Vine - Datalake Format base on Rust (WIP)

Status: Work in Progress

This project aimes 'datalake table format' optimized for streaming data writes, built on Rust.

Quick Start

Build

./build.sh

This builds:

• vine-core: Rust library for Vine
• vine-spark: Spark DataSource V2 connector

Usage with Spark

// Write streaming data
spark.readStream
  .format("vine")
  .load("input-path")
  .writeStream
  .format("vine")
  .option("path", "/data/my-table")
  .start()

// Read with Spark SQL
val df = spark.read.format("vine").load("/data/my-table")
df.show()

Architecture

┌─────────────────────────────────────┐
│   Query Engines (Spark, Trino)     │
└──────────────┬──────────────────────┘
               │ DataSource API
┌──────────────▼──────────────────────┐
│  Connectors (vine-spark/vine-trino) │
└──────────────┬──────────────────────┘
               │ JNI
┌──────────────▼──────────────────────┐
│  Rust Core (vine-core)              │
│  - Fast Parquet writes              │
│  - Date-based partitioning          │
└──────────────┬──────────────────────┘
               │
┌──────────────▼──────────────────────┐
│  Storage (Parquet files)            │
│  2024-12-26/data_143025.parquet     │
│  2024-12-27/data_091500.parquet     │
└─────────────────────────────────────┘

Components

Component	Language	Status	Purpose
vine-core	Rust	WIP	Write-optimized datalake table format
vine-spark	Scala	WIP	Spark DataSource V2 connector
vine-trino	Java	Planned	Trino connector (not started)

Storage Format

• Files: Apache Parquet (columnar)
• Partitioning: Date-based directories (YYYY-MM-DD/data_HHMMSS.parquet)
• Metadata: JSON schema file (vine_meta.json)
• Types: integer, string, boolean, double

Documentation

• Presentation

Development

Build Components Individually

Rust Core

cd vine-core
cargo build --release
cargo test

Spark Connector

cd vine-spark
sbt clean assembly

Requirements

• Rust 1.70+
• Scala 2.13, sbt 1.x
• Java 11