[HUDI-3221] Support querying a table as of a savepoint

• Type: New Feature
• Status: Closed
• Priority: Blocker
• Resolution: Done
• Fix Version/s: 0.11.0
• Component/s: hive, reader-core, spark, writer-core
• Labels: pull-request-available user-support-issues
• Story Points: 1
• Epic Link: Hudi Spark SQL
• Assignee: Forward Xu
• Reporter: Ethan Guo

Reviewers

Shiyan Xu, Yann Byron

• Created: 11/Jan/22 22:58
• Updated: 09/Mar/22 05:00
• Resolved: 09/Mar/22 05:00

Description

Right now point-in-time queries are limited to what's retained by the cleaner. If we fix this and expose via SQL, then it's a gap we close.

Dataframe read path support this option but not for SQL read path

https://hudi.apache.org/docs/quick-start-guide/#time-travel-query

SparkSQL Syntax

// code placeholder
SELECT * FROM A.B TIMESTAMP AS OF 1643119574;
SELECT * FROM A.B TIMESTAMP AS OF '2019-01-29 00:37:58';
SELECT * FROM A.B VERSION AS OF 'Snapshot123456789';

Issue Links

• duplicates: HUDI-109 RFC-7 Ability to query older snapshots of data
• Testing discovered: HUDI-3574 Improve maven module configs for different spark profiles
• links to: GitHub Pull Request #4720
• links to: GitHub Pull Request #4885
• links to: related GH issue