Encoding

According to ECMA-404,

(in Introduction) JSON text is a sequence of Unicode code points.

The earlier RFC4627 stated that,

(in §3) JSON text SHALL be encoded in Unicode. The default encoding is UTF-8.

(in §6) JSON may be represented using UTF-8, UTF-16, or UTF-32. When JSON is written in UTF-8, JSON is 8bit compatible. When JSON is written in UTF-16 or UTF-32, the binary content-transfer-encoding must be used.

RapidJSON supports various encodings. It can also validate the encodings of JSON, and transcoding JSON among encodings. All these features are implemented internally, without the need for external libraries (e.g. ICU).

Unicode

From Unicode's official website:

Unicode provides a unique number for every character, no matter what the platform, no matter what the program, no matter what the language.

Those unique numbers are called code points, which is in the range 0x0 to 0x10FFFF.

Unicode Transformation Format

There are various encodings for storing Unicode code points. These are called Unicode Transformation Format (UTF). RapidJSON supports the most commonly used UTFs, including

  • UTF-8: 8-bit variable-width encoding. It maps a code point to 1–4 bytes.
  • UTF-16: 16-bit variable-width encoding. It maps a code point to 1–2 16-bit code units (i.e., 2–4 bytes).
  • UTF-32: 32-bit fixed-width encoding. It directly maps a code point to a single 32-bit code unit (i.e. 4 bytes).

For UTF-16 and UTF-32, the byte order (endianness) does matter. Within computer memory, they are often stored in the computer's endianness. However, when it is stored in file or transferred over network, we need to state the byte order of the byte sequence, either little-endian (LE) or big-endian (BE).

RapidJSON provide these encodings via the structs in rapidjson/encodings.h:

namespace rapidjson {
template<typename CharType = char>
struct UTF8;
template<typename CharType = wchar_t>
struct UTF16;
template<typename CharType = wchar_t>
struct UTF16LE;
template<typename CharType = wchar_t>
struct UTF16BE;
template<typename CharType = unsigned>
struct UTF32;
template<typename CharType = unsigned>
struct UTF32LE;
template<typename CharType = unsigned>
struct UTF32BE;
} // namespace rapidjson

For processing text in memory, we normally use UTF8, UTF16 or UTF32. For processing text via I/O, we may use UTF8, UTF16LE, UTF16BE, UTF32LE or UTF32BE.

When using the DOM-style API, the Encoding template parameter in GenericValue<Encoding> and GenericDocument<Encoding> indicates the encoding to be used to represent JSON string in memory. So normally we will use UTF8, UTF16 or UTF32 for this template parameter. The choice depends on operating systems and other libraries that the application is using. For example, Windows API represents Unicode characters in UTF-16, while most Linux distributions and applications prefer UTF-8.

Example of UTF-16 DOM declaration:

typedef GenericDocument<UTF16<> > WDocument;
typedef GenericValue<UTF16<> > WValue;

For a detail example, please check the example in DOM's Encoding section.

Character Type

As shown in the declaration, each encoding has a CharType template parameter. Actually, it may be a little bit confusing, but each CharType stores a code unit, not a character (code point). As mentioned in previous section, a code point may be encoded to 1–4 code units for UTF-8.

For UTF16(LE|BE), UTF32(LE|BE), the CharType must be integer type of at least 2 and 4 bytes respectively.

Note that C++11 introduces char16_t and char32_t, which can be used for UTF16 and UTF32 respectively.

AutoUTF

Previous encodings are statically bound in compile-time. In other words, user must know exactly which encodings will be used in the memory or streams. However, sometimes we may need to read/write files of different encodings. The encoding needed to be decided in runtime.

AutoUTF is an encoding designed for this purpose. It chooses which encoding to be used according to the input or output stream. Currently, it should be used with EncodedInputStream and EncodedOutputStream.

ASCII

Although the JSON standards did not mention about ASCII, sometimes we would like to write 7-bit ASCII JSON for applications that cannot handle UTF-8. Since any JSON can represent unicode characters in escaped sequence \uXXXX, JSON can always be encoded in ASCII.

Here is an example for writing a UTF-8 DOM into ASCII:

using namespace rapidjson;
Document d; // UTF8<>
// ...
StringBuffer buffer;
d.Accept(writer);
std::cout << buffer.GetString();

ASCII can be used in input stream. If the input stream contains bytes with values above 127, it will cause kParseErrorStringInvalidEncoding error.

ASCII cannot be used in memory (encoding of Document or target encoding of Reader), as it cannot represent Unicode code points.

Validation & Transcoding

When RapidJSON parses a JSON, it can validate the input JSON, whether it is a valid sequence of a specified encoding. This option can be turned on by adding kParseValidateEncodingFlag in parseFlags template parameter.

If the input encoding and output encoding is different, Reader and Writer will automatically transcode (convert) the text. In this case, kParseValidateEncodingFlag is not necessary, as it must decode the input sequence. And if the sequence was unable to be decoded, it must be invalid.

Transcoder

Although the encoding functions in RapidJSON are designed for JSON parsing/generation, user may abuse them for transcoding of non-JSON strings.

Here is an example for transcoding a string from UTF-8 to UTF-16:

#include "rapidjson/encodings.h"
using namespace rapidjson;
const char* s = "..."; // UTF-8 string
StringStream source(s);
bool hasError = false;
while (source.Peek() != '\0')
if (!Transcoder<UTF8<>, UTF16<> >::Transcode(source, target)) {
hasError = true;
break;
}
if (!hasError) {
const wchar_t* t = target.GetString();
// ...
}

You may also use AutoUTF and the associated streams for setting source/target encoding in runtime.