[SPARK-27424] Joining of one stream against the most recent update in another stream

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.0
• Component/s: Structured Streaming
• Assignee: Unassigned
• Reporter: Thilo Schneider
• Created: 10/Apr/19 06:45
• Updated: 16/Mar/20 22:51

Description

Currently, adding the most recent update of a row with a given key to another stream is not possible. This situation arises if one wants to use the current state, of one object, for example when joining the room temperature with the current weather.

This ticket covers creation of a stream_lead and modification of the streaming join logic (and state store) to additionally allow joins of the form

SELECT *
FROM A, B
WHERE 
    A.key = B.key 
    AND A.time >= B.time 
    AND A.time < stream_lead(B.time)

The major aspect of this change is that we actually need a third watermark to cover how late updates may come.

A rough sketch may be found in the attached document.