There are myriads of [JSON](http://json.org) 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 [examples below](#examples) and you know, what I mean.
- **Trivial integration**. Our whole code consists of a single header file `json.hpp`. 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.
- **Serious testing**. Our class is heavily [unit-tested](https://github.com/nlohmann/json/blob/master/test/src/unit.cpp) and covers [100%](https://coveralls.io/r/nlohmann/json) of the code, including all exceptional behavior. Furthermore, we checked with [Valgrind](http://valgrind.org) that there are no memory leaks.
- **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**. We currently implement the parser as naive [recursive descent parser](http://en.wikipedia.org/wiki/Recursive_descent_parser) with hand coded string handling. It is fast enough, but a [LALR-parser](http://en.wikipedia.org/wiki/LALR_parser) with a decent regular expression processor should be even faster (but would consist of more files which makes the integration harder).
The single required source, file `json.hpp` is in the `src` directory or [released here](https://github.com/nlohmann/json/releases). All you need to do is add
to the files you want to use JSON objects. That's it. Do not forget to set the necessary switches to enable C++11 (e.g., `-std=c++11` for GCC and Clang).
- GCC 4.8 does not work because of two bugs ([55817](https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55817) and [57824](https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57824)) in the C++11 support. Note there is a [pull request](https://github.com/nlohmann/json/pull/212) to fix some of the issues.
- 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.
The code compiles successfully with [Android NDK](https://developer.android.com/ndk/index.html?hl=ml), Revision 9 - 11 (and possibly later) and [CrystaX's Android NDK](https://www.crystax.net/en/android/ndk) version 10.
- For GCC running on MinGW or Android SDK, the error `'to_string' is not a member of 'std'` (or similarly, for `strtod`) 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 [this site](http://tehsausage.com/mingw-to-string) and [this discussion](https://github.com/nlohmann/json/issues/136) for information on how to fix this bug. For Android NDK using `APP_STL := gnustl_static`, please refer to [this discussion](https://github.com/nlohmann/json/issues/219).
Note that in all these cases, you never need to "tell" the compiler which JSON value you want to use. If you want to be explicit or express some edge cases, the functions `json::array` and `json::object` will help:
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.
We designed the JSON class to behave just like an STL container. In fact, it satisfies the [**ReversibleContainer**](http://en.cppreference.com/w/cpp/concept/ReversibleContainer) requirement.
Any sequence container (`std::array`, `std::vector`, `std::deque`, `std::forward_list`, `std::list`) whose values can be used to construct JSON types (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 how the elements are ordered in the respective STL container.
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 are can construct an `std::string` and whose values can be used to construct JSON types (see examples above) can be used to 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.
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.
## Thanks
I deeply appreciate the help of the following people.
- [Teemperor](https://github.com/Teemperor) implemented CMake support and lcov integration, realized escape and Unicode handling in the string parser, and fixed the JSON serialization.
- [elliotgoodrich](https://github.com/elliotgoodrich) fixed an issue with double deletion in the iterator classes.
- [kirkshoop](https://github.com/kirkshoop) made the iterators of the class composable to other libraries.
- [wancw](https://github.com/wanwc) fixed a bug that hindered the class to compile with Clang.
- Tomas Åblad found a bug in the iterator implementation.
- [Aaron Burghardt](https://github.com/aburgh) 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.
- [Eric Cornelius](https://github.com/EricMCornelius) pointed out a bug in the handling with NaN and infinity values. He also improved the performance of the string escaping.
- [dariomt](https://github.com/dariomt) fixed a subtlety in MSVC type support and implemented the `get_ref()` function to get a reference to stored values.
- [twelsby](https://github.com/twelsby) 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.
- The code contains numerous debug **assertions** which can be switched off by defining the preprocessor macro `NDEBUG`, see the [documentation of `assert`](http://en.cppreference.com/w/cpp/error/assert).
- As the exact type of a number is not defined in the [JSON specification](http://rfc7159.net/rfc7159), 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 [**floating-point exceptions**](https://github.com/nlohmann/json/issues/181) 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.