[SPARK-13720] SQLBuilder.toSQL is broken when there is a sort by after having

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Incomplete
• Component/s: SQL
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: Wenchen Fan
• Created: 07/Mar/16 11:59
• Updated: 21/May/19 04:36
• Resolved: 21/May/19 04:36

Description

parse this SQL "SELECT COUNT(value) FROM parquet_t1 GROUP BY key HAVING MAX(key) > 0 SORT BY key" to logical plan and generate SQL string again, the generated SQL string can't be parsed.

This is because, we will add extra `SubqueryAlias` while converting logical plan to SQL string(the `RecoverScopingInfo` rule in SQLBuilder). However, we only update qualifiers in the `Project` operator that just above the added `SubqueryAlias`, logically we also need to update qualifiers in all ancestor plans. For example, if there is a Project under a Sort, if we add `SubqueryAlias` under Project, we need update qualifiers for Sort too.

Issue Links

• links to: [Github] Pull Request #11555 (cloud-fan)