Implement Smile codec with encoder, decoder, comprehensive tests, and improved architecture

[Copilot]

This PR implements a complete Smile binary format codec according to the official specification v1.0.6. Smile is an efficient JSON-compatible binary data format that provides significant space and parsing efficiency improvements over JSON while maintaining full compatibility with the JSON data model.

Implementation Overview

The implementation includes:

Core Components

• SmileEncoderBase: High-performance base encoder implementing the core Smile encoding logic
• SmileEncoder: Main encoder class extending SmileEncoderBase following the established codec pattern
• SmileDecoderBase: Robust base decoder implementing the core Smile decoding logic
• SmileDecoder: Main decoder class extending SmileDecoderBase following the established codec pattern
• Utility Functions: VInt encoding/decoding, ZigZag encoding, safe binary encoding, IEEE 754 float encoding
• Type Definitions: Complete TypeScript interfaces with proper type safety integrated into respective classes

Architecture Improvements

• Inheritance Pattern: Follows the same inheritance structure as CborEncoder/CborDecoder for consistency
• Interface Compliance: Implements BinaryJsonEncoder and StreamingBinaryJsonEncoder interfaces
• Performance Optimizations: Uses const enum declarations for better tree-shaking and performance
• Clean Separation: Encoder/decoder options and types are co-located with their respective classes

Key Features

• Full support for all Smile token types (strings, numbers, arrays, objects, binary data)
• Shared string optimization for both values and property names to reduce output size
• Safe 7-bit binary encoding to avoid reserved byte values
• Proper 4-byte header validation and version checking
• Graceful handling of JavaScript limitations (very large integers beyond safe range are converted to strings)

Usage Example

import {SmileEncoder, SmileDecoder} from '@jsonjoy.com/json-pack/smile';

// Encode JavaScript value to Smile binary format
const encoder = SmileEncoder.create({
  sharedStringValues: true,    // Enable string deduplication
  sharedPropertyNames: true    // Enable property name deduplication
});
const encoded = encoder.encode({name: 'John', age: 30, city: 'NYC'});

// Decode back to JavaScript value
const decoder = SmileDecoder.create(encoded);
const decoded = decoder.decode();
console.log(decoded); // {name: 'John', age: 30, city: 'NYC'}

Quality Assurance

The implementation includes comprehensive testing with 104 test cases covering:

• Round-trip integrity tests ensuring perfect data fidelity
• Fuzzer tests with randomly generated data structures
• Edge case handling for Unicode strings, large numbers, and binary data
• Shared string optimization verification
• All data types supported by the Smile specification

Specification Compliance

This implementation follows the official Smile format specification v1.0.6, including:

• Correct token encoding for all value and key modes
• Proper shared string reference management with table rotation at 1024 entries
• VInt encoding with ZigZag for signed integers
• IEEE 754 float encoding using 7-bit chunks to avoid reserved bytes
• Long string handling with end markers

The codec integrates seamlessly with the existing json-pack architecture and follows the same patterns as other codecs (CBOR, MessagePack, etc.) in the repository.

Fixes #34.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[streamich]

Only types in main index.ts. Remove this line.

[Copilot]

Fixed the main index.ts to only export types, not the smile implementation. Removed the export line as requested. (commit 5cece23)

[streamich]

Revert changes to this file.

[Copilot]

Reverted all changes to package-lock.json as requested. (commit 5cece23)

[streamich]

Follow the patter in other codecs. See CborDecoder. The SmileDecoder should follow similar inheritance and name the methods similarly.

[Copilot]

Refactored SmileDecoder to follow the CborDecoder pattern with proper inheritance. Created SmileDecoderBase that extends the base functionality and SmileDecoder that extends SmileDecoderBase, following the same structure and method naming as CborDecoder. (commit 643ccfd)

[streamich]

Use const enum.

[Copilot]

Converted all enums to const enum in constants.ts for better performance as requested. (commit 5cece23)

[streamich]

Use const enum.

[Copilot]

Converted all enums to const enum in constants.ts for better performance as requested. (commit 5cece23)

[streamich]

Move this interface into SmileDecoder file.

[Copilot]

Moved the SmileDecoderOptions interface into the SmileDecoder file as requested. The types.ts file has been removed and all interfaces are now in their respective files. (commit 5cece23)

[streamich]

In this file:

• Make sure all methods are as performant as possible.
• Do not add unnecessary or trivial comments.
• Remove all newlines inside functions.

[Copilot]

Started optimizing the util.ts file by removing unnecessary comments and some newlines inside functions. More optimization is in progress to ensure all methods are as performant as possible. (commit 5cece23)

[streamich]

Follow inheritance and code style and method naming pattern as in CborEncoder.

[Copilot]

Refactored SmileEncoder to follow the CborEncoder inheritance and code style pattern. Created SmileEncoderBase that implements the base functionality and SmileEncoder that extends SmileEncoderBase, following the same structure and method naming as CborEncoder. (commit 643ccfd)

[streamich]

• Move these tests to codec.spec.ts.
• In this file, instead, use existing fixtures. See other automated.spec.ts test files. They import documents and binary documents from fixture files.

[streamich]

Do not implement a random JSON generator here. See other fuzzer.spec.ts files, they use an existing random JSON generator.