Provides functionality to encode Elixir terms into Smile binary format.

The encoder handles all JSON-compatible data types and provides optimization features such as back-references for field names and string values to reduce output size.

Supported Types

The encoder supports the following Elixir types:

• nil - Encoded as Smile null token
• true, false - Encoded as Smile boolean tokens
• Integers - Encoded with variable-length encoding (small int, 32-bit, 64-bit)
• Floats - Encoded as 64-bit IEEE double precision
• Strings (binaries) - Encoded with length optimization (tiny, small, long)
• Lists - Encoded as Smile arrays
• Maps - Encoded as Smile objects
• Atoms - Converted to strings and encoded

Encoding Options

The encoder accepts the following options:

• :shared_names (boolean, default: true) - Enable back-references for field names. When enabled, repeated object keys are stored once and referenced in subsequent occurrences, reducing output size.

• :shared_values (boolean, default: true) - Enable back-references for short string values (64 bytes or less). Similar to shared names, but for string values.

• :raw_binary (boolean, default: false) - Allow raw binary data in the output. When enabled, binary data can be included without escaping.

Examples

# Basic encoding
iex> Smile.Encoder.encode("hello")
{:ok, <<58, 41, 10, 3, ...>>}

# Encoding with options
iex> Smile.Encoder.encode(%{"key" => "value"}, shared_names: false)
{:ok, <<58, 41, 10, 0, ...>>}

Implementation Details

The encoder implements the following optimizations:

• Small integers (-16 to +15) are encoded in a single byte
• Variable-length integer encoding (VInt) for efficient number representation
• ZigZag encoding for signed integers
• String encoding optimization based on length and character set (ASCII vs Unicode)
• Back-references for repeated field names and values