What’s New in StarRocks 3.2? Key Features and Usability Enhancements

Usability Improvements

Table creation now supports optimized random bucketing, faster schema evolution via fast_schema_evolution property, and an OPTIMIZE TABLE command for restructuring tables.

Data import adds a PIPE command for large‑scale ingestion from S3 or HDFS, automatically splitting large jobs into smaller tasks and continuously monitoring source directories. The FILES table function is enhanced with support for Azure/GCP Parquet/ORC files, columns_from_path extraction, and complex types (ARRAY, MAP, STRUCT).

Export syntax aligns with import using INSERT INTO FILES() to write data to S3 or HDFS in Parquet format, enabling a unified import/export workflow.

Query layer introduces an HTTP SQL API for executing SELECT, SHOW, EXPLAIN, and KILL statements over HTTP, and a new Runtime Profile with text‑based analysis commands ( SHOW PROFILELIST, ANALYZE PROFILE, EXPLAIN ANALYZE).

Storage‑Compute Separation

The architecture continues to converge integrated storage and compute, adding persistent index support for primary‑key tables to reduce memory usage. Future releases will store primary‑key indexes in cloud storage for fault tolerance.

Data Lake Analysis

Performance optimizations for ORC, Parquet, and CSV reading, including predicate rewrite, dictionary decoding, adaptive I/O merging, and faster COUNT operations.

Support for Hive, Iceberg, and Unified Catalogs, allowing mixed‑format tables within a single Hive Metastore or AWS Glue instance.

Statistics collection via ANALYZE TABLE for Hive and Iceberg tables improves CBO planning.

Unified Catalog simplifies management of multiple table formats and external system schema discovery.

Materialized Views

Asynchronous materialized views now support partition‑level incremental refresh for Iceberg and Paimon catalogs, automatic activation of stale views, and configurable consistency‑performance trade‑offs.

Developer tools added: Trace Rewrite and Query Dump for debugging view rewrite failures.

Synchronous materialized views now accept WHERE clauses and expression features such as CASE‑WHEN, CAST, and arithmetic.

Row‑Column Hybrid Storage

Future versions will allow primary‑key tables to use a hybrid storage format by setting STORE_TYPE='column_with_row', benefiting high‑concurrency point queries and partial column updates.

Other Enhancements

Prepared statements for higher concurrency and SQL‑injection protection.

Optimized primary‑key index persistence and multi‑disk data balancing.

New built‑in functions for strings, dates, windows, etc.

Improved compatibility with Metabase, Superset, and external catalogs.

Technical references: https://github.com/StarRocks/starrocks/blob/main/docs/zh/release_notes/release-3.2.md