This release (Release 2026.03.30) introduces a series of new features, enhancements, and fixes.

• Domestic Regions

  • Alibaba Cloud (Shanghai)
  • Tencent Cloud (Shanghai / Beijing / Guangzhou)
  • AWS (Beijing)

• International Regions

  • Alibaba Cloud (Singapore)
  • AWS (Singapore)

New Features

• Added Catalog and Storage Connection connectivity validation functionality, which allows verifying the correctness of storage connection configurations using the VALIDATE_STORAGE_CONNECTION function.
• Full-text search now supports stopword removal and lemmatization capabilities, effectively improving text retrieval precision and efficiency.
• Incremental computation capabilities expanded with new support for reading Hudi-format external tables, enriching the data source types for incremental data processing.
• Data sharing functionality upgraded: UDFs can now be added to share objects and shared with other Singdata account instances, suitable for encapsulating specialized data analysis workflows, compliance templates, and other professional knowledge and capabilities, enhancing the service value of data sharing.

SQL Enhancements

Functions

• Improved compatibility for ClickHouse migration

  • Added a ClickHouse-compatible alias HASANY_CLICKHOUSE for the existing ARRAYS_OVERLAP function.
  • Added the groupArray function, which is an alias for COLLECT_LIST and has all the features of COLLECT_LIST.

• Compatibility with Doris functions

  • Added EXPLODE_JSON_ARRAY_JSON, EXPLODE_JSON_ARRAY_STRING, EXPLODE_JSON_ARRAY_INT, and EXPLODE_JSON_ARRAY_DOUBLE functions. These accept a JSON array parameter, convert it to the corresponding typed array, and then process it with the EXPLODE function. Note: These functions only support JSON objects containing elements, and must be used with LATERAL VIEW.
  • Added the EXPLODE_JSON_OBJECT table function, which expands a JSON object into multiple rows, each containing a key-value pair. Typically used to expand JSON objects into an easily queryable format. Note: This function only supports JSON objects containing elements, and must be used with LATERAL VIEW.
  • Added the ARRAY_FILTER_DORIS function, which filters array elements based on specified conditions and returns a new array of matching elements.
  • Added the ARRAY_MAP_DORIS function, which applies a lambda expression to array elements and returns a new array with the processed results.

• Added ngram tokenization index, a tokenization method for text segmentation scenarios that breaks a sentence or text segment into multiple adjacent token groups using ngram tokenization.

Behavioral Changes

• The compute cluster cancel jobs statement now includes cleaning up all queued jobs.

  • ALTER VCLUSTER <vcluster_name> CANCEL ALL JOBS; — the effective scope of this statement has been adjusted. After execution, it will cancel all currently running and currently queued jobs within the specified VCluster.

Bug Fixes

• Fixed an issue where WITHIN GROUP ORDER BY produced inconsistent results when ASC and DESC appeared together.
• Optimized COUNT DISTINCT to use a single-path aggregation when the distinct column is already in the GROUP BY fields, providing better performance.
• Fixed a sort order instability issue in complex SQL queries, controllable via the parameter set cz.optimizer.pass.through.enabled=false.