Parallel CREATE INDEX for GIN indexes

[grantzhou]

--- As of 2026-01-26T23:33:55.000Z ---

Tomas Vondra fixed two issues in parallel GIN index creation. First, he corrected tuplesort_begin_index_gin() which was incorrectly using b-tree operators instead of GIN opclass comparison functions for sorting. Second, he added regression test coverage for parallel GIN builds by modifying existing tests to create GIN indexes in parallel, improving code coverage significantly. Both patches were pushed and backpatched to PostgreSQL 18. The discussion confirmed that the original comment about "serially executing a scan that was planned to be parallel" was incorrect and not applicable to current implementation.

Tomas Vondra修复了并行GIN索引创建中的两个问题。首先，他纠正了tuplesort_begin_index_gin()函数，该函数错误地使用了b-tree操作符而不是GIN操作类的比较函数进行排序。其次，他通过修改现有测试来并行创建GIN索引，为并行GIN构建添加了回归测试覆盖，显著提高了代码覆盖率。两个补丁都已推送并反向移植到PostgreSQL 18。讨论确认了关于"串行执行计划为并行的扫描"的原始注释是错误的，不适用于当前实现。

participants - 参与者:

• boekewurm+postgres@gmail.com
• hlinnaka@iki.fi
• michael@paquier.xyz
• reshkekirill@gmail.com
• tomas@vondra.me