Improve the side input materialization for the DirectRunner/ULR from iterable to storing the multimap directly
- Dominant language
- Java
- Stars
- 8.7k
- Forks
- 4.7k
- Avg merge
- 2d 5h
- Merged PRs (30d)
- 204
Description
https://github.com/apache/beam/pull/4011 migrated to using a multimap as the materialization format for side inputs.
The migration used a trivial multimap -\> iterable -\> multimap conversion within the DirectRunner for first pass implementation purposes. Note that this is no different then the current materialization from a performance perspective it just moves this logic within the purview of the runner.
Imported from Jira [BEAM-3080](https://issues.apache.org/jira/browse/BEAM-3080). Original Jira may contain additional context.
Reported by: lcwik.
Contributor guide
Research direction
Start by reading PR 4011 and tracing the DirectRunner/ULR side-input materialization path. Compare the current multimap-to-iterable-to-multimap conversion with storing the multimap directly; done means the conversion is removed without changing side-input behavior or materialization semantics.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java
- Domain
- distributed-systems
- Issue type
- Refactor
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100