A memory-optimized and data-oriented JSON library written in C++.

Includes benchmarks and Google Test support.

• 16-Bytes JSON values (12 on x86)
• 24-Bytes JSON members (16 on x86)
• Short string optimization up to 14 characters (10 on x86)
• Specialized array types for bool, int64_t and double
• Custom StringPool for string deduplication
  • based on an optimized intrusive hash table
  • shareable for easy reuse on consecutive parsing
  • with zero-copy support for const strings
• Allocator-aware PoolAllocator for memory management
  • designed as a slab allocator with dead-cells recycling
  • composable with Heap and StackAllocator
• Fast and easy-to-use reference-based API

This library doesn't include a JSON parser and relies on external 3rd-party for this purpose. RapidJSON is integrated as a default backend to demonstrate usage in the benchmark example.

Benchmark configuration:

• Platform: Windows 10 64-bits, i7-10875h
• 3rd-parties: xxHash 0.8.1, RapidJSON 1.1.0
• Compiler: MinGW 8.1, -O2 -DNDEBUG

Notes:

• Memory benchmarks were done with default chunk size for both libraries (RapidJSON: 65K, LFJSON: 2*32K)
• Speed benchmarks were done with MinGW-64 on Windows 10. Results may vary depending on platform/compiler.

This project uses git submodule to include Google Test repository:

$ git clone https://github.com/gaujay/lfjson.git
$ cd lfjson
$ git submodule init && git submodule update

LFJSON also uses xxHash as an optional 3rd-party to speed up its StringPool hashing requirements. To disable this dependency, just set the flag LFJ_NO_XXHASH.

Support GCC/MinGW and MSVC (see 'CMakeLists.txt').

You can open 'CMakeLists.txt' with a compatible IDE or use command line:

$ mkdir build
$ cd build
$ cmake ..
$ make <target> -j

LFJSON - Apache License 2.0

xxHash - BSD 2-Clause License

RapidJSON - MIT License + 3rd-parties