You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@arrow.apache.org by "Vibhatha Lakmal Abeykoon (Jira)" <ji...@apache.org> on 2022/09/06 01:32:00 UTC

[jira] [Created] (ARROW-17622) [C++] Order-aware non-sink Fetch Node

Vibhatha Lakmal Abeykoon created ARROW-17622:
------------------------------------------------

             Summary: [C++] Order-aware non-sink Fetch Node
                 Key: ARROW-17622
                 URL: https://issues.apache.org/jira/browse/ARROW-17622
             Project: Apache Arrow
          Issue Type: Improvement
          Components: C++
            Reporter: Vibhatha Lakmal Abeykoon
            Assignee: Vibhatha Lakmal Abeykoon


Considering the existing sink nodes and newly introduced Fetch node with sort capability, we will only need two nodes, "sort", and "fetch" in the long run. Because once the ordered execution is integrated, some features could be removed. Right now, there are three nodes doing somewhat closely related things which is redundant work assuming unordered execution. Namely they are, "order_by_sink", "fetch_sink", and "select_k_sink". So one of them will need to go away at some point and all of them will no longer be sink nodes and sorting behavior will need to be removed from "fetch".

The task breakdown needs to be determined. Better to keep a few sub-tasks. 



--
This message was sent by Atlassian Jira
(v8.20.10#820010)