Timelaketimelake
DocsBenchmarksRoadmapSpecContribute
Apache 2.0 ·
TimelaketimelakeThe open time-series lakehouse.
v0 · Apache 2.0 · Contribute · GitHubBuilt on Apache Arrow, Apache DataFusion, Apache Parquet, and Apache Iceberg.
Roadmap

Milestone

Exact answers before compaction

Done0.5.0Updated 2026-10-17View on GitHub

Rows sent twice count once, from the moment they land: dedup on read, an insertion time per row, and cold partitions compacted sooner.

Specs

  • 12Late data and dedup on readDone0.5.0
  • 19CompactionDone0.1.0, 0.5.0, 0.6.0, 0.7.0, unreleased

Summary

Devices resend data, and compaction runs later. The Timelake reader keeps only the latest write of each reading from the moment it lands, each row records when it arrived, compaction never lets an older write win, and partitions that are no longer written are compacted sooner so dedup on read only touches recent data.

History

  • 2026-10-16: dedup on read, insertion time, cold partitions, compaction fix (0.5.0)
  • 2026-10-19: milestone written when the specs were regrouped one per feature.