General

Cross-platform
- Compilers: Visual Studio, gcc, clang, etc.
- Architectures: x86, x64, ARM, etc.
- Operating systems: Windows, Mac OS X, Linux, iOS, Android, etc.
Easy installation
- Header files only library. Just copy the headers to your project.
Self-contained, minimal dependences
- No STL, BOOST, etc.
- Only included <cstdio>, <cstdlib>, <cstring>, <inttypes.h>, <new>, <stdint.h>.
Without C++ exception, RTTI
High performance
- Use template and inline functions to reduce function call overheads.
- Internal optimized Grisu2 and floating point parsing implementations.
- Optional SSE2/SSE4.2 support.

Standard compliance

RapidJSON should be fully RFC4627/ECMA-404 compliance.
Support Unicode surrogate.
Support null character ("\\u0000")
- For example, ["Hello\\u0000World"] can be parsed and handled gracefully. There is API for getting/setting lengths of string.

Support UTF-8, UTF-16, UTF-32 encodings, including little endian and big endian.
- These encodings are used in input/output streams and in-memory representation.
Support automatic detection of encodings in input stream.
Support transcoding between encodings internally.
- For example, you can read a UTF-8 file and let RapidJSON transcode the JSON strings into UTF-16 in the DOM.
Support encoding validation internally.
- For example, you can read a UTF-8 file, and let RapidJSON check whether all JSON strings are valid UTF-8 byte sequence.
Support custom character types.
- By default the character types are char for UTF8, wchar_t for UTF16, uint32_t for UTF32.
Support custom encodings.

SAX (Simple API for XML) style API
- Similar to SAX, RapidJSON provides a event sequential access parser API (rapidjson::GenericReader). It also provides a generator API (rapidjson::Writer) which consumes the same set of events.
DOM (Document Object Model) style API
- Similar to DOM for HTML/XML, RapidJSON can parse JSON into a DOM representation (rapidjson::GenericDocument), for easy manipulation, and finally stringify back to JSON if needed.
- The DOM style API (rapidjson::GenericDocument) is actually implemented with SAX style API (rapidjson::GenericReader). SAX is faster but sometimes DOM is easier. Users can pick their choices according to scenarios.

Recursive (default) and iterative parser
- Recursive parser is faster but prone to stack overflow in extreme cases.
- Iterative parser use custom stack to keep parsing state.
Support in situ parsing.
- Parse JSON string values in-place at the source JSON, and then the DOM points to addresses of those strings.
- Faster than convention parsing: no allocation for strings, no copy (if string does not contain escapes), cache-friendly.
Support 32-bit/64-bit signed/unsigned integer and double for JSON number type.
Support parsing multiple JSONs in input stream (kParseStopWhenDoneFlag).
Error Handling
- Support comprehensive error code if parsing failed.
- Support error message localization.

RapidJSON checks range of numerical values for conversions.
Optimization for string literal
- Only store pointer instead of copying
Optimization for "short" strings
- Store short string in Value internally without additional allocation.
- For UTF-8 string: maximum 11 characters in 32-bit, 15 characters in 64-bit.
Optionally support std::string (define RAPIDJSON_HAS_STDSTRING=1)

Support rapidjson::GenericStringBuffer for storing the output JSON as string.
Support rapidjson::FileReadStream and rapidjson::FileWriteStream for input/output FILE object.
Support custom streams.

Minimize memory overheads for DOM.
- Each JSON value occupies exactly 16/20 bytes for most 32/64-bit machines (excluding text string).
Support fast default allocator.
- A stack-based allocator (allocate sequentially, prohibit to free individual allocations, suitable for parsing).
- User can provide a pre-allocated buffer. (Possible to parse a number of JSONs without any CRT allocation)
Support standard CRT(C-runtime) allocator.
Support custom allocators.