V3.1.48 release notes - Hologres - Alibaba Cloud Documentation Center

These release notes describe the enhancements and bug fixes in Hologres V3.1.48.

Enhancements

Enhancement for retention functions: Hologres now supports intermediate result type inference for the Range Retention Metric under adaptive query execution (AQE).
Performance optimization for ANALYZE on MaxCompute archive storage partitions: Optimized the Auto Analyze behavior for MaxCompute partitioned tables with many archive storage partitions by reducing unnecessary statistics collection.
Support for V1-to-V2 algorithm upgrade for row count statistics: When the row count for a column-oriented table in table_info is inconsistent with the actual number of rows returned by queries, you can upgrade the statistics algorithm from V1 to V2 to improve the accuracy of row count statistics.

Severity	Title	Description	Affected version	Workaround
P2	Fixed a schema deduction error during the final stage of a multi-stage min/max_by aggregation when AQE is enabled.	When adaptive execution was enabled, the result type was incorrectly inferred during the final merge stage of a multi-stage min/max_by aggregation. This issue caused operations like COALESCE to fail with a `CASE expression should have the same type` error.	V3.1	Disable adaptive execution: `set hg_experimental_enable_adaptive_execution = off;`.
P2	Fixed a coredump issue caused by ORC DataBuffer resizing.	When reading data in ORC format, resizing the DataBuffer could trigger a coredump.	V2.0	Upgrade to this version.
P2	Fixed incorrect estimation in Auto Analyze for MaxCompute partitioned tables with archive storage partitions.	When a MaxCompute partitioned table contained more than 100 archive storage partitions, Auto Analyze produced incorrect statistical estimates.	V3.0 and later	Run ANALYZE manually.