[
https://issues.apache.org/jira/browse/HIVE-17576?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16193858#comment-16193858
]
Mithun Radhakrishnan edited comment on HIVE-17576 at 10/5/17 10:44 PM:
-----------------------------------------------------------------------
Hey, [~owen.omalley]. I'm trying to move the {{ReflectiveProgressHelper}} to
the shims layer, as per your suggestion. It looks like I'll have to add a
{{ProgressHelper}} interface to {{HadoopShims}}, and a
{{ReflectiveProgressHelperImpl}} in {{HadoopShimsSecure}}. It's beginning to
look a little convoluted, to go through a Shim, and then reflection, for this
one call. Add to this the fact that we're mixing concerns by putting Tez stuff
in {{HadoopShims}}.
Edit: We should also consider that the reflection will likely be changed to a
direct call, once Hive {{branch-2x}} is switched to a more current Tez version.
I can post a patch to illustrate, but I'm beginning to wonder if this is
overkill. :/
was (Author: mithun):
Hey, [~owen.omalley]. I'm trying to move the {{ReflectiveProgressHelper}} to
the shims layer, as per your suggestion. It looks like I'll have to add a
{{ProgressHelper}} interface to {{HadoopShims}}, and a
{{ReflectiveProgressHelperImpl}} in {{HadoopShimsSecure}}. It's beginning to
look a little convoluted, to go through a Shim, and then reflection, for this
one call. Add to this the fact that we're mixing concerns by putting Tez stuff
in {{HadoopShims}}.
I can post a patch to illustrate, but I'm beginning to wonder if this is
overkill. :/
> Improve progress-reporting in TezProcessor
> ------------------------------------------
>
> Key: HIVE-17576
> URL: https://issues.apache.org/jira/browse/HIVE-17576
> Project: Hive
> Issue Type: Bug
> Affects Versions: 2.2.0, 3.0.0, 2.4.0
> Reporter: Mithun Radhakrishnan
> Assignee: Mithun Radhakrishnan
> Attachments: HIVE-17576.1.patch, HIVE-17576.2-branch-2.patch,
> HIVE-17576.2.patch
>
>
> Another one on behalf of [~selinazh] and [~cdrome]. Following the example in
> [Apache Tez's
> {{MapProcessor}}|https://github.com/apache/tez/blob/247719d7314232f680f028f4e1a19370ffb7b1bb/tez-mapreduce/src/main/java/org/apache/tez/mapreduce/processor/map/MapProcessor.java#L88],
> {{TezProcessor}} ought to use {{ProgressHelper}} to report progress for a
> Tez task. As per [~kshukla]'s advice,
> {quote}
> Tez... provides {{getProgress()}} API for {{AbstractLogicalInput(s)}} which
> will give the correct progress value for a given Input. The TezProcessor(s)
> in Hive should use this to do something similar to what MapProcessor in Tez
> does today, which is use/override ProgressHelper to get the input progress
> and then set the progress on the processorContext.
> ...
> The default behavior of the ProgressHelper class sets the processor progress
> to be the average of progress values from all inputs.
> {quote}
> This code is -whacked from- *inspired by* {{MapProcessor}}'s use of
> {{ProgressHelper}}.
> (For my reference, YHIVE-978.)
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
