Skip to content

Releases

Release Policy

Apache Paimon Rust follows Semantic Versioning. All releases are published to crates.io.

Upcoming

0.6.0 (In Development)

See GitHub Issues for the current roadmap.

Past Releases

0.5.0

Released on November 13, 2026.

Source, signature, and SHA-512 checksum · KEYS.

The minimum supported Rust version for 0.5.0 is 1.95.

Key features:

  • Append, primary-key and data-evolution compaction, with lookup changelog production and incremental primary-key compaction
  • Prepared and captured mutation plans, native MERGE execution, distributed dynamic-bucket execution, and expanded Arrow row-ID inputs
  • Native batch and streaming read improvements, REST query authorization, catalog branch and version management, and external table engine time travel
  • DataFusion 55.2 integration with expanded SQL functions, JSON operations, lateral collection generators, comments, and tag, branch and consumer procedures
  • Built-in full-text search, BTree-accelerated upsert key lookups, global-index maintenance, and scheduled hybrid search execution
  • Broader Java-compatible schema evolution, nested Parquet projection, temporal and decimal casts, VARIANT and ROW reads, and manifest codec support
  • Node-shared disk block caching, cached Parquet metadata, coalesced cache reads, and reuse of BTree and ROW sidecar metadata
  • Expanded C commit-message and schema APIs, Python streaming SQL results, temporal matching kernels and compaction APIs

Compatibility changes:

  • The release lockfile uses DataFusion 55.2.0, Arrow 59.3.2 and Parquet 59.3.2.
  • Removed Lumina vector index support; use vindex DiskANN via create_global_index with index_type => 'diskann'. Rebuild existing Lumina indexes in the vindex format.
  • Full-text search is now built in. Remove the former fulltext Cargo feature from downstream dependency feature lists.

0.4.0

Released on October 12, 2026.

Source, signature, and SHA-512 checksum · KEYS.

The minimum supported Rust version for 0.4.0 is 1.94.

Key features:

  • Expanded primary-key vector and hybrid search with vector/full-text fusion, streamed index builds, DiskANN and quantized IVF backends, scalar pre-filtering, and externally planned bucket splits
  • Java-compatible file and global indexes, including Bloom, Bitmap V1/V2 reads and V2 writes, multi-value array indexes, and exact FM substring search
  • Native read-only object tables and TableType-based routing to pluggable external table engines
  • Batch incremental-diff reads, watermark-based time travel, and deletion-vector merge-on-read for uncompacted primary-key snapshots
  • Expanded DataFusion SQL with ALTER COLUMN, database lifecycle statements, configurable runtime resources, and the Java-compatible $consumers system table
  • Expanded Go table-write support plus postpone-table fixed-bucket write primitives for Rust, C, and Go, with Python commit rollback
  • Standalone BlobDescriptor batch reads and seekable streams, with Go batch and stream APIs and MAP<STRING, BLOB> descriptor reads
  • Extensible and cached I/O through caller-provided OpenDAL operators, an in-memory local block cache, and C cache callbacks

API compatibility changes:

  • BTreeIndexReader::null_bitmap() is now asynchronous and returns io::Result<&RoaringTreemap>. Downstream Rust callers must replace reader.null_bitmap() with reader.null_bitmap().await? and handle load errors.
  • BTree readers load the null bitmap on first access and reuse successful loads within the reader. Null bitmap I/O, CRC, and deserialization errors now surface on first access rather than when opening the reader. Failed loads can be retried.

Go Linux amd64 prebuilt binaries require glibc 2.38 or later, OpenSSL 3, and libffi.so.8. See Go prerequisites.

0.3.0

Key features:

  • Primary-key changelog writes plus aggregation and partial-update merge improvements
  • Java-compatible Mosaic, Vortex, schema-evolution, and data-evolution read paths
  • Lumina and vindex vector search, hybrid search, and BTree, bitmap, and full-text global indexes
  • Dedicated vector storage plus BLOB and VARIANT type support
  • DataFusion 54 integration, REST catalog views and functions, and a filesystem-backed REST server
  • Expanded Python DataFrame-style read/write APIs and C and Go binding coverage

0.2.0

Key features:

  • Primary-key table read/write support with sort-merge deduplication
  • DataFusion DML support (INSERT OVERWRITE, TRUNCATE TABLE, DROP PARTITION, CALL procedures)
  • System tables ($snapshots, $tags, $manifests, $schemas, $partitions, $files, $table_indexes)
  • Multi-catalog support, session-scoped dynamic options (SET/RESET), and temporary tables
  • Lumina vector index read infrastructure and Vortex columnar file format support
  • Exact COUNT(*) pushdown via partition statistics
  • Partial-update merge engine support (fixed-bucket and dynamic-bucket)

0.1.0

The first release of Apache Paimon Rust.

Key features:

  • Paimon table format reader (Parquet, ORC, Avro) with schema evolution
  • Filesystem and REST Catalog support
  • Apache DataFusion integration (catalog, predicate/projection push-down, time travel, partition pruning, statistics)
  • Deletion vectors, data-level stats pruning, BTree global index, Tantivy full-text search index
  • Commit pipeline with SnapshotCommit abstraction
  • Go / Python bindings