[C++] MapGenerator should optionally forward reentrant pressure
- Dominant language
- C++
- Stars
- 17.1k
- Forks
- 4.3k
- Avg merge
- 3d 13h
- Merged PRs (30d)
- 88
Description
Currently the map generator will allow the map function to run in parallel but it will not forward reentrant pressure onto the source generator. Instead it queues requests.
In some cases this is the right decision (if source is not async reentrant) but in some cases we want it to forward the pressure (so that the entire chain can run in parallel).
By making it an option when the mapped generator is created we can allow pressure to be forwarded where appropriate.
Phrasing it another way. If we have source, map function A, map function B, map function C, and then a reentrant pull we would currently only run C in parallel.
**Reporter**: [Weston Pace](https://issues.apache.org/jira/browse/ARROW-13155) / @westonpace
#### PRs and other links:
- [GitHub Pull Request #10795](https://github.com/apache/arrow/pull/10795)
**Note**: *This issue was originally created as [ARROW-13155](https://issues.apache.org/jira/browse/ARROW-13155). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*
Contributor guide
Research direction
Start by reviewing the MapGenerator creation path and the linked pull request #10795, then trace how reentrant pulls move from the mapped generator to its source. Done means an option controls whether reentrant pressure is forwarded through the source/map A/map B/map C chain instead of being queued.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- data
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100