issue or the result of a design decision. Please also have a look at the
before you
. Please provide as much information as possible to help us understand and reproduce your issue.
There is also a
for the documentation browsers
,
, and
that contains the full
as an offline resource.
Quick reference
Constructors
,
,
,
Object inspection:
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
Value access;
,
,
,
,
,
Element access:
,
,
,
,
Lookup:
,
,
Iterators:
,
,
,
,
,
,
,
,
Capacity:
,
,
Modifiers:
,
,
,
,
,
,
,
,
Lexicographical comparison operators:
,
,
,
,
,
,
Serialization / Dumping:
Deserialization / Parsing:
,
,
JSON Pointer functions:
,
JSON Patch functions:
,
,
,
Static functions:
,
Binary formats:
,
,
,
,
,
,
,
,
,
Non-member functions:
,
,
Literals:
Helper classes:
,
Examples
Here are some examples to give you an idea how to use the class.
Besides the examples below, you may want to:
→ Check the
→ Browse the
→ Read the full
with self-contained examples for every function
Read JSON from a file
The json class provides an API for manipulating a JSON value. To create a json object by reading a JSON file:
#include<fstream> #include<nlohmann/json.hpp>using json = nlohmann::json; // ... std::ifstream f("example.json"); json data = json::parse(f);If using modules (enabled with NLOHMANN_JSON_BUILD_MODULES), this example becomes:
import std; import nlohmann.json; using json = nlohmann::json; // ... std::ifstream f("example.json"); json data = json::parse(f);Creating json objects from JSON literals
Assume you want to hard-code this literal JSON value as a json object:
{ "pi": 3.141, "happy": true }There are various options:
// Using (raw) string literals and json::parse json ex1 = json::parse(R"( { "pi": 3.141, "happy": true })"); // Using user-defined (raw) string literalsusingnamespacenlohmann::literals; json ex2 = R"( { "pi": 3.141, "happy": true })"_json; // Using initializer lists json ex3 = { {"happy", true}, {"pi", 3.141}, };JSON as a first-class data type
Here are some examples to give you an idea how to use the class.
Assume you want to create the JSON object
{ "pi": 3.141, "happy": true, "name": "Niels", "nothing": null, "answer": { "everything": 42 }, "list": [1, 0, 2], "object": { "currency": "USD", "value": 42.99 } }With this library, you could write:
// create an empty structure (null) json j; // add a number stored as double (note the implicit conversion of j to an object) j["pi"] = 3.141; // add a Boolean stored as bool j["happy"] = true; // add a string stored as std::string j["name"] = "Niels"; // add another null object by passing nullptr j["nothing"] = nullptr; // add an object inside the object j["answer"]["everything"] = 42; // add an array stored as std::vector (using an initializer list) j["list"] = { 1, 0, 2 }; // add another object (using an initializer list of pairs) j["object"] = { {"currency", "USD"}, {"value", 42.99} }; // instead, you could also write (which looks very similar to the JSON above) json j2 = { {"pi", 3.141}, {"happy", true}, {"name", "Niels"}, {"nothing", nullptr}, {"answer", { {"everything", 42} }}, {"list", {1, 0, 2}}, {"object", { {"currency", "USD"}, {"value", 42.99} }} };Note that in all these cases, you never need to "tell" the compiler which JSON value type you want to use. If you want to be explicit or express some edge cases, the functions
and
will help:
// a way to express the empty array [] json empty_array_explicit = json::array(); // ways to express the empty object {} json empty_object_implicit = json({}); json empty_object_explicit = json::object(); // a way to express an _array_ of key/value pairs [["currency", "USD"], ["value", 42.99]] json array_not_object = json::array({ {"currency", "USD"}, {"value", 42.99} });Serialization / Deserialization
To/from strings
You can create a JSON value (deserialization) by appending _json to a string literal:
// create object from string literal json j = "{ \"happy\": true, \"pi\": 3.141 }"_json; // or even nicer with a raw string literalauto j2 = R"( { "happy": true, "pi": 3.141 })"_json;Note that without appending the _json suffix, the passed string literal is not parsed, but just used as JSON string value. That is, json j = "{ \"happy\": true, \"pi\": 3.141 }" would just store the string "{ "happy": true, "pi": 3.141 }" rather than parsing the actual object.
The string literal should be brought into scope with using namespace nlohmann::literals; (see
).
The above example can also be expressed explicitly using
:
// parse explicitlyauto j3 = json::parse(R"({"happy": true, "pi": 3.141})");You can also get a string representation of a JSON value (serialize):
// explicit conversion to string std::string s = j.dump(); // {"happy":true,"pi":3.141}// serialization with pretty printing// pass in the amount of spaces to indent std::cout << j.dump(4) << std::endl; // {// "happy": true,// "pi": 3.141// }Note the difference between serialization and assignment:
// store a string in a JSON value json j_string = "this is a string"; // retrieve the string valueauto cpp_string = j_string.get<std::string>(); // retrieve the string value (alternative when a variable already exists) std::string cpp_string2; j_string.get_to(cpp_string2); // retrieve the serialized value (explicit JSON serialization) std::string serialized_string = j_string.dump(); // output of original string std::cout << cpp_string << " == " << cpp_string2 << " == " << j_string.get<std::string>() << '\n'; // output of serialized value std::cout << j_string << " == " << serialized_string << std::endl;
returns the originally stored string value.
Note the library only supports UTF-8. When you store strings with different encodings in the library, calling
may throw an exception unless json::error_handler_t::replace or json::error_handler_t::ignore are used as error handlers.
To/from streams (e.g., files, string streams)
You can also use streams to serialize and deserialize:
// deserialize from standard input json j; std::cin >> j; // serialize to standard output std::cout << j; // the setw manipulator was overloaded to set the indentation for pretty printing std::cout << std::setw(4) << j << std::endl;These operators work for any subclasses of std::istream or std::ostream. Here is the same example with files:
// read a JSON file std::ifstream i("file.json"); json j; i >> j; // write prettified JSON to another file std::ofstream o("pretty.json"); o << std::setw(4) << j << std::endl;Please note that setting the exception bit for failbit is inappropriate for this use case. It will result in program termination due to the noexcept specifier in use.
Read from iterator range
You can also parse JSON from an iterator range; that is, from any container accessible by iterators whose value_type is an integral type of 1, 2, or 4 bytes, which will be interpreted as UTF-8, UTF-16, and UTF-32 respectively. For instance, a std::vector<std::uint8_t>, or a std::list<std::uint16_t>:
std::vector<std::uint8_t> v = {'t', 'r', 'u', 'e'}; json j = json::parse(v.begin(), v.end());You may leave the iterators for the range [begin, end):
std::vector<std::uint8_t> v = {'t', 'r', 'u', 'e'}; json j = json::parse(v);Custom data source
Since the parse function accepts arbitrary iterator ranges, you can provide your own data sources by implementing the LegacyInputIterator concept.
structMyContainer { voidadvance(); constchar& get_current(); }; structMyIterator { using difference_type = std::ptrdiff_t; using value_type = char; using pointer = constchar*; using reference = constchar&; using iterator_category = std::input_iterator_tag; explicitMyIterator(MyContainer* tgt = nullptr) : target(tgt) {} MyIterator& operator++() { target->advance(); return *this; } booloperator!=(const MyIterator& rhs) const { return rhs.target != target; } reference operator*() const { return target->get_current(); } MyContainer* target = nullptr; }; MyIterator begin(MyContainer& tgt) { return MyIterator{&tgt}; } MyIterator end(const MyContainer&) { return MyIterator{}; } voidfoo() { MyContainer c; json j = json::parse(begin(c), end(c)); }SAX interface
The library uses a SAX-like interface with the following functions:
// called when null is parsedboolnull(); // called when a boolean is parsed; value is passedboolboolean(bool val); // called when a signed or unsigned integer number is parsed; value is passedboolnumber_integer(number_integer_t val); boolnumber_unsigned(number_unsigned_t val); // called when a floating-point number is parsed; value and original string is passedboolnumber_float(number_float_t val, conststring_t& s); // called when a string is parsed; value is passed and can be safely moved awayboolstring(string_t& val); // called when a binary value is parsed; value is passed and can be safely moved awayboolbinary(binary_t& val); // called when an object or array begins or ends, resp. The number of elements is passed (or -1 if not known)boolstart_object(std::size_t elements); boolend_object(); boolstart_array(std::size_t elements); boolend_array(); // called when an object key is parsed; value is passed and can be safely moved awayboolkey(string_t& val); // called when a parse error occurs; byte position, the last token, and an exception is passedboolparse_error(std::size_t position, const std::string& last_token, const detail::exception& ex);The return value of each function determines whether parsing should proceed.
To implement your own SAX handler, proceed as follows:
Implement the SAX interface in a class. You can use class nlohmann::json_sax<json> as base class, but you can also use any class where the functions described above are implemented and public.
Create an object of your SAX interface class, e.g. my_sax.
Call bool json::sax_parse(input, &my_sax); where the first parameter can be any input like a string or an input stream and the second parameter is a pointer to your SAX interface.
Note the sax_parse function only returns a bool indicating the result of the last executed SAX event. It does not return a json value - it is up to you to decide what to do with the SAX events. Furthermore, no exceptions are thrown in case of a parse error -- it is up to you what to do with the exception object passed to your parse_error implementation. Internally, the SAX interface is used for the DOM parser (class json_sax_dom_parser) as well as the acceptor (json_sax_acceptor), see file
.
STL-like access
We designed the JSON class to behave just like an STL container. In fact, it satisfies the
requirement.
// create an array using push_back json j; j.push_back("foo"); j.push_back(1); j.push_back(true); // also use emplace_back j.emplace_back(1.78); // iterate the arrayfor (json::iterator it = j.begin(); it != j.end(); ++it) { std::cout << *it << '\n'; } // range-based forfor (auto& element : j) { std::cout << element << '\n'; } // getter/setterconstauto tmp = j[0].get<std::string>(); j[1] = 42; bool foo = j.at(2); // comparison j == R"(["foo", 1, true, 1.78])"_json; // true// other stuff j.size(); // 4 entries j.empty(); // false j.type(); // json::value_t::array j.clear(); // the array is empty again// convenience type checkers j.is_null(); j.is_boolean(); j.is_number(); j.is_object(); j.is_array(); j.is_string(); // create an object json o; o["foo"] = 23; o["bar"] = false; o["baz"] = 3.141; // also use emplace o.emplace("weather", "sunny"); // special iterator member functions for objectsfor (json::iterator it = o.begin(); it != o.end(); ++it) { std::cout << it.key() << " : " << it.value() << "\n"; } // the same code as range forfor (auto& el : o.items()) { std::cout << el.key() << " : " << el.value() << "\n"; } // even easier with structured bindings (C++17)for (auto& [key, value] : o.items()) { std::cout << key << " : " << value << "\n"; } // find an entryif (o.contains("foo")) { // there is an entry with key "foo" } // or via find and an iteratorif (o.find("foo") != o.end()) { // there is an entry with key "foo" } // or simpler using count()int foo_present = o.count("foo"); // 1int fob_present = o.count("fob"); // 0// delete an entry o.erase("foo");Conversion from STL containers
Any sequence container (std::array, std::vector, std::deque, std::forward_list, std::list) whose values can be used to construct JSON values (e.g., integers, floating point numbers, Booleans, string types, or again STL containers described in this section) can be used to create a JSON array. The same holds for similar associative containers (std::set, std::multiset, std::unordered_set, std::unordered_multiset), but in these cases the order of the elements of the array depends on how the elements are ordered in the respective STL container.
std::vector<int> c_vector {1, 2, 3, 4}; json j_vec(c_vector); // [1, 2, 3, 4] std::deque<double> c_deque {1.2, 2.3, 3.4, 5.6}; json j_deque(c_deque); // [1.2, 2.3, 3.4, 5.6] std::list<bool> c_list {true, true, false, true}; json j_list(c_list); // [true, true, false, true] std::forward_list<int64_t> c_flist {12345678909876, 23456789098765, 34567890987654, 45678909876543}; json j_flist(c_flist); // [12345678909876, 23456789098765, 34567890987654, 45678909876543] std::array<unsignedlong, 4> c_array {{1, 2, 3, 4}}; json j_array(c_array); // [1, 2, 3, 4] std::set<std::string> c_set {"one", "two", "three", "four", "one"}; json j_set(c_set); // only one entry for "one" is used// ["four", "one", "three", "two"] std::unordered_set<std::string> c_uset {"one", "two", "three", "four", "one"}; json j_uset(c_uset); // only one entry for "one" is used// maybe ["two", "three", "four", "one"] std::multiset<std::string> c_mset {"one", "two", "one", "four"}; json j_mset(c_mset); // both entries for "one" are used// maybe ["one", "two", "one", "four"] std::unordered_multiset<std::string> c_umset {"one", "two", "one", "four"}; json j_umset(c_umset); // both entries for "one" are used// maybe ["one", "two", "one", "four"]Likewise, any associative key-value containers (std::map, std::multimap, std::unordered_map, std::unordered_multimap) whose keys can construct an std::string and whose values can be used to construct JSON values (see examples above) can be used to create a JSON object. Note that in case of multimaps, only one key is used in the JSON object and the value depends on the internal order of the STL container.
std::map<std::string, int> c_map { {"one", 1}, {"two", 2}, {"three", 3} }; json j_map(c_map); // {"one": 1, "three": 3, "two": 2 } std::unordered_map<constchar*, double> c_umap { {"one", 1.2}, {"two", 2.3}, {"three", 3.4} }; json j_umap(c_umap); // {"one": 1.2, "two": 2.3, "three": 3.4} std::multimap<std::string, bool> c_mmap { {"one", true}, {"two", true}, {"three", false}, {"three", true} }; json j_mmap(c_mmap); // only one entry for key "three" is used// maybe {"one": true, "two": true, "three": true} std::unordered_multimap<std::string, bool> c_ummap { {"one", true}, {"two", true}, {"three", false}, {"three", true} }; json j_ummap(c_ummap); // only one entry for key "three" is used// maybe {"one": true, "two": true, "three": true}JSON Pointer and JSON Patch
The library supports JSON Pointer (
) as an alternative means to address structured values. On top of this, JSON Patch (
) allows describing differences between two JSON values -- effectively allowing patch and diff operations known from Unix.
// a JSON value json j_original = R"({ "baz": ["one", "two", "three"], "foo": "bar"})"_json; // access members with a JSON pointer (RFC 6901) j_original["/baz/1"_json_pointer]; // "two"// a JSON patch (RFC 6902) json j_patch = R"([ { "op": "replace", "path": "/baz", "value": "boo" }, { "op": "add", "path": "/hello", "value": ["world"] }, { "op": "remove", "path": "/foo"}])"_json; // apply the patch json j_result = j_original.patch(j_patch); // {// "baz": "boo",// "hello": ["world"]// }// calculate a JSON patch from two JSON valuesjson::diff(j_result, j_original); // [// { "op":" replace", "path": "/baz", "value": ["one", "two", "three"] },// { "op": "remove","path": "/hello" },// { "op": "add", "path": "/foo", "value": "bar" }// ]JSON Merge Patch
The library supports JSON Merge Patch (
) as a patch format. Instead of using JSON Pointer (see above) to specify values to be manipulated, it describes the changes using a syntax that closely mimics the document being modified.
// a JSON value json j_document = R"({ "a": "b", "c": { "d": "e", "f": "g" }})"_json; // a patch json j_patch = R"({ "a":"z", "c": { "f": null }})"_json; // apply the patch j_document.merge_patch(j_patch); // {// "a": "z",// "c": {// "d": "e"// }// }Implicit conversions
Supported types can be implicitly converted to JSON values.
It is recommended to NOT USE implicit conversions FROM a JSON value. You can find more details about this recommendation
. You can switch off implicit conversions by defining JSON_USE_IMPLICIT_CONVERSIONS to 0 before including the json.hpp header. When using CMake, you can also achieve this by setting the option JSON_ImplicitConversions to OFF.
// strings std::string s1 = "Hello, world!"; json js = s1; auto s2 = js.get<std::string>(); // NOT RECOMMENDED std::string s3 = js; std::string s4; s4 = js; // Booleansbool b1 = true; json jb = b1; auto b2 = jb.get<bool>(); // NOT RECOMMENDEDbool b3 = jb; bool b4; b4 = jb; // numbersint i = 42; json jn = i; auto f = jn.get<double>(); // NOT RECOMMENDEDdouble f2 = jn; double f3; f3 = jn; // etc.Note that char types are not automatically converted to JSON strings, but to integer numbers. A conversion to a string must be specified explicitly:
char ch = 'A'; // ASCII value 65 json j_default = ch; // stores integer number 65 json j_string = std::string(1, ch); // stores string "A"Arbitrary types conversions
Every type can be serialized in JSON, not just STL containers and scalar types. Usually, you would do something along those lines:
namespacens { // a simple struct to model a personstructperson { std::string name; std::string address; int age; }; } ns::person p = {"Ned Flanders", "744 Evergreen Terrace", 60}; // convert to JSON: copy each value into the JSON object json j; j["name"] = p.name; j["address"] = p.address; j["age"] = p.age; // ...// convert from JSON: copy each value from the JSON object ns::person p { j["name"].get<std::string>(), j["address"].get<std::string>(), j["age"].get<int>() };It works, but that's quite a lot of boilerplate... Fortunately, there's a better way:
// create a person ns::person p {"Ned Flanders", "744 Evergreen Terrace", 60}; // conversion: person -> json json j = p; std::cout << j << std::endl; // {"address":"744 Evergreen Terrace","age":60,"name":"Ned Flanders"}// conversion: json -> personauto p2 = j.get<ns::person>(); // that's itassert(p == p2);Basic usage
To make this work with one of your types, you only need to provide two functions:
using json = nlohmann::json; namespacens { voidto_json(json& j, const person& p) { j = json{{"name", p.name}, {"address", p.address}, {"age", p.age}}; } voidfrom_json(const json& j, person& p) { j.at("name").get_to(p.name); j.at("address").get_to(p.address); j.at("age").get_to(p.age); } } // namespace nsThat's all! When calling the json constructor with your type, your custom to_json method will be automatically called. Likewise, when calling get<your_type>() or get_to(your_type&), the from_json method will be called.
Some important things:
Those methods MUST be in your type's namespace (which can be the global namespace), or the library will not be able to locate them (in this example, they are in namespace ns, where person is defined).
Those methods MUST be available (e.g., proper headers must be included) everywhere you use these conversions. Look at
for errors that may occur otherwise.
When using get<your_type>(), your_typeMUST be
. (There is a way to bypass this requirement described later.)
In function from_json, use function
to access the object values rather than operator[]. In case a key does not exist, at throws an exception that you can handle, whereas operator[] exhibits undefined behavior.
You do not need to add serializers or deserializers for STL types like std::vector: the library already implements these.
Simplify your life with macros
If you just want to serialize/deserialize some structs, the to_json/from_json functions can be a lot of boilerplate. There are
to make your life easier as long as you want to use a JSON object as serialization.
Which macro to choose depends on whether private member variables need to be accessed, a deserialization is needed, missing values should yield an error or should be replaced by default values, and if derived classes are used. See
this overview to choose the right one for your use case
.
Example usage of macrosThe to_json/from_json functions for the person struct above can be created with
NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE
. In all macros, the first parameter is the name of the class/struct, and all remaining parameters name the members.
namespacens { NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(person, name, address, age) }If you want to inherit the person struct and add a field to it, it can be done with:
namespacens { structperson_derived : person { std::string email; }; NLOHMANN_DEFINE_DERIVED_TYPE_NON_INTRUSIVE(person_derived, person, email) }Here is another example with private members, where
NLOHMANN_DEFINE_TYPE_INTRUSIVE
is needed:
namespacens { classaddress { private: std::string street; int housenumber; int postcode; public:NLOHMANN_DEFINE_TYPE_INTRUSIVE(address, street, housenumber, postcode) }; }Or in case if you use some naming convention that you do not want to expose to JSON:
namespacens { classaddress { private: std::string m_street; int m_housenumber; int m_postcode; public:NLOHMANN_DEFINE_TYPE_INTRUSIVE_WITH_NAMES(address, "street", m_street, "housenumber", m_housenumber, "postcode", m_postcode) }; }How do I convert third-party types?
This requires a bit more advanced technique. But first, let's see how this conversion mechanism works:
The library uses JSON Serializers to convert types to JSON. The default serializer for nlohmann::json is nlohmann::adl_serializer (ADL means
).
It is implemented like this (simplified):
template <typename T> structadl_serializer { staticvoidto_json(json& j, const T& value) { // calls the "to_json" method in T's namespace } staticvoidfrom_json(const json& j, T& value) { // same thing, but with the "from_json" method } };This serializer works fine when you have control over the type's namespace. However, what about boost::optional or std::filesystem::path (C++17)? Hijacking the boost namespace is pretty bad, and it's illegal to add something other than template specializations to std...
To solve this, you need to add a specialization of adl_serializer to the nlohmann namespace, here's an example:
// partial specialization (full specialization works too)namespacenlohmann { template <typename T> structadl_serializer<boost::optional<T>> { staticvoidto_json(json& j, const boost::optional<T>& opt) { if (opt == boost::none) { j = nullptr; } else { j = *opt; // this will call adl_serializer<T>::to_json which will// find the free function to_json in T's namespace! } } staticvoidfrom_json(const json& j, boost::optional<T>& opt) { if (j.is_null()) { opt = boost::none; } else { opt = j.get<T>(); // same as above, but with// adl_serializer<T>::from_json } } }; }How can I use get() for non-default constructible/non-copyable types?
There is a way if your type is
. You will need to specialize the adl_serializer as well, but with a special from_json overload:
structmove_only_type { move_only_type() = delete; move_only_type(int ii): i(ii) {} move_only_type(const move_only_type&) = delete; move_only_type(move_only_type&&) = default; int i; }; namespacenlohmann { template <> structadl_serializer<move_only_type> { // note: the return type is no longer 'void', and the method only takes// one argumentstatic move_only_type from_json(const json& j) { return {j.get<int>()}; } // Here's the catch! You must provide a to_json method! Otherwise, you// will not be able to convert move_only_type to json, since you fully// specialized adl_serializer on that typestaticvoidto_json(json& j, move_only_type t) { j = t.i; } }; }Can I write my own serializer? (Advanced use)
Yes. You might want to take a look at
in the test suite, to see a few examples.
If you write your own serializer, you'll need to do a few things:
use a different basic_json alias than nlohmann::json (the last template parameter of basic_json is the JSONSerializer)
use your basic_json alias (or a template parameter) in all your to_json/from_json methods
use nlohmann::to_json and nlohmann::from_json when you need ADL
Here is an example, without simplifications, that only accepts types with a size <= 32, and uses ADL.
// You should use void as a second template argument// if you don't need compile-time checks on Ttemplate<typename T, typenameSFINAE = typename std::enable_if<sizeof(T) <= 32>::type> structless_than_32_serializer { template <typename BasicJsonType> staticvoidto_json(BasicJsonType& j, T value) { // we want to use ADL, and call the correct to_json overloadusing nlohmann::to_json; // this method is called by adl_serializer,// this is where the magic happensto_json(j, value); } template <typename BasicJsonType> staticvoidfrom_json(const BasicJsonType& j, T& value) { // same thing hereusing nlohmann::from_json; from_json(j, value); } };Be very careful when reimplementing your serializer, you can stack overflow if you don't pay attention:
template <typename T, void> structbad_serializer { template <typename BasicJsonType> staticvoidto_json(BasicJsonType& j, const T& value) { // this calls BasicJsonType::json_serializer<T>::to_json(j, value)// if BasicJsonType::json_serializer == bad_serializer ... oops! j = value; } template <typename BasicJsonType> staticvoidto_json(const BasicJsonType& j, T& value) { // this calls BasicJsonType::json_serializer<T>::from_json(j, value)// if BasicJsonType::json_serializer == bad_serializer ... oops! value = j.get<T>(); // oops! } };Specializing enum conversion
By default, enum values are serialized to JSON as integers. In some cases, this could result in undesired behavior. If an enum is modified or re-ordered after data has been serialized to JSON, the later deserialized JSON data may be undefined or a different enum value than was originally intended.
It is possible to more precisely specify how a given enum is mapped to and from JSON as shown below:
// example enum type declarationenum TaskState { TS_STOPPED, TS_RUNNING, TS_COMPLETED, TS_INVALID=-1, }; // map TaskState values to JSON as stringsNLOHMANN_JSON_SERIALIZE_ENUM( TaskState, { {TS_INVALID, nullptr}, {TS_STOPPED, "stopped"}, {TS_RUNNING, "running"}, {TS_COMPLETED, "completed"}, })The NLOHMANN_JSON_SERIALIZE_ENUM() macro declares a set of to_json() / from_json() functions for type TaskState while avoiding repetition and boilerplate serialization code.
Usage:
// enum to JSON as string json j = TS_STOPPED; assert(j == "stopped"); // json string to enum json j3 = "running"; assert(j3.get<TaskState>() == TS_RUNNING); // undefined json value to enum (where the first map entry above is the default) json jPi = 3.14; assert(jPi.get<TaskState>() == TS_INVALID);Just as in
above,
NLOHMANN_JSON_SERIALIZE_ENUM() MUST be declared in your enum type's namespace (which can be the global namespace), or the library will not be able to locate it, and it will default to integer serialization.
It MUST be available (e.g., proper headers must be included) everywhere you use the conversions.
Other Important points:
When using get<ENUM_TYPE>(), undefined JSON values will default to the first pair specified in your map. Select this default pair carefully. If you desire an exception in this circumstance use NLOHMANN_JSON_SERIALIZE_ENUM_STRICT() which behaves identically except for throwing an exception on unrecognized values.
If an enum or JSON value is specified more than once in your map, the first matching occurrence from the top of the map will be returned when converting to or from JSON.
Binary formats (BSON, CBOR, MessagePack, UBJSON, and BJData)
Though JSON is a ubiquitous data format, it is not a very compact format suitable for data exchange, for instance over a network. Hence, the library supports
(Binary JSON),
(Concise Binary Object Representation),
,
(Universal Binary JSON Specification) and
(Binary JData) to efficiently encode JSON values to byte vectors and to decode such vectors.
// create a JSON value json j = R"({"compact": true, "schema": 0})"_json; // serialize to BSON std::vector<std::uint8_t> v_bson = json::to_bson(j); // 0x1B, 0x00, 0x00, 0x00, 0x08, 0x63, 0x6F, 0x6D, 0x70, 0x61, 0x63, 0x74, 0x00, 0x01, 0x10, 0x73, 0x63, 0x68, 0x65, 0x6D, 0x61, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00// roundtrip json j_from_bson = json::from_bson(v_bson); // serialize to CBOR std::vector<std::uint8_t> v_cbor = json::to_cbor(j); // 0xA2, 0x67, 0x63, 0x6F, 0x6D, 0x70, 0x61, 0x63, 0x74, 0xF5, 0x66, 0x73, 0x63, 0x68, 0x65, 0x6D, 0x61, 0x00// roundtrip json j_from_cbor = json::from_cbor(v_cbor); // serialize to MessagePack std::vector<std::uint8_t> v_msgpack = json::to_msgpack(j); // 0x82, 0xA7, 0x63, 0x6F, 0x6D, 0x70, 0x61, 0x63, 0x74, 0xC3, 0xA6, 0x73, 0x63, 0x68, 0x65, 0x6D, 0x61, 0x00// roundtrip json j_from_msgpack = json::from_msgpack(v_msgpack); // serialize to UBJSON std::vector<std::uint8_t> v_ubjson = json::to_ubjson(j); // 0x7B, 0x69, 0x07, 0x63, 0x6F, 0x6D, 0x70, 0x61, 0x63, 0x74, 0x54, 0x69, 0x06, 0x73, 0x63, 0x68, 0x65, 0x6D, 0x61, 0x69, 0x00, 0x7D// roundtrip json j_from_ubjson = json::from_ubjson(v_ubjson);The library also supports binary types from BSON, CBOR (byte strings), and MessagePack (bin, ext, fixext). They are stored by default as std::vector<std::uint8_t> to be processed outside the library.
// CBOR byte string with payload 0xCAFE std::vector<std::uint8_t> v = {0x42, 0xCA, 0xFE}; // read value json j = json::from_cbor(v); // the JSON value has type binary j.is_binary(); // true// get reference to stored binary valueauto& binary = j.get_binary(); // the binary value has no subtype (CBOR has no binary subtypes) binary.has_subtype(); // false// access std::vector<std::uint8_t> member functions binary.size(); // 2 binary[0]; // 0xCA binary[1]; // 0xFE// set subtype to 0x10 binary.set_subtype(0x10); // serialize to MessagePackauto cbor = json::to_msgpack(j); // 0xD5 (fixext2), 0x10, 0xCA, 0xFECustomers
The library is used in multiple projects, applications, operating systems, etc. The list below is not exhaustive, but the result of an internet search. If you know further customers of the library, please let me know, see
.
Ecosystem
Beyond projects that use the library, there are third-party projects that build on top of it - schema validators, language bindings, format converters, and the like. See the curated
page.
Supported compilers
Though it's 2026 already, the support for C++11 is still a bit sparse. Currently, the following compilers are known to work:
GCC 4.8 - 14.2 (and possibly later)
Clang 3.4 - 21.0 (and possibly later)
Apple Clang 9.1 - 16.0 (and possibly later)
Intel C++ Compiler 17.0.2 (and possibly later)
Nvidia CUDA Compiler 11.0.221 (and possibly later)
Microsoft Visual C++ 2015 / Build Tools 14.0.25123.0 (and possibly later)
Microsoft Visual C++ 2017 / Build Tools 15.5.180.51428 (and possibly later)
Microsoft Visual C++ 2019 / Build Tools 16.3.1+1def00d3d (and possibly later)
Microsoft Visual C++ 2022 / Build Tools 19.30.30709.0 (and possibly later)
I would be happy to learn about other compilers/versions.
Please note:
GCC 4.8 has a bug
: multiline raw strings cannot be the arguments to macros. Don't use multiline raw strings directly in macros with this compiler.
Android defaults to using very old compilers and C++ libraries. To fix this, add the following to your Application.mk. This will switch to the LLVM C++ library, the Clang compiler, and enable C++11 and other features disabled by default.
APP_STL := c++_shared NDK_TOOLCHAIN_VERSION := clang3.6 APP_CPPFLAGS += -frtti -fexceptionsThe code compiles successfully with
, Revision 9 - 11 (and possibly later) and
version 10.
For GCC running on MinGW or Android SDK, the error 'to_string' is not a member of 'std' (or similarly, for strtod or strtof) may occur. Note this is not an issue with the code, but rather with the compiler itself. On Android, see above to build with a newer environment. For MinGW, please refer to
and
for information on how to fix this bug. For Android NDK using APP_STL := gnustl_static, please refer to
.
Unsupported versions of GCC and Clang are rejected by #error directives. This can be switched off by defining JSON_SKIP_UNSUPPORTED_COMPILER_CHECK. Note that you can expect no support in this case.
See the page
on the compilers used to check the library in the CI.
Integration
is the single required file in single_include/nlohmann or
. You need to add
#include<nlohmann/json.hpp>// for convenienceusing json = nlohmann::json;to the files you want to process JSON and set the necessary switches to enable C++11 (e.g., -std=c++11 for GCC and Clang).
You can further use file
for forward-declarations. The installation of json_fwd.hpp (as part of cmake's install step) can be achieved by setting -DJSON_MultipleHeaders=ON.
CMake
You can also use the nlohmann_json::nlohmann_json interface target in CMake. This target populates the appropriate usage requirements for INTERFACE_INCLUDE_DIRECTORIES to point to the appropriate include directories and INTERFACE_COMPILE_FEATURES for the necessary C++11 flags.
External
To use this library from a CMake project, you can locate it directly with find_package() and use the namespaced imported target from the generated package configuration:
# CMakeLists.txtfind_package(nlohmann_json3.12.0REQUIRED) ...add_library(foo...) ...target_link_libraries(fooPRIVATEnlohmann_json::nlohmann_json)The package configuration file, nlohmann_jsonConfig.cmake, can be used either from an install tree or directly out of the build tree.
Embedded
To embed the library directly into an existing CMake project, place the entire source tree in a subdirectory and call add_subdirectory() in your CMakeLists.txt file:
# Typically you don't care so much for a third party library's tests to be# run from your own project's code.set(JSON_BuildTests OFFCACHEINTERNAL"") # If you only include this third party in PRIVATE source files, you do not# need to install it when your main project gets installed.# set(JSON_Install OFF CACHE INTERNAL "")# Don't use include(nlohmann_json/CMakeLists.txt) since that carries with it# unintended consequences that will break the build. It's generally# discouraged (although not necessarily well documented as such) to use# include(...) for pulling in other CMake projects anyways.add_subdirectory(nlohmann_json) ...add_library(foo...) ...target_link_libraries(fooPRIVATEnlohmann_json::nlohmann_json)Embedded (FetchContent)Since CMake v3.11,
can be used to automatically download a release as a dependency at configure time.
Example:
include(FetchContent) FetchContent_Declare(json URL https://github.com/nlohmann/json/releases/download/v3.12.0/json.tar.xz) FetchContent_MakeAvailable(json) target_link_libraries(fooPRIVATEnlohmann_json::nlohmann_json)Note: It is recommended to use the URL approach described above, which is supported as of version 3.10.0. See
https://json.nlohmann.me/integration/cmake/#fetchcontent
for more information.
Supporting Both
To allow your project to support either an externally supplied or an embedded JSON library, you can use a pattern akin to the following:
# Top level CMakeLists.txtproject(FOO) ...option(FOO_USE_EXTERNAL_JSON"Use an external JSON library"OFF) ...add_subdirectory(thirdparty) ...add_library(foo...) ...# Note that the namespaced target will always be available regardless of the# import methodtarget_link_libraries(fooPRIVATEnlohmann_json::nlohmann_json)# thirdparty/CMakeLists.txt...if(FOO_USE_EXTERNAL_JSON) find_package(nlohmann_json3.12.0REQUIRED) else() set(JSON_BuildTests OFFCACHEINTERNAL"") add_subdirectory(nlohmann_json) endif() ...thirdparty/nlohmann_json is then a complete copy of this source tree.
Package Managers
Use your favorite
to use the library.
The library is part of many package managers. See the
for detailed descriptions and examples.
Pkg-config
If you are using bare Makefiles, you can use pkg-config to generate the include flags that point to where the library is installed:
pkg-config nlohmann_json --cflagsLicense
The class is licensed under the
:
Copyright © 2013-2026
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
———
The class contains the UTF-8 Decoder from Bjoern Hoehrmann which is licensed under the
(see above). Copyright © 2008-2009
The class contains a slightly modified version of the Grisu2 algorithm from Florian Loitsch which is licensed under the
(see above). Copyright © 2009
The class contains a copy of
from Evan Nemerson which is licensed as
.
The class contains parts of
which is licensed under the
.
The library is compliant to version 3.3 of the
:
Every source file contains an SPDX copyright header.
The full text of all licenses used in the repository can be found in the LICENSES folder.
File .reuse/dep5 contains an overview of all files' copyrights and licenses.
Run pipx run reuse lint to verify the project's REUSE compliance and pipx run reuse spdx to generate a SPDX SBOM.
Contact
If you have questions regarding the library, I would like to invite you to
. Please describe your request, problem, or question as detailed as possible, and also mention the version of the library you are using as well as the version of your compiler and operating system. Opening an issue at GitHub allows other users and contributors to this library to collaborate. For instance, I have little experience with MSVC, and most issues in this regard have been solved by a growing community. If you have a look at the
, you will see that we react quite timely in most cases.
Only if your request would contain confidential information, please
. For encrypted messages, please use
.
Security
and
are signed with this
.
Thanks
I deeply appreciate the help of the following people.
implemented CMake support and lcov integration, realized escape and Unicode handling in the string parser, and fixed the JSON serialization.
fixed an issue with double deletion in the iterator classes.
made the iterators of the class composable to other libraries.
fixed a bug that hindered the class to compile with Clang.
Tomas Åblad found a bug in the iterator implementation.
fixed a bug in the floating-point serialization.
implemented code to parse streams incrementally. Furthermore, he greatly improved the parser class by allowing the definition of a filter function to discard undesired elements while parsing.
fixed a bug in the compilation with GCC 5.0.
fixed a bug in and improved the performance of the comparison operators.
pointed out a bug in the handling with NaN and infinity values. He also improved the performance of the string escaping.
implemented a conversion from anonymous enums.
patiently pushed forward the support for Microsoft Visual Studio.
simplified the implementation of reverse iterators and helped with numerous hints and improvements. In particular, he pushed forward the implementation of user-defined types.
fixed a bug in the Unicode handling.
fixed some typos in the examples.
cleaned up some pointers and implemented exception-safe memory allocation.
took care of a small namespace issue.
corrected a variable name in the documentation.
overloaded parse() to accept an rvalue reference.
fixed a subtlety in MSVC type support and implemented the get_ref() function to get a reference to stored values.
added a workaround that allows compilation using Android NDK.
replaced a function that was marked as unsafe by Visual Studio.
fixed two small warnings.
noted a potential portability problem in the has_mapped_type function.
fixed some typos in the contribution guidelines.
fixed the array subscript operator, an issue that failed the MSVC build, and floating-point parsing/dumping. He further added support for unsigned integer numbers and implemented better roundtrip support for parsed numbers.
fixed a link in the README file.
added support for American Fuzzy Lop.
fixed an example in the README file.
noted a wrong URL in the README file.
fixed a namespace issue with int64_t and uint64_t.
analyzed the issues with GCC 4.8 and proposed a
.
added useful notes to the README file about Android.
added a fix to use move iterators and improved the integration via CMake.
cleaned up the CMake files.
fixed a subtle bug with MSVC 2015 which was also proposed by
.
fixed a small typo.
found a really embarrassing performance regression in the 2.0.0 release.
fixed one of the last conversion warnings.
fixed a warning in a test case and adjusted MSVC calls in the CI.
patiently and constructively oversaw the long way toward
. He also implemented the magic behind the serialization/deserialization of user-defined types and split the single header file into smaller chunks.
fixed a minor issue in the documentation.
fixed the documentation regarding conversions from std::multiset.
overworked the CMake files to ease project inclusion.
made a SFINAE hack more readable and added Visual Studio 17 to the build matrix.
fixed a grammar issue in the README file.
found a subtle bug in the dump() function.
pointed to
to avoid too much locale joggling, found some nice performance improvements in the parser, improved the benchmarking code, and realized locale-independent number parsing and printing.
had an idea how to fix the Coverity scan.
silenced a nasty documentation warning.
fixed an integer overflow check.
merged two iterator classes into a smaller one.
helped to get Travis to execute the tests with Clang's sanitizers.
fixed an example in the README file.
supported the implementation of user-defined types.
helped to get the user-defined types working with Visual Studio.
supported the implementation of user-defined types.
noted a typo in an example.
found a way for a 2x speedup for the compilation time of the test suite.
proposed an improvement for the examples section.
noted a typo in the README.
fixed a bug in the comparison with nullptrs.
added
support to speed up the compilation.
noted a typo in the README, removed unnecessary bit arithmetic, and fixed some -Weffc++ warnings.
made exceptions more visible.
fixed a compiler warning.
made sure all pushed warnings are properly popped.
found a bug in the documentation.
implemented a Meson build description.
fixed a warning in ICC and improved the iterator implementation.
maintains a package for the Conan package manager.
fixed a potential issue with MSVC and std::min.
fixed some typos.
noted misleading documentation about comparison of floats.
reduced the memory consumption by replacing <iostream> with <iosfwd>.
cleaned up the CMake files to simplify including/reusing of the library.
allowed for moving values from initializer lists.
fixed a typo.
fixed a typo.
fixed an issue related to the Intel OSX compiler.
fixed a typo.
fixed a subtle error in a precondition check.
noted an error in a code sample.
reported some warnings with ICC and helped to fix them.
simplified reading from input streams.
fixed a small compilation error.
fixed all MSVC warnings.
added a Doxygen tag file.
helped to fix a warning in ICC.
helped to fix some warnings in MSVC.
avoided unnecessary string copies in find() and count().
fixed some typos.
updated the Hunter package links.
added a .natvis for the MSVC debug view.
fixed some C++17 deprecation warnings.
fixed some MSVC warnings.
integrated the Grisu2 algorithm for proper floating-point formatting, allowing more roundtrip checks to succeed.
fixed a Markdown issue in the README.
fixed a compiler warning.
allowed to template the string type in the serialization and added the possibility to override the exceptional behavior.
helped fix an ICC error.
fixed links in the README file.
found a bug in the implementation of CBOR's indefinite length strings.
added a note on the cget package manager.
made the integration section of the README more concise.
detected and fixed a memory leak in the parser callback.
allowed dumping JSON to an alternative string type.
overworked the C++11 compiler checks in CMake.
simplified a CMake check and added support for the
.
fixed a typo.
fixed a version number in the compilers section.
adjusted the CMake files to the CMake packaging guidelines and provided documentation for the CMake integration.
fixed a typo.
fixed a typo.
removed the dependency from std::stringstream.
added code to use alternative string implementations.
allowed to use more algorithms with the items() function.
fixed the Meson include directory and fixed the links to
.
fixed the compilation with MSVC 2015 in debug mode.
fixed the test suite and re-enabled several test cases.
introduced the macro JSON_INTERNAL_CATCH to control the exception handling inside the library.
fixed a compiler warning.
fixed a subtle compilation error with Clang 3.4.2.
allowed to call find_package without installing the library.
fixed an issue with a double macro definition.
made some error messages more understandable.
fixed a compilation problem with the Intel C++ compiler.
fixed a compilation problem.
added version and license information to the Meson build file.
added support for GCC 4.8.
made sure the test suite does not stall when run in the wrong directory.
fixed an MSVC 2017 warning.
implemented the NLOHMANN_JSON_SERIALIZE_ENUM macro to quickly define an enum/JSON mapping.
added line and column information to parse errors.
added BSON support.
added support for structured bindings.
added support for Clang 5.0.1 (PS4 version).
implemented an input adapter to read from FILE*.
fixed a link in the documentation.
fixed a typo in the documentation.
fixed an MSVC warning.
added code to avoid an issue with MSVC.
fixed a bug when JSON was parsed from an input stream.
allowed to install the library via Meson.
found an issue with a missing namespace.
fixed a compilation issue with libc 2.12.
fixed the endian conversion.
fixed a warning in MSVC.
added operator/ for JSON Pointers.
noted a missing header.
fixed compilation with GCC 9.0.
fixed compilation with GCC 9.0.
helped to reduce the CMake requirement to version 3.1.
updated the Buckaroo instructions.
fixed a compilation issue with GCC 7 on CentOS.
improved the integer serialization performance and implemented the contains() function.
suppressed an unfixable warning.
improved Meson support.
fixed an example in the README.
fixed a typo.
fixed a bug in the serializer.
implemented push_back() and pop_back() for JSON Pointers.
added support for Conda.
fixed links in the README.
documented how to install the library with NuGet.
fixed a typo.
helped to reduce the CMake requirement to version 3.1.
maintains a package for the MSYS2 software distro.
added GNUInstallDirs to the CMake files.
fixed a unit test.
implemented the to_string method.
fixed a Clang warning.
switched the unit tests from
to
fixed a typo.
fixed a bug in the contains function.
fixed some cppcheck warnings.
fixed some typos.
added a const version of json_pointer::back.
made the items() function work with custom string types.
updated fixed a bug in Hedley and updated this library accordingly.
fixed a lot of typos.
fixed an issue in the conversion from std::pair and std::tuple to json.
fixed a compile error in an enum deserialization.
noted a subtle bug in a preprocessor check.
fixed numerous issues in the library.
added a CI step for GCC 10.1.
fixed an MSVC warning.
proposed an improvement in the floating-point serialization in CBOR.
made performance improvements in the input adapters.
documented how the library can be included via FetchContent.
fixed an error message.
fixed some examples and a link in the README.
made CMake's version config file architecture-independent.
implemented the binary values for CBOR, MessagePack, BSON, and UBJSON.
fixed a compilation issue with GCC 10 and fixed a leak.
integrated the library to the wsjcpp package manager.
fixed a compiler warning.
fixed the year in the copyright.
fixed a compilation issue with MSVC.
fixed an example in the README.
fixed some typos in the documentation.
updated links to the Hunter package.
fixed a link in the README.
added instruction for using Build2's package manager.
fixed an example in the README.
fixed a warning.
updated the Conan package source.
fixed the MSYS2 package documentation.
improved the CMake tests.
fixed MSVC warnings.
fixed an example in the documentation.
fixed a compiler warning.
fixed links in the README.
improved the fuzzer coverage for UBSAN input.
fixed a compiler warning.
made NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE inline.
improved the upper bound of arguments of the NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE/NLOHMANN_DEFINE_TYPE_INTRUSIVE macros.
fixed a bug in the CBOR parser for binary and string values.
fixed a bug in the new hash implementation.
adjusted the CBOR writer to create tags for binary subtypes.
implemented an ordered map container for nlohmann::ordered_json.
added support for pkg-config.
proposed an implementation for the NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE/NLOHMANN_DEFINE_TYPE_INTRUSIVE macros.
implemented string_view support and allowed C++20 support.
improved CMake support for FetchContent.
provided a GDB pretty printer.
Lars Wirzenius reviewed the README file.
fixed a compiler path in the CMake scripts.
fixed typos in the documentation.
fixed a move constructor and the Travis builds.
added CPM.Cmake support.
fixed a warning.
updated doctest and implemented unit tests.
fixed a bug in the CMake files.
fixed a bug in the input adapters.
fixed some Markdown issues in the README file.
fixed an example from the README.
fixed an example from the README.
fixed a warning.
fixed the documentation.
updated doctest.
fixed pkg-config.pc.
fixed a warning.
added the possibility to compile the library without I/O support.
fixed a typo.
allowed treating the library as a system header in CMake.
fixed the indentation of the CMake file.
added a link to Conan Center to the documentation.
updated the links in the documentation to use HTTPS.
fixed the Google Benchmark default branch.
fixed a conversion operator.
made the examples in the README more consistent.
suppressed some -Wfloat-equal warnings.
fixed -Wswitch-enum warnings.
made the GDB pretty-printer robust against unset variable names.
updated the Homebrew command as nlohmann/json is now in homebrew-core.
fixed some -Wextra-semi-stmt warnings.
fixed -Wunused warnings on JSON_DIAGNOSTICS.
hosts the
for offline documentation viewers.
fixed an assertion error when using JSON_DIAGNOSTICS.
provided an important fix to compile C++17 code with Clang 9.
fixed a warning for shadowed variables.
fixed typos in the operator[] documentation.
fixed spelling mistakes in comments.
fixed typos in documentation.
corrected the parameter name in the parse documentation.
fixed a link to the FAQ.
implemented std::string_view support for object keys and made dozens of other improvements.
implemented the Binary JData (BJData) format.
added macros NLOHMANN_DEFINE_TYPE_INTRUSIVE_WITH_DEFAULT and NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE_WITH_DEFAULT.
adjusted to code to not clash with Arduino defines.
fixed the output of meta() for MSVC.
fixed a check for std::filesystem.
fixed a typo.
fixed a typo.
fixed a typo.
fixed a typo.
fixed the CITATION.CFF file.
added a clarification on macro usage to the documentation.
refactored the tests to use CHECK_THROWS_WITH_AS.
fixed a typo.
fixed a whitespace error.
fixed a build error when including <windows.h>.
moved .pc and .cmake files to share directory.
added the patch_inplace function.
highlighted common usage patterns in the README file.
added the Visual Studio output directory to .gitignore.
improved the performance of the vector output adapter.
fixed the std::iterator_traits.
added macro JSON_NO_ENUM to disable default enum conversions.
fixed tests when compiling with C++20.
fixed an example in the README.
added more files to the include.zip archive.
fixed a compilation issue when typedefs with certain names were present.
improved the description of an example.
updated the year in the README file.
fixed a warning.
fixed a typo.
added a note to an example in the README file.
fixed a typo.
fixed the Clang detection.
added a Doozer badge.
fixed the string conversion with Clang.
fixed a Doxygen error.
removed an invalid parameter from CMake.
fixed a link in the README file.
fixed a warning.
added ensure_ascii to the dump function.
fixed the sed discovery in the Makefile.
implemented SFINAE-friendly iterator_traits.
fixed a typo in the README.
added operator/= and operator/ to construct JSON pointers.
added support for afl-fuzz testing.
fixed a typo in the README.
improved the CMake testing.
fixed a typo.
added the possibility to define a custom base class.
fixed typos in the documentation.
added Bazel build support.
fixed typos in the documentation.
added a CIFuzz CI GitHub action.
fixed the debug pretty-printer.
bumped the years in the README.
cleaned up the badges of used services.
fixed a build error.
fixed a typo in a CMake file.
fixed the custom allocator support.
fixed some security issues in the GitHub workflows.
add vcpkg version badge.
added tests.
fixed the use of get<> calls.
fixed a typo in the CODEOWNERS file.
fixed a typo.
fixed a typo in the README.
fixed a parenthesis in the documentation.
fixed the examples to catch exception by const&.
fixed a parenthesis in the documentation.
fixed a compilation error.
fixed a compilation error.
fixed a deprecation warning in CMake.
added macros for serialization-only types.
fixed typos.
fixed the MinGW workflow.
added support for Apple's Swift Package Manager.
fixed the installation path in CMake.
clarified the parse error message in case a file cannot be opened.
fixed the enum conversion.
fixed a version in the documentation.
fixed the amalgamation call.
fixed a version in the documentation.
fixed an example.
fixed a warning in a test.
fixed a function name in the documentation.
fixed some typos.
fixed the GDB pretty printer.
fixed an example for JSON Pointer.
fixed some typos.
updated the Natvis file.
fixed a typo in the documentation.
fixed a URL in the contribution guidelines.
added NLOHMANN_DEFINE_DERIVED_TYPE_* macros.
allowed overriding the CMake target name.
made iterator_proxy_value a std::forward_iterator.
added type conversion support for std::optional.
added exceptions when nullptr is passed to parse.
fixed number parsing when EINTR set in errno.
generated a pkg-config file that follows the pkg-config conventions.
optimized the binary get_number implementation.
added type conversion support for multidimensional arrays.
added ONLY_SERIALIZE for NLOHMANN_DEFINE_DERIVED_TYPE_* macros.
removed alwayslink=True Bazel flag.
added diagnostic positions to exceptions.
fixed NLOHMANN_DEFINE_TYPE_INTRUSIVE_WITH_DEFAULT with an empty JSON instance.
added support for BSON uint64 serialization/deserialization.
updated the documentation.
added support for BJDATA optimized binary array type.
added support for diagnostic positions.
templated to NLOHMANN_DEFINE_TYPE macros to also support ordered_json.
added support for VisionOS in the Swift Package Manager.
fixed a typo.
added CPack support.
made GDB pretty printer work with Python 3.8.
fixed a compiler warning.
made std::filesystem::path conversion to/from UTF-8 encoded string explicit.
Thanks a lot for helping out! Please
if I forgot someone.
Used third-party tools
The library itself consists of a single header file licensed under the MIT license. However, it is built, tested, documented, and whatnot using a lot of third-party tools and services. Thanks a lot!
amalgamate.py - Amalgamate C source and header files
to create a single header file
for fuzz testing
for
on Windows
for automatic source code indentation
for compilation with code sanitizers
for build automation
for further
to measure
for
for static analysis
for the unit tests
to generate the
to implement the benchmarks
to avoid re-inventing several compiler-agnostic feature macros
to process coverage information and create an HTML view
to implement fuzz testing for OSS-Fuzz
for the style of the documentation site
for the documentation site
for continuous fuzz testing of the library (
)
for automating maintainer tasks such as closing stale issues, requesting missing information, or detecting toxic comments.
to check for correct memory management
Notes
Standards compliance
The library targets strict conformance with
. Both the original
and its updated revision are exercised in CI; their test data is downloaded from
at configure time rather than committed to this repository (see
):
The updated revision runs all mandatory y_ (must-accept) and n_ (must-reject) cases through the strict
entry point; the original suite runs its n_ cases through parse() and its y_ cases through
.
The i_ (implementation-defined) cases are, by RFC 8259, free to be accepted or rejected, so "passing all i_ cases" is not a meaningful conformance metric. The library makes deliberate, documented choices there: nesting depth is not artificially limited, a leading UTF-8 byte order mark is silently ignored,
are forwarded unchanged, invalid UTF-8 and lone/unpaired UTF-16 surrogates are rejected (stricter than required), and a number that cannot be stored without becoming NaN/INF raises
.
One behavioral nuance is worth calling out, because a superficial test often misreads it as non-compliance:
is strict and rejects trailing data after a value, whereas
follows relaxed iostream semantics — it parses a single value and leaves the stream positioned right after it. Feeding "a valid document followed by trailing bytes" through operator>> reports success; the same input through parse() is rejected. This is a documented two-API design, not a conformance gap. See
for details.
Character encoding
The library supports Unicode input as follows:
Only UTF-8 encoded input is supported, which is the default encoding for JSON according to
.
std::u16string and std::u32string can be parsed, assuming UTF-16 and UTF-32 encoding, respectively. These encodings are not supported when reading from files or other input containers.
Other encodings such as Latin-1 or ISO 8859-1 are not supported and will yield parse or serialization errors.
will not be replaced by the library.
Invalid surrogates (e.g., incomplete pairs such as \uDEAD) will yield parse errors.
The strings stored in the library are UTF-8 encoded. When using the default string type (std::string), note that its length/size functions return the number of stored bytes rather than the number of characters or glyphs.
When you store strings with different encodings in the library, calling
may throw an exception unless json::error_handler_t::replace or json::error_handler_t::ignore are used as error handlers.
To store wide strings (e.g., std::wstring), you need to convert them to a UTF-8 encoded std::string before, see
.
Comments in JSON
This library does not support comments by default. It does so for three reasons:
Comments are not part of the
. You may argue that // or /* */ are allowed in JavaScript, but JSON is not JavaScript.
This was not an oversight: Douglas Crockford
in May 2012:
I removed comments from JSON because I saw people were using them to hold parsing directives, a practice which would have destroyed interoperability. I know that the lack of comments makes some people sad, but it shouldn't.
Suppose you are using JSON to keep configuration files, which you would like to annotate. Go ahead and insert all the comments you like. Then pipe it through JSMin before handing it to your JSON parser.
It is dangerous for interoperability if some libraries would add comment support while others don't. Please check
The Harmful Consequences of the Robustness Principle
on this.
However, you can set parameter ignore_comments to true in the parse function to ignore // or /* */ comments. Comments will then be treated as whitespace.
Trailing commas
The JSON specification does not allow trailing commas in arrays and objects, and hence this library is treating them as parsing errors by default.
Like comments, you can set parameter ignore_trailing_commas to true in the parse function to ignore trailing commas in arrays and objects. Note that a single comma as the only content of the array or object ([,] or {,}) is not allowed, and multiple trailing commas ([1,,]) are not allowed either.
This library does not add trailing commas when serializing JSON data.
For more information, see
JSON With Commas and Comments (JWCC)
.
Order of object keys
By default, the library does not preserve the insertion order of object elements. This is standards-compliant, as the
defines objects as "an unordered collection of zero or more name/value pairs".
If you do want to preserve the insertion order, you can try the type
. Alternatively, you can use a more sophisticated ordered map like
(
) or
(
).
See the
for more information.
Memory Release
We checked with Valgrind and the Address Sanitizer (ASAN) that there are no memory leaks.
If you find that a parsing program with this library does not release memory, please consider the following case, and it may be unrelated to this library.
Your program is compiled with glibc. There is a tunable threshold that glibc uses to decide whether to actually return memory to the system or whether to cache it for later reuse. If in your program you make lots of small allocations and those small allocations are not a contiguous block and are presumably below the threshold, then they will not get returned to the OS. Here is a related issue
.
Further notes
The code contains numerous debug assertions which can be switched off by defining the preprocessor macro NDEBUG, see the
. In particular, note
implements unchecked access for const objects: If the given key is not present, the behavior is undefined (think of a dereferenced null pointer) and yields an
if assertions are switched on. If you are not sure whether an element in an object exists, use checked access with the
. Furthermore, you can define JSON_ASSERT(x) to replace calls to assert(x). See the
documentation on runtime assertions
for more information.
As the exact number type is not defined in the
, this library tries to choose the best fitting C++ number type automatically. As a result, the type double may be used to store numbers which may yield
in certain rare situations if floating-point exceptions have been unmasked in the calling code. These exceptions are not caused by the library and need to be fixed in the calling code, such as by re-masking the exceptions prior to calling library functions.
The code can be compiled without C++ runtime type identification features; that is, you can use the -fno-rtti compiler flag.
Exceptions are used widely within the library. They can, however, be switched off with either using the compiler flag -fno-exceptions or by defining the symbol JSON_NOEXCEPTION. In this case, exceptions are replaced by abort() calls. You can further control this behavior by defining JSON_THROW_USER (overriding throw), JSON_TRY_USER (overriding try), and JSON_CATCH_USER (overriding catch). Note that JSON_THROW_USER should leave the current scope (e.g., by throwing or aborting), as continuing after it may yield undefined behavior. Note the explanatory
string of exceptions is not available for MSVC if exceptions are disabled, see
. See the
for more information.
Execute unit tests
To compile and run the tests, you need to execute
mkdir build cd build cmake .. -DJSON_BuildTests=On cmake --build . ctest --output-on-failureNote that during the ctest stage, several JSON test files are downloaded from an
. If policies forbid downloading artifacts during testing, you can download the files yourself and pass the directory with the test files via -DJSON_TestDataDirectory=path to CMake. Then, no Internet connectivity is required. See
for more information.
If the testdata is not found, several test suites will fail like this:
=============================================================================== json/tests/src/make_test_data_available.hpp:21: TEST CASE: check test suite is downloaded json/tests/src/make_test_data_available.hpp:23: FATAL ERROR: REQUIRE( utils::check_testsuite_downloaded() ) is NOT correct! values: REQUIRE( false ) logged: Test data not found in 'json/cmake-build-debug/json_test_data'. Please execute target 'download_test_data' before running this test suite. See <https://github.com/nlohmann/json#execute-unit-tests> for more information. =============================================================================== In case you have downloaded the library rather than checked out the code via Git, test cmake_fetch_content_configure will fail. Please execute ctest -LE git_required to skip these tests. See
for more information.
Some tests are requiring network to be properly execute. They are labeled as git_required. Please execute ctest -LE git_required to skip these tests. See
for more information.
Some tests change the installed files and hence make the whole process not reproducible. Please execute ctest -LE not_reproducible to skip these tests. See
for more information. Furthermore, assertions must be switched off to ensure reproducible builds (see
).
Note you need to call cmake -LE "not_reproducible|git_required" to exclude both labels. See
for more information.
As Intel compilers use unsafe floating point optimization by default, the unit tests may fail. Use flag
then.