[
https://issues.apache.org/jira/browse/ARROW-17837?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sasha Krassovsky updated ARROW-17837:
-------------------------------------
Component/s: C++
Description:
As we begin to include more and more plan-global stuff (such as Scheduler,
AsyncScheduler, IO thread pool, ExecContext), ExecPlan's definition is becoming
cluttered. ExecPlan should store and organize nodes, and store these structures
elsewhere. We thus introduce QueryContext which will store any information
related to the currently executing plan.
An added bonus is that now more components can be decoupled from ExecPlan and
rely only on QueryContext (useful for e.g. testing, or just reusing a component
in a non-ExecPlan context).
Also, this QueryContext can introduce options for the query such as maximum
memory used and this legacy batching thingy.
was:
As we begin to include more and more plan-global stuff (such as Scheduler,
AsyncScheduler, IOScheduler, ExecContext), ExecPlan's definition is becoming
cluttered. ExecPlan should store and organize nodes, and store these structures
elsewhere. We thus introduce QueryContext which will store any information
related to the currently executing plan.
An added bonus is that now more components can be decoupled from ExecPlan and
rely only on QueryContext (useful for e.g. testing, or just reusing a component
in a non-ExecPlan context).
Also, this QueryContext can introduce options for the query such as maximum
memory used and this legacy batching thingy.
> [C++] Create ExecPlan-owned QueryContext that will store a plan's shared data
> structures
> ----------------------------------------------------------------------------------------
>
> Key: ARROW-17837
> URL: https://issues.apache.org/jira/browse/ARROW-17837
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Sasha Krassovsky
> Assignee: Sasha Krassovsky
> Priority: Major
> Labels: pull-request-available
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> As we begin to include more and more plan-global stuff (such as Scheduler,
> AsyncScheduler, IO thread pool, ExecContext), ExecPlan's definition is
> becoming cluttered. ExecPlan should store and organize nodes, and store these
> structures elsewhere. We thus introduce QueryContext which will store any
> information related to the currently executing plan.
> An added bonus is that now more components can be decoupled from ExecPlan and
> rely only on QueryContext (useful for e.g. testing, or just reusing a
> component in a non-ExecPlan context).
> Also, this QueryContext can introduce options for the query such as maximum
> memory used and this legacy batching thingy.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
