[SPARK-48588] Fine-grained State Data Source

• Type: Epic
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 4.0.0
• Component/s: Structured Streaming
• Labels: pull-request-available
• Epic Name: Fine-grained State Data Source
• Assignee: Unassigned
• Reporter: Yuchen Liu
• Created: 11/Jun/24 19:25
• Updated: 10/Jul/24 17:48

Description

The current state reader API replays the state store rows from the latest snapshot and newer delta files if any. The issue with this mechanism is that sometimes, the snapshot files could be wrongly constructed, or user want to know the change of state across batches. We need to improve the State Reader so that it can handle a variety of fine-grained requirements. For example, reconstruct a state based on arbitrary snapshot; support CDC mode for state evolution.

Issue Links

• links to: GitHub Pull Request #46944