Skip to content

v0.6.0 — Turbo Mode

Latest

Choose a tag to compare

@rushikeshmore rushikeshmore released this 06 Apr 20:14
· 4 commits to main since this release

What's New

Turbo Mode (--turbo)

30-55x faster encode speed with ~2% ratio tradeoff. Same .dcx format — decompression is unchanged.

datacortex compress logs.ndjson --turbo        # 99 MB/s encode
datacortex bench corpus/ -m fast --turbo       # benchmark turbo

Benchmarks (Apple M-series):

File Normal Encode Turbo Encode Speedup
GH Archive 10MB 2.7 MB/s 169 MB/s 55x
NDJSON 10K rows 3.3MB 2.3 MB/s 68 MB/s 30x
Twitter API 617KB 2.2 MB/s 87 MB/s 40x

Improved Compression Ratio

  • Raised brotli quality to 10 for 1-16MB files (was 9)
  • GH Archive: 12.9% → 12.5% ratio (closed 78% of the brotli-11 gap)
  • Overall corpus: 9.8% → 9.6% ratio

Full Changelog

  • feat: add turbo mode for 30-55x faster encode speed
  • perf: raise brotli quality to 10 for 1-16MB files
  • 3 new turbo mode tests (393 total)
  • Updated README with turbo docs and benchmarks