[FLINK-20670] Support query hints for Flink SQL

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Table SQL / API
• Assignee: Unassigned
• Reporter: Shuo Cheng
• Created: 18/Dec/20 07:17
• Updated: 10/Jan/22 14:00

Description

Now Flink has supported dynamic table options based on the Hint framework of Calcite. Besides this, Hint syntax is also very useful for query optimization, as there is no perfect planner, and query hints provide a mechanism to direct the optimizer to choose an efficient query execution plan based on the specific criteria.

Currently, almost most popular databases and big-data frameworks support query hint, for example, Join hints in Oracle are used as following:

SELECT /*+ USE_MERGE(employees departments) */ *
FROM employees, departments 
WHERE employees.department_id = departments.department_id; 

Flink also have several join strategy for batch job, i.e., Nested-Loop, Sort-Merge and Hash Join, it will be convenient for users to produce an efficient join execution plan if Flink supports Join hint. Besides Join in batch job, it's also possible to use join hints to support partitioned temporal table join in future. In a word, query hint, especially join hint now, will benefit Flink users a lot.

Issue Links

• is related to: FLINK-25596 Specify hash/sortmerge join in SQL hint
• is related to: FLINK-25595 Specify hash/sort aggregate strategy in SQL hint
• is related to: FLINK-22896 Flink sql supports parallelism at the operator level

Sub-Tasks

• Introduce partitioned lookup join to enforce input of LookupJoin to hash shuffle by lookup keys: In Progress