Skip to content

CHANGELOG.md

Latest commit

 

History

History
211 lines (99 loc) · 15.5 KB

CHANGELOG.md

File metadata and controls

211 lines (99 loc) · 15.5 KB
Raw

Changelog

All notable changes to timeseries-table-format are documented here beginning with the unified 0.3.0 release. This is the only changelog updated for current Rust library, CLI, and Python releases.

0.7.0

Bug Fixes

  • Preserve segment metadata error discriminants (#423) (1418593)

Features

  • Return structured reports from append (#419) (032f609) - Breaking: return structured reports from append (#419)

  • Expose physical row group statistics in append diagnostics (#421) (f84cfd7)

0.6.1

Features

  • Expose phase-level telemetry for streaming append (3fce372)

  • Add retention-aware vacuum for interrupted writes (0a85239)

0.6.0

Features

  • Add bounded compressed append writer settings (971aa64)

0.5.0

Bug Fixes

  • Make streaming append cleanup safer and APIs clearer (#371) (48da5b5)

  • Reject duplicate index intervals within one append (0f07ebf)

Code Refactoring

  • Extract notebook display stylesheet (9586974)

  • Remove legacy path-first append APIs (b967711) - Breaking: remove legacy path-first append APIs

  • Rename ordered index granularity and intervals (0f8a6fd) - Breaking: rename ordered index granularity and intervals

  • Unify index granularity across Python and CLI (828204c) - Breaking: unify index granularity across Python and CLI

  • Clarify index granularity interval APIs (b3b5b4c) - Breaking: clarify index granularity interval APIs

  • Rebuild subsystem error foundations (7e71170)

  • Preserve typed append error sources (a07f7f0) - Breaking: preserve typed append error sources

  • Organize table errors by operation (7365c59) - Breaking: organize table errors by operation

  • Harden public error APIs and preserve typed causes (f9183ed) - Breaking: harden public error APIs and preserve typed causes

  • Organize metadata modules (#405) (ed63cbf)

  • Preserve typed error sources across table operations (2633392) - Breaking: preserve typed error sources across table operations

Documentation

  • Make the documentation site canonical for Python package guidance (01f2747)

Features

  • Expose native Rust diagnostics through standard Python logging (9e8a147)

  • Add Python Arrow stream append bindings (#366) (791f7ef)

  • Upgrade DataFusion to 55 and Arrow/Parquet to 59.2, align public APIs, and document migration changes (8c36850) - Breaking: Public Arrow and DataFusion API types now use the upgraded dependency versions. Empty Parquet directories are rejected at registration, and the Rust MSRV is now 1.94.0.

  • Add structured diagnostics for table operations, recovery, transaction commits, and scan planning (ec264d1)

  • Surface configurable structured diagnostics from the tstable CLI (17c8d7f)

  • Add core Rust streaming append transaction (#364) (0ab1812)

  • Route CLI Parquet imports through streaming append (#365) (4902723)

  • Add per-append Parquet row-group sizing (5d47600)

  • Support lossless scalar widening during append (de01516)

  • Preserve typed read operation errors (#396) (be27912)

  • Establish protocol v7 metadata foundation (#402) (edd8fbd)

  • Enforce protocol compatibility for table reads and writes (1199c7a)

  • Add strict protocol v7 migration tool (#407) (94d66e4)

Testing

Perf

  • Benchmark streaming append against path-first pipeline (#367) (61cbe2e)

  • Benchmark direct lossless widening during append (752ed44)

0.4.0

Bug Fixes

  • Upgrade PyO3 to 0.29 (#301) (888fd9a)

  • Prevent failed concurrent appends from leaking or deleting table artifacts (#291) (e36915e)

  • Remove orphaned Parquet copies after failed appends (#293) (fdd6d73)

  • Upgrade CLI dependencies (#304) (740dca6)

  • Upgrade SNAFU to 0.9 (#305) (d5f129d)

  • Prevent ordered index columns from being configured as entity identities (6a0ef72)

  • Render logical overlap bucket ranges (#342) (0ad9e58)

Code Refactoring

  • Remove the obsolete timestamp pruning engine after the DataFusion migration (#308) (aacea8e)

Documentation

  • Overhaul project documentation (#317) (bd971a4)

  • Clarify unordered row behavior and entity-aware coverage (4eb3dac)

Features

  • Expose multi-entity tables and structured overlap diagnostics in Python (902507c)

  • Inspect Parquet files without full payload buffering (#282) (7056997)

  • Finalize bounded Parquet append profiling and RSS proof (#284) (823edc3)

  • Stream time-series range scans incrementally with bounded memory (#290) (df76bbb)

  • Introduce ordered-index metadata and 64-bit coverage (#295) (1d9bb52)

  • Support signed and unsigned integer indexes for Parquet appends (#296) (8002ec0)

  • Support signed and unsigned integer range scans and coverage queries (#297) (467b2ed)

  • Speed up signed and unsigned integer index queries by skipping irrelevant segment files (5eecce1)

  • Improve timestamp query pruning with shared metadata predicates (#307) (4cbffec)

  • Create and query Timestamp, Int64, and UInt64 ordered-index tables from CLI and Python (57a6f20)

  • Add per-entity coverage model with deterministic sidecar encoding (8d17c54)

  • Compute exact entity-scoped coverage for multi-entity Parquet segments (#320) (e4a1e8a)

  • Support entity-scoped append, overlap detection, and recovery (3ee339c)

  • Query coverage and gaps safely for individual entities (4227763)

  • Persist exact segment entity layouts for reliable optimization planning (d6625c5)

  • Prune conflicting single-entity segments from DataFusion scans using log metadata (da04bb9)

  • Add explicit entity layout optimization to the CLI and Python API (5a12258)

  • Support string and integer entity columns across the full table lifecycle (6ff0e80)

Testing

  • Characterize timestamp segment pruning across SQL transforms and DST (#306) (cc8e9aa)

Perf

  • Add bounded-memory and first-batch regression coverage for range scans (#292) (63e7676)

Legacy package history

Before 0.3.0, the repository released several packages independently. Their historical changelogs remain available at the last tag in each release stream: