MogDB 5.0.9

1. Version Description

MogDB 5.0.9 is a patch version of MogDB 5.0.0, released on 2024-10-31. It adds a new feature and fixes some defects based on MogDB 5.0.8, with the following details:

2. New Features

Statement Level PLSQL Function Cache Support

When executing a query, if the statement contains a PLSQL function, the execution of the function will take up a lot of time. If the function is stable/immutable and the function parameters are repeated, caching the result can greatly improve the query efficiency.

This feature supports that in the process of SQL execution, for the first time encountered parameters of stable/immutable type function, the result will be put into cache after calculation, and then if the same parameters are encountered, the result can be taken out from cache directly, which can save the calculation process and improve the query performance.

Related Page: Statement Level PLSQL Function Cache Support

3. Defect Fixes

1. [9464] Fixed a low probability downtime issue due to inappropriate handling of the guc parameter when comparing because of the occurrence of the name NULL scenario.

2. [9431] Fixed the issue that queries with fetchsize are not automatically committed after select-o auto-transaction commit is enabled.

3. [9355] Fixed the issue that MogDB will fail to upgrade again after helicoptering from MogDB 3.0.x to MogDB 5.0.7 and later and performing a rollback operation during the upgrade.

4. [9178] Fix a bug that MogDB will go down when creating a partition table, creating a function, creating a trigger, inserting at least 256M data, and performing an update operation after the compression is completed.

5. [9296] Fix a memory overflow issue when querying a large table with more than 10GB of variable-length fields after single-column sorting.

6. [6889] Fix the problem that autovacuum is not done for INSERT-only tables.

   Related Page: autovacuum_vacuum_insert_threshold, autovacuum_vacuum_insert_scale_factor

7. [6740] Fixed memory buildup in calls to SELECT * FROM pg_logical_slot_peek_changes('xxx_slot',NULL,1024,'include-timestamp','1','skip-empty-xacts','1'); which would result in memory buildup. issue.

8. [6130] Fixed the issue that executing insert operation on partitioned table while executing vacuum full leads to deadlock.

9. [5642] Fixed the issue that using aggregate function will report error when the result set of order by is empty in B mode.

10. [6357] Fix the issue that after gs_dump exports data, some tables cannot be restored due to the B-mode does not support default(-1) way to set the default value of columns.

11. [6472] Fix the issue that the amount of data queried in parallel is inconsistent with the amount of data queried out serially when querying data using indexes in descending order with ustore scenario enabled.

12. [6785] Fix the problem that querying data with low probability after modifying the default values of columns under b-bank will lead to database downtime.

13. [6048] Fixed a security vulnerability that allowed authenticated database users to execute arbitrary code due to lack of overflow checking during SQL array value modification.

14. [5825] Fixed an issue where gs_dump reports an error when exporting tables with self-incrementing primary keys after renaming the user and schema.

15. [6166] Fix the issue that gs_probackup does not report error reasonably when specifying the wrong format of j parameter, j=0, during backup.

16. [9081] Fix the problem that after enabling char_coerce_compat parameter, the length of nvarchar2 type is judged in a wrong way.

17. [9122] Fix the problem of case sensitivity when gs_dump supports package, type, trigger export.

18. [9267] Fix the problem of inconsistency of data types returned by union all versions in dolphin.

19. [9453] Fix the problem that select current_schema() is not supported in B-compatible mode.