vortex-go - source code release

I'm happy to share:

A native Go implementation of the Vortex columnar file format.

Vortex is a next-generation columnar file format (Linux Foundation project) designed for analytical workloads. It achieves significantly faster scans and better compression than Apache Parquet through adaptive encoding selection and lazy materialisation.

vortex-go provides full read and write support from pure Go — no CGo, no Rust FFI.

Key Features

• 25 encoding decoders — ALP, BitPacking, Dict, FoR, FSST, Pcodec, RLE, Zstd, and more
• Compressed writer — automatic encoding selection (dict, FoR+BitPacking, Pcodec, zone maps)
• Column projection — read only the columns you need, skip the rest
• Zone map pruning — skip entire chunks based on min/max filter predicates
• Full cross-compatibility — files readable by the Rust reference, Python vortex, and DuckDB
• CLI tools — vtxinfo for file inspection, vtxconvert for Parquet-to-Vortex conversion
• Zero external dependencies at runtime — only Go standard library plus FlatBuffers/Protobuf

Performance

On a 7-column schema at 1M rows (Intel 14th Gen, Go 1.24):

Zone map pruning on 100K rows (10 chunks, filter selects 10%): 8.5× faster than full scan.

The source code is available at:  https://github.com/LaurieRhodes/vortex-go