[
https://issues.apache.org/jira/browse/ARROW-13532?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michal Nowakiewicz updated ARROW-13532:
---------------------------------------
Component/s: C++
Fix Version/s: 6.0.0
> [C++][Compute] Join: add set membership test method to the grouper
> ------------------------------------------------------------------
>
> Key: ARROW-13532
> URL: https://issues.apache.org/jira/browse/ARROW-13532
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Michal Nowakiewicz
> Priority: Major
> Fix For: 6.0.0
>
>
> Hash table used in group by provides one main method: map.
> This method will find an existing matching key in the hash table and output
> the corresponding group id, it the key already has been inserted in the hash
> table. Otherwise it will insert a new key and assign a new group id value to
> it.
> This interface is tailored for the group by. In order to reuse the same hash
> table implementation in join, there must be a way to skip insertion of new
> keys into the hash table when looking up existing keys. When join processes
> probe side it needs to filter input rows based on finding a match in the hash
> table, but keeping hash table immutable and not automatically adding missing
> keys to it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
